. : Approach to make a LED pumped Solid State Laser : .

An idea to make a solid state laser (SSL) with pumping from common light emitting diodes (LEDs) has came to every mind connected with lasers. Nevertheless to amateur or to specialist. The preliminary considerations, accompanying this idea are very tempting: "I should take a pair of flood light diodes, 100 W each, then I should attach a laser rod somehow. And probably five to ten watts of power will be provided." Why 5..10 W? Because a human orients to a widely advertised value, that a lamp pumped laser yields 1% of efficiency typically. And one thinks that with LEDs it will be surely more...

If one comes from the preliminary statements to some, maybe rough, estimations, it will look like this:

1. Let's assume that YAG:Nd is used. YAG:Nd has the lowest threshold among available solid media. Let's take that rod is 3 mm in diameter and 50 mm long. Let's suppose that the LEDs used are in standard (for powerful LEDs) casing, that has diameter 8 mm and a dome shaped transparent lens.

   

2. Elementary geometrical considerations show, that one can place up to 24 pcs of LEDs around the rod. 6 belts, each having 4 LEDs:

   

   It seems like for now there are no single-crystal diode more powerfull than 10 W. (Examples of single crystal 10W diodes are Cree Xlamp T6, Cree XML XM2.) One should note that LEDs (on the contrary to laser diodes) are marked by their supply power, not by their optical output. How much from the applied 10 W of power a LED will be able to give out as a light is unknown. Yes, LED efficiency is usually provided in their datasheets, but still... It is usually given in lumens per watt, and lumen is a complicated value, and direct calculation of lumens to watts of light is usually out of DIYer scope.

   For now let's take that LEDs emit to light something like 20% of the supply energy. It will mean then, that during upper laser level lifetime (250 mcs for YAG:Nd) those 24 diodes having been placed around the rod will emit:

   24 pcs * 10 W * 20% *250e-6 sec = 0.012 J = 12 mJ.

   How much of those 12 mJ will reach the rod and how much of energy will be
   absorbed there?

   To estimate, how much of light will reach the rode, one can take the angle of view for the rod from the center of diode (marked yellow on the next figure). Once taken this angle should be divided by the full angle of diode emittance. (Something like 180 degrees or Pi radians)

   

   The angle of view can be roughly estimated as a ratio of rod diameter to the distance between the rod's axis and the center of crystal of LED. In our case it is 3mm/4mm = 0.75 rad. It means that the energy that reaches the rod will be 0.75/3.1415 = 24% i.e. 2.88 mJ.

3. How much energy will be absorbed? It depends on spectrum. For white diodes one can suppose that about 10% will be absorbed. And such diodes, that emit strictly to the bands of absorption can provide 30..50% of absorption:

   2.88 mJ * 40% = 1.152 mJ.

   Further on it is simple: one should multiply the absorbed energy by the ratio of wavelengths (pump to lasing), for example: 1.152 mj*(590nm/1064nm) = 0.638 mJ, - we got the energy accumulated on the upper laser level. (Here the pump wavelength was taken of 590 nm - Yellow LED).

   Let's use the resulting stored energy as the input for laser gain calculator, which will show us that in a YAG:Nd rod, 3 mm in diameter, the gain will be 1.04347, i.e. 4% per pass.

   It is understandable now, that if we haven't got a 95% mirror, we have no hope to see any lasing at all. (One could note, that if the back mirror reflects 100%, then per two passes we will have 1.04^2 = 1.08, and it seems that 92% mirror would be enough. But the smart one may have forgotten that even AR coated ends of rod make losses. Moreover there also are diffraction losses and non-zero self-absorbtion in the laser crystal. Yes, all those non-idealities are negligible usually, but when You have only 4% per pass any of these things can be fatal).

4. However let's assume that one can feed the LEDs with the current 10 times higher than nominal one, and in order to not overheat them, wi will use pulses with a very low duty cycle. The question, whether the LED's will endure this or not we will leave for future.

   If now we propose that the light yield of LED is proportional to the supply current, the energy stored in the laser rod will be increased by 10 times: from 0.638 mJ to 6.38 mJ. And it corresponds to gain of 1.53, i.e. 53% per pass, and LED pumped laser turns from something evanescent to something real. But how many unproved propositions we made for this?

Let's count:

• How much energy (not lumens) LED gives out is uncertain
• How much of the emitted energy can be absorbed by Nd:YAG rod is unknown
• What current will LED be able to endure - incomprehensible
• And what pulsed power will correspond to that limiting current - who knows?

These four questions I try to answer in this web-report. Will it end up with a working model of LED pumped SSL - it is unknown even for me. However even in case that it won't, I hope that the data on LED operation in pulsed modes will be interesting for most, and may be useful for somebody.

LYRICAL RECESSION #1. About namings.

Before telling out which parameters were measured and how, one should stop on the description of the object of investigation. Any serious work on radio components research should start with something like this: "The object of the current research was 2n222 transistor, produced by Phillips, such and such production bunch, such and such year of production..." However the first pitfall awaits us exactly here. All I can say on this topic is: "PARAMETERS WERE MEASURED OF CHINA LEDs DIFFERING BY THEIR COLOR AND NOMINAL POWER." The reason is that our Chinese brethren don't bother with accurate naming their production. The negative side here is that when one buys the LEDs even from reliable distributors, the only available info sounds like: "LEd white, 1W, China." The positive thing is that the author of this web-report is free of playing the role of voluntary advertisement agent, like most scientists do in their articles.

One could avoid the problem of LED identification using the 'white' production of 'white' manufacturers. like Cree. OSRAM, etc. Such products are usually named well, and their parameters are also better than ones of "noname" products. Except one: the price. But exactly the lowest price (per emitted watt) will be of most importance for You when designing a LED pumped SSL. Indeed if the price did not bother You, You would go and make laser diode pumped SSL like all "white" people in the "civilized" world.

Having made this lyrical recession I can name the subjects of our consideration with a clean conscience.

These are:

• Noname LED of yellow emitted colour, rated to 1W.
  Will be further designated as "LED Yellow 1W".
  
• Noname LED of green emitted colour, rated to 1W.
  Will be further designated as "LED Green 1W".
  
• Noname LED of white emitted colour, rated to 1W.
  Will be further designated as "LED White 1W".
  
• Noname LED of deep red emitted colour (~730 nm), rated to 3W.
  Will be further designated as "LED 730nm 3W".
  
• Noname SMD LED of deep red emitted colour (~740 nm), rated to 3W.
  Will be further designated as "LED 740nm 3W".
  

POWER IN CW MODE

The power emitted by LEDs was measured by homemade Peltier calorimeter. The emitting part of the diode was placed in close proximity of the Peltier sensing surface, but without a physical touch (air gap of 0.2..0.5 mm). Since the sensing area of the calorimeter is rather large (30x40mm) this way of LED placement allowed to intercept almost all its radiation (gathering angle is 158x162 degrees, and it is commonly more than the LED divergence angle.)

During measurements the LEDs were powered by double transistor analogue constant current driver. One can find its description here.

The results are given in table #1. Don't be confused with many significant digits. The main error of measurements was due to zero drift, and the latter was about 2 mV in this session.

Table #1. The results of LED power tests.

It is interesting that some diodes have very decent efficiency, For example the voltage drop over the deep red (730 nm) diode was 2.4 V at working conditions. At current of 1.2 Amp it gives 2.9 Wt of supply power. And the LED emits 0.86 Wt at this current. It gives efficiency of 29%. One should note here that a half of emitted radiation goes to the opposite side and does only heat the diode's casing and crystal support.

Not less interesting, that despite 740 nm diodes were marked as 3W, their allowed current (typed on their package) was 700 mA. The package also has an insignation, that voltage drop should be from 2 to 2.4 V. Several times I tried to multiply 0.7 Amp by either 2 V or by 2.4 V. I never got 3W. What a deception.

ABSORPTION BY NEODIMIUM IN LASER ROD

Even if one has many many of very very powerful LEDs it is senseless to try to use them to pump a laser if their light is not absorbed by the laser rod.

As it was said above a YAG:Nd was choosen as a laser rod, because it has the lowest threshold among available laser media. At present time it is not difficult to get such a rod from, eb@y, amaz0n, a1iexpress and other web trading portals. One only needs the money and the desire.

The absorption bands in ruby are wider and stronger than ones in YAG. It means that it is simplier to get an appropriate LED for ruby. It is sorrowful however, that the ruby is three level laser medium at room temperature, and it, in its own turn, means that it needs a HALF of all active centers to be in excited state for only to reach zero gain. Even for small rods, to pump a half of their ions, means to input several of joules into the rod. And as we've seen from the estimations in the introduction, LEDs don't promise several joules. Several millijoules at the very best case. So the ruby does not form a decent alternative to YAG:Nd.

Nd glass is four level medium, but its gain per unit of input energy is lower than in YAG:Nd by an order of magnitude.
Nd in YAG (and in other media) has 5 bands of absorption (see table 2).

Table 2. Estimated parameters of absorption bands of neodymium (from[1])

Note that the strength of absorption in the table is shown in arbitrary units. The real value of absorption will of course depend on the Nd concentration and size (diameter) of the rod. So the table itself does not answer the question: "How much light will stay in the rod?", but does only allow to orient in "which LEDs are better, and which are completely useless?"

There was found no LEDs emitting in either of infrared bands (808 and 880 nm). It is strange, since the 808 nm line is so beloved by laser diode manufacturers.

In literature they recommend to use LEDs with emission line center at 750 nm [2]. Despite the statement of authors of [2] that these diodes are widely available and affordable, there was no success in finding them on market. There are however several diode types that emit not exactly at 750 nm, but rather close to. These are 730 nm diodes and 740 nm. Both are called "deep red" and 730 nm ones are slightly less exotic than 740 nm ones. Light of these diodes is brightly seen by bare eye, and this fact had determined the first test: one can take the LED and look at it through a laser rod. If the visible brightness of the diode becomes lower, it might sign that there is a good absorption.

The result was a bit unexpected: even when looking at LED (730 nm 3W) throughout a YAG:Nd along its axis (!) there was no visible dimming of light. The same was observed with green and white diodes. The sole type of LEDs that demonstrate visual detectable dimming, when being observed through Nd containing media, are yellow LEDs.

After a bit of discourage from this result, it was supposed that neodymium eats rapidly the part of radiation, that fit its lines of absorption, and the other part of radiation travels easily through the rod without any obstacle. It means that one can not trust in visual observations and one should measure the absorption by instruments.

The next was an attempt to 'stupidly' insert the rod between the LED and calorimeter, and to try to determine the absorption by the difference of readings. A 3 mm rod was used.

Result: in the limits of measurement errors no absorption was detected. It is understandable in principle. Let the absorption in the rod be 100%. What if the rod intercepts only 10% of light, and the measurement errors are also about 10%? The difference between readings with and without the rod would be not higher than 10% in this case, and it will be well inside the limits of errors.

Finally a slit diaphragm was made. The diaphragm is to be placed over the calorimeter in order for the light, going aside the rod, won't give a contribution to the measurements. The resulting layout is shown on the picture below.

Model of the diaphragm for 3D printing for Peltier element 40x40 mm can be downloaded by this reference: Peltier_Slit.stl

The results obtained with the diaphragm are given in table 3.

Table 3. Results of measurements of absorption of LED light by 3 mm YAG:Nd rod. Made by Peltier calorimeter.

One can see that due to the diaphragm the useful signal has dropped down, and despite the fact that zero drift was below 0,5 mV in these tests, the final uncertainty of the result appears to be high. For the comparison one can say that Fresnel reflection on the boundary surface between YAG and air is 8..9%, i.e. 16..18% for both borders. It means that if one uses pessimistic estimations from the confidence range (see table 3) then even for 730 nm diode one cannot say anything good about the absorption.

In order for lowering the discrepancy and to get meaningful estimations of absorption, an attempt had been made to replace Peltier calorimeter by a solar cell element, working as a light sensor. However the solar cell appeared to be strongly non-linear both in measurements by photo current and in photovoltaic mode. Somewhere at the very beginning of the characteristic (a few millivolts and microamps) there exists an area of linearity, but for our purposes it misses the point. It's worth to note here, that in (rather) recent experiments with (rather) powerful lasers the linearity of Peltier calorimeter was proven for up to 30W (see for example here).

On flea market a photodiode was found, that has a comparatively large sensitive area. Here is its photo:

There are no marks on its casing, and the seller had difficulties to name its type. Unlike the solar cell the photodiode appeared to be able to demonstrate decent area of linearity. In photovoltaic mode it was non linear at all conditions, but in the mode of photo current it was linear up to currents of 1 mA.

For the measurements the photodiode (as like as calorimeter and solar cell) was used with slit diaphragm. The general view of the sensor is shown on the next photo:

Results of the measurements with photodiode as the sensor are given in table 4.

Table 4. Results of measurements of absorption of LED light by 3 mm YAG:Nd rod. Made by photodiode.

The last column of the table contains the value of absorption having been corrected with taking into account the Fresnel losses, but without correction for fluorescence of the rod itself.

Fresnel losses per one face were calculated in the next way:

Absorption was calculated as:

               Irod
abs = 1 - ---------------  
            Io*(1-F)^2

where Io - photocurrent when the photodiode was lit by LED through the slit diaphragm. Irod - the same as above, but when YAG:Nd rod was inserted between LED and sensor. F - fresnel losses.

Dark current of the photodiode was negligible for these measurements. The linearity can be additionally proven by the fact that almost for all cases (excluding green LED) the photo current given by LED @ 0.64 Amps of supply current was almost twice as large as one given by LED @ 0.32 Amps of supply current. As for the green LED - it has decent non linearity itself (see table 1).

A bit astonishing is that the absorption looks like to be almost equal for deep red diode and for the yellow one. Visually it looks completely different. The obviously overestimated value of absorption for the white does disturb too.

To turn the visual feelings into something documented, some photos were made, where the luminescence of the rod is shown under the irradiation by the light of different LEDs. The snapshots were taken by Nikon Coolpix 4600 camera looking through HWB850 glass filter.

Snapshot of luminescence of YAG:Nd rod lit by yellow LED light.

Snapshot of luminescence of YAG:Nd rod lit by light of a LED with the center of emission near 730 nm.

Snapshot of luminescence of YAG:Nd rod lit by white LED light. (warm white).

One can see that in the rod lit by yellow LED light only a small area of fluorescence exists near the place of LED light entrance. When lit by deep red (730 nm) the area of fluorescence is significantly longer, which gives the evidence of weaker absorption. The light of white LED is being absorbed even more weakly.

OVERCLOCKING IN PULSED MODE

The next question to answer is "what is the maximal energy, that a LED can emit during 250 mcs pulse?" Lemme remind, that 250 mcs is the upper laser level lifetime for YAG:Nd. If duration of the light pulse LED emits is shorter than 250 mcs, it means that only the full energy (per pulse) is important for SSL pumping. Power is arbitrary value in this case. On the contrary if LED pulse duration is much longer than 250 mcs, the power becomes important. And the total energy per pulse becomes of low interest. (Actually in both cases only the energy per each 250 mcs does matter, but in the second case this energy is directly proportional to the emission power.)

From the said above it is clear that for our certain problem it has no sense to feed LED by kiloamp rated nanosecond pulses.

A pulse forming network was assembled to produce 250 mcs long pulses.
Its schematics is given on the next figure.

The circuit does actually not use batteries as power supplies. Signal part of the circuit (the one using 555-timer) is fed by a small 12 V power supply, powered from the mains (lets assume that the left battery on the schematics represents this PSU). The power part of the circuit (the one based on IRFP460) is fed by a variac with a diode bridge rectifier.

L1C3 tank circuit is tuned so that its main frequency corresponds the pulse duration of 250 mcs. L2C4 is tuned to its third harmonics.

When only L1C3 is working, the current through LED has the shape of bell-like pulses (half of sine wave). When L2C4 is added the top of the pulse becomes flat (or even with some pit, dependently on exactC4 value) and the pulse becomes more close to rectangle.

When tuning the circuit one should adjust L2 value so that the pit of oscillations in L2C4 tank circuit sit exactly on the top of L1C3 pulse. If this condition was not fulfilled, addition of L2C4 may worsen the shape of the sum pulse rather than to make it better.

When the impedance (the value being equal to sqrt(L/C) of L1C3 circuit is higher than the resistance of the LED (or to be more precise when it is higher than the ratio of voltage drop over LED to current through it) the circuit works like constant current driver with reactive ballast.

The monitoring of current through the LED is made by oscilloscope connected to the terminals of Rcurrent_sense. By variation of C3 charging voltage one can vary the amplitude of current through the diode. When ferrite cores of L1 and/or L2 become saturated the shape of the pulse may become disturbed. So the higher current You want from the driver the larger cross section of the ferrite cores You should choose.

Peak output power of the LED was monitored by a sensor, consisted of BPW-34 photodiode, loaded by 300 Ohm resistor. Such a detector is simple, but does provide decent speed and decent linearity up to signals of 1.1 V. If the output signal became higher, it was needed to move the sensor further apart of the LED, and then to "sew" the results of measurements.

Table 5. The results of measurements of LED emission power in dependence on the supply current in pulsed mode of operation.

Yellow 1W rated:
I, Amp	sensor readings, V
0.72	0.44
1.04	0.6
2.32	1.2
3.2	1.56
4.48	1.8
4.6	1.848214286
6	2.008928571
7	2.169642857
8	2.25
9	2.410714286
10	2.491071429
11	2.571428571
12	2.591517857	dead in 10 sec

soft UV (404nm) 1 W rated:
I, Amp	sensor readings, V
0.5	176
0.64	214
0.88	272
1.02	320
1.2	352
1.42	408
1.88	540
2.08	560
2.48	640
3.04	760
4	940
5.04	1080
6	1207.058824
7	1270.588235
8	---	dead in 30 sec

Green 1W rated:
I, Amp	sensor readings, V
0.56	240
0.68	260
0.8	300
1.04	340
1.24	380
1.5	440
1.76	500
2.2	560
2.52	620
3.12	700
3.6	760
4	820
4.48	860
5.2	940
6	1015.2
7.2	1090.4
8.2	1165.6
9	1240.8
10	1316	worked flawlessly for 5 min
11.2	1391.2
12.4	1428.8
14	1541.6	dead in 30 sec

White 1W rated:
I, Amp	sensor readings, V
0.66	200
1	260
1.52	340
2	400
2.5	460
3	520
3.5	580
4	620
4.56	660
5.04	700
5.68	760
6.08	780
6.5	800
7	840
8	880	dead in 10 sec

730nm 3W rated:
I, Amp	sensor readings, V
0.76	168
0.88	192
1.04	224
1.4	272
1.76	328
2.48	408
3	448
4	520
6	580
7	620
7.4	620	suddenly dead when no one expected

SMD 740nm 3W rated:
I, Amp	sensor readings, V
0.6	304
0.92	432
1.32	592
1.88	778
2.5	964
3.08	1150
4	1406
5	1597
6	1725
7	1885
8	1949
9	2045
10	2109
12	2167
13	2226
15	2284
17	2343
22		dead as expected

Cree White XML T6 10W rated:
I, Amp	sensor readings, V
0.6	120
1.12	184
1.76	256
2.48	328
3.6	424
4.96	528
5.84	576
7.6	680
9.2	760
13	832
15.4	896
18.6	960
18.8	960
23.6	1024
24		dead

The values given in table 5 are natural only up to 1.1 V. The values higher than 1.1 V were recovered by division of the real readings by the attenuation coefficient.

It is understandable, that these readings do not reflect the real power of LED. However if I've succeeded in keeping the linearity by means of signal attenuation, then the data obtained, make one able to judge how the power in pulsed high current mode relates to the one at normal current. LED emission power have already been measured (see table 1.) So in order to discover the top limiting power, one needs to take from table 5 a reading that corresponds to some value of current in table 1 (take 0.64 Amp for simplicity). If necessary value of current is absent in table 5 one should use extrapolation or interpolation.

EXAMPLE: The reading for yellow LED, extrapolated to 0.64 Amp of current, will be 0.44*0.64/0.72 = 0.39 V. From table 1 we can see that Yellow LED gives 173.4 mW when fed by 0.64 Amps current. I.e. 1 Volt of readings corresponds to 444 mW. Or in other words the calibration is 444 mW/mV. From table 5 we get that maximal readings were 2.6 V. I.e. 1.144 W of power were obtained. This procedure should be made independently for every diode, since different diodes were at different distances from sensor during measurements. It means that sensor readings (in millivolts) are not to be compared for different measurement sessions with different LEDs.

FINALLY:

LED type Power achieved

Yellow 1 W : 1.14 W

Green 1 W : 2.02 W

White 1 W : 2.39 W

730nm 3 W : 2.17 W

740nm 3 W : 4.9 W

Cree 10 W : 5.2 W

20.04.2021

We've finally got, that white Cree and 3W SMD 740nm LED can be forced to emit more than 4W in pulsed mode. In principle this information, accompanied by absorption value, could be enough to calculate the number of LEDs needed to create a laser. However a huge discrepancy in values of absorption, that were obtained by measurements with calorimeter and by measurements with photodiode makes an alert. There's a feeling that measurements with Peltier do understate the result and ones with photodiode - overestimate it.

What could it be connected with?

First of all, in measurements we take that the energy of LED light, that had been absorbed by the rod, would stay in the rod and would not give a deposit into the calorimeter's readings. It is not true in reality. Being a laser medium with good quantum yield of luminescence, YAG:Nd is a good secondary emitter of the absorbed energy into surrounding space. And since the sensitivity of calorimeter does barely depend on the wavelength (on the contrary to photodiode) the light emitted by the rod can give a deposit into the readings.

This deposit can be estimated from the next considerations:

1. If a pumping light having power W falls onto the crystal, then W*(1-F) of it enters, where F - Fresnel losses at the wavelength of pumping.
2. If we designate the absorption in the crystal as "A", then W*(1-F)*(1-A) of the light comes to the output facet
3. From all that have fallen onto the output facet only W*(1-A)*(1-F)^2 will go out; and this whole power makes deposit into heating of calorimeter.
4. In addition the crystal emits energy/power of W*(1-F)*A*(750/1064)*0.8, where (740/1064) is the ratio of pumping and working wavelengths (or in other words - so called "quantum defect"),
   0.8 - is the quantum yield. This number was got from a handwaving, but the experience gives evidence that long living things have a tendency to have it lower than 100%. (If anyone likes other number, he/she can reproduce the computations with any other value at gustum.)
5. W*A*(740/1064)*0.8*(1-F)^2 goes out of the crystal (Fresnel losses at output were added). And this energy goes in all ways evenly. If we take that calorimeter is large enough, we will get that useful rays are all those go to the correspondent half space. So in order to take the luminescence deposit into account, one needs to add a half of this value:
   (1/2)*W*A*(740/1064)*0.8*(1-F)^2
6. We've finally got that the calorimeter is being heated by power:
   W*(1-A)*(1-F)^2+W*A*(740/1064)*0.4*(1-F)^2
   Let's modify a bit:
   W*(1-A)*(1-F)^2+W*A*(740/1064)*0.4*(1-F)^2 = W*(1-F)^2*[(1-A)+A*(740/1064)*0.4] =
   = W*(1-F)^2*[(1-A)+A*0.28] = W*(1-F)^2*(1-0.72*A)
7. I.e. if earlier we assumed that calorimeter is heated by W*(1-A)*(1-F)^2, and thus we calculated the absorption as A = 1 - (W1/W0)/[(1-F)^2], where
   W1 - calorimeter readings with rod installed
   W0 - calorimeter readings without the rod.
   Now the calorimeter is heated by W*(1-0.72*A)*(1-F)^2 and the absorption should be determined as: A = {1 - (W1/W0)/[(1-F)^2]}/0.72

IN SIMPLY WORDS THE ABSORPTION, HAVING BEEN CORRECTED FOR THE LUMINESCENCE, IS EQUAL TO THE ONE WITHOUT CORRECTION, HAVING BEEN MULTIPLIED BY 1.39.

That's all with the correction.

Besides, for the yellow one should multiply by 1.29 instead of 1.39 and for the green: by 1.25.

As we can see, even with the correction for the luminescence, the values from the calorimeter (table 3) are too small to compare with the ones from photodiode (table 4). One should not multiply the result from photodiode by the luminescence correction factor because sensitivity of silicon photodiodes @1064 nm is very low in comparison to the sensitivity at the wavelength of pumping.

So there was no success in the attempt to explain the understated absorption by taking into account the luminescence. What else could work here? It is possible that the heat, that goes to the rod during absorption just can not go out from the box of the slit diaphragm. So the only way for it to go out is through the Peltier element, In this case all the heat absorbed buy the rod will yield in Peltier heating and thus will be taken as readings.

After this guess, some ventilation holes were added to the casing of the slit diaphragm. It didn't help. The result was still understated. Nevertheless during the measurements it was found out that in the first 30 seconds of measurements the absorption looks rather relevant, and only with further waiting the calorimeter readings with the rod begin to approach the ones without rod.

Temperature of a body (in our case - the surface of the calorimeter) being heated by a constant power and having a heat sink in the form of constant heat resistance, can be shown as this:

T(t) = To*[1-exp(-t/tauT)], where To - is a constant temperature (one can say that it is equal to the heat addition power divided by the heat sink heat resistance). tauT - is some time constant, that depends on the body's heat capacity and on the heat sink heat resistance, t - is the time variable. Since the calorimeter is the same in all measurements, tauT should not depend on whether the rod have been installed or not. If we designate the stationary temperature of calorimeter's surface with the rod as T1 and if we designate the same but without rod as T2, then:

  T1*[1-exp(-t/tauT)]   T1
  ------------------- = --
  T2*[1-exp(-t/tauT)]   T2

It means that if one is more interested in ratio of values rather than in the absolute ones, one can omit the waiting for the calorimeter readings to become stable. It is enough that both measurements (one with the rod and one without) should be taken at the same interval of time after the LED had been turned on. It means no need to wait until the LED or the rod will be heated enough for some uncontrolled factors to interfere with the measurements.

For the 3W 740nm SMD LED the attempt to make the measurements in the new way was undertaken. The readings with and without the rod were taken on 30-th second after the LED's switching on. Before LED switching on it was waited for the calorimeter readings relax to under 0.5 mV.

And here are the results:

Table 6. Results of measurement of absorption of LED 3W 740 nm radiation by using Peltier calorimeter and method of shortened measurement time In the last but one column of the table - absorption value calculated as earlier (with a correction for Fresnel losses). In the last column - the same absorption value, having been additionally corrected for luminescence (refer to text above)

I agree, that all the above considerations remind an attempt to drag the result "by the ears". On the other hand they at least look consistently and allow to make buckle and tongue meet. On the third hand I got used to trust the calorimeter much more than any other photodiode. So to my mind the absorption values from the table 6 are the least unreliable.

With more or less known value of the absorption we can now proceed to the calculations of how many diodes are needed for the laser. As earlier let's propose that the rod as 3 mm diameter. (To use rod of larger size means definitely ask for the mountain to bring forth a mouse.)

Let's assume further that the (laser) gain should be 1.5 times per pass.

Using the laser gain calculator we can know that the energy stored on the upper laser level must be not less than 6.08 mJ.

The necessary energy, absorbed by the rod, will be then:

6.08*1064/740=8.74 mJ.

The necessary energy required to be delivered to the surface of the rod may be got as the previous result having been divided by the absorption:

8.74 mJ / 0.32 = 27.32 mJ.

The angular size of the rod (when being observed from the center of LED crystal) is about 80 degrees for SMD LED when the rod touches its dome. The beamwidth of the LED (known from its advertisement) is 146 degrees. It means that the factor of delivery is: 80/146 = 54%

It means the LEDs are needed to emit 27.32/54% = 50.6 mJ of light in order for 27.32 mJ of it be delivered to the rod.

Since the pulse duration is 250 mcs, it relates to the power of: 50.6e-3/250e-6 = 202 W.

Let's take some "safety margin" and suppose that the LEDs can emit 4 W of light more or less safely. With this we will get that the laser needs 202 W/ 4 W = 50.6 pcs of LEDs.

Taking that store is no sore, and since these SMD 740 nm diodes are sold in bunches of 20 pcs each, let's "round" the necessary number to 60 pcs.

The next question: can we place 60 LED's tightly around the rod or not? From pure geometry the answer is : YES. For example as 5 columns having 12 diodes each. Width of SMD LED is 3.5 mm, so the full assembly can be 42 mm long... Or can not?

To check it out, a PCB board was made. Its stencil for laser printer - hot iron method is here:

Onto this board with the use of tin-lead (40/60) solder and with a help of the God 12 LEDs were soldered. This is how it looks like:

Full length of the assembly is 50 mm, the length of its light emitting part is 42 mm. The bias of LED positions from rectilinear placement looks acceptable.

So it looks like it is really possible to place 60 LEDs around the rod. It means that it has sense to try to make the laser real.

The diodes have been ordered. Awaiting them. After a month or two, when they arrive, here will be the results of testing for lasing.
...

30.12.21

First of all I want to make an apology that the continuation of the project comes so late. The 2021-th appeared to become anomaly hot (at least at Russia), the summer was long and all the laser things had to be set aside for the sake of summer deeds. Moreover the bunch of NIR LEDs having been bought in spring were successfully lost during days of hot weather, and I had to order a new ones. They finally arrived and this time I can present the result.

In short it is simple: the laser was assembled and agreed to lase. The set up was the next: YAG:Nd rod d=3mm l=50 mm with flat anti-reflective endings. The back mirror was flat high reflective one (R>99%), the output coupler was also flat (R=70%). The rod was closely surrounded by five handmade assemblies (bars) containing 12 LEDs each (SMD LEDs, 3W 740 nm). The diodes are fed by 240 mcs pulses given by a homemade power supply. The threshold current was 6..7 amps.

And now the details.

The bars were soldered in the same way as the previous testing bar. Some inconvenience during the assemblage was due to the fact that a part of the diodes had 1.4 mm high dome (lense) and others had 1 mm high dome. In surplus the new ones have "reverse polarity" in comparison with the old ones. As the result, when the laser is being assembled, and when a bar becomes abut to the laser rod, it becomes a bit distorted. The reverse polarity was also found not instantly. Initially the bars consisting of some old and some new diodes refused to conduct at all. It was necessary to flip solder some diodes to achieve conductivity. The fable is: don't trust the datasheets. Test the parts with multimeter whenever possible.

The assembled bars were connected in series and placed around the laser rod (by hands) to form a pentagonal prism. The whole assembly was fixed up by wrapping with a sticky tape.

The circuit of the power supply has already been shown above. Here are the photos from the inside and outside.

Need to say that after the power supply assemblage initially I was unable to get the proper current pulse shape (as modelled in LTSpice). The cause appeared to be in the aluminum electrolytic capacitors, used as storage ones (C3 and C4). C3 was then replaced by some motor start cap, but the shape of the pulse was still bad. I had to specially purchase some mylar film capacitors to make the storage bank. Here is how they look like:

In the reality the power supply gives the pump pulses, each about 250 mcs long, with the repetition rate of 9.5 Hz. The pump current can be adjusted by altering the charge voltage of C3 and C4 storage capacitors. In its own turn it can be done with a help of variac, being used to feed the "high voltage" part of the scheme. One can measure the pump current by attaching a scope to the leads of current sensing shunt resistor Rcurrent_sense. In the present configuration Rcurrent_sense=1 Ohm, so each volt on the trace corresponds to one Amp of pump current.

In process of tests with single diodes it was found out that this power supply can easily yield enough current to burn out any of tested diodes (over 20 Amps). However when loaded by all five bars in series (60 diodes) the power supply was able to press only 8 Amps through them when the charging voltage was at the available maximum (250V by the variac's dial). 8 amps is a bit low even in relation to the saturation of the dependence of the LED power to its current, however it was decided to make a test with what we currently have. Here is how the laser looks like when being installed over its power supply:

And here is how its lasing looks like:

The video was taken by a camera looking through IKS-3 (HWB850 analog) filter in order to avoid interference from bright red flashes of pumping light. The laser spot is not very bright due to low sensitivity of the camera in NIR range of spectrum.

Need to say that at present days not every other camera sees Nd laser radiation good. Actually silicon detectors can easily operate at this area, but this is prevented by a so called Bayer filter - a thin colored film that is placed directly on the face of camera sensor to provide a good color reproduction. In the Internet one can easily find the description of its removing, but the process is complicated and has not 100% probability of success. I used simplier way: tested all the affordable cameras for the sensitivity and chose the most suitable one.

Here are the results:

• iPhone 5 - sees almost nothing through IKS-3 filter. One may distinguish only the hottest hart of the filament of incandescent lamp.
• Nikon D3100 - almost the same as iPhone 5.
• Nikon Coolpix 4600 - one can see not only the filament but also the flask of incandescent lamp. One cannot see the surroundings of the lamp.
• Old fashioned black-and-white camera with output of analogue video signal through coaxial cable. Through IKS-3 filter one can easily see all the surroundings when a room is lit by some incandescent lamp. Brightness and contrast are really good.
• Noname cheap chinese automotive video registrator, that should be able (if trust the advertisement on its box) to do it at 720x576 pixels with FPS rate at 30. In reality its resolution is worth of something like 320x200 while its digital circuits stretch the image to the said 720x576. It sees in infrared a bit worse than the BW camera with analogue output, but much better than all other cameras I have tested.

The attempts to make a video of working laser by the BW camera with analogue output were successful in general. However either the camera or its video capture board appeared to be unable to work correctly with repeatingly pulsed signal. The vertical position of the laser spot on video frame was erratic. In principle one could say that the laser lases even with this camera, but the video could easily fool unprepared user and thus it was not suitable to present it here.

The video made by the cheap chinese automotive video registrator has already been shown above.

Variations of pump current have shown that the laser spot on the screen disappears when the current becomes below 6..7 amperes. Thus the laser threshold lies somewhere between these values. It means that at 8 amps of pump current the laser is 14% - 30% above the threshold.

In addition it was found out that in a minute from start the laser spot looses its brightness. If one stops the laser at this moment and touches the diode bars it may be found, that they are too warm (about 40..50 Celcium degrees). Most probably exactly the overheat of the diodes causes loss of laser power. It means that such a simple design of laser is not intended for any prolonged operation. One needs to take care of cooling.

Power measurements with a homemade Peltier calorimeter haven't yield anything trustworthy - the laser degrades before the calorimeter readings become stable. So it is.

Concerning the laser power, i think that one should not judge by that simple model. One certainly needs to go up further from the threshold. Otherwise there will never be any efficient lasing. And secondly one needs to equip the laser with heat sink and increase repetition rate.

20.01.22

It lases. One can easily see this using a camera, and with proper focusing it is able to make tiny spots on carbon paper. The energy is about 300 mcJ per pulse. Let me note that milliwatt powers and microjoule energies are pretty out of range with my measurement equipment, so the errors are comparable with the very value to be measured.

The most evil here is that the laser sits almost exactly on the threshold. Indeed, it begins lasing between 6 and 7 amperes. To be definite let's assume that it is 6.5 Amps. And laser yields 0.3 mJ at 8 Amps. If we take data from the curve of power to current dependency for 740 nm diode, we'll get that at 8 amps it emits 1948 arbitrary units of power, and at 6.5 amps it does (1836+1725)/2 = 1780.5. Dividing one by another we'll get 1948/1780.5 = 1.094. It means that the laser is only 9.4% above its threshold. It is obvious that it can not emit more than this exceedence as a light. So it emits not more than 9% of the total energy stored in the rod.

So we need to increase the pumping. But how? Even at 15 amps we can expect only 2284/1780.5 = 1.28. Maybe we can increase the number of diodes? Again: how? The geometry does not allow to place more than 5 12-diode bars around some 3 mm rod. If one took a more thick rod it would have less laser gain and may be unable to reach the threshold at all.

However there exists another mystic jitsu. One can try to increase not the real diameter of the rod but the perceived one. Top do this one needs to place the rod into a pipe made of highly refractive material. If the refraction index of the material is n, than the perceived rod diameter will be increased by n times. If n=1.5 (refractive index of glass, polystyrene, silicone) and real diameter is 3 mm, then the visible diameter may be made up to 4.5 mm. The question is only in getting some suitable tubing and highly refractive liquid, to fill the spacing between the tubing and the rod. (Let me note, that in case of the spacing between the tubing and the rod was filled with air, there would be no increasement of visible diameter at all.)

It was not easy (uneasy?) to find some proper tubings. A piece of drip-bottle hosepipe appeared to be suitable more or less. Its outside diameter was not exactly 4.5 mm but 4 mm only. 4mm/3mm = 1.333 - so one can fill the tube with any transparent liquid, even with water.

As it appeared one can place 7 diode bars around 4 mm rod (with some crunch and with ~0.5 mm of spacing between domes of diodes and rod's surface.) Two new 12-diode bars were made in addition to the previous ones, and also a heatsink/holder was mold (using common tin solder).

All 7 bars were connected in series. More than expected, that the power supply was unable to press more than 8 amps through the bars even at the highest voltage setting on the variac. To simplify the things the most straightforward solution was choosen: a voltage doubling rectifier was placed in between the power supply unit and the variac. As the result, now this power supply can provide up to 15 amps of pump current even when loaded by all 7 LED bars.

Testing of the laser (3 mm rod inside of 4 mm jacket, 70% rated output coupler and 99+% other mirror) in the newly made luminaire have shown that lasing threshold is slightly below 5 Amps. Let it be 4.9 amps to be definite. The 740 nm LED gives 1577 arbitrary units of light at this current. So at 15 amps we can expect 2284/1577 = 1.45, i.e. 45% above the threshold.

An attempt of measurements of the output power gave something like 1.4+-0.3 mJ at 15 amps of pump current. And by feel the laser has become much more stable. When focused it burns carbon paper "to white spots". It is still not enough to light a match or to make anything even more useful.

Besides, any attempts to put 5 mm rod into the luminaire, have failed. No lasing. The mirrors were the same (70%/99%), resonator length was varied from 100 mm to 150 mm, since 5 mm rods are usually long.

16.02.2022

To check the ability of the laser to work with other types of diodes, 60 pcs of 730 nm LEDs were purchased (at the same 'ali') They were ordered when the laser consisted of five bars 12 LEDs each. Today 60 pcs is not enough to replace all the LEDs in current version of the luminaire. However even if 5 bars of 7 total were changed from 740 nm to 730 nm ones it should clearly show the tendency. By tracing the threshold and output power one can clearly say whether the laser works better with 740 nm diodes or with 730 nm ones.

The plan was the next: to measure the threshold and output power of the laser with only old (740 nm) bars, then to replace five 740-nm bars by 730 nm ones, and then to to measure the threshold and output power again.

However from the very beginning it has gone criss cross.

The threshold on old bars has been measured without any troubles. It appeared to be 4.7 Amps (compare to 4.9 Amps estimated earlier). When the power was about to be measured the troubles began. Earlier the working current of the laser was choosen to be 15 Amps, due to the fact that at 20+ Amps the LEDs fail. The laser was pretty able to work at this current. Only once in a while a random LED was shot out (not more often than once a pair of days, which correspond to sum of 10 minutes of continuous run at rep rate of 9.5 Hz). In this case the diode was replaced by a new one and the laser was able to operate further. But today it bluntly refused to get the current over 12 Amps. And even at 12 Amps diodes were shot out too frequently (each 30-50 seconds of continuous run). It is still incomprehensible what the evil has happened to it? The laser was stored for a month. It was not turned on. It was at room temperature, not at winter cold. It was not kicked or dropped down...

As the result it appeared to be necessary to take 10 Amps as the limiting top working current. At this current the output power was measured. It appeared to be 11 mW. Of course all dead diodes were replaced by working ones before the measurements. (5 pcs were replaced.)

Then the laser was dismantled, 5 bars were taken out, and 5 bars with 730 nm diodes were put in their place. An attempt was undertaken to measure the threshold. Up to the current of 10 Amps there was no lasing observed. Since the laser was able to run at 12 Amps before, the attempt was made to increase the current further. However somewhere near 11 Amps the LEDs died. As the result of flaw detection it was found out that all new (730 nm) LEDs were dead as a single one. Here is the photo of the dead bars. (Sorry, I remembered to photograph them when it was already too late, so I have no shots of the alive bars.)

However visible damages are small, thus the dead bars on the shots look almost like alive ones. Note the definite red color of crystals, whereas in the old 740-nm LEDs the crystals are to the most extent "colorless" (see photos above).

The continuity test has shown why 740-nm LEDs were shot out one by one, while 730-nm LEDs died all at once. The reason is that 740-nm diodes go into disconnection when dead and 730-nm LEDs go into
short-circuit.

But still the essence of the result is not in how much of damage can bring the death of these or that diodes. The sense is that

LASER DOES NOT WORK WITH 730nm DIODES.

And generally it is of little importance whether it does not work due to the lousy coincidence of spectra or due to lack of power. Not working means not working.

January 2023

An attempt was made to light a match with this laser. Obviously there;s nothing to expect at 10 Hz repetition rate, so the laser was modified for work at 100 Hz. LED bars were assembled on thin (0.01") cladded textolite and then mounted on a metal heat sink (tin mold with heptahedral cavity).

The power supply, being overclocked to 100 Hz, began to show deep depression of current pulses (when, say, it yields 10 Amp at 10 Hz, this sinks to 8 Amps at 100 Hz). Due to the fact that the laser sits near the threshold, it is easy to understand that such a depression in current would produce unbearably low output power (may be of the order of magnitude). As the result yhe power supply unit was rearranged into a classic analogue current regulator, which keeps current at constant level independently to the repetition rate until there exists any excess of voltage.

It is sorry, but the laser was not able to light a match even when overclocked. But at least a video was made, with its quality much better than the previous videos in this guide. Here You can observe the operation of this laser in details.

The diodes in bars burn much more easily than the match. Apparently there's nothing more to expect from the laser, so the

PROJECT WAS CLOSED

THE END

1. Handbook of lasers. Edited by A.M. Prokhorov. In 2 volumes. Vol1. M.:Sov.Radio, 1978.
2. Kuan-Yan Huang, Cheng-Kuo Su, Meng-Wei Lin, et al. Efficient 750-nm LED pumped Nd:YAG laser. OPTICS EXPRESS, Vol 24, No 11, OSA 2016 (Thanx to Pawel Woznyak for kindly provided reference)

<< HOME PAGE