Laser Kids

. : X-Charge (cross charge) waveguide CO2 laser : .

Cross charge (x-charge) laser design is rather widely known and used. A good example of its application is technology laser LANTAN. Its schematics is given at the figure below (cited from [1]).


The scheme of gas-discharge chamber of "LANTAN-2M" laser. 1 - the body of the chamber, fiberglass plastic; 2 - ionization electrode; 3 - ionization electrode terminal,; 4 - insulator (silica, glass); 5 - cathode of the main discharge, 6 - anode of the main discharge. (The picture is cited from [1]).

Another good example is a laboratory laser designed by authors of [2]. I think it makes sense to put its scheme here too (see the picture below).


Scheme of the pulsing ionization.

B) Discharge chamber. 1 - main electrodes (bronze); 2 - coolers (contain water); 3 - dielectric walls of the chamber (alund ceramic, 1 mm thick).

The idea of the scheme is that the barrier discharge is responsible for the ionization of the gas, so the main discharge can have non self sustainable form. (Reminder: Barrier discharge is that kind of gas discharge, where the electrodes are insulated from the plasma by a dielectric layer. Obviously this type of the discharge can exist only when being powered by alternating or pulsed current.)

In the more wide sense the scheme has two modes of operation:

A) The mode with self sustained discharge, preionized by a barrier discharge;

B) The mode, when the main discharge is completely incapable to self sustain and it is fully controllable by the external ionization formed by the barrier discharge.

In the mode A the scheme looks attractively to be applied in some TEA laser. But in practice the multiplication ratio (the value equal to the energy deposition of the main discharge divided by the energy deposition in the preionizing discharge) appears to be small and degrades seriously with minor additions of electronegative gases to the mixture, Moreover the main discharge strongly tends to form a sliding discharge along one of the preionizing electrodes. One can note that there exist methods of struggle those drawbacks, but if we apply them the scheme becomes complicated, requiring precise timing of the discharges, and still it remains very sensitive to the gas mixture contaminations. I can say for example, that I was unable to get any lasing in the x-charge lasers designs on air with CO2 mixtures (and even on nitrogen with CO2) at the pressures over 250 torr.

Another interesting field of application may be found among the high (average) power continuous wave (CW) or quasi continuous wave (QCW) lasers working at low pressure. These lasers have more appropriate conditions for the x-charge design. Apriori they use low pressure and helium. The helium is unavoidable if one hopes for the decent output power. It provides the gas mixture with the high thermal conductivity, needed for the cooling. The pressure should be low just due to internal laser circumstances: the lower the pressure - the higher is the upper laser level lifetime. In a certain pressure range the emission cross section rises with the reduction of pressure too. One and another allow us to reach the lasing threshold at the reasonable levels of pumping, when the laser gas mixture is still at reasonable temperature, and the laser chamber is still not molten. (This limitation has no effect in the short pulse (TEA) systems, where the megawatts of pumping power are on only for a few microseconds, and can not produce a sufficient overheat. On the other hand in CW and QCW systems each kilowatt of pumping power put the construction into heavier and heavier conditions, more and more hard to survive.)

Since the pressure is low and the gas mixture is full of helium, there are no principal difficulties for the use of the x-charge design. The suitable mode is B, moreover than A mode is just inapplicable here, due to the fact that the discharge glows continuously of for hundreds of microseconds, and these periods of time are more than enough for the contraction of the most uniform (at the initial stage) discharge.

Why one should seek such an exotic design for the continuous reduced pressure laser if we have ye olde tubular longitudinal scheme?

First of all the feeding voltage. The meter long longitudinal laser requires almost kilowatt of power supply at 15..20 kV for a commercial tube and at over 35..40 kV for the amateur one (due to the gas contamination). And the laser with the transverse discharge, having the width of 30 mm, promises to be operational at 1..2 kV. It means the use of affordable microwave oven transformer instead of the sophisticated high voltage unit. An additional bonus will be that one can slightly increase the working pressure and thus reduce the quality of the vacuum pumps being used, and also prolong the lifetime of the gas mixture. Need to say that the mentioned transversal discharge has the stability out of its distinguishing features even at low pressures, It means that one needs some means of stabilization. Such as sectioned electrodes with resistive ballast, or the usage of non self sustain discharge type, exactly as it's offered here.

The last but not the least fact is that the geometry of the laser channel in the form of narrow gap between two heat conducting planar walls is well suited for the gas cooling. Let's for example take a common tube-shaped laser: the ratio of the temperature difference deltaT between the center and the wall divided by the total power Q of the uniformly distributed heat source Q (in other words the thermal resistance of the cylindrical gas column) does not depend to the tube diameter and is equal to:

Rt = 1/(4*pi*L*lambda)

where lambda [W/(m K)] is the heat conductivity of the gas, pi=3.14 is the well known "pi" number, L is the length of the gas-discharge tube. (This formula is quite easy to be derived from the basic heat conductivity law, so there's no sense to make reference to any sources).

The independence of the thermal resistance to the diameter of the cylinder may be comprehended from the next considerations: the area of the cooling surface grows with the increasement of the diameter, but the distance for the heat to pass does grow with the same rate. As the result the total resistance does not change. If we now fix the top allowed gas temperature Tmax, we'll get that the top allowed energy deposition should not exceed the next value:

Qmax = (Tmax - Tout)*(4*pi*lambda)*L/(1-eff)

where Tout - is the temperature of the (outer cooled) wall, eff - the efficiency of the laser.

The obtainable laser output will then have the next form:

Wmax = eff*Qmax = eff/(1-eff)*(Tmax - Tout)*(4*pi*lambda)*L

This relation explains the well known limitation of the (diffusion cooled) CO2 lasers: "the output power depends only to the length of the tube, and the factor of this dependency describes the quality of the laser, but generally does not exceed some limiting value."

One may make the numeric estimations. First of all lets neglect the dependency of the efficiency to the gas temperature and assume that eff = 0.1 = 10%. Then we'll postulate that the gas heat conductivity is determined by the buffer gas only. It may be helium (lambda = 0.147 W/(m K)) or nitrogen (lambda = 0.0315 W/(m K)) or either argon (lambda = 0.0164 W/(m K)). Lets take that Tout = 20oC and Tmax = 200oC (the people's opinions on the Tmax differ strongly, but here we take rather "classic" value [3]):

buffer gas He N2 Ar
Wmax/L W/m 37 8 4.1

The layout is totally different in case the active zone of the laser has the shape of the planar plate (with thickness a, wideness b, and length L), being cooled from both sides. From the hat conductivity law one can get:

Rt = a/(8*b*L*lambda), or Rt = a/(8*S*lambda)

where S is the area of the projection of the plate "on the top view".

Hence that energy input into the plasma will be:

Qmax = (Tmax - Tout)*(8*S*lambda)/(1-eff)/a

and the laser yield:

Wmax = eff*Qmax = (Tmax - Tout)*(8*S*lambda)*eff/(1-eff)/a

The obtained formula can be interpreted as the proportionality of the laser output to the area of the surface being cooled. But one may go further and divide the obtainable laser power for the planar geometry by the obtainable laser power for the tube design:


(Tmax - Tout)*(8*S*lambda)*eff/(1-eff)/a     2*S/a   2*b*L
----------------------------------------  =  ----- = ------- = (2/pi)*(b/a)
eff/(1-eff)*(Tmax - Tout)*(4*pi*lambda)*L    pi*L    pi*L*a

I.e. the laser, having its active medium in a shape of thin planar layer, allows to get power from the unit of length greater by (2/pi)*(b/a) times than the one, obtainable from the common pipe-like laser.

The (2/pi) factor essentially describes the difference between planar and cylindrical shapes in the relation to the heat conductivity, and the (a/b) factor (also named as "aspect ratio") describes the elongations of the laser channel cross section.

For example if the thickness of the plasma column is 3 mm and the wideness is 3 cm, the aspect ratio will be equal to ten, and this tube will allow 6.3 times more power from the unit of length than the common cylindrical one. E.g. a meter long tube will allow to get 230 W of power in the contrast to 37 W from the common one. (Of course if other sources of losses won't interfere) One may comprehend the nature of this gain in the next way. Let's assume that we want a huge power from the unit of the laser length, but for some reason we have to stick to the cylindrical geometry. The readily coming solution is to take N laser tubes and to put it into one harness in parallel. The batch of tubes not necessary must be three dimensional. Lets put the tubes on the table top tightly one near another. The next step is to eliminate the walls between the adjacent tubes. We'll loose some some cooling (which otherwise would take place on that walls) but we will gain the integrity of the laser beam. Maybe it will have an uncommon shape (rectangular section) but still it will be the single beam, coherent to itself. On the contrary in case of N tubes we would have N independent beams with all related problems in focusing them into one point.

Trust me, I had no intention to bother anyone with considerations and derivations, but it was necessary to point out the essence, for one to understand the reasons forcing to test the lasers of uncommon shapes. Now this is done and we could proceed to the results of the tests, but one more notice before this. I should say at the very beginning that the task of "outperforming the commercial lasers" is not considered seriously here. Indeed I got used to obtain efficiencies being one order of magnitude lower than ones of the commercial grade devices. Since that I would be glad if my tubes (abusing the benefits of the planar design) could even approach to the parameters of the commercial cylindrical lasers.


1. Laser cells.

The idea and scheme of the laser cell are more than clear when looking at the pictures given above. Here we'll stop only on the materials and the specialties of the assemblage.

The dielectric walls of the cells were made of inorganic (window grade) glass, 2 mm thick. Such a glass is affordable from some frames for photos, or either it can sometimes be ordered in glass shops. In the latter case it makes things much easier, because one can also order the appropriate cutting.

The smaller photo frames can contain some 1 mm thick glass, but their size is less than 200 mm, and if we used such a glass, we'd have to make segmented walls, hard to be sealed vacuum tight.

The glass walls are 400 mm long (made of photo frame glass 400 x 300). The width of the glass is determined from the sizes of the electrodes and the value of the spacing between them.

Two different types of electrodes have been made. The first pair of electrodes is made of T-shaped aluminium extrusion directly. Another pair of electrodes has more sophisticated design (initially intended to reduce the total size of the cell). Both types of the electrodes have the thickness of 3 mm. Of course all edges and angles were rounded, and the working surface was polished. The inter electrode spacing is 30 mm (the aspect ratio of 1:10).

cuvette01_electrodes_a cuvette01_electrodes

cuvette02_electrodes_a cuvette02_electrodes_b

With the sizes of the active area at 3 x 30 x 400 mm with commercial quality of the gases and assemblage one could hope to get about 80 Watts of output, but I expect only 8..10 W.

The vacuum cell, having the wall 30 x 50 cm, feels the atmospheric pressure of 120 kg. It looks like too much for the 2 mm thick glass. Therefore the glass walls of the cell are supported by some aluminium parts, having been glued upon the walls. These parts play the role of coolers and at the same time serve as the electrodes of the external ionization.

The coolers-ionizers are made of rectangular aluminium tubes, 25 x 10 mm cross sectioned, 380 mm long. The ends of the tubes are sealed off. It is supposed that at some point they can be filled with cooling water, provided that the cell could live long enough to achieve the mentioned point of necessity. The sharp edges and angles are rounded as usual. The coolers-ionizers are glued to the glass walls using a silicone sealant, in the hope that it will provide some freedom for the parts to move, and thereby, will reduce the tension, arising due to the different heat expansion of the glass and aluminium.


The endpieces. Somebody among the great physicists has said: "If the God has created the volume, the Devil has created the surface." Being applied to our problems it means that "Devil has created the ends". It is almost easy to seal off the joint of the electrodes with the glass walls, but You will be very unhappy with the endpieces.

The cells are meant to be sealed with silicone caulk, and thus one should choose the material for the endpieces more carefully. The silicone sticks to the glass well. It is also good for aluminium/ However the endpieces are hard to be made as a single piece of inorganic glass. And metals are bad choice too, due to the possibility to shorten the discharge. On the other hand organic glass, polystyrene, PVC or polyethylene are prone to detach easily when glued with silicone. Fortunately it was found out that epoxy molds are very compatible with silicone, so the endpieces were made of epoxy resin.

mirr_mount01 mirr_mount02

mirr_mount03 mirr_mount04

The endpieces and mirror mounts have an unusual shape. It is due to the specific shape of the laser cell and channel. As it was said above the endpieces are made of epoxy resin and the mirror mounts are made of organic glass 10 mm thick.

Finally all this husbandry was glued up into a complete construction.


cuvette02b cuvette01a

cuvette01b cuvette02a


Interlude 1. Technological. VACUUM TIGHT SEALING.

The "adult" technique knows only three ways of permanent vacuum sealing: those are welding, soldering and brazing. All types of flanges and rubber gaskets are apriori considered as leaking and aging. The question is only at which rate.

Welding brazing and soldering require the device to be assembled in the glass or metal vessel. The first requires a hardcore glassworking. The second not only requires machining, but also rises problems with electric strength. Both variants require some solution to entry the electric current into the sealed volume. It may be metal-ceramic terminals, glass-metal solds and so on. All those things are expensive, rare and hard to make at home.

A DIY'er has to rely upon temporary solution an reconcile to some leaking. In this area the choice is quite wider. One may use glues, plastics, sealants etc. The most known and widely used ways of sealing are the next three ones:

  • Sealing with a hot glue ("glue-gun", low molecular polyethylene)
  • Filling with an epoxy resin
  • Filling with a silicone caulk.

One can also remind that there also exist sealing of rubber parts with a rubber glue and sealing metal parts with tin, but those methods are rarely met in practice.


• Hot glue works best with "fat" plastics: polyethylene, polypropylene, PVC, etc. The fluoroplastic is a sad exclusion. The hot glue works good also with metals (aluminium, copper, steel, including the galvanized one). However in order to get good results with metals one must literally blanch them with the hot glue, baking it in by some torch or a heat gun. Gluing the inorganic or organic glasses gives good results too. In case of inorganic glass one should bake in the hot glue too. Be careful in this case for not to crack the glass. Gluing of epoxy molds or textolite is also good.

Being a fat hydrocarbon the low molecular polyethylene dissolves grease, and thus is tolerant to some grease contamination. In the reasonable amounts of course. Some oil dipped parts are very unlikely to be able to be glued together.

The hot glue is good for vacuum or low aggressive gases. It can work underwater. Much worse it endures alcohols and acetone (flakes off in several weeks or even days). Petrol, kerosene and gear oil do dissolve the hot glue. (Not instantly, but for a long time operation with these liquids the hot glue is inappropriate.)

It develops brittleness under the sun rays or even after a long storage. Due to the appearing fractures the lasers, having been sealed off by the hot glue, loose the vacuum tightness in two-three years.

In the home conditions the hot glue is exclusively good for sealing. Small vessels, several cubic centimeters in volume, can be sealed to have leaks less than 100 torr per year. However it usually requires several attempts.

When heated it is able to self cure the leaks. If the seam does not allow the hot glue to get into the vessel, one can evacuate the vessel and heat the seam. In this case the hot glue can self cure small leaks.


• Epoxy resin. Sticks excellently to itself or to materials based on it (textolites, epoxy plastics). Not able to glue "fat" plastics (polyethylene, mylar, PVC, etc). Poorly holds the inorganic glass, even if its surface was grinded. Better, but still poorly, glues the organic glass. The results with metals depend strongly to the type of the metal being glued and to the type of the epoxy resin (look at the writings on its box if it is intended to glue the chosen metal) and to the condition of the surface. Before gluing the metal surface should be grinded or, better, grooved. Gluing the mirror like polished metals is the way to fail.

The most special property of the epoxy resin is that it is able (on the contrary to the hot glue or silicone) to stick to zinc selenide parts vacuum tightly. So if You are using selenide mirrors or window_s You may have to use the intermediate parts made of epoxy or textolite. The ZnSe things will then be glued to those intermediate parts by means of epoxy resin and the intermediate parts in their own turn will be glued to the rest of the construction by any other sealant of Your choice.

Epoxy resin is intolerant to grease. In fact most of failures are oftenly explained as the affect of the contamination. Indeed the grease is the best excuse for the resin manufacturers. From the practice of glass silvering we know, that in order to completely degrease the surface one should boil it for several hours in a concentrated nitric acid. Rare parts are able to endure that so the excuse works almost always.

Epoxy resin is capable to hold vacuum and gases. Most of the epoxy resins (excluding the ones, specially intended for wet conditions) are prone to flake off when used in water or high humidity. Even more fast the flaking takes place under the action of alcohols. Epoxy resins tend to soften when exposed to acetone for a long time (several days). This fact can be used to disassemble the connections, having been glued with epoxy. Epoxy resins are usually durable to the action of gasoline and mineral oils. However no warranties here.

The connections, having been glued by epoxy resins have exclusively long lifetime. Most of them continue to work even after 10 years of use. However the seam is rather brittle. And if it is exposed to thermomechanical loads or bendings, its lifetime shortens seriously.

My personal experience says that the results of epoxy resins usage for the tasks of vacuum tight sealings give average results. I've heard about some commercial He-Ne (or even argon) lasers, where the mirrors were sealed with epoxy. Evidently they use some very exotic resin or very special gluing technology. On my memory the best epoxy glued things had leakages of several torrs a day. It is understandable that there's no sense to speak about years.

The seam is hard, brittle and is not capable for self-curing.


• Silicone caulk. Ideally suitable for gluing inorganic glass. Also suits good for aluminium. Completely incapable to stick to the "fat" plastics (polyethylene, mylar, PVC, etc). With other materials it gives unpredictable (and usually bad) results. For example silicone caulk flakes happily from organic glasses. Recently I've been able to discover that the adhesion of silicone to organic glass (and many other materials) can be highly improved if the organic glass surface was painted by a rhodamine marker before gluing up. It is still unclear whether the rhodamine itself is responsible for the phenomenon, or either the surface active additives do that. It also was found out that silicone sticks well to epoxy molds. However it is still unclear how long such a connection will live. Silicone caulk (as like as epoxy resin) completely hates grease. So the surfaces to be connected must be cleaned thoroughfully.

Silicone is durable to water, vacuum and gases. Longer than other (known to me) sealants endures alcohols and acetone (however it finally fails too). Automotive silicones are durable to gasoline and oils.

Silicone seam has usually excellent lifetime. It can keep integrity for tens of years, A fly in the ointment is that it keeps its integrity all by itself. Apart from the parts that've been sealed. Silicone is prone to flake off anything but inorganic glass, With other materials the moment, when it decides to do it, is unpredictable. Anyways none of my lasers (the ones having been sealed with silicone) have stayed vacuum tight longer than a year.

In the short time perspective silicone gives an illusion of an ideal thing for sealing. The vessels, having been sealed with silicone, leak exclusively slow (it is easy to get less than 1 torr a day). With some experience one can easily succeed in vacuum tight sealing from the first attempt.

Silicone has a property to self-cure the leaks. It works in the next way: there is a little hole somewhere, and there is a silicone snot above it. Certainly when the pressure is lowered the snot is sucked to the hole and shuts it. Not always this is for good. Because if You shake the laser all leaks can open once again. And it is double pain to hunt for something unstable.

Generally silicone is good for the projects of the "make-work-dispose" style. Only one fact complicates everything: the silicone cures for indecently long time. Thick seams with restricted air access can cure for several weeks. Even longer they can contaminate atmosphere in the laser by their fumes.


And finally some general recommendations:

  • The number of places for sealing should be as low as possible.
  • The seams should be as short as possible.
  • The seams should have a simple and closed shape.



  • A mirror on one of the ends of laser. Or a looking window on its side. The total length of the seam is small and it has a shape of closed ring. This sealing is usually easy and produces good results.
  • A glued up box having a shape of a rectangular parallelepiped. The total length of the seams is huge. The corners of the box are the places, where three seams meat each other - too complicated shape. And also the shape of the seams is open (one usually can not glue all the box in one action).
    As the result this is ne of the most complicated designs to be sealed off. There's low probability of the success. The leaks are usually situated in the corners.
  • Planar (low inductance) terminals of nitrogen lasers, The seam is long, but has a simple shape. In some designs it can also be closed. Result: the terminals are usually easy to seal and provide much less problems than
    the endpieces.


2. Power Supply Unit.

The power supply unit is made following the KISS principle ("keep it simple").

It consists of:

  • microwave oven transformer
  • electronic transformer for halogen lamps (ye olde Feron 250 W rated)
  • two ignition coils with closed core (E - shaped).
  • a head from some small milliammeter
  • a bunch of diodes for assemblage of a diode stack (note that here's no need to rectify the voltage from the ignition coils, it is used for ionizing and should stay alternating)

There was also used a ballast resistor in the "low voltage" feeding circuit. There are also some kilowatt rated halogen lamps in series with mains cords - the standard precaution for works with high power devices fed by mains. One can use fuse here, but it would bother to replace them each time.


WARNING! It should be mentioned that the output ratings of the microwave oven transformer are similar to ones of the electric current used in the electric arm chairs. The only difference is that electric arm chairs use direct current, while here we have an alternating one. However this difference will mean nothing if one touches the output terminals.


Yes, lethal. And still handy, robust, cheap and affordable when we speak about a source rated to several kilovolts and several hundreds of milliamperes.

The scheme of connections is the next:

                                double horned
                                  ignition +-------+
            +-------------------+   coil   |       |
   o--------|         )|        |---o--- |(        |  laser channel
  ~ 220 V   |  Feron  )|( 16    |       )|(      =====
            |  250 W  )|( turns |       )|(   +--)   (--+
   o--------|         )|        |---o--- |(   |  =====  |
            +-------------------+          |  |    |    |
                                           +--)----+    |
  microwave                       Rb          |         |
    oven           8 x HER308     6k8         | Cp 4 nf |
 transformer   ---|>|--...--|>|--/\/\/--------+---||----+
   4kV peak ||(                                         |
   o------- ||(                                         |
           )||(                                         |
           )||(                                         |
   o------- ||(     0..100 mA                           |
            ||(        /                                |

The microwave oven transformer has the output of 4 kv peak. (This sample has the said voltage. Generally I've seen ones giving from 2.8 kv to 4.3 kV.) A shunt was designed specially for the milliammeter head to make its limit of measurement to be 100 mA (constant current). The resistance of the shunt appeared to be 2 Ohms. Here's no smoothing, at least for now, since in 50Hz pulsed mode we have lower average power and hence can avoid water.

At first there was no Cp capacitors. Those capacitors, connected in parallel with the laser cell - so called "peak capacitors", or shortly "peakers". Later it was found, however, that connection of the peakers to the laser cell allows to rise the top current, at which there's still no arcing (surge current).

Initially the ignition coil was also single. However the tests have shown a serious deficiency of the ionization. Increasing the ionization voltage was not acceptable due to breakdowns in the laser cell, so it was decided to reduce the series resistance of the transformer (ignition coil). The simplest way to do this is to connect two identical ignition coils in parallel. There were doubts if the coils would agree to operate in this mode, but the tests were successful (see below in the results section).

The general view of the power supply unit with and without the top plate with the laser cell installed are shown below.

mot_psu_with_laser1 mot_psu_with_laser2


Later it was understood that at the design pressures the laser was unable to reach threshold. And for the lower pressures the voltage of 4 kv microwave oven transformer is excessive. There could be one rated to 2.8 kv, but who knew it beforehand? After this the microwave oven transformer was connected not directly to the mains but through a variac. This provides an ability to finely tune the voltage and to approach smoothly to the surge current (the limit of arc formation.)


3. The Results.



The first of the laser cells (the one having foil terminals) was installed onto the initial variant of the power supply unit (the one with the single ignition coil, without voltage regulation by variac, and without peaking capacitors).

The gas discharge in the cell was tested. During the tests the cell was
filled by welding grade argon at different pressures.

  1. At argon pressures above 2.8"Hg (inches of mercury) and without any external ionization, the output voltage of the microwave transformer is just not enough to cause the breakdown in the cell. The current is zero.
  2. When the external ionizer is on, there appears a current in the "low voltage" part of the circuit. By gradual decreasing of the pressure and smooth approaching to the breakdown threshold (2.8"Hg) one can (rather reliably) achieve currents of two divisions of the milliammeter scale (10 mA).
  3. With the increasement of the pressure the current drops rapidly and it becomes immeasurably small already at 3.5"Hg.
  4. When reducing the pressure, somewhere between 2.5 and 2.7 inches of s mercury the current suddenly jumps (up to 25-40 mA) and the discharge contracts. One can note, however, that here the applied voltage exceeds the one of the cell breakdown.
  5. In parallel to the milliammeter (shunt resistance 2 Ohm) an oscilloscope was attached. The oscillograms show that the current through the laser cell is formed by series of bell-shaped pulses with their tops being flattened slightly. The duration of the pulses is 3.5-4.5 ms at the half height, the repetition frequency is 50 Hz (period of 20 ms). The amplitude of the current pulses is up to 60 mA, when the milliammeter shows two divisions.
  6. Addition of two peaking capacitors (2 pcs of Murata 2 nf x 40 kv doorknob capacitors, connected with wide stripes to make the inductance low) does not change anything by itself, but allows to obtain 1.5-2 times greater current.
  7. In the uniform glow mode (current as measured by milliammeter at 10 mA) the cell stays almost cool. It means if I was able to get lasing, I could do it without cooling water.
  8. The is no success in current increasement. Either we have that we have, or the discharge drops to arcing. And this is in the inert gas, where the lifetime of free electrons should be long, and the multiplication ratio should be top most.



The cell was equipped with mirrors. The rear one is the rearview automotive one. Aluminium on glass, concave with 2 m radius. The front mirror is commercial one, ZnSe coated by dielectric. Plano, 10 mm in diameter. Reflection is 94% (I have no larger mirrors with such a high reflection).

The cell was filled with mixture CO2:N2:He = 2:3:15 (or either 0.4:0.6:3, the share of molecular gases was one third).

The first thing that was noted - the drop of current. If one could achieve average current of 10 mA with argon or argon mixed with 10% of nitrogen, then here one could get only 5 mA without surging to arcing. (Here I use the term "average" current to designate the current as shown by milliammeter. If the shape factor was kept unchanged it means only 30 mA of peak current). Need to say that the helium was taken from a party balloon with unknown storage time in the shop. The nitrogen was welding grade, but it stood a long storage in an intermediate vessel (PVC ball). CO2, as usual, was taken from a Crossman's cartridge, and the quality of these cartridges does not grow year to year.

There were some runs on CO2:N2 = 2:3 mixture. With this mixture there was no range of operation at all. At higher pressures the current is immeasurable (less than 1 mA avg) and when trying to lower the pressure it immediately goes to arcing.

Nevertheless the laser was aligned and got a run. No lasing was observed. Neither in the mode of low current (I=5 mA avg) nor in the mode with arcing (I=40 mA avg).

In attempts of increasement of the preionization power the number of turns was increased in the winding that feeds the ignition coil (inside the Feron electronic transformer). As the result the voltage caused a breakdown of silicone near the cell endings. The charred silicone was removed and the remaining silicone was covered with epoxy.



Ionization voltage was returned to the previous level. And to increase the ionization power the second ignition coil was installed in parallel with the first one. Also the wires (the ones connecting ignition coils with the ionization electrodes) were replaced. Previously the automotive type high voltage wires were used. Their drawback is that they are designed specially to suppress radio interference and thereby can suppress the most useful part of our ionization current too. At their place just common copper wires were installed, that should reduce RF losses. Ballast resistor was also reduced (from 6.8 kOhm to 3.4 kOhm).

Test run with argon:N2 = 1:5 mixture:

  1. The glow inside the cell became much brighter. And not only the glow from the ionization source became brighter, but the brightness increasement given by turning on the main current became much more prominent.
  2. Average current (the top achievable without dropping to arcing) stayed unchanged. It was 10 mA and it remains 10 mA. It is especially incomprehensible when taking into account the increased glow.
  3. The corner of the laser cell (silicone sealant between the electrodes and endpieces) was charred again. After this an attempt was made to clean the ill place ant to fill it with epoxy. However it is clear that the error is in the design/ One should leave much greater reserve between the cooling - ionizing parts and the endpieces. Generally the cell should be redesigned.
  4. With a variac one can succeed in using the cell at higher vacuum (than the previous 3 inches of mercury) with corresponding reduction of feeding voltage. With decreasing of the pressure the allowable current decreases too but slower than the pressure. At one inch of mercury one could get to somewhere between 5 mA avg and 10 mA avg without arcing.



The cell with the foil terminals is now dead. the breakdown of the silicone near the endpieces appeared to be fatal. In attempts to increase current and drive the cell to lasing its glass wall was overheated and cracked.

fail01a fail01b

The cell with T-shaped terminals (the one with its electrodes being made directly of T-shaped aluminium extrusion) has agreed to lase. With the mixture of CO2:N2:He = 2:3:15 havin been mixed 31.01.2017. Resonator is also the same. Rear mirror is automotive rear view mirror (washed out from the protective paint), aluminium on glass, concave R=2 m, the front one is ZnSe plano, dielectric coated, ro = 94%, 10 mm in diameter.

The measured output power was 6 mV by the readings on the homemade Peltier calorimeter. Calibration of the calorimeter was checked once again but it constantly gives 6.4 mW/mV. It means the laser gave 38 mW of output. (Note that the integration spacing was 5 and the 10 mm mirror intercepts only one third of the laser channel width).

To increase the area of the resonator a hybrid front mirror was made. In a glass concave mirror (automotive one, R=2 m, aluminium coated) a 10 mm hole was drilled and then the hole was covered by the mentioned above ZnSe 10 mm 94% mirror. Care was taken to keep alignment between the concave glass part of the hybrid mirror and its ZnSe part.

hybrid_mirror_a hybrid_mirror_b

With the same gas mixture and with the new mirror the maximum output was 12.5 mV (80 mW), but the result is not directly comparable, because here the laser was allowed to go deeper into the mode with arcing.


The results of observations (it hardly can be called as "measurements"):

  • The pressure optimum is sharp and lies between 1.1 and 1.2 inches of mercury. At 1 inch and at 1.5 inches the output drops to one half of the optimal one. At 0.5 inch and below and at 2 inches and above no lasing was attained.
  • That lasing threshold is about 10 ma (by the indications of the milliammeter) and it depends weakly to the pressure (in the range of pressures, where the lasing was attainable). The threshold of arcing lies at the same 10 mA when the pressure is optimal. I.e. when completely without sparks there's no lasing too.
  • Going further into the area with arcing gives the optimal output at the average current of 20 mA. Even higher current leads only to drop of power. The arm of the milliampere here jumps strongly (due to the random arcing) so the correct readings are impossible. The variac appears to be set to
    100..110 V here.
    The highest laser output is when the arc is "running" (randomly appears in different places of the laser channel). When the arc sticks to a certain place on the electrodes, the laser output drops significantly.
  • The cell heats strongly. The tests were made without cooling water, and in these conditions the coolers become hot to touch (about 40oC) in less than a minute of operation.

To the end of the series of the tests something went wrong and the mode of "running arc" became unattainable. The output power of 12 mW became unattainable too. The arc has stuck to a point on the electrodes and refuses to leave the place. The top power in this mode was 4.5 mV only (28.8 mW).



To understand further reading we should discuss the structure of the laser cells in more details. The schematic section of cells in ASCII graphics is shown below:

           25 mm  2.5 mm
     |   ________   |
     |  [________]  |  
 ////)              (////  
     |  [________]  |
     |              |
     |    30 mm     |

For things to be more clear I place here a picture too (however the bad thing with the picture is that it is stored separately from the text file, and when You've downloaded the file for the offline reading at the very moment it may appear to be lost.)


The gap, designated as "delta" on the drawings has appeared here not accidentally. Its presence or absence affects the discharge shape strongly. In the pulsed lasers (not actually TEA but similar) the introduction of this gap allows to suppress sliding discharges along the ionizing electrodes, and thus to reach the lasing action at pressures up to 250 torr with mixtures of air with CO2. Without this gap the top pressure, at which the volumetric discharge still exists in this design in oxygen-containing mixtures, was below 40 torr.

When there was a hope to get a working laser at higher pressures this gap was excusable. But now, when one can see that the laser can not reach threshold even at 50 torr this gap may harm, because it leaves an area of non ionized gas that can cause improper breakdown when under the main voltage.

The gaps were covered with aluminium foil (put it above the glass of course).

The behavior of the system has changed. Previously the top pressure at which the uniform glow was attainable was 3 inches of mercury, and the attainable current was almost independent to the pressure and was equal to 10 mA (by the milliammeter gauge indications). Now the limiting current depends to the pressure an is equal to (with mixture CO2:N2:He = 2:3:15)

P Imax
0.5"Hg 5 mA
1.0"Hg 10 mA
1.4"Hg 15 mA
1.5"Hg 10 mA
1.6"Hg 5 mA


Above 1.6 inches of mercury there was no success in obtaining the uniform glow.

Tests for the lasing (resonator is the same as was in the record of 06.02.17):

  • The dependence of the output to the pressure has very sharp maximum. When we approach the top pressure, at which the uniform glow is still obtainable, the output power grows rapidly. When the arcing appears the power drops by 2-3 times, but the lasing is still present.
  • Somewhere a bit below 1.5 inches of mercury (probably 1.45"Hg - I cannot say precisely due to the lack of the vacuometer precision) the laser yielded 30 mV by the indications of the calorimeter (192 mW). When its output is focused by a lens (F=75 mm) the laser rapidly burns carbon paper and it is able to ignite a match (even when it is hold by hands - not very firm ones).



The output mirror (hybrid one) was replaced by another planar ZnSe mirror having reflectance of 85% and diameter 15 mm. It was installed alone - without an attempt to make another hybrid mirror.

At 1.45 inches of mercury the laser with this mirror is unable to reach the threshold. The top pressure, where the threshold is attainable is 1.1..1.2"Hg. The maximal power is 12 mV = 77 mW; the optimal pressure is 0.9..1"Hg.

It is interesting that the dependence of the power to the pressure has lost its sharpness and its maximum has shifted to the lower pressures. The maximum is noticeably lower than the limit of the discharge stability.

Then the rear mirror was replaced. A planar glass mirror coated with aluminium was installed. It is an old (ab)used mirror from a nitrogen laser that has noticeably darkened due to aging. The resulting laser power has dropped further on. It became 6..8 mV = 38..51 mW. Optimal pressure became even lower and the maximum itself became even more soft.

What do the words "optimal pressure became even lower" mean? In fact when the 94% hybrid mirror was replaced by 85% simple one the laser could not reach the threshold in the area of stable discharge (15 mA) but still could lase in the area of the discharge with some arcing (20 mA). With lowering the pressure one could return the threshold to the area of the stable discharge, but with the corresponding lowering of the output.

When the front mirror was replaced too, the threshold raised once again and could not be returned to the area of stable discharge by any changes of pressure. (Need to notice that the current border of the stable discharge shifts to lower values with the reduction of the pressure.) However the one could still move the threshold into an area, where the discharge contains some arcing (<20 mA). Of course it caused further drop of the output.

Also need to notice that the border of the discharge stability shifts with time. It is the lowest at the first turn on of the laser. It is about 10 mA. After some baking of the electrodes the border moves somewhere to 15 mA, sometimes it can reach 20 mA. If there appear a bright arc, it burns a spot on the surface of the electrodes and the border of stability drops to 10 mA again. Then the cycle repeats. (Here always we speak about the "average" current as indicated by milliammeter).

Further on an idea to remove peaking capacitors had stroke my head. The achievable current (without dropping to arcing) has decreased strongly (to about 7 mA) and refuses to rise back. Of course no lasing is here.

I.e. the peakers are essential for the laser operation. It may be due to they form a relaxation oscillator together with the ballast resistor and with the laser channel itself. Its frequency should be about 73 kHz, so the laser channel is actually fed by RF current.

Then an assembly of four K15-4 capacitors (4.7 nF x 12 kV each) was placed instead of the peaking Murata's. The area of the stable discharge just disappeared. When using the variac to increase the voltage, initially there is no discharge at all and then suddenly there appears white spark in the cell. The spark is fat, strong and has a hearable sound.

Then the K15-4 assembly was removed, and a single Murata (2nF) was placed in. The area of the stable discharge has reappeared. The lasing was even got there (4..6 mV) but the things were much worse than with two Murata's.

Then in parallel with the laser electrodes three Murata's were placed (1/RC = 49 kHz). The power became 21 mV = 134 mW. (One can suppose that if we return the previous mirrors there will be 400..500 mW, but it usually does not work in such a way).

The question on the optimal value of the peaking capacitors Cp should be studied further. It also has a sense to check how the laser will react to the capacitors of different types. E.g. to the single K15-4. Or to mylar film
rolled K73-13.



The results with different mirrors.


With the mixture of CO2:N2:He = 2:2:15

  • With the front mirror, plano, 85% reflectance; with the rear mirror plano, aluminium coated glass; and with 3 x 2 nF as the peakers the laser as it was already written above reaches 21 mV (134 mW)
  • With hybrid mirror as the front one (a hole in the center of Al coated glass concave mirror, enclosed with plano ZnSe 94% mirror); with the back mirror plano, Al coated glass; and with 3 x 2 nf peaker the laser yields 25 mV ( 160 mW)
  • With the hybrid mirror as the front one and with rear mirror concave, Al coated glass, and with 2 x 2 nf as the peaker the laser yields 30 mV (192 mW). All concave mirror were spherical with the curvature radius of 2 meters.


Further on (in the resonator having the front hybrid mirror and the rear one concave) the next gas mixtures were tested:

  • CO2:N2 = 1:2
    • the highest pressure, at which the lasing was observed 0.6"Hg
    • the optimal pressure 0.5"Hg
    • the highest current (by milliammeter) achievable without dropping to arcing: 3 divisions (15 mA)
    • threshold current (by milliammeter) a bit less than 3 divisions (~14 mA)
    • top registered output power 6.4 mV (40 mW).
  • CO2:N2:Ar = 1:2:6
    • the highest pressure, at which the lasing was observed 1.5"Hg
    • the optimal pressure 1.3"Hg
    • the highest current (by milliammeter) achievable without dropping to arcing: 4 divisions (20 mA)
    • threshold current (by milliammeter) 2.5 divisions (~12.5 mA)
    • top registered output power 25 mV (160 mW)
      A tiny droplet of xylene was added into the tyre, that contains the mixture. Gases then were stirred vigorously. As the result the surge current and the lasing yield became lower (3 divisions and 16 mV correspondently).
  • CO2:N2:Ar = 1:3:6
    • the highest pressure, at which the lasing was observed 1.4"Hg
    • the optimal pressure 1.4"Hg
    • the highest current (by milliammeter) achievable without dropping to arcing: 3.5 divisions (17.5 mA)
    • threshold current (by milliammeter) 2.5 divisions (~12.5 mA)
    • top registered output power 37 mV (236 mW)
  • CO2:N2:He = 1:2:6
    • the highest pressure, at which the lasing was observed 1.5"Hg
    • the optimal pressure 1.3"Hg
    • the highest current (by milliammeter) achievable without dropping to arcing: 4 divisions (20 mA)
    • threshold current (by milliammeter) 2 divisions (~10 mA)
    • top registered output power 25 mV (160 mW)


Hardly comprehendible why increasement of nitrogen to CO2 ratio has caused output drop when with helium and caused power rise when with argon. Is it the native properties of the mixtures or just the impurity of the experiment. (Helium was stored in automotive tyre from the may of the previous year. By the way, if stored in party balloon it would ten times be lost and dead. And, judging by the discharge current and lasing, the tyre did really contain helium after all this time. However its purity might be low.)


There also were runs (on CO2:N2:Ar = 1:2:6 mixture, after xylene addition) in gas - static mode. The lasing does probably exist at the first moment after turning on, but no continuous lasing was observed.

Need to say that all previous runs with lasing were in the slow gas flow mode (somewhere at 0.1 - 0.3 liter*bar per minute). To achieve this, the exhaust pipe (the one that goes to the vacuum pump) is clamped slightly. The gas/vacuum valve on the incoming pipe (the one connecting the laser cell with the tyre) is to be set to a position that provides the proper pressure (e.g. at 1.5 inches of mercury column). After this one may turn the ionizer on, and then turn on the power of the main discharge.

Having read in articles and web-reports how fast do the low pressure continuous lasers burn out the gas mixture, I have never (until now) tried to get the lasing in a gas static mode. Ant it is very expectable, that the first try is fail. (Naturally one could use gas regenerators, special gas mixture and special materials for the tube, but I had neither desire to do that, nor the possibilities.)


The web-report genre is rather free and one of its drawbacks is that one usually needs to read tons of unnecessary info in order to understand what was meant in one or another place. However need to note that the guides are best to be read as the entirely whole too. Sometimes somebody, having read three paragraphs, thinks that he've understood anything and rushes to production. But no - it does not work. And the one(body) becomes frustrated and rings all over the mail and internet... And this is when he(she) could just read the guide a bit further, or a bit more attentively, or a bit more thoughtfully, or a bit more patiently and get all the answers.


Back from lyrics... Another test run was with CO2:N2:He = 1:2:6 mixture and another output coupler. This one has diameter 25 mm (ZnSe, dielectric coated) and reflection of 75% or 50% (The marks on it have disappeared in the process of epoxy-mounting. Occasionally i will measure its reflectance and write down here). The laser haven't reached threshold with this mirror at all.



Turning On the Brains.

Finally the top power, that have been achieved to the moment, is 236 mW. With taking into account the fact that the resonator does possibly gather the light poorly (say one half) and taking into account the duty cycle (suppose that it was one fifth) we'll get that this laser can give 236 mW x 5 x 2 = 2.36 W in continuous mode.

It have already noted above about the caution to be taken when making such an estimations. For example considering the light gathering by the resonator - all affordable mirrors and their combinations have already been checked (plano-plano, plano-concave, concave-plano, concave-concave/hybrid) and I already use the best one (concave-concave/hybrid). So here's a little hope to move forward in this area. Or I should try to install mirrors directly into the waveguide? The reference waveguide parameter B of the mirror set up is B = pi*w^2/wavelength, where wavelength = 10.6e-6 m, a = 0.7*(a/2) = 0.7*1.5e-3 m - for the planar channel, pi = 3.14... Substituting the values we get B = 0.326 m. And the condition of low losses due to modes mismatch, which has the form [3]: h < B/10, where h - is the distance from the waveguide end to the mirror. We can see that this condition is fulfilled, because in the current design h is less than 2 cm.

On the other hand we can remember a pocket-sized-lamp-pumped-dye-laser, where this condition was fulfilled with even more warranty and still the closer mirrors were to the capillary the better it worked. So in reality here we have the possibilities unknown...

Coming from the pulsed power to the continuous does promise five times up, but in reality we can meet gas burning out, or developing of the misalignment due to overheat. And the promised five can turn out to be three or even two.

Even if we hope for better the estimated two watts with a half sound a bit small to the wishes of 8-10 W. So I have to research the reasons "why things go bad" and "how to set it right". Fortunately this laser is a peaceful one in the contrast to the pulsed (TEA) systems with their tens or hundreds of kilovolts, tens or hundreds of kiloamperes, their nanoseconds and their killing electromagnetic pulse. In those systems the measurements are almost impossible and we have to rely on the theory only. But here we can use all the arsenal of our measurers and gauges to squeeze the info about what goes wrong.

The current and the voltage on the cuvette were measured by an oscilloscope during the lasing/ A mixture of CO2:N2:He = 1:2:6 was used. The measurements were made when the indications of the milliammeter were 3 divisions. The lasing threshold was 2.5..3 divisions.

There is the oscilloscope trace of the current through the laser cell (in the reality the oscilloscope can not measure current directly, naturally this is the trace of the voltage drop over a shunt having 2 Ohms of the resistance. It was just the shunt of the milliammeter,) For better visibility I made the image black-and-white and stressed the trace by a thick black line. You can take a look at the original image by clicking the picture.



And here is the trace of the voltage (the oscilloscope was connected to the terminals of the laser cell through an 1:1000 voltage divider):



The voltage is a bit unstable, so the superimposed pulses show that scatter. The mean line is also stressed by a black thick line. And also one can take a look at the original by clicking the image.


One can see that the average current at the top of the pulses is 50-55 mA. The current pulse duration is about 6 ms at its bottom and 2,5 ms at its top. So the duty cycle is 1/3.3 for the bottom and 1/10 for the top. One can also get the graduation mark: the measurements were taken at three divisions of the milliammeter scale, so one division corresponds to ~17 mA of peak current (providing that the duty cycle remains constant).

The voltage has comparatively short rise (2 ms) flat top (5-7 ms) and long smooth tail (10..12 ms) due to the remaining charge on the peaking capacitors. The height of the flat top over the zero level is 1.5 - 1.9 kV.

Interesting to compare the measured voltage with calculations. Vitteman [4] gives the field tension in self-sustained discharge in CO2:N2:He = 1:1:3 mixture being equal to 13 kv/cm at 760 torr. Here we have non self sustained discharge, but in any case we have to force it as close to self-sustainity as possible. For 1.5"Hg it will be 13 kV/cm *1.5"/29.9" = 0.65 kV/cm. And the full voltage drop over 3 cm of the inter electrode spacing will be 1.95 kV. In comparison with measured 1.7 kV and with taking into account the limited precision (mostly of the pressure) one can say that the coincidence is excellent.


Using the obtained data let's try to comprehend what takes places in the laser (how much of useful energy it has, and how much of it yields the light). Let the threshold current is 3 divisions (51 mA peak), working current is
4 divisions (68 mA peak). And the useful power is all that is above the threshold. I.e. the useful power = 1.7 kV*(68-51) mA = 29 W.

On the other hand the period of time, during which the current exceeds the threshold one (77% level) is 2.5-3 ms. The duty cycle is then one seventh - one eighth, so the measured 200 mW of lasing correspond to 1400 - 1600 mW in each pulse.

The efficiency = 1.5 W (lasing) / 29 W (useful, electric) = 5.2%.

So the differential efficiency is 5%. It is even more than it was expected at the start of the project. (Naturally using the best measured values one can pull its ears up to 8-10%, but I like more reliable values.)

From the estimations above it becomes clear what goes wrong. The power goes to "the subthreshold". Indeed, if we increase current, say twice, the useful power will be 1.7 kV*(68*2-51) mA = 144.5 W and the emitted light will be 7.2 W - almost the desired value for the design.

Lets see if the laser will be overheated with this current (136 mA). In the continuous mode the total power (including the "useless" one) will be: U*I = 136 mA * 1.7 kV = 231 W. With the heat remover cross section of 0.03 * 0.35 = 0.01 square meters; with the thickness of the laser channel of a = 3e-3 m and with the heat conductivity lambda = 0.147 W/(m K) the thermal resistance will be Rt = a/(8*S*lambda) = 0.25 K/W. and the temperature difference will be deltaT = 231 W * 0.25 K/W = 58 K.

To estimate the temperature difference in the glass wall we should get that one wall conducts only a half of the total power (i.e. 115 W). Let the heat conductivity of the glass be 1 W/(m K) and the thickness of the glass be 2 mm. Here the temperature difference in the glass wall will be:
115 W * 2e-3 m / 0.01 sq.m / 1 W/(m K) = 23 K.

The full temperature difference is 58 + 23 = 81 K and the gas temperature in the center of the tube : 113 oC. All seems to be right.

For the helium-less mixtures the things are worse. The temperature difference in the gas will be about 500 K and the gas will certainly be overheated.

The considerations above mean that in order to drive the laser to the desired power we should (keep) use helium based mixture and rise the feeding current twice. Sounds easy... Until one remembers that the current is already limited by the discharge instability... The ghost of the Bealieau scheme has appeared with its multiple of the resistors by the sides of the laser cell... Are there less infernal ways to do it? The authors of [2] could somehow drive their laser up to half of an ampere of the supply current and the sizes of their cell were almost twice smaller.

It is possible that the power supply pulses are bad for the laser and with going to the constant current the loading characteristics will become softer. But I am still not ready to suddenly increase the supply power by eight times. The risk to burn the laser cell is too high. And I still have some questions
to it.

The material of the electrodes may play a bad role. The aluminium oxidizes easily and it causes strong change in the voltage drop in the discharge near the electrodes. In its own turn it may provoke arcing.

It is possible that the power of the ionizer is just low. But the most direct way of its increasement has already met an obstacle in form of limited electrical strength of the cell. And there is lack of ideas how to increase the ionization power without making the complicated radio-frequency oscillator.


For now it was decided to build a new laser cell with copper electrodes and look what it will change.




While the new laser cell is curing (and it can take as long as two weeks) some more measurements were taken. I measured the tolerance of the laser to the mirror misalignment and to the type of peaking capacitors being used.


On the misalignment

To simplify the description lets choose a coordinate system. Let the Z-axis be directed along the optical axis of the laser (along the laser channel. Let X axis to lie in the plane of the waveguide and be orthogonal to Z-axis. (The x-axis is essentially collinear to the direction of the electric current or either to the direction from one electrode to another). And finally let Y axis to be directed orthogonal to the waveguide plane.

Here is a picture (hopefully it will clarify the things rather than mess them up even more):

          ^ Y
          +--------> X

Let's remember that in the process of the alignment we direct the alignment laser beam throughout the resonator of our CO2-laser and observe the spots, produced by the beams reflected by the front and rear mirrors. Let's call these spots as "alignment spots."

In the taken coordinate system it was obtained that moving the alignment spot (related to the front mirror) by 1 cm along the Y-axis (both upwards and downwards) drives the laser to the threshold with correspondent reduction of power by 10 times. Even slightly more movement of the front mirror and the laser can not reach the threshold.

Along the Y axis the threshold is related to the alignment spot motion of 3 cm.

It means that the required tolerance of the alignment is ±10 mrad in the plane of the waveguide and ±3 mrad in the direction normal to the plane of the waveguide. (The distance from the alignment laser to the front mirror was 150 cm and we are noting that the beam deflection is twice of the mirror deflection.)

The resonator was of the stable type. Having both concave mirrors. The rear one was spherical with R=2m radius; the front one was hybrid one, and the radius of its spherical part was also 2 m.

The sensitivity to the alignment in the vertical plane appeared to be unexpectedly high. It makes me think that I need to expand the waveguide a bit. For example to 4 mm.


On the tolerance to the capacitors type.

In practice capacitors are described not only by their capacity and endurable voltage. There are at least ESR and ESL parameters:

  • ESR - equivalent series (active) resistance;
  • ESL - equivalent series inductance.

The ability of the capacitor to heat, when alternating current flows through it, is determined by its ESR. And the lowest possible discharge time (or the short circuit current, as You wish) is determined by ESR and ESL both.

Here is some other thing, that should not be related with the current laser, but to the full extent determines the behaviour of TEA lasers. It is the dependence of the dielectric permeability of segnetoelectrics to the applied electrical field. As the result the accumulated electric charge q quits to depend linearly to the applied voltage U. In other words the law q = C * U (where C is the capacity of the capacitor) changes its status from the precise to roughly approximate. It also means that even all three parameters (C, ESR and ESL) are not enough for the complete description of the behaviour of the capacitor. E.G. TEA CO2 lasers work best with ceramic doorknob pulsed capacitors and tend to fail with (seeming to be) low inductance Maxwell ones.

So the theory as well as practice tells us that different capacitors can lead to different results even in case they have the same capacity. Needless to say it is interesting to see how the type of the peaking capacitors can affect the present laser.

The tests were made using the mixture of CO2:N2:He = 1:2:6. The mixture was aged seriously. The resonator was: rear mirror concave, the front mirror - the hybrid one.

The initial test (the reference point) was made with using 3 pcs of ceramic doorknob pulsed capacitors rated to 2 nF x 40 kV each. They were connected to the laser cell by wide aluminium stripes.


Threshold current (as indicated by the gauge) 2.5 divisions (42 mA peak)
Current of drop to arcing 4 divisions (68 mA peak)
Lasing average power 25 mV (160 mW)


The peaking capacitors were then replaced by the ones having plastic mold body insulation and wire terminals. The capacitors were of the ceramic type, and rated to 40 kV 2 nF. Their capacity was verified by a multimeter. However even the first glance at them can tell You that they would definitely fail under the advertised voltage - the gap between the terminals is less than 15 mm. Many of them come with cracks in the ceramic discs with corresponding degradation of the performance. Their seller named them as "Muratas" but I refuse to hope that such a serious manufacturer could throw such a bad thing into the market.

Three of such capacitors were connected in parallel to the laser cell. The connection was made by copper wire 1 mm in diameter. Subjectively the laser began to work a bit worse. But the sensation is not verified by figures:

Threshold current (as indicated by the gauge) between 2.5 and 3 divisions (48 mA peak)
Current of drop to arcing a bit less than 4 divisions (68 mA peak)
Lasing average power 20 mV (130 mW)


The results should be taken as "nothing has changed in the limits of errors."


And finally the peakers were replaced by three K73-13 (mylar film, rolled type) rated to 2 nF x 10 kV. This time even the figures can prove that it became worse:

Threshold current (as indicated by the gauge) a bit over 3 divisions (51 mA peak)
Current of drop to arcing 3.5 divisions (60 mA peak)
Lasing average power 16 mV (100 mW)


After the laser had been returned to its initial state (after connecting the doorknob Muratas) it gave 24 mV - i.e. the output has returned too.



An attempt was made to increase the power of the ionizer by means of finer match of the AC-oscillator (Feron) to its load. In order to do this, different numbers of turns in the secondary winding (the one feeding the ignition coil) were tested. In the range from 10 to 20 turns the difference was barely notable. The limiting current, the threshold and the laser output were essentially the same. Finally the winding was returned to its original 15 turns.

Another thing that was done this weekend was a little upgrade to Feron electronic transformer, used as the ionizing oscillator here. Namely a 1000 uF x 450 V electrolytic capacitor was attached to the output of the internal diode bridge inside the Feron unit. So now the ionizer works in continuous mode, not in the pulsed one. Still testing, and now I can say only that the Feron unit is still alive but the laser seems to have lost some power.

feron_a feron_b




The cell with copper electrodes is now ready. Frankly speaking it should stand for another week, but lets call it a day.



Here's how the laser looks like with the new cell:

laser_Cu_cuvette_a laser_Cu_cuvette_b

The electrodes are made of copper wire (3mm dia). There were real problems with its straightening. The most simple way to make it straight is to stretch it until it breaks. The diameter of the wire after stretching will be slightly smaller, but this is allowable. Near the very point of the break the diameter will be considerably smaller, but one can always discard that part.

To pull the 3 mm copper wire until it breaks by hands only is a Herculean task. And if the piece is ling enough (like 10 meters) any jacks and handles are out of the job too. Finally it was decided to tie the wire to a trailing hook of one car and pull the wire by another car. It appeared to be unexpectedly hard to break the wire even by force of car. Fortunately no harm to health and life was inflicted. I will abstain from describing all the fun here. Lets leave it for the imagination of Yours.

For the ones, who wants to reproduce the experience, i have only one hint: You better sand the wire from the oxides before trying to stretch it. Otherwise You can easily bend so precious straight piece of the wire just by rubbing it with a sandpaper.

The wire was straightened, and a pair of suitable pieces were cut from it. After some sharpening and cleaning those pieces became the future electrodes. They were glued in between two flat glass plates. The sizes of the laser sell were kept intact: laser channel width 30 mm; waveguide thickness 3 mm; length of the active part 350 mm; length of the glass walls 400 mm and length mirror-to-mirror 450 mm.


Aside the different material of the electrodes the other difference of the new cell is that the coolers-ionizers are made not of the hollow rectangular aluminium tube, but they are made of E-shaped aluminium extrusion. It is supposed to operate as an air-cooled heat sink. I'm still wanting to avoid using water as long as possible. Ask why? Remember the written above about the microwave oven transformer and electric arm-chair.

The new cell was tested on argon containing mixtures: CO2:N2:Ar = 2:3:15 and CO2:N2:Ar = 1:2:3. It was also tested with some helium containing mixture, which initially was CO2:N2:He = 2:3:15.

Let me remind, that the cell with the aluminium electrodes almost always allowed to reach the discharge current of 3 divisions as indicated by milliammeter, and it was possible to get up to 4 divisions from time to time. The interesting property is that the top obtainable current depends weakly to the N2 to CO2 ratio and to the type of the buffer gas being used.

On argon mixtures the cell with copper electrodes appeared to be able to bear up to 5 divisions of current (as indicated by milliammeter) usually and up to 6 divisions - occasionally. However the output growth il less considerable. The threshold has grown up to 4 divisions. It seems like the main factor of the average current increasement is the increasement of the pulse duration rather than the increasement of its amplitude. But I've still not bothered to use the oscilloscope to check it up.

With helium based mixture CO2:N2:He = 2:3:15 the top current, limited by the arcing, appeared to be the same 3..3.5 divisions as it was with aluminium electrodes. In this particular case that laser haven't reached threshold. The helium was (obviously) from a party balloon (newly bought however).

After a proposition that the mixture still lacks helium, it was diluted with helium twice by volume. I.e. it became something like CO2:N2:He = 2:3:35. Nevertheless, the current of dropping to arcing was still equal to 3 - 3.5 divisions of milliammeter scale (at pressure of 1.2"Hg). One might note that with this mixture the laser was able to reach the threshold, but the output was lamentable.

Carefully (for not to let the air in) into the vessel with the gas mixture a little droplet of xylene was added. And still the surge current did not change. But the threshold became unreachable again.

It was supposed that argon may be somehow responsible for higher current on argon based mixtures. So argon was added to the mixture. The amount of the addition was ~25% of the total mixture volume. It became then something like CO2:N2:Ar = 2:3:35:10. And the surge current did again make us happy with its constantness. There was no lasing too.

Finally, almost by fortune, the mixture was diluted by nitrogen. The amount of the nitrogen was not measured - a part of the volume that had been expended in the tests was filled. Kinda like 10 - 20 % of the volume. I.e. it became CO2:N2:He:Ar = 2:8-13:35:10. (Or if we divide it by two: CO2:N2:He:Ar = 1:4-6:18:5).

After the addition of the nitrogen the mixture has suddenly stopped to behave bad. The surge current has risen to the values already obtained on the argon mixtures: 5-6 divisions of milliammeter scale, There appeared lasing. And not only just appeared... The power has surpassed 240 mW. The mixture with its almost unknown composition has shown nearly the best result.

Difficult to name the exact value of the optimal CO2 to N2 ratio for this laser, but it definitely likes plenty of nitrogen.

With the argon based mixture and with this (unknown composition) one a video was taken, how the laser lights some matches:

With CO2:N2:Ar = 1:2:3 the highest obtained power was 25 mV (160 mW).

With unknown He based mixture the top power was 42 mV (269 mW).


The heat builds up during the laser's operation. The thermal inertia of the laser cell is just enough for five minutes before the output starts to drop due the overheat.



Usage of the copper electrodes has really allowed to increase the limiting current (achievable before dropping to arcing). But it takes place with "good" mixtures only and the amount of this increasement didn't even reach 2 times. With average mixtures the cell with copper electrodes behaves almost like the one with aluminium ones.

The further point in our checklist is to see whether the discharge characteristic will become more soft when using the constant current rather than pulsed one. Let me remind that now the laser is fed by a half-period rectifier, i.e. by pulses having 50 Hz repetition rate. The ionization generator has already been prepared for the constant operation (see the record of 19.02.2017). Lets go further.

Three new diode stacks were soldered up. Each of them contains 8 x HER308 diodes, connected in series. The new stacks plus the old one were assembled into the full rectifying bridge.


There is still no smoothing capacitor, so the laser cell is now fed by pulses with 100 Hz repetition rate. (It should be considered when trying to interpret the values of current indicated by the milliammeter. Its gear does the time averaging of the acting forces, so the pulses with the same amplitude but with the doubled repetition rate will probably lead to doubled indication values. But this is in theory. In practice it just means that recalibration is needed.)

After the modifications the laser was assembled, aligned and got a run.

With CO2:N2:Ar = 1:2:3 the power has grown to 45 mV (288 mW).

With the helium mixture of the unclear composition (see the record of 24.02) it reached 70 mV (450 mW).

The power rise with the argon mixture was 1.8 times so far, and with the helium mixture - 1.7 times. Ideally the output power should be exactly doubled in both cases. And we got a bit less. But I think that it is too early to panic, because in range of errors and instabilities it more looks like the things go right.

Such a power allows not only to burn up a match. But also to burn through a sheet of paper. Even a white one. It seems that CO2 laser should ignore the type of paper - the black one as well as the white one should absorb 10.6 mcm well. But in practice the difference is more similar to the one we observe when using a neodymium laser - the white paper is much harder to burn. Here's the video:

There's still no water cooling and the overheat of the laser cell comes less than in 3 minutes. And if one wants to measure something or to have some fun, he needs to move his... fingers.. quick to have a time to increase the current gradually (not letting the arcing to develop) and for the rest stuff.

And finally the output of the rectifying bridge was loaded with a smoothing capacitor. A microwave oven HV capacitor rated to 0.9 mcf was used. Its value was intentionally taken lower than needed - it should keep the duty cycle below 100% keeping the heating at the reasonable level.

The result is that with smoothing capacitor laser failed to reach threshold on argon mixture at any pressures. And on the "good" hellium mixture it barely reached threshold at optimal pressure. Lemme describe this in more details. After the smoothing capacitor was connected the limiting current (the one when the laser drops to arcing) became lower by 1..2 divisions of the milliammeter scale. Moreover, if earlier the approach to arcing was smooth - the laser behaved more and more unstable while approaching the surge current, and here the laser suddenly drops to heavy arcing (dangerous for the cell integrity) without any warning signs. The approach became sharp and dangerous. And if the arcing took place even for a short time the laser appears to be completely overheated and needs a half of an hour to cool down.

It seems like the current pulses (when feeding without smoothing) play kinda arc-quenching role. When the arc tries to develop, just in a couple of milliseconds the voltage vanishes and the discharge becomes unable to sustain itself. And it takes place 100 times per second. This allows more safe and more close approach to arcing, and even (dynamically) to slightly surpass it.

On the other hand when we use a smoothing capacitor nothing can stop the development of the arc once it happened. The arc does not quench even in the valleys between pulses, because of the non zero current there.



(Abt myself:) Did You got what You said?

In the introduction to this web-report i named the kind of the discharge, which ionizes this laser, as "barrier type". Why not use the old good term: "electrodeless discharge of the capacitive type"? Additionally I used a bit
lousy definition, saying that "the electrodes are insulated from plasma". Naturally on the contrary to the "true" electrodeless discharge the barrier one has no need to be electrodeless. The insulation layer is still present
there, though. The schematics below show three types of discharges, where some dielectric between the electrodes makes an obstacle for the direct discharge - makes a "barrier" in its way:


1) both electrodes are insulated (true electrodeless discharge of the capacitive type)

2) one of the electrodes is insulated (barrier type discharge)
//    |    //
//  --+--  //
//         //
//         //
//  --+--  //
//    |    //
3) none of the electrodes is insulated the dielectric blocks the direct way for the discharge


The barrier may be set at both electrodes (here the discharge becomes "true" electrodeless), at one of the electrodes, and finally away from the electrodes. The last two types of the discharge are not electrodeless, but still the current there is determined by the displacement current, that arises during the charging process of the distributed capacity of the dielectric surface. (Certainly the dielectric has to be strong enough. If it was broken down the discharge becomes the common one, based on the conductivity currents.)

If one looks attentively at the scheme of the laboratory laser (see the second picture in the introduction to this web-report), one can discover that there is an electrical connection between one of the ionizing electrodes
and one of the electrodes of the main discharge. The connection goes through the grounding. I.e. respectively to the external high frequency ionization the laser is connected not as the electrodeless discharge circuit.

The said is not related to the "LANTAN" laser. It cannot be seen from the picture in the introduction, so i need to cite another picture from the same book [1]:


Here one can clearly see that the LANTAN uses namely the electrodeless scheme. (By the way if we are speaking about a non-sustainable form of the discharge, it is curious, why the authors of the cited works, do continuously use the term "preionization"? Looks like a bad habit.)

What's the difference? First of all the type of the connection does affect the external ionization discharge power.

The high frequency discharge is connected to the external high frequency (or the pulsed one) oscillator through a capacitor. The first plate of this capacitor is the cooling - ionizing electrode, the second plate is the plasma
itself, and the dielectric layer is the glass wall of the laser cell. When the plate has sizes 35 x 5 cm, the glass wall is 2 mm thick and the dielectric permeability is 5, the capacity of the planar capacitor will be:

Ci = 8.85e-12[F/m]*5*0.03[m]/0.002[m] = 232 pF.

In case of the "true" electrodeless discharge the laser is connected to the HF or pulsed oscillator as this:

 _/\_  Ci||     +-----+     || Ci
         ||     +-----+     ||

And the energy, depositing in plasma with each recharging of the capacities is: Ei = q * Ui, where q - electric charge, having passed through the plasma, ui - voltage drop on the plasma.

The electric charge q is equal to the one, that accumulates on the capacitor plates, when the capacitor charges to the voltage Uc (the voltage that feeds the circuit). I.e. q = (Ci/2)*Uc. A half of the Ci is taken due to the fact
that here we have two similar capacitors connected in series.

The voltage drop on the plasma can also be estimated easily. The measured voltage drop over the main discharge was measured and appeared to be 1.7 kV. And the length of the ionization discharge (3 mm)is ten times less than the length of the main one (30 mm). So Ui = 170 V. The charging voltage is harder to estimate, so lets take it from handwaving: Let it be Uc = 10 kV. Then

Ei = (232e-12/2)*1e4*170 = 1.97e-4 J. At the frequency f = 20 kHz the ionization power will be 8 W (with taking into account that the recharging takes place twice a period). And those 8 watts are from 250 watt rated oscillator (Feron electronic transformer).

Actually the question whether our oscillator is capable to develop the said voltage on the said capacity is unclear. To illustrate this let's estimate the required power as the energy, needed to charge the capacitor to the given voltage (Ci*Uc^2/2) multiplied by the frequency of recharges (2f):


Wi = 2*2e4[Hz]*232e-12[F]*(1e4[V])^2/4 = 232 [W]


The estimation shows clearly that it is not easy even for 200 W rated Feron to recharge the ionizing capacities at 20 kHz oscillation frequency. However fine load matching allows to approach the mentioned 10 kV.

In order to use the power of the oscillator effectively one should aim to the equality Us ~ Ui. However the lower voltage allows only lower currents, and thus lower powers of the ionization. One could increase the current by increasing the frequency, but it may affect the simplicity of the design (and many people consider the radio-frequency technique as being hard and complicated one).

Setting aside the increasement of the voltage and increasement of the frequency, what's left? Increasement of the Ci capacity. And exactly in this point the schematics of LANTAN laser circuit and of the laboratory laser
circuit do differ strongly. If we use the same simplification to draw the ionization circuit of the laboratory laser as we did it for the electrodeless discharge, we'll end with a strange picture:

                   |  ||    |
                   |        |
 _/\_  Ci||     +--+--+     |
         ||     +-----+

Despite it is strange, the circuit is distinguished by the fact that the laser is connected in series with Ci capacity rather than with Ci/2 as it took place before. And it is twice as much.

One can develop this idea further and draw:

 _/\_  Ci||    
     |   ||   |
     |        |  +-----+     
     |        +--|LASER|-------------o
     |        |  +-----+
     | Ci||   |

And it corresponds to the connection through 2*Ci capacity already - i.e. four times greater than it was in the initial variant. And this is absolutely for free. It does not require exotic materials or very thin walls. It even does not require to build a new laser cell. Just reconnect the old one in the new way and presto...

One needs to connect both cooling/ionizing electrodes by means of some wire. (Care must be taken to avoid flashovers from this wire to other parts of the circuit.) And to connect one terminal of the high frequency/high voltage oscillator to any of the ionizing electrodes, and to connect the another terminal of the oscillator to one of the main discharge electrodes.

And finally we get the next circuit:

                                      two ignition coils
                                       in parallel +-------+----------+
            +---------------------------+          |       |          |
   o--------|  +---+         )|         |---o--- |(        |  Laser   |
  ~ 220 V   | /_\ ---  Feron )|( 16     |       )|(      =====        |
            | \^/ ---  250 W )|( turns          )|(   +--)   (--+     |
   o--------|  +---+         )|         |---o--- |(   |  =====  |     |
            +---------------------------+          |  |    |    |     |
                                                   |  |    +----)-----+
                                                   |  |         |
  microwave                                    Rb     |         |
  oven              8 x HER308                 3k4    | Cp 4 nf |
  transformer  --+--|>|--...--|>|--+-------+--/\/\/---+---||----+
   4kV      ||(  |                 |       |                    |
   o------- ||(  |                 |       |                    |
           )||(  |  8 x HER308     |       |                    |
           )||(  +--|<|--...--|<|--)---+   |                    |
   o------- ||(                    |   |   |                    |
            ||(                    |   |   |                    |
               |  8 x HER308       |   |   |                    |
               +--|>|--...--|>|----+   |  --- Cf                |
               |                       |  ---                   |
               |  8 x HER308           |   |          /         |

The scheme has marks that the electronic halogen lamp transformer Feron has suffered the next modifications: the secondary of its internal toroidal transformer was replaced and a smoothing capacitor was added to its internal diode bridge rectifier. It is also shown that now the main discharge is fed by full bridge rectifier rather than by a half-wave one.

The laser was assembled and aligned. The cell with copper electrodes was used. And there was still no smoothing (Cf=0).

On the CO2:N2:Ar = 1:2:3 mixture it was obtained 50 mV (320 mW)

With a (new fresh) CO2:N2:He = 2:5:21 mixture it was obtained 110 mV (700 mW)

Most of the "powerful blue" (450 nm) laser pointers, that are positioned as 1 Watt rated on the market, do really not more than 600 mW. And after a couple of years of storage they drop to 0.3-0.5 W. So the laser being described here did successfully surpass that powerful laser pointers. And i'd like to hope that it is still far from its limits.

As one can see from the schematics, now the high frequency (HF) power is applied between the cathode and both (connected one with each other) coolers - ionizers. Other types of the connection were also tested. Among them the one, when the HF voltage is applied between the anode and both coolers-ionizers. The one when one of the ignition coils is connected between anode and first cooler- ionizer, and the second ignition coil is connected between cathode and the second cooler-ionizer. The latter case was tested with the coils connected
in-phase and in opposite phases. Any of tested alternatives was worse than the variant, shown on the figure above, but still better than the "true" electrodeless connection.

Just another interesting thing to note is that with "true" electrodeless connection the discharge was almost uniform, while with the barrier discharge connection it has stratified into multiple of streamers. Since the pressure is low the streamers are wide - almost 3 mm wide. The dark gaps between the streamers are ~ 1 mm wide. It might be bad, but apparently does not affect lasing.

And finally to the output terminals of the rectifier, that feeds the main discharge, a smoothing capacity was attached. Initially it was a humble 1 mcF, which has grown to Cf = 4 mcF later.

With the CO2:N2:Ar = 1:2:3 mixture it gave 90 mV (570 mW)

With the fresh CO2:N2:He = 2:5:21 mixture it did 170 mV (~1100 mW).

The current of surging to arcing (as indicated by the milliammeter) has reached 70 mA. The lasing threshold was 50 mA.

The top achievable current does not depend to the mixture type in the limits of +-5 mA. The same can be said about the lasing threshold.

It is interesting that with the old circuit ("true electrodeless") using the smoothing capacitor caused that any arcing undoubtedly meant the stop of the laser. 30-50% reduction of the supply voltage by the variac was not sufficient to suppress arcing. And with the new scheme (barrier discharge) the arcing can be stopped by 15-20% voltage decreasement. And the approach to the arcing limit itself has become soft. - The laser warns its user by sudden and short flashes inside the cell and by waving of the milliammeter arm. In addition the arc lost its tendency to sit at the single (chosen) spot. It became possible to enter the mode of "running arc" again.

It feels like the additional alternating voltage of the external ionizer blows the arc off. It is possibly due to the nulls of the current flow, or even due to the repolarizing (which probably causes quenching of the cathode spots).

The power of the laser has finally grown to the 1 watt level. It is still an order of magnitude lower than it was expected, but it already can perform some interesting things.



Since the capacity of the external ionization circuit has grown by 4 times the transformer ratio of the ignition coils has become too excessive. As an attempt to match the load once again the ignition coils were replaced by a television flyback transformer. Its name is TVS110-PC15, but You are unlikely to find one outside of Russia. So i'll say that its secondary has 1500 turns and it is wound on a ferrite with high saturation level. (Power ferrite.) Magnetic permeability of the ferrite is 3000. Its section is 1.8 square cm. The primary of the flyback transformer is of low use here. A new winding was added instead. It has 4 turns of thick (3 sq. mm.) wire in PVC insulation. The winding is made with 2 wires simultaneously (totally 6 sq. mm.).


The secondary of the internal toroidal transformer inside the Feron unit was also replaced. Now it has 4 turns of the same (double) wire (totally 6

                                     TV flyback 
                                       transformer +-------+----------+
            +---------------------------+          |       |          |
   o--------|  +---+         )|         |---o--- |( 1500   |  Laser   |
  ~ 220 V   | /_\ ---  Feron )|( 4      |   4   )|(turns =====        |
            | \^/ ---  250 W )|( turns    turns )|(   +--)   (--+     |
   o--------|  +---+         )|         |---o--- |(   |  =====  |     |
            +---------------------------+          |  |    |    |     |
                                                   |  |    +----)-----+
                                                   |  |         |
  microwave                                    Rb     |         |
  oven              8 x HER308                 3k4    | Cp 4 nf |
  transformer  --+--|>|--...--|>|--+-------+--/\/\/---+---||----+
   4kV      ||(  |                 |       |                    |
   o------- ||(  |                 |       |                    |
           )||(  |  8 x HER308     |       |                    |
           )||(  +--|<|--...--|<|--)---+   |                    |
   o------- ||(                    |   |   |                    |
            ||(                    |   |   |                    |
               |  8 x HER308       |   |   |                    |
               +--|>|--...--|>|----+   |  --- Cf = 4uF          |
               |                       |  ---                   |
               |  8 x HER308           |   |          /         |

Spark test of the new oscillator can be disappointing at the first glance. It gives only 4 mm of spark between two thin wires (read - between sharp objects). More attentive inspection shows that the wires melt in these tests. It means rather low voltage but decent power.

With this external ionization generator:

With CO2:N2:Ar - 1:2:3 the laser did 110 mV (700 mW)

With fresh CO2:N2:He = 2:5:21 the laser did 328 mV (2100 mW)

The pressure was 1.2 inches of mercury in both cases. The laser cell had copper electrodes and the rectifier of the main discharge circuit had smoother capacitor of 4 mcf (4pcs of microwave oven capacitors in parallel).

By turning the variac's handle gradually one can easily put the arm of the milliammeter onto its limiter (the average current has surpassed 100 mA without dropping to arcing). The threshold of lasing is somewhere between 60 and 70 mA.

One can see that the circuit in its present embodiment has reached its limits. The ballast resistor Rb is needed to be reduced, the milliammeter's shunt is needed to be reduced (with the following calibration of course), the smoothing capacity Cf is already too small, and so on.

• the power of the laser has surpassed 2 W;
• full efficiency of the laser has surpassed 1%;
• differential efficiency is about ~3%



The scheme was upgraded. Now its ballast resistor consists of three 6.8k ones in parallel (resulting to 2.26 kOhm). The milliammeter's shunt is now formed by two 2 Ohm resistors (resulting to 1 Ohm). Feron halogen lamp transformer unit was moved into "the chassis of the power supply". The smoothing capacity was kept unchanged - I had no more capacitors to attach there.


For some reason after the reassemblage the old type of the ioniator connection (between the cathode of the cell and the pair of its cooling parts) is no longer optimal. It allows the current to be only 5 divisions of the milliammeter scale before dropping to arcing. While the connection to between anode and the pair of the cooling parts allows to reach 7 divisions. By the way it ruins the idea that the alternating current of ionizer helps to suppress arcing by dimming the cathode spots. Apparently the reason is not in the polarity, but rather in the parasitic coupling.

Generally the parasitic coupling plays a great role in high voltage high frequency circuits. For example even Alfonso Rodriguez has noted in his blog that the next two circuits give completely different results when used to feed a TEA nitrogen laser.

   |  Cs||  |           |
   |        |  LASER    |
   V        +--)  (--+  )
SG          |        |  ) Lc
   ^       ---       |  )
   |       ---       |  |
   |        |        |  |
 +--> <----+--------------+   
 |         |              |
 |         |  LASER       |
 |Cs       +---)  (---+   )
---        |          |   ) Lc
---       ---         |   )
 |        --- Cp      |   |
 |         |          |   |


And it despite the fact that they are essentially identical from the point of view of the electricity. The reason here is in the parasitic capacities between the parts of the circuit and its common (earth) wire. If one takes them into account the schemes become completely different.

Another example is the well known Marx bank circuit. The circuit looks simple, and works in reality. But if one attempts to model it in any of the electronic modeling software (i.e. Microcap et. al.) the computer will be glad to inform that the circuit refuses to do anything good. And only the careful addition of the (parasitic) capacities to the ground wire and other parts allows finally to set it alive.

Our case looks similar. The geometry of the circuit has changed, the parasitic capacities have changed too, - so the very other variant of the connection has become optimal.

The maximal registered laser power with the new circuitry was 365 mV (2300 mW). It was wit CO2:N2:He = 2:5:21 mixture and of course in the gas flow mode. The pressure, when this maximal power was attained was equal to 1 inch Hg. It is lower than it was in previous tests.

The power level of 2 Wt allows not only to light a match but also to cut it into halves (see the short video below):



Naturally, the experience of laser pointers use shows that with some skills and persistence one can cut a match even with 1 W rated laser. However the 2 W one does it easier, as You could see in the video clip.

At the working laser the oscilloscope traces of current and voltage were taken.

The voltage (when the current was 6 divisions of milliammeter scale and the gas pressure was 1 inch Hg):



The voltage (when the current was 6 divisions of milliammeter scale and the gas pressure was 1.2 inch Hg):



In both cases a voltage divider 1:1000 was used, so one volt on the trace means one kilovolt a[[lied to the laser cell electrodes. The voltage drop on the cell became lower. If earlier it was 1.7 kV at 1.4"Hg pressure, that meant 1.2 kilovolts per inch, then now it is almost equal to 1 kv/inch. The possible explanation: now the power level is higher, so the gas is hot and has lower density of its molecules at the same pressure. In its own turn it means lower electric strength and lower voltage.

One may also conclude that the filtering capacity Cf is enough - there are no pits in the voltage having the frequency of 100 Hz. Some insignificant (probably) oscillations are present, but their frequency has nothing to deal with the mains and the rectifier.

If the things are generally clear with the voltage, one can not say the same about the current. For example here is the trace of the current (voltage over the 1 Ohm shunt) at the gas pressure 1 inch of mercury and when the milliammeter indicates 6.5 divisions:



At the beginning I thought that those are interferences of the glitches of the oscilloscope software, that connects the sampling points with solid lines (yes it happens, if You are using a digital oscilloscope). But the frequency of the oscillations (~400Hz) does not allow to relate them to any of the known sources of interference. The mains - 50 Hz, the rectifier - 100 Hz, the Feron electronic unit - several tens of kilohertz, RC oscillations of the Rb+Cp chain repose even higher. The proposition that these are the glitches does also fail, because the traces are reproducible from test to test, with may be 10% variations of frequency and amplitude.

With changing of the applied current the shape and frequency do change too. Here is the current when the milliammeter shows 6 divisions:



This is probably may be somehow related to the 100 Hz pulsations of the rectified voltage, however the traces of voltage make this doubtful. But here is a couple of traced taken at 7 divisions of milliammeter:




Neither the frequency nor the shape does conserve from test to test. Only the mean current value does. However it is evidently due to the fact that it is set manually by turning the variac handle and looking onto the milliammeter indications.

From those traces I can only say that 7 divisions of the milliammeter scale correspond to nearly 140 ma of the average current through the laser cell and to about 300 mA of the peak one. 7 divisions in this tests are almost exactly at the dropping to arcing limit. The lasing threshold is 4.5..5 divisions.


Interlude 2. Radiotechnical.


One can see that the laser is starving for ionization power. If the power being put into the gas from the external source could be risen, say, 5 times, the laser would give all that we want from it and maybe with some surplus. However the external generator in the form of the Feron with some ignition coil or with a flyback transformer has almost approached to its limits. Naturally there exists some reserve here: e.g. one could cut the cooling-ionizing electrodes into segments and apply two Ferons or more. However this solution is not beautiful. And it will stick when the voltage on the electrodes will reach 30 kV and the insulation will start to fail.

It would be good if we increased the external ionizer frequency towards several megahertz, and the cuvette would be included into some tank circuit in order for not to force the generator to produce hundred of watts of the reactive power in vain.

The search of some cheap and powerful high frequency transistors has forced me not to rely upon my own abilities and to make some internet surfing. After a long doubt whether the results should be shown here or not I finally decided to do it. It may be interesting for laser hobbyists even if it can not be applied directly to enhance the laser.


SUDDENLY: MANY OF THE "POWER SUPPLY" TRANSISTORS APPEAR TO BE ABLE TO OPERATE AT SEVERAL TENs of MEGAHERTZ! And since for the purposes of the external ionization (btw for the direct pumping too) it is more than enough to reach only 10 MHz, there is no need to buy the expensive UHF transistors having the shape of golden screws or plates. One can easily adopt the power supply MOSFETS that can be easily bought or even scavenged from the olden power supply units. Looking at the results that the internet people achieve one can decide that most of the "take-it-first" transistors can operate at 10 MHz and above. Not only the fast and light-gated irf510 and irf840 can do it, but also the heavy weighted IRFP460 or pondering IRF3205 can.


2.1. Slayer Exciter.

Maybe the first schematics, one faces to, when searching the web for the HV HF things, is so called "Slayer Exciter". This term has its own author and operation principle, but the socium has its own ways. The thing being widely called as the "slayer exciter" has low relation with the original author and principle. As far as i could get the term "Slayer Exciter" means the whole class of schematics, that can be determined as:


In the simplest case it can be formed by a single stage circuitry having only one transistor. More complicated variants can contain multi stage amplifiers and phase correctors.

A simple slayer exciter circuit is shown below:

           30 mm dia |
             250mm   )
             long    )  -----------------+
             coil    ) (                 |
             of      ) ( 5 turns         |
             0.18 mm )  --------------+  |
             wire    |                |  |
       R1 1k         |       R2 5.1k  |  |
   +---/\/\/---------+-------/\/\/----+  |
   |                 |                |  |
   |  20V   / 20V    |                |  |
   +---|>|--|<|------+                |  |
   |     /           |                |  |
   |                 |  VT1 IRF840    |  |
   |                 ---              |  |
   |                -------           |  |
   |                |  ^  |           |  |
   |   R3 3 Ohm     |  |  |           |  |
   +---/\/\/---+----+--+  +-----------)--+
   |           |                      |
   |  c1 10n   |                      |
   +----||-----+                      |
   |           |                      |
   |  C2 10u   |                      |
   |    ||     |                      |
   +----||-----+                      |
   |    ||+                           |
   |                                  |
   |           Cf1 2 x 470n           |
   |                                  |
   |                || Cf2 10u        |
   |                || +              |
   |                                  |
   |                                  | VL1 protective lamp
   |                                 (X) 
   |                                  |
   O GND                              O +20 V and more

The automatic biasing circuit R3C1C2 provides the negative feedback for the direct current and hereby helps to get rid of the classical pots (variable resistors) in the gate circuit. Moreover when the slayer exciter oscillates the voltage drop on the R1 resistor (occasionally) allows to make the source to be more positive than the gate and thus to move the transistor into "C" class for higher efficiencies and lower heat dissipation.

The salt of the circuitry is that it has intrinsic property to excite the tank exactly at its resonant frequency. And it does not depend to the coil or whatever. The scheme does either operate resonant or does not operate at all. To comprehend this fact lets redraw the feedback signal path in the next way:

                       |     |
                       )     |
 ...-+--|              )     |
        |              )     |
     +->|L--...----+---      | C "Secondary-to-ground"
     +--|    C     |        ---
     |    "gate to---       ---
     |     source"---        |
     |             |         |

Here we've drawn the stray capacity from the top of the secondary coil to the earth in the explicit form. One can note that the feedback circuit now looks like the classic "resonant capacitive transformer", being widely used in the output stages of radio transceivers for the impedance matching with antennas. The capacitive transformer here does transform the low current and high voltage at the secondary top into the high current low voltage near the gate of the transistor. Because the gate capacity of some powerful FET is usually large the high current is really welcome here.

The bipolar transistor slayer exciter principle is generally the same, with the single difference that the base-emitter junction with the protective diode attached in the opposite direction do work as the short circuit here rather than as the capacitor. (Usually Your want to protect Your transistor from being killed by the overvoltage. In case of MOSFET you will use a pair of anti-parallel zener diodes connected between the gate and the source. And in case of bipolar transistor You will use some fast diode, attached between the base and the emitter in the direction opposite to this junction conductivity.) So with the bipolar transistor we actually can not use the capacitive transformer to imagine how it works. And in order to comprehend how the high voltage and low current is being transformed into the low voltage and high current needed to control the base we need to refer to the terms of "standing wave", "node" and "antinode" of current, with taking into account that the "reflection" of the secondary coils "in the earth surface" complements the quarter-wave circuit to the half-wave one.

The schematics is simple as an anecdote and if You have a droplet of a (radiotechnical) humour, You will want to make it and check. Even despite the fact that its practical use is somewhere near zero. The frequency is low, the efficiency is low, and the loading capacity is low (the oscillations tend to stop when trying to harvest any reasonable power).

When replicating the scheme it is bad to get rid of the auto-biasing circuit connected to the source (R3, C1, C2). Without it the MOSFET's begin to behave like matches. Bipolar ones are more tough, but the magical RC circuit will be useful even with them. If the device refuses to operate from the first try - one should interchange connections of the primary coil. The fresh assembled circuit should have a probability of being made correctly of 50%. Its mystics, but whenever I've did slayer exciters or blocking generators or similar ones it always requires to exchange the connection of the feedback coil. Even when (knowing this fact) I had exchanged the leads beforehand. Here's a snapshot of a brush discharge produced by a single-transistor slayer exciter when the feeding voltage was 30 V and the current was 1 Amp.

kacher fitonka


2.2 The Classique.

Three schemes were compared. (The protective zener diodes are not shown for the sake of clarity and simplicity).


A) Meyssner-Armstrong Oscillator with the external tank circuit. (Resonant oscillator with an inductive feedback in the form of a separate coil). The idea was to make the coupling coils L2 and L3 to have small number of turns while the main tank coil L1 has large number of turns, providing high Q-factor and high voltages.

   +----------------+------------------+--------o +12..24 V
   |                |                  |
   |      +-------  ) L3               |
   |   C1 |    L1 ) )  ---------+      |
   |     ---      ) ) (         |      |
   |     ---      ) ) ( L2      |      | 
   |      |       ) ) (         |      /
   |      +-------  |  -----+   |      \ R2
   |                |       |   |      / 2k2
   |        VT1     |       |   |      \
   |        IRF510  +--|    |   |      /
  ---        ...       |    |   |      |
  --- C5    IRF840  +->|L---+   |      |
   | 100n           +--|        |      |
   |                |           +------+
   |     +-----+----+           |      |
   |     |     |    |       C4  |      |
   |     |     |    /      10n ---     /
   |  C2 |+  C3|10n \ R1       ---     \ R3
   |    ---   ---   / 3.0       |      / 1k
   |    ---   ---   \           |      \
   |     |     |    /           |      /
   |     |     |    |           |      |
   +-----+-----+----+-----------+------+-------o GND


B) The capacitive tri-spot circuit (Clapp-Colpitz oscillator.) The resonant oscillator with a feedback through the capacitive voltage divider C1a,C1b, being included into the main tank circuit. Here we can again observe a pi-network (capacitive transformer) in the main tank circuit. If C1a and C1b are large and C1c is small then we should be able to achieve high voltages on the coil.

   +-----------------------+-----------+--------o +12..24 V
   |                       |           |
   |                   +---+---+       |
   |                C1a|       |       |
   |                  ---      |       |
   |                  ---      )       | 
   |                   |    L1 )       |
   |      +------------+       )       |
   |      |         C1b|       )       |
   |      |           ---      )       |
   |      |           ---      |       |
   |      |            |       |       /
   |      |         +--+       |       \ R2
   |      |         |  | C1c   |       / 2k2
   |      |         | ---      |       \
   |      |         | ---      |       /
   |      |         |  |       |       |
   |      |         |  +-------+       |
   |      | VT1     |                  |
   |      | IRF510  +--|               |
  ---     |  ...       |               |
  --- C5  | IRF840  +->|L-------+      |
   | 100n |         +--|        |      |
   |      |         |           +------+
   |      +---------+           |      |
   |                |       C4  |      |
   |                /      10n ---     /
   |     +-----+    \ R1       ---     \ R3
   |     |     |    / 3.0       |      / 1k
   |   L2)     |    \           |      \
   |     )     |    /           |      /
   |     )     |    |           |      |
   |     |     +----+           |      |
   |     |                      |      |
   +-----+----------------------+------+-------o GND


C) The inductive tri-spot circuit. (Hartley oscillator - the resonant one with a feedback having the form of a tap of the main coil).

   +---------------------+----------------------+--------o +12..24 V
   |                     |                      |
   |                     |                      /
   |                     |                      \ R2
   |                     |                      / 2k2
   |            VT1      |                      \
   |            IRF510   +--|                   /
  ---             ...       |                   |
  --- C5        IRFP460  +->|L---+--------------+
   | 100n                +--|    |              |
   |                     |       |              |
   |          +-----+----+       |  C4          |
   |          |     |    | R1   --- 10n         |
   |          |     |    / 3.0  ---             /
   |       C2 |+  C3|10n \       |              \ R3
   |         ---   ---   /    +--+---+          / 1k
   |         ---   ---   \    |      |          \
   |          |     |    /    |      |          /
   |          |     |    |    |C6    )  ---+    |
   |          +-----+----+   ---   L2) ( L1|    |
   |                     |   ---     ) (  ---   |
   |                     |    |      ) (  ---   |
   |                     +----)------+ (   |C1  |
   |                          |      )  ---+    |
   |                          |      )     |    |
   +--------------------------+------+-----+----+--------o GND


Judging by the output power, by the variety of the transistors that agreed to cooperate, and by the easiness of starting, the undoubtful leader is the Heartley scheme. This result is more than expected, however. With the HF bipolar transistors it was the same - if wanting highest power from the simplest circuit, get ready to make the inductive tri-spot.

The looser is the capacitive tri-spot. The heavy or slow MOSFEt's don't want to oscillate here at the highest frequencies. The lighter ones (a la irf510 - irf840) are happy to work here but the power does not impress. The network suffers from the skew of the main tank circuit, when trying to increase the oscillations voltage by reducing of C1c and increasing C1a and C1b.

The Meyssner's one did give intermediate results. Ideally it should not be worse than the inductive tri-spot, but in practice, when one tries to optimize the turns numbers and positions of the L1 and L2 coils, the scheme has a strong tendency to switch the modes. The mode of oscillations having the L1C1 tank circuit as the frequency determining one is often less stable than the one that Uses L2 and the source to gate capacity of the transistor or the one that uses L3 with the drain-to-gate capacity.

Why to want the external tank circuit rather than rely on the internal one? It looks much simpler to attach some coil in parallel to the laser cell to form a tank circuit (of the frequency unknown:) and to couple an exciter to it, that can oscillate at the native frequency of the laser cell + coil system. The alternative is to make an oscillator with a pre-set frequency, and expend tons of efforts in order to tune the laser cell + coil to resonance with that pre-set frequency. And one is lucky enough if having only one tank circuit in his oscillator.

The Hartley circuit (at least in the form pictured above) has at least three modes:

  • the one having L1C1 tank as the frequency determining circuit,
  • the one having L2C6 tank as the frequency determining circuit,
  • the one that determines its frequency by means of some part of L2 coil and the internal capacities of the transistor.

Amazing, but the mode with the external frequency determining tank appears to be rather stable. Especially when the Q-factor of the external tank is high and the ratios of frequencies of the external tank and the internal one are close to some integral number.

The photo and video below show the example of the inductive tri-spot oscillation "on the external mode". The frequency is 1.6 MHz. The native frequency without that "Tesla column" was 3.65 Mhz. The power supply params were: voltage ~50V, current ~1.5 Amps.


The internal microphone of the camera suppresses bass tones and the impression after watching the video is poor when compared to the one when looking at this with one's own eyes. Naturally this "brushy" hisses and growls rather scary. Note the needle pin, cooling from the white hot state after switching the oscillator off. Also note the splashes of the molten metal (looking like orange sparks) when it operates.


2.3. ZVS push-pull.

           R1             +---+----+
         50..100 Ohm      |   |    |
      +---/\/\/--------+  |   |    ) L1 2x0.5microHenry
+24V  |                |  |   |C1  ) 
 <----+    R2          |  |  ---   +----+
      |  50..100 Ohm   |  |  ---   )    |
      +---/\/\/---+    |  |   |    )    |
 <-+              |    |  |   |    |    ) L2 >100 microHenry
GND|    +---------)----)--+   +----+    )
  ---   |         |    |           |    )
   -    +---|<|---+    +----|>|----+    )
        |  D1      \  /    D2      |    |  +20..100V
        |  HER506   \/    HER506   |    +---->
        |           /\             |
        |    U1    /  \      U2    |    +---->
 U1,2   +--|      /    \        |--+    | GND
IRF540     |     /      \       |      ---
...     +--||   /        \     ||--+    -
IRFP460 +--|+--+          +----+|--+
        |      |          |        |
        |      \ R3   R4  \        |
        |      / 10k  10k /        |
        |      \          \        |
        |      |          |        |
               | GND

The schematics is interesting by the fact that it is not a pure self-sustaining oscillator. A 24V external direct current source serves as the gate driver here. It charges the gates through R1 R2 resistors and then they discharge through D1 D2 diodes and very low drain-to source resistance of the open transistor of the opposite shoulder.

While the diodes stay alive and the auxiliary power source does not fail the voltage at the gates won't exceed the preset one (24V) and hence one can omit the protection zener diodes.

The main part of the circuit (drain network) can be fed by a sufficiently higher voltage and current. Do note, however, that the L1C1 tank circuit tends to oscillate with a high amplitude and the total voltage at the sources can exceed the feeding one by 2..3 times. Therefore if one uses, say, a transistor with 400V source-to drain allowed voltage, then the circuit should not be fed by a voltage higher than 130V. If one uses lower voltage rated FET's, like irf 540 or irf3205 the limitation to the power supply becomes like 30 V.

The output is suitable to be made by means of an inductive coupling. One can wind a coupling coil (1..2 turns) above the main coil and connect loading there. By using high voltage transistors (irf840, irfp460), by precise choice of the load and by adding a small ballast resistor (of about 10 Ohms) one can attain the operation of this circuit directly from 220v mains (with some rectifier). It allows not to use a powerful transformer or PSU.

The circuit is proof to the loosing of oscillations. One can note that if we exchange the inductive load by a resistive one the circuit will turn out to be a well known trigger. And this one can have only two stable states. In any of them one of the transistors is fully open while another is fully shut. In any case the power dissipation is about zero. So there would be no overheat in case of oscillation fail. Apparently with no oscillations there will be no overvoltage. The sole thing that can suffer from this situation is the power supply unit (in most cases it load ability is lower than the top current those mosfert's are rated to).

The L1C1 tank circuit makes the trigger astable and turns it into an oscillator. The circuit can operate at any frequency of the L1C1 tank network provided that the gates are charged in time through the R1 R2 resistors. In my tests a small 24V power supply (actually another FERON with a rectifier) was used to drive the gates. In those conditions IRFP460 succeeded to do 480 kHz and IRF540 did 1 MHz.

A curious fact: the circuit is fully operational when fed by auxiliary power supply only. I.e. with the main power unit being disconnected. In this mode the circuit oscillates (at the natural frequency of L1C1) and produces noticeable power - several watts of output. However it is too small for this circuitry. And when one connects the main power supply the output rises drastically.

The circuit can also operate when fed without L2 choke. The choke does simply limit the consuming power.

The most interesting property of the circuit is indeed the low heat of the transistors. At any tested levels of power they stood cool. The switching mode allows them transfer tens of volts and amperes without notable heat. It makes an impression that one can easily get it to very high power at reasonably high frequency.

However the impression is fallacious. The fact that the transistors stay cool does not mean that everything else is cool too. The resistors R1 R2 for example are hot enough to boil a cup of tea. But this is still not the main problem. A kettle is finally a kind of resistor too. The most unwanted thing is that D1 D2 diodes are hot to crackle and smoke. And the overheated diodes tend to increase the reverse current. And it causes leakage of the high voltage from the oscillating powerful tank circuit to the tender gates of the transistors. The most pitiful thing is that the diodes appear to be alive after all (after cooling down) and the transistors are not.

Among the things in access the best results were with HER type diode (HER506 etc.). Even when their plastic was scorched and their legs were cindered they continued to work. On the contrary the (seeming to be) powerful 10A10's caused the death of many MOSFET's. As the mockery they have survived themselves. Shotckey diodes from computer PSU's (like 20FQ040) have low reverse voltage and still play hot dogs.

As the result ZVS push-pull has shown itself like beast-powerful but low frequency device. And the limitation to the frequency is set mostly by the diodes. Transistors with light-weighted gates (2nf IRF510) can operate up to 1 MHz. The heavy weighted ones (4nF, IRFP460) did it only up to 0.5 MHz. But it is only for a short time. The prolonged operation is possible only at the frequencies 2-3 times lower.


2.4. The common push-pull oscillator.

           || C4 47nf
    |      ||   |        Q1 IRF510, IRF540N, IRFML8244, etc.
    |           |
    +----/\/\/--+---+ +  +-----------------------------+------+
    |   R1 3 Ohm    | V  |       C2 1 nf               |      |
    |              --------         ||                 |      |
    |                --         +---||----+            | L1   |
    |   R3 1k        |          |   ||    |            |      |
    +----/\/\/-------+          |         |       2 trn )     |
    |                \          |         |             )     |
    |                 \       --+--/\/\/--+-------------+     |
    |                  \     /                    2 trn )     | C1
    |                   \   /    R6 3k3                 )    ---
    |                    \ /                       +----+    ---
    |                     \      R4 3k3            | 2  )     |
    |                    / \                       |trn )     |
    |                   /   \---+--/\/\/--+--------)----+     |
    |   R5 1k          /        |         |        | 2  )     |
    |                 /         |   ||    |        |trn )     |
    +----/\/\/-------+          +---||----+        |   |      |
    |                |              ||             |   |      |
    |                --          C3 1 nf           |   |      |
    |              --------                        |   |      |
    |   R2 3 Ohm    | ^  |  Q2 IRF510, IRF540N...  |   |      |
    +----/\/\/--+---+ +  +-------------------------)---+------+
    |           |                                  |
    |      ||   |                                  |
    +------||---+                                  |
    |      || C5 47nF                              |
    |                     ||  C6 47nf              |
    |                     ||                       |
    o                                              o
   GND                                           +12..24V

The circuitry resembles the ZVS one shown above. But the diodes are changed to capacitors C2 C3. And the gates are "hanged between the ground and the sky" by means of biasing resistors R3/R4 and R5/R6. As usual the LC tank circuit tends to swing to double or triple of the feeding voltage. So the feedback capacitors C2 C3 are better to be attached not to the drains of the opposite shoulder transistors, but rather to some tap of the L1 coil. As the bonus You will get lesser influence of the gate capacity onto the resonant frequency. As the result the L1 appears to consist of four sections. (In my case each section contained two turns of 1 mm copper wire. It means 8 turns total. And all this thing is wound over a rod, having the diameter of 11 mm. Over a glue gun rod to be more precise. After winding the glue gun rod was extracted from the bore of the coil. With C1 capacitor rated to ~1nf and irf510 transistors this oscillated between 11,,13 MHz. I've heard that a technological radio band is placed there. By reducing C1 and L1 the frequency could be pulled to about 22 MHz. Further on the circuit goes to a blocking-oscillator mode and the resonant frequency of L1C1 has low effect on the resulting frequency. The natural frequency of this HF blocking oscillator appears to be at 7.5 MHz with irf510.

On the contrary to the ZVS circuitry, here the transistors do heat. If one sets the biasing on the gates to be lowest possible, still providing the self start of the oscillations, one can have the efficiency of 60..70%. Nevertheless with the subtle irf510-s the RF output of the oscillator was enough for the full incandescence of some 12V 20W bulb. Transistors' temperature was acceptable.

THE RESULT: the circuit can operate at the frequencies needed. But 20W are kinda not enough for a good external laser ionizing source. It seems good to make the similar scheme but with irf540N or irfp460A, however I've got none of them available with the necessary letter. And the simple ones (540 and 460 without proper letter) aren't capable to operate above 3..3.5 MHz. It takes too much time to wait for them from the Internet and the local stores again have no transistors with the proper letter. Moreover in the local stores the price for a piece is like the one for ten pieces if buying from Ali. ...And also in addition to the transistors one should get the protective zener diodes (1.5KE12...1.5KE350 suppressors).

The said above makes one to think that this approach is not affordable enough for a DIYer. Mabbe those ones, who say that the RF technique is too complicated, were right... I will certainly continue to build a good RF source and occasionally will put here the results, but I can not recommend to follow this way for those ones who have not enough experience with high frequency power electronics. And for those ones who have this experience, maybe it would be simpler to take their transmitter, to attach a 200 W antenna power amplifier, to set a simple impedance matching transformer between the laser cell and the amplifier and finally to press the talk-listen button to send the power into the laser.


<continued 04.11.2017>

The parts from the e_b_@_y take long time to arrive. And it is hard to order exactly the needed ones from the first try. One needs to foresee everything in advance. Or else it appears that while the parts for 20 wt rated scheme are still on their way the laser already asks for 40 wt. And while the parts for 40 Wt scheme are coming it appears that 100 Wt needed.

This way it turns out that the power electronics is a slow work. Nevertheless from march till november a way from watts to kilowatts was found. Generally if one had said me at the very beginning that it would take a kilowatt rated RF power source to ionize this laser, I would send all this branch of the research into the erotic style journey with the special wish for it to stay thee forever.

But what's done is done, and at the end of this (a bit longer than wanted) radiotechnical interlude I'm presenting the




In principle the same scheme was already presented above, but that time I had no proper parts and lack of the experience, so the output and the reliability were far from the needed ones. Further an understanding had came that for the stable operation when powered from the mains, one really needs kilovolt rated transistors. Yes, fine ballasting and loading can help to reduce oscillations and to keep alive even irfp460-s. But the laser cell (with and without the discharge) is far from the ideal type of the load. It tends to suddenly change its impedance more than by one order of magnitude. And using some parallel resistive loading or some snubber-type schemes give intolerable heat and even more requirements for the output power.

Generally as the understanding has come that the circuit has to be working at the free run, immediately there came the success.

                R1              +---+----+------------------------+ 
                 25 Ohm         |   |    | C1 = 3n3 KVI-3 x 10 kV |
            +---/\/\/--------+  |   |    ) L1 10 microHenry       |
+15V        |                |  |   |C1  )                        |
 <----------+    R2          |  |  ---   )   +------+             |
            |    25 Ohm      |  |  ---   )   |      |             |
            +---/\/\/---+    |  |   |    )   |      |             |
 <-+                    |    |  |   |    |   |      ) L2 > 100uH  )
GND|          +---------)----)--+   +----+---+      ) 100uH < L3  )
  ---         |         |    |           |          )             )
   -          |         |    |           |          )             )
              |         |    |   ||C4    |          |             |
              +---------)----)---||------+          |             |
              |         |    |   ||      |          |             |
              |         |    |           |          +-------------+
              |         |    |           |          |
              +-|<|-|<|-+    +-|>|--|>|--+          |
              |  D1, D2  \  /   D3, D4   |          |  +310V
              |  STTH806  \/    STTH806  |          +---->
        +-----+           /\             +----+
        |     |    U1    /  \      U2    |    |     +---->
 U1,2   |     +--|      /    \        |--+ C3---    | GND
IXFK   ---C2     |     /      \       |      ---   ---
24N100 ---    +--||   /        \     ||--+    |     -
        |     +--|+--+          +----+|--+    |
        |     |      |          |        |    |
        |     |      V D5     D6V        |    |
        +-----+      -          -        +----+
              |      ^          ^        |
D5, D6:       |      |          |        |  C2=C3=800pf, Teflon diy
1.5KE18CA     +------+----------+--------+
                     | GND                  C4=2200 pF Teflon diy

You can see that the circuit is simple enough for the first glance. One only has to keep in mind that for its building the printed board is just not enough. It takes to use a drill, a hacksaw, a filer, screwdriver and wrench...

The transistors are any MOSFETS rated to not less than 1 kilovolt of drain voltage and for the currents not less than 20 Amps. One should also pay attention to the channel resistivity in the open state. It should be less than 1 Ohm and generally the lower the better.

In the reality there are not so many types of the suitable transistors especially having the suitable price. Practically any of them are products of the Ixys corp. The transistors are mounted onto some heatsinks from the old Pentium processors. The coolers are installed but weren't used yet.


The diodes have 15 ns recovery time and I was able to find only 600V rated ones. And those are tandem type i.e. they contain two diodes in series inside one unit. In order for them to hold the free running voltage of 1 kV I connected two in series, meaning the real stack of 4 physical diodes in each shoulder. The resistors of the gates feeding circuit are four white ceramic rectangular bricks having 100 Ohms x 5 wt for each gate. Such an assembly heats in reasonable limits and has suitable inductance. The gates are connected with sources by TVS protective diodes, rated to 18 v. This helps to keep the gates alive in conditions of high voltage leaks and interferences.

That's all the commercially available parts. Other parts are diy homemade. A pair of words on the capacitors. In principle the C1, C2, C3 and C4 capacitors together do form the sole tank circuit capacity. However its a bad idea to unite them into the one. At those currents the scheme operates with, the tremendous affect is given by the stray inductance (of the current path from the transistor's crystal to the nearest capacitor). Have it 40..60 nH and You will drown in the gate ringing with the corresponding heating and so on. On the other hand its hard to find the proper low ESL high voltage capacitor rated to several kilovolt-amperes and having the suitable size to mount it directly between the drains with the lowest possible inductance. Therefore the tank circuit capacity was divided into three parts. The first "snubber" part - C2 and C3 capacitors - has to be attached directly between drains of the transistors and the common wire (which should certainly be wide and thick). The second "constant" part - C4 capacitor - is to be mounted directly between drains of the transistors. And only the third "tuning" part - C1 capacitor - can be placed more freely.

C2, C3, C4 capacitors should be mounted DIRECTLY TO THE LEGS OF THE TRANSISTORS using the WIDEST POSSIBLE metal shims, soldered with a wide seam. These capacitors must have the lowest possible stray inductance and the working voltage not less than 1 kV. I could find no commercial product of reasonable price that could suit. Neither mica nor ceramic ones. So I had to roll the capacitors by hands. Using the aluminium foil and teflon ribbon. If You are interested in how to do this You may refer to the pocket sized dye laser guide where the "candy" type capacitors are described. However one should avoid mylar here, since it has noticeable dielectric losses at ~1 MHz, which lead to heating melting and breakdown. On the contrary using the teflon allows the capacitors to stay only warm (i.e. not "hot") at the nominal power. The free rims near the border of dielectric should be not less than 10 mm. Yes here is "only one kilovolt" but also "the whole million of Hertz ", so here are the good conditions for any kind of sliding or brush discharges.


L2 and L3 chokes are to endure 5 Amps of current and have more than 100 microHenries of inductance. Luckily one may avoid ferrite cores here. Wind somewhere about 100..200 turns onto some suitable ceramic or glass tubing. Both chokes can be easily placed on the single tube (2 sections 100 mm long if the wire has 0.3 mm diameter and the tube has 25 mm diameter). In practice the chokes stay almost cool. However, just in case, don't wind them on a plastic tubing.

And finally here is the description of the main coil. The coil is the most complicated element of the circuit and the most hard to be done. "The state of the diy art and the miracle of the garage-kitchen technology." This is due to the fact that You can not choose to large diameter - it will lead to the increasement of the length of the wire and the losses will grow intensively (worth to mention, that the radiation losses increase with the diameter too and it means the interference that bother others and even problems with the law). The second reason is that it would be hard to obtain high mutual inductance when You are trying to collect the power by some secondary coil, when their diameter is large and the length is short, On the other hand You can not choose the diameter to be small. First of all due to the breakdowns, and second - because the coil would melt from heat.

The main problem is certainly the heat. In the inductor ovens they solve it simply by making the coil of some copper tubing and by letting water to flow through this tubing. For the obvious reason it is better to avoid any dependence to the tap water, so it means very very thick wire, coreless coil with its turns being "transparent" to the natural air flow.

The coil is wound with aluminium wire having 16 square millimeters of section. There are two layers of winding, The first layer contains 9 turns and can be wound on some plastic 2 inch sewer's tube. The outer layer contains 9 turns too and can be wound on some glass or bottle having 65 mm diameter. (Keep in mind that direction of turns is to be the same in both layers.) Than the parts of the coil are to be removed from their frames and to be assembled into the single double layered coil. The electric connection is made by bolting together the leads of the parts by means of M2 bolts and nuts. To ensure the layers do not touch each other insert mica insets between the layers.

AlumeCoil_1 AlumeCoil_2

This is how the complete coil does look like.


In the process of the development many samples of coils were burned out. (No transistor was lost (!) by the way.) And even such a monstrous coil stays hot over 100 celsius degrees during the free run of the scheme.



The circuit requires two power supplies. One of them is made as rectifier of the mains line voltage and the other on is some independent unit rated to 12..15 volts and 2..4 amps.

Here is the scheme of the mains rectifier:

          ----  Fuse 5-7Amp
      |   ----  |
 <----+         /
 ~220V 50 Hz    \ thermistor 
 Wall plug      /   7 Amp
 <----+         +
      |        / \  Diode Bridge: KBU1009 or 4 x 10A10
      |       /   \
      |   +--+ |>| +------+------------------> to +310 v powering wire
      |   |   \   /       | +
      |   |    \ /       ---
      |   |     +        === 1000 uF x 450 V 
      |   |     |         | 
      |   +-----)---------+------------------> to GND of the generator
      |         |         
      |         |

The rectifier consists of diode bridge rated to more than 10 Amps and more than to 500 Volts (one may solder it using four suitable diodes, e.g. 10A10) and of ripple filtering capacitor. The latter should be 500..1000 microfarads x 450 volts. Good advice: don't choose electrolytic capacitor rated to 300..350 volts here. The rectifier operates in rather heavy conditions and such a capacitor may just overheat and blow.

Thermistor is needed for the capacitor charging current not to blow Your fuse plugs and not to kill the diode bridge. IT IS A MUST TO USE A FUSE IN YOUR POWER LINE! The thermistor can be obtained from some old ATX computer power supply unit. Here's how it looks like:



In principle the fuse + thermistor combo may be replaced by a sole halogen lamp, rated to 220 V and 1000..1500 W. However the power output would be lower this case.


The gates' power supply can be made of some halogen lamp electronic transformer unit. Use some 40..60 W one. The electronic transformer is used here without any upgrades or modifications. Just load its output to a diode bridge and ripple filtering capacitor rated to 500..1000 microfarads x 50 volts.

 <-------------| Halogen lamp      |------------+  4 x HER308
  ~220V 50 Hz  | electronic        |           / \                 to +15v wire
  wall plug    | transformer       |          /   \            of the generator
 <-------------| 12v x 60 W output |---+  +--+ |>| +---+---------------------->
                -------------------    |  |   \   /    |                 
                                       |  |    \ /     | +
                                       +--)-----+     ---
                                          |           ===
                                          |            | 1000 uF x 50V
                                          |            |
                                                        to GND of the generator

The bridge diodes better be fast enough. Choose HER or FR series. Yes, they seem brakish when compared with STTH, but still fast enough when compared to KBU's or 10A10.


Tuning up the circuit is easy if You have an oscilloscope and, say, variac, or another 50..100 V power source. Initially attach the oscilloscope in parallel to one of the protective surge diodes connecting gate and source of one of the transistors. Then switch the gate-feeding power supply on. The gate voltage trace should have the shape of "shark fin" - a quarter of sinewave being distorted slightly by Miller effect.

shark_fin1 shark_fin2

Connect then the main (but lowered) power voltage. The Frequency and amplitude of the gate voltage oscillations will increase, but generally the shape of the trace should stay smooth. If there appears some "bloody knead", You should better not connect the circuit to the full voltage of the mains (310V).

Having ensured than the voltage on the gates has normal shape, attach the probe to one of the drains. Use a resistive divider here (1:100 is good) or else You will end up with dead oscilloscope. The correct drain voltage has the bell-like shape (well spaced halves of sinewave).


If there are glitches in the form of stray high frequency oscillations or inductance interrupt peaks, the circuit is also not ready to operate at full power.

What can one do with the incorrect shape of pulses? The main way is to revise the geometry of the circuit. Lessen the stray inductances. Any those snubbers or resistive ballasts are out of help here - the power is too high.


If You have no oscilloscope - well, just design the scheme to intrinsically have the lowest possible stray inductances. To wire the circuit use shims of the lowest possible length and the widest possible width. Solder the parts with wide seam, don't use pinpoint soldering.


When You turn the circuit on for the first time, use only the gates' feeding power supply. Put an incandescent lamp probe into the main coil. Such a probe is just a low power incandescent lamp (like 3 V x 0.12 Amps) soldered to a pair or three loops of a copper wire. Kinda like that:

    |          |
   (           |
   (   +--(X)--+
   (   | Incandescent lamp ~3V ~0.12 Amp
    |  |


The lamp should glow when only the gates feeding power supply is switched on. If it does not, - look for an error in the circuit. If it does, try to add the drains supply power carefully. Keep an eye to the heating of the transistors. If they begin to get hot the scheme probably requires rebuilding. Just for information, my sample of this device showed the temperature of the heat sinks just below 30oC just after heating the movable part of the tabletop vise to red hot state. And it is still without turning on the fans.



One may touch the heats sinks only after (all) the plug(s) have been taken out from the receptacles. It seems dangerous, but when we use the circuit to feed even more dangerous device (like high voltage laser) it should not bother too much. Just keep in mind that the device is not only "untouchable", but in most cases one should treat it as "unapproachable".


The device can be used as an induction heater. However to reach higher temperatures its coil should be revised towards more compact design and more dense field. It may require water cooling though.

red_hot_bolt02 red_hot_bolt03 red_hot_bolt04 red_hot_bolt05 red_hot_bolt06


One may used it as a powerful high voltage power supply. In order to get out the power just put a secondary coil over the main one. The circuit can incandescent a halogen lamp rated to 220vx1000W. Here is the sequence of shots, taken while the device heats a kilowatt lamp. Do note that room lights weren't darkened. The images get darker and darker due to the fact that camera reduces its exposition in trying to compensate the enormous brightness of the kilowatt lamp. I.e. the darkness of the last photos can give You indirect sense how bright the lamp was.

kilowatt01 kilowatt01 kilowatt04 kilowatt05 kilowatt06 kilowatt07



Keep in mind, however, that the lamp of such a power behaves just like a shortcircuit when cold. In order for not to break oscillations one will need to limit the coupling between primary and secondary at first, and then
gradually increase the coupling.


When using the circuit as a kind of high voltage power supply one may want to make a "spark test" of the secondary coil. I should warn You that such a test is very dangerous. In a second inches of wire turn to vapours and their place takes a hot arc spaying dozens of droplets of molten copper. I won't place a photo of this because I have no wish to reproduce this again. And do note: after such an explosion not only the transistors stay alive, even the fuse stays not blown off. It means that this fireworks takes place at nominal level of output power.

<end of interlude>



Hereby the new generator of external ionization is present. But the laser lives not so happily. The last of the laser cells (the one with sectioned copper electrodes 4 mm thick) has dried up shrunk and began to leak. Now I'm trying to cure the leaks with silicone. And while the silicone is drying, another cell was scavenged from old stock - the one with aluminium T-profiled electrodes.

For the try with this cell The external ionization was connected using "electrodeless" schematics (not the "barrier" one). The RF voltage is taken from the megahertz kilowatt ZVS push-pull unit by a secondary coil having 90 turns of 0.5 mm enameled copper wire, wound on a frame 85 mm in diameter (a core from some sticky tape bobbin).

The mixture was CO2:N2:He = 2:5:14.

As the result it was got that even only the RF power was able to heat the laser cell up to the temperature painful to touch (~60oC) in less than two minutes. If it is due to the heat deposition in the plasma (not in aluminium electrodes or either in silicone seals) it means that the laser should be able to operate when being powered by RF only.

The laser was aligned and tested. No lasing was observed when fed by RF only. When the constant current from microwave oven transformer with its rectifier was connected to the laser cell it was easy to reach the limits of the milliammeter scale (~200ma) without falling to arcing.

The shunt of the milliammeter was replaced by one half of an Ohm (was one Ohm). After this its calibration in statics gives scale division value equal to 46 mA/div. The ballast resistor was also reconfigured. Now it consists of two resistors (1 kOhm and 1.3 kOhm) in parallel. It yields 565 Ohm.

Using the mixture CO2:N2:He = 2:5:14 at the pressure of 1.1 inch of mercury under the current of 6 divisions (~276 mA average if trusting the static calibration) the laser gave 550 mV by the homemade Peltier calorimeter indications. Taking its calibration factor as 6.4 mW/mV we can estimate the output power as 3.5 W.

Here we should note that the calorimeter is not intended for use with so high powers and one should not treat this value as precise. However due to the fact that most of the factors of possible non-linearity tend to decrease indications, this value is most probably a kind of lower estimation.



In addition to the written above there was one more modification of the circuit, not mentioned in my previous record. The smoothing capacitor of the rectifier (the one, rectifying MOT output) was increased to 50 mcf. Those four microwave oven caps, earlier used, are just not enough for the current currents.

The described above tests showed that the current of dropping to arcing was 6 divisions of milliammeter and the laser threshold current was 4 divisions. Curious situation: the current was nearly 10 times higher than the one in early experiments but there was no success in trying to go far away from the threshold. To comprehend what's going on a diy kilovoltmeter was installed in parallel to the laser cell. It was found out that the voltage across the cell barely reaches 800 volts even at 2 inches of mercury of pressure. I.e.

RF discharge from the kilowatt rated ZVS generator ionizes the gas to such an extent, that its resistance becomes low and the voltage drop becomes too low too.

What can be done with that? To try to feed the main discharge with RF current too? OK, a second sample of the kilowatt rated ZVS generator was built. Quite similar to the first one.



It was attached as follows:

                       |                |
                       |              =====
                       |         +--|-     -|---+
                +--+   |         |    =====     |
    ---------   |  |  (          |      |       |
  | KW & MHz |--+   ) (     +----)------+       |
  | rated    |      ) (     |    |              |
  | ZVS #1   |--+   )  |    |    |              |
    ---------   |  |   +----+    |              |
                +--+             |              |
                                 |              |
          -------------          |              |
         | KW & MHZ    |---------+              |
         | rated       |                        |
         | ZVS #2      |------------------------+

Its clear that the shunt capacitors across the cell were removed (because they'd short the RF currents). And also there was no ammeter or voltmeter, since they're not able to measure the RF values.

A fresh mixture of CO2:N2:He = 2:3:15 was prepared for the tests. And a tiny
droplet of xylene was added to the gas vessel. It was obtained:

feeding type RF/RF;

mixture type 2:3:15 + xylene droplet

Pressure gauge readings
0.8"Hg 150 mV
1.5"Hg 220 mV
2.0"Hg 180 mV
3.0"Hg no lasing at all


Simply saying two ZVS'es did not allow to surpass the result obtained with feeding by direct current. For the comparison here are results, obtained with direct current using the following circuit:

                                 |      Cp=12nF |
              +----------+  +----)------+       |
              |          |  |    |      |       |
       -------+-------   )  (    |    =====     |
      | KW x MHZ unit |  )  (    +--|-     -|---+
       -------+-------   )  (    |    =====     |
              |          |  |    |      |       |
              +----------+  +----)------+       |
                                 |              |
                                 |       /      |
                   1..3 kV from        /        |/
               Microwave Oven Transformer     (mA)
                  with Rectifier               /|
                                  Rb=565 Ohm

All other conditions were the same (the same position of the alignment, the same laser cell, the same speed of the gas flow):

feeding type RF/DC;

mixture type 2:3:15 + xylene droplet

Pressure gauge readings
1.5"Hg 480 mV (3.1W)
1.6"Hg 460 mV
2.0"Hg 360 mV
1.5"Hg 180 mV - too slow gas flow


Here the top achieved laser cell current was 6-6.5 divisions (276-300 mA). The voltage drop across the cell was not seriously measured (too fast tests, due to the lack of water cooling), bit it had the order of 800 volts. At the same conditions but with 2:5:15 mixture having also 0.1ml of xylene it did 300 mv at 1.0"Hg of pressure. And it was almost the optimum.


The most important problem of the RF ionization is difficulties in the attempts to hit the resonance with the tank circuit formed by the self capacity of the laser cell and the inductance of the secondary coil of the oscillator. When there is initially no plasma in the cell, and it ignites only due to the applied ionizing voltage (so the capacity is proportional to the charge of the capacitor), one can show (by calculations in modeling programs) that this circuit has no resonance for harmonic oscillations. And if the plasma glows continuously, the capacity depends to the area of surface covered with this plasma, and to the parameters of this plasma. When one approaches the resonance, this area increases and the resonant frequency changes. And I'm not sure that a phase lock circuit could help here. Moreover the complexity of the phase lock loop contradicts the "KISS" principle.

Exactly due to the difficulties with resonance match it was needed to build a kilowatt rated RF oscillator, when it needs only 20-50 W of the absorbed power to reach the ionization needed.

Another way of getting the RF voltage for the external ionization is to connect some inductance to the laser cell, to kick this tank circuit by some high voltage pulse and to allow it to oscillate freely. Exactly this approach was taken by the authors of the works cited in the beginning of this web-report. I can not say this way is more effective. Its just another. When the authors write that the energy deposition of the auxiliary discharge was from 20 to 50 W, so the initiating energy was not more than 10% - it still does not mean that the energy consumption of the auxiliary discharge generator was also low.

Indeed, let's take the capacity of the laser cell to be 600 pF. And as it is said in [2] the amplitude of the ionizing pulses was 10 kV. It means that each pulse the cell was charged up to 10 kv and accumulated CU^2/2 = 30 mJ of energy. If we multiply this energy by the repetition rate we'll get

30 mJ x 30 kHz = 900 W.

Yet again when reading scientific articles the things look admirable, but they become not, when one tries to reproduce. However in our case this is ignorable, because no one gave promise that the efficiency would be high. And it is interesting what results can give the usage of the "RF interference generator." (Yes, the pulses of high power and fast decaying oscillations are the classic form of the interferences. I warned You - keep an eye on the integrity of Your TV sets and cell phones while working with this thing.)

The schematics of the decaying RF pulses generator is as follows:

                                        C3-C6 mica 0.05 uF x 500 V  
                  R1 300 Ohm x 300W   ||       ||
          |         |             |   ||   |   ||   |
 ~220V    |        --- C1         |        |        | Transform ratio 
  50Hz    |        === 220uF      |   ||   |   ||   | 1:10           
          |         |  450V       +---||---+---||---+     +---------> _/\_
   o------)---------+             |   ||       ||   |     |       Output pulses
          |         |  C2         |                 |  W (        up to 2 kV.
          |        --- 220uF      +--|<|-------+     ) W (
          |        === 450V      /  HER608     |     ) W (
          |         |          ||              |     ) W (
          +--|<|----+      +---+|              |    |  W (
                    |      |     \ IGBT 20N120 |    |     |
                    |      |      |            |    |     |
            _    -> >------+
          _| |_ Rectangular control pulses:
                30us-off  10us-on
                9-12V amplitude


It is rather simple. A wide spread and kinda cheap FGA25N120AN transistor is used as IGBT here. I won't draw the control circuit - it is trivial and can be made easily using some 555 timer, some NOT gate and some IGBT/MOSFET driver (like TC4420). The control circuit yields rectangle pulses with the amplitude of 9-12 V (dependently to the feeding voltage of the control circuit). The duration of mark is about 10 mcs and the duration of space is about 30 mcs, giving together the frequency of ~25 kHz.

The transformer is wound on two U-shaped halves of a TV-set flyback transformer core. Both halves are set IN PARALLEL, i.e. form a LARGE U-shaped core, not a small O-shaped one. The wire is rather thick (~0.5 mm in diameter). The primary has 5 turns and the secondary has 50 turns.

Energy storing capacitors are of mica transmitting type. These ones have low RF losses. The ballast resistor is 3 water heaters (rated to 500 W each) connected in series.

When the transistor turns on, the output of the transformer carries a pulse of fast decaying RF oscillations, having their frequency ~200 kHz. During the full period the amplitude of oscillations falls by about three times. It is somewhat faster than desired bit sic.

The laser cell was connected using the next schematic:

                                       Cp 12 nF
                                 |                    |
             +----------+   +----)------+----------+  |
             |          |   |    |      |          |  |
     --------+--------   ) (     |    =====        |  |
    |RF Pulse Generator| ) (     +--|-     -|---+--)--+
     --------+--------   ) (     |    =====     |  |
             |          |   |    |      |       |  |
             +----------+   +----+      +-------)--+
                                 |              |
                                 |       /      |
                   1..3 kV from        /        |/
               Microwave Oven Transformer     (mA)
                  with Rectifier               /|
                                 Rb=565 Ohm


It did:

feeding type RF pulse/DC;

mixture type 2:3:15 + xylene tiny droplet;

Laser cell with Al T-shaped electrodes

Pressure Calorimeter readings
1.2"Hg 230 mV
1.5"Hg 330 mV
1.8"Hg 510 mV (3.2 W)
2.0"Hg 460 mV


At the pressure of 1.8 inch of mercury the lasing threshold was 160 mA, the top achievable current was 300 mA, and it was limited not by arcing, but by power supply possibilities. The voltage drop across the laser cell was about 800 V.

I.e. usage of the decaying RF pulses generator allows to obtain almost the same laser power, as it takes place when using the continuous RF generator, rated to kilowatt. But the generator of decaying pulses is simpler and cheaper. Concerning the power consumption, both schemes eat from the mains the powers of the same order of magnitude.



From the old stock a laser cell with copper electrodes was extracted.

sha_tube02 sha_tube01

The possibility of water cooling is completely omitted in the design of the cell, but still there was no water cooling with the aluminium cell too - due to leakages of the cooling vessels. So both cells give equal possibilities, with the exclusion of larger heat capacity of the aluminum one. On the other hand the previous tests had shown that the copper cell generally gives greater power. So it is possibly worth trying.

The cell was installed, aligns and connected using the next electric scheme:

                                       Cp 27 nF
                                 |                    |
             +----------+   +----)------+----------+  |
             |          |   |    |      |          |  |
     --------+--------   ) (     |    =====        |  |
    |RF Pulse Generator| ) (     +--|-     -|---+--)--+
     --------+--------   ) (     |    =====     |  |
             |          |   |    |      |       |  |
             +----------+   +----+      +-------)--+
                                 |              |
                                 |       /      |
                   1..3 kV from        /        |/
               Microwave Oven Transformer     (mA)
                  with Rectifier               /|
                                 Rb=565 Ohm

There were attempts to run it using the mixture of the next contents: CO2:N2:He = 2:3:15. Either with xylene or without it, but the laser refused to give the power. When peaker capacitor was of 12 nF one could achieve the current of 6 divisions of milliammeter and the output was about 200 mW. Then the peaker's capacity was decreased to 6 nf. After this the falling to arcing began to take place at the current of 3.5 divisions. The laser was just under the threshold. Then the peaker capacitor's capacity was increased up to 27 nF. But the current of falling to arcing did not rise significantly (when compared to the initial variant) - 6.5 divisions. The power was 500 mW.

And then exactly in this web-report I found a record that this cell has already refused to work with 2:3:15 mixture at least once. The mixture was diluted by nitrogen till the composition became ~2:7:15. And - YES - with this mixture the problems of the discharge stability just vanished. The cell yielded 2 watts and after some correction of the alignment it did 3 watts. In addition it was noticed that with this mixture there was no optimum for pressure. Ether no normal discharge (immediate fall to arcing) or three watts already.

In an attempt to increase the heat conductivity of the gas the mixture was diluted by helium. Till its content became CO2:N2:He ~ 2:7:25. Neither the power output nor the limiting current had changed. It was still three watts and 280 mA. No apparent optimum of pressure had appeared either.

Finally a new mixture of CO2:N2:He = 2:5:21 was mixed. (Initially one takes two volume parts of carbon dioxide and mixes them with five volume parts of nitrogen. Further one one takes one volume part of this mixture and adds three parts of helium. And finally one should add a small droplet of xylene per one automotive tyre of the gas.) The things became better. Here are the results:

feeding type Pulse/DC;

mixture type 2:5:21;

the cell with copper electrodes 3mm thick

Pressure Peltier readings Current
1.8"Hg 614 mV 6.5 div
1.6"Hg 660 mV 7.0 div (with some arcing)
1.6"Hg 740 mV 7.0 div (without any arcing)
1.2"Hg 620 mV 7.5 div
1.8"Hg 780 mV 7.2 div
2.0"Hg 320 mV 7.0 div


To illustrate what this laser is able to, a small video was made:


And here are some additional photos:

lazer_zzhet01 lazer_zzhet02


I think that the sausage has brilliantly shown that one should not put his fingers into the beam. And concerning the useful applications... In principle there are a lot of CNC engraving and cutting machines having 5 W rated laser diodes in there. They are able to cut thin leafs of organic substances and engrave metal surfaces. However this is rather long process and this laser is still not able for the long runs. It is because it has no water cooling yet. Well, if not trying to achieve higher powers, one could apply even some air fans as the coolers here.

Need to say that in the process of testing the rear mirror of the resonator was broken. It cracked due to the heating by light. It is too far from the main discharge to receive enough heat from plasma.

zerkalo_lomanoe_a zerkalo_lomanoe_b


In principle such a behaviour of the mirrors was more than expected from the very beginning. And there were some thoughts how to avoid this. It is more amazing, how long it took for the laser to grow up to approach this phenomenon.

Finally its power has reached 5 watts, i.e. about a half of the expected one from the very start of the project. One can certainly try to increase current and energy deposition further (by applying good cooling and good ionizer), but...

...But its time to unclose one more reserve of power: we need to increase the efficiency of the laser cavity (resonator). The deal is that the diameter of the semi-transparent ZnSe mirror is only 10 mm, while the width of the laser channel is as large as 30 mm. It means that the most simple scheme (design A on the figure below) can utilize not more than 30% of light.

The simplest solution would be the design B (again on the figure below) if I had a 30 mm ZnSe mirror with transparency of 5%...10%. However those mirrors are expensive, and moreover - they are too rare.



As a simple but compromise solution now I use the design C (see the picture above). The semi-transparent mirror is installed in a hole inside some larger and non-transparent mirror. In comparison with the design A it gives a 50%..100% surplus of the output with all the other conditions are the same. And nevertheless, due to the problems of alignment of the semi-transparent mirror with its enframing one; due to the losses on the borders of the enframing mirror; and due to comparatively large distance between the waveguide ends and the mirrors I believe that this design of resonator does still utilize not more than 50% of the useful light power. I.e. just by upgrading the resonator one can turn this 5W laser into a 10W one.

Great hopes are placed on the design D (see the figure above). It uses comparatively small mirrors and still the light sweeps all the active volume. However the design has its own problems. First of all one will have to learn how to install the mirrors directly into the laser cell. It is needed to avoid the end losses between the mirrors and the waveguide. And, say... four little alignment plates for a single laser, is it normal? Secondly: it is still unclear how to align such a laser, and whether it could be aligned at all.


• The power has reached full 5 watts. It means, the laser can be compared with most elite and expensive laser pointers.
• There is a reserve how to increase the output without any increase of the consumed power: just by redesign of the resonator.
• And there is still some reserve of stability, allowing to further increase the energy load. However a good cooling and more powerful external ionizer will be needed here.



SUDDENLY: the laser appeared to be able to operate with heliumless mixture. After the tests described above, the remains of CO2:N2 = 2:5 mixture (used on the first stage of mixing the helium containing one) were let into the laser cell. Here are the results:

feeding type Pulse/DC;

mixture type 2:5:0;

the cell with copper electrodes 3mm thick.

Pressure Peltier readings Current
1.2"Hg 360 mV 7.5 div
1.35"Hg 510 mV 7.0 div with gradual rise up to 9 div
1.5"Hg 360 mV 8.0 div


It means that now the generator of external ionization is powerful enough even to let the current pass through heliumless gas mixtures. 3.2 Watts of output look a bit humble when compared to 5 Watts, but no helium means more affordable.



Happy New Year to all of You.
Two new laser cells were assembled. The first one has copper electrodes, the other one - aluminum ones. Both cells have electrodes ~3 mm thick. The main feature of the new cells is that the glass walls of their laser channels protrude through the endpieces and mirror mounts almost to the mirrors of the resonator.


laser_Cu_cuvette_long_glass01 laser_Al_cuvette_long_glass02

laser_Al_cuvette_long_glass01 laser_Al_cuvette_long_glass02



Another feature is that the electrodes of external ionization have planar shape. This allows to mount large external heatsinks. (It was found unreasonable to dispose the heatsink together with each broken laser cell.)




The main sizes are the same for all the cells, beginning from the first ones described in this report and ending with these being described now. Namely: the width of the laser channel is 30 mm, the thickness of the waveguide (the spacing between the glass walls) is about 3 mm, the active length (the length of the zone occupied by the plasma) is 350 mm and the full length of the resonator (mirror-to-mirror) is 450 mm.


At first the cells were tested with resonator scheme C (see the picture above) with the same ZnSe mirror (92% reflectance) having 10 mm diameter and being placed in the center of the enframing aluminum concave mirror. The external ionization power supply was the pulse generator based on IGBT transistor (it was also fully described above).


The main results of the tests are the next:


  • The cell with copper electrodes had refused to bear current above 270 mA with CO2:N2:He = 2:5:14 gas mixture. In order to reach 350 mA and above it was needed to enrich the mixture with nitrogen. So the new mixture became CO2:N2:He = 2:7:18.
  • With 2:7:18 mixture the cell with copper electrodes appeared to be operational at pressures up to 2.1"Hg. The optimal pressure was 1.7"Hg. At this pressure and at 350 mA of pumping current the laser output was ~6W.
  • The cell with aluminum electrodes with 2:7:18 mixture was able to operate at pressures up to 5.5"Hg. The optimal one was 4.5"Hg. At this pressure and at 350 mA of current this cell was able to give ~6W too. In an attempt to increase power further the output mirror was killed:

    completely_broken_mirror01 completely_broken_mirror02


One can clearly see the molten crater in the glass. The size and shape of the crater implies that inside the cell there was much higher light power than those 6 watts detected outside.

Moreover the glass walls of the cell had crackled too:




And there is a probability that this coracle is not due the electric heat, but due to the heating by the laser beam. The crackle was sealed by silicone RTV and left aside for curing.


The output mirror was replaced by a germanium one having the reflectance of 70%. The germanium mirror has full diameter of 25 mm, but its reflecting part covers only a disc 15 mm in diameter. The enframing concave aluminum mirror was not used this time.


With the new mirror the cell with copper electrodes reached ~7W. The top and optimal pressures stayed unchanged: 2.5"Hg and 1.7"Hg correspondently. The result looks like good but...


The silicone, used to repair the cell with aluminum electrodes, had finally dried up, and the cell was tested with the new mirror. The result surpassed all expectations: the calorimeter readings were 1910 mV. 12.2 W! It was at the pressure of 4.5"Hg (114 torr or 0.15 bar) and with 350..380 mA of current. The voltage drop across the cell was ~900 V. The gas mixture was still 2:7:18.


It is interesting that in these tests a shortcircuit bridge between upper and lower ionizing electrodes was forgotten. I.e. the cell run essentially with only one ionizing electrode being connected. So the circuit was the next:

                                       Cp 27 nF
                                 |                    |
             +----------+   +----)------+             |
             |          |   |    |      |             |
     --------+--------   ) (     |    =====           |
    |RF Pulse Generator| ) (     +--|-     -|---+-----+
     --------+--------   ) (     |    =====     |  
             |          |   |    |      |       |  
             +----------+   +----+              |  
                                 |              |
                                 |       /      |
                   1..3 kV from        /        |/
               Microwave Oven Transformer     (mA)
                  with Rectifier               /|
                                 Rb=565 Ohm

The attempt to recover the error - to connect the second electrode - lead only to that the top current, achievable without arcing, became 270 mA and the laser barely reached the threshold.

Apparently with this kind of connection the circuit formed by the inductivity of the secondary winding of the transformer and the capacity of the electrode had its resonant frequency more close to the one of the circuit formed by the primary winding inductance and the storage capacitor of the pulse generator.

Need to say that the previous step-up transformer has unscrupulously burned out. the insulation between the windings has molten with the consequent breakdown and breakover. The transistor has survived however. The present step-up transformer is wound on the same ferrite core (double core from a TV flyback transformer of medium size). The primary has 6 turns of thick copper wire and the secondary has 55 turns of enameled copper wire having 0.5 mm diameter. The primary is wound above the secondary. The insulation between layers and windings is now made of sticky kapton ribbon (so called "heat resistant duct tape").


• The laser output has reached 12 watts; This is a bit more than it was expected at the very beginning of the project. And this is the level of small commercial laser tubes.
• The necessary vacuum is in range of operationability of affordable DIY vacuum pumps like reversed fridge or automotive compressors
• There probably still exists a reserve to increase the output even further


Considering the comparison to laser pointers. I admit that there is a video on Y_o_u_t_u_b_e that demonstrates a power of a laser pointer based on 7 blue laser diodes rated to 7w each. Totally 49 watts. It is not so easy to surpass this result for DIY laser even if it is a CO2 one. Another thing is that those laser diodes are essentially ready-made lasers and there is a little hope that one can produce them at home in the predictable future. And they also can not be treated as affordable and cheap parts easy to obtain by anyone. The things are not better with the powerful laser pointers themselves. The ones with output over 10 W are exclusive goods, they cost damned high and they are piece-work produced. (The scums are out of consideration. Only their labels cry out about their damned high power. The truth is that most of them give less than one watt.)


The power of the laser can be illustrated by the next short video:


DSC_0001 DSC_0004

DSC_0006 DSC_0017



By its unfocused beam it can:

  • Charr and burn wood. The output beam spot has rectangular shape, 18 mm wide and 3 mm high. So the power density is 22 W/square centimeter.


By its focused beam (lens with FL=85 mm) it can:

  • Can puncture a finger. Almost throughout in less than a minute.
  • In two seconds (after applying full current) it can puncture a 4 mm thick wooden board.
  • It can do a poker-work with as high speed as You can draw.
  • With proper mechanics and with proper mode it should be able to cut thin (up to 6 mm thick) wooden board. When done by hands it gives bad results, as You could see in the video. It is hard to keep a smooth and even move of focal spot. And when You miss, the wood becomes embraced with flames making it even harder, since it obscures the view.


Amazing, but the laser COULD NOT:

  • Drill a razor blade. And moreover, the reflected beam was so strong, that it was able to damage the plastic mirror mount of the laser. So it was decided to stop the headless experiment.
  • Drill through an 1.5 mm textolite from a printed circuit board. (Without the copper foil of course.) The textolite charrs, swells, hisses and stinks. But no via hole appears. So this laser still can not be used for PCB drilling.
  • Drill a microscope slider. There shines a little sun on the glass. The glass melts, flows down, crackles. But there's no via hole again.


It is interesting to note here that the previous experience shows that it takes only 350 mW of average power (50 mJ at 7 Hz) and 4 minutes of time for a TEA laser to drill through the similar microscope slider. The hole is large enough to thrust a 0.1 mm copper wire through it. It means that for the TEA laser 0.35*4*60 = 84 Joules were enough to drill the glass, and the CW laser was unable to do the same expending as large as 720 Joules (=12 W x 60 sec). Need to note that TEA laser in that experiment was fed by a "slow" gas mixture having the N2 to CO2 molar ratio above 3. It allowed to produce rather long pulses and to some extent avoid the unwanted energy dissipation in the plasma. (On the contrary if You aim is to obtain the laser spark, You should aim to as high power as possible and use a "fast" mixture with N2 to CO2 molar ratio of unity and below.)


The failure to drill the razor blade has shown two points:

First: KEEP YOUR EYES PROTECTED! The reflecting properties of materials for the deep infrared radiation of CO2 laser differ strongly from ones for visible light. And You can SUDDENLY get powerful and well shaped reflected beam from the surfaces not expected to produce it.
Second: The generic error in the design of the very first laser cell (from the ones described here) was discovered. Its electrodes had only narrow chamfers. Moreover they were polished. When the pumping power is high and the output mirror is too transparent it leads to development of parasitic lasing in the parasitic cavity formed by the reflecting surfaces of the electrodes and by the glass walls of the cell. This parasitic lasing depletes the energy without any use. In case of high reflectance of the output mirror one can succeed in obtaining lasing in external cavity, but the losses are too high In the new laser cell with aluminum electrodes this error is recovered. The working surface of the electrodes is rounded (by filer) and sanded. There are no planar surfaces and no polishing. Naturally this explains the better results: the absence of parasitic lasing and reduction of light losses on the way between the laser channel and mirrors gives the opportunity to use 70% mirror, more close to the optimal one.


The laser was tested with heliumless mixture, but the results were not so impressive. After some optimisation of gas mixture, pressure and current it was able to yield 620 mv by Peltier calorimeter indications (3.9W). It looks humble on the background of 12 W with helium.

The optimal heliumless mixture appeared to be CO2:N2 = 2:5 with the addition of a small droplet of xylene and a droplet of water per full automotive tyre of gas mixture (6..8 liters). Without the water there was only 520 mv (3.3W). The optimal pressure was 2..2.5"Hg.


Additionally on the way to assimilate the schemes with internal mirrors the laser was tested with a homemade mirror. This mirror was made by hand polishing of some aluminum plate. The mirror being tested was installed in place of back mirror of the resonator. The output coupler remained as 30% transmitting Ge mirror.


In this configuration and with helium based mixture the laser gave 112 mv (0.7 W).

A guess was made that the handmade polishing had not allowed to keep a needed shape (of course I failed to make the perfect plane) and the irregular shape hinders laser power. The mirror was replaced by a commercial one. This time it was first surface aluminum on glass mirror from some industrial optical device.

A new test run had shown 22 mv (140 mW).

Both runs were made using CO2:N2:He = 2:7:18 gas mixture at the pressure of 4 inches of mercury.

It is probably incorrect to say that the laser can not operate with plano-plano resonator. I'd say that plano-plano requires more precise alignment and in these tests there was no success in achieving the necessary grade of the alignment. But anyways do note this fact and try not to use plano-plano configuration if trying to reproduce this laser.





Wide output couplers for CO2 lasers are not only expensive, they are also rare. I failed to find one, having transmission 30% and the width of the working area 25 mm, at any reasonable (for DIY-er) cost.

It was decided to try to make a homemade output coupler. Because the DIY conditions generally do not allow to make a vacuum dielectric coating with good properties, it was decided to use a mirror with a number of holes. The transparency of such a mirror is approximately equal to the total area of all holes divided by the total (working) area of the mirror.

Obviously one should take one of those household or automotive mirrors (aluminum coated glass) because their operationability in CO2-lasers was many times proven. One could drill there a couple of holes, but... The diamond bores of the necessary diameter had suddenly appeared to be unaffordable too.

It was finally decided to make a handwork polishing of some aluminum plate and drill there the needed holes by a common drill and bit.

Grinding and polishing are responsible processes and can not be reduced much. For example one can not use here a drill with a felt disc and fine abradant. Or else he will end up with a ruined shape. (We need the mirror to be as perfectly planar as possible and with the felt disc we will have random hills and pits there.) One can not shorten a grinding stage - or every non-uniformities will show up in polishing.

One can, for example, reduce number of polishing stages. I.e. use 3-mcm abradant directly after 10 mcm one. But in this case it will be needed to prolong each stage, or the traces of the previous abradant work will be seen as scratches.


The detailed description of the DIY polishing of optics one can find at the astronomy hobbyists sites. I will only give a few hints:

The mirror when being grinded and polished should emit a weak and even hissing sound. If no sound at all - it means the work in vain - You are not removing metal. The sudden crunch sound means You may stop the work, wash the mirror, and start from the very beginning - from the most rough abradant.

Do not underestimate the importance of the chamfers on the edges of the blank. In the process of polishing the edges can become very sharp. And not only they can tear the polishing pad, they can also chip, and their pieces will ruin the polished surface.

Water, when being used as a lubricant, can make the work faster, but it has a nasty property to dry out. When it dries out the abradant particles glue up into one large conglomerate and it ends up with that fatal crunch sound. It often happens suddenly - You see that the polishing pad is still evenly wet, but several seconds later it appears to be dry with the all consequences. So if You are not intended to design a dewdrop, You probably should better use a mineral oil.

The material (metal) of the blanks is also of the great importance. Often the "hatred crunch sound" appears not due to the outer dust, but due to the fact that another particle of the "intrinsic sand" was opened and detached from the metal.

The process of grinding and polishing is rather short, when taken alone. About a half of an hour per each abradant grit (and there are about ten of them), about a half of an hour for washing and cleaning between the stages. Totally it takes about ten hours of work to get a really good mirror... However the sudden tearings of the surface and shape disturbances will force You to repeat the work again and again. So my final advance here:

If You have even a minimal possibility to avoid the polishing of a homemade laser mirror - AVOID IT AT ANY COST!


...The first three polished plates were just shameful to look at. It was senseless to try them in the laser. Well, its my fault - I tried to "deceive the nature" - to reduce efforts and time needed for the process by some tricks. The conclusion is: if You want a good result, don't try any tricks, - just honestly, patiently and accurately perform every stage.


The fourth sample looks like that:


Al_mirror01 Al_mirror02


One can see that the reflection of the ruler is straight enough, what means the mirror is more or less planar. It can be barely seen on the photo, but the mirror surface is covered by a net of tiny tiny scratches. This is the result of the work of my finest abradant. I have no abradant finer grit than 1.5 mcm and this one leaves "hair-thin scratches", seen in the visible light. But it should be OK for the laser, since its working wavelength is 15 times longer than the one of the visible light, and the surface of the mirror should behave as ideal for the infrared.


Then the holes were drilled. Diameter of the most large IR windows I have is equal to 25 mm. It means that the working area of the mirror is 3x25 mm (remember: 3 mm is the height of the laser channel/waveguide). 30% of transparency corresponds to 22.5 of the total area of all the holes. In its own turn it corresponds to 12 holes, each having 1.5 mm diameter. Before drilling the mirror was painted with protective varnish, and during the drilling it was pressed between two wooden planks. And despite all protective measures, a narrow mat area appeared around the holes. The width of the mat hem was about a quarter of millimeter. I.e. from the point of view of the light losses in the resonator, one may treat the holes as being 2 mm rather than 1,5 mm wide. So the reflectivity of the mirror appeared to be 50% instead of the expected 70%.

Its a pity but the holes can not be drilled before grinding and polishing. Otherwise one will end up with a wide deep pit around each hole.


Despite the shortcomings, the mirror was installed onto the mirror mount and sealed with using a kalium chloride window above all the holes:




Test run of the laser on CO2:N2:He = 2:7:18 gas mixture has even shown a bit of output: about 120 mW. At the highest current and at the pressure of 3"Hg the laser was barely above the threshold.


The mirror was removed and polished to repair the mat areas around the holes. The mat has gone, but deep and well polished pits appeared instead.




With this mirror the laser completely refused to lase.

Further on the mirror was shifted to the position, where the laser channel "sees" only one row of holes. This is a simple way to reduce the transparency of the mirror from 30% to 15%.

With this position of the mirror the laser agreed to lase, and, after some tuning of the alignment and pressure it gave about two watts (309 mV by Peltier calorimeter).



It was checked out that the replacement of this mirror by a germanium one with the reflection of 70% allows to get more than 10 Watts easily.


The obtained result causes ambiguous feelings. It is good that hand working allows to make a mirror, with which the laser not only barely works, but also gives a decent output. On the other hand the quality of the homemade mirror appears to be low enough to kill all the expectations to improve the output in comparison with the commercial grade mirror. Even with the one being not well suited for the task.


  1. Laser technique and technology. In 7 books. Book 2: Engineering basics of
    technological lasers design: High school textbook. / V.S. Golubev,
    F.V. Lebedev; Editor: A.G. Grigoryants. - Published: Moscow, Vishaya shkola,
  2. A.A. Kuznetsov, M.Z. Novgorodov, V.N. Ochkin, et al. Compact Slit-Type
    CO2-laser Excited by non Self Sustain Direct Current Discharge, Maintained
    by Short Pulses. Preprint of Physical Institute of Academy of Sciences,
    named by P.N. Lebedev, N15, Moscow, 1997.
  3. V.N. Ochkin. Waveguide gas lasers. M.: Znanie, 1988.
  4. V. J. Vitteman. CO2-laser. (russian edition) M.: Mir. 1990.
    ISBN 5-03-001351-2 (rus)
    ISBN 3-540-17657-8 (eng) by Springler-Verlaf Berlin Heidelberg, 1987