Differences & Similarities of Red and Blu Diodes

[IgorT]

Since i often have to test large batches of diodes to sort them by efficiency, i was preparing an improoved method of plotting them, which would allow for a better comparison of diodes of the same type, and also allow comparing completelly different diodes, be it a different model or even a different wavelength..

I also wanted to compare Reds to BluRays in a meaningful way, something that is not possible with the usual PI or PIV plots..

For now i tried comparing a high efficiency 8x and a high efficiency LOC - something i've been meaning to do for a while, especially since BluRay diodes are suddenly reaching powers higher than even the toughest Reds..

I found the results quite interesting, and thought they deserved their own thread, so here goes...

***
​

Until now we have mostly been using the slope efficiency plots (mW per mA) to compare one diode to another...

Slope efficiency comparison is often enough, but diodes vary in their forward voltage even among the same model. The variations in Vf can be even greater between diodes from different manufacturers, while different wavelength diodes work in completelly different voltage ranges.. And Vf matters a lot, when it comes to actual diode efficiency!

For example:
- If one diode produces less power at a certain current, but has a substantially lower forward voltage, it could actually be more efficient than a diode that produces more power at that current but with a much higher Vf...
- If two diodes produce the same power at the same current, but one of them has a lower Vf, that diode is in reality more efficient, and will produce less heat at that current..


When testing 6x's, i found the Vfs to vary quite a bit, and because of that, i wanted to use a more accurate method for comparing 8x's.. I would still make the usual slope efficiency plots (mW per mA), but for a proper comparison, i'd use Po vs. Pe plots (Optical Power OUT vs. Electrical Power IN)...


Since the differences are small among diodes of the same type, i wanted to demonstrate the importance of Vf by comparing an 8x to a LOC...


I started by collecting the data and plotting it in the usual way.

In every plot i used:
- a red line for the LOC's primary axis (showing output power (Po) in all graps)
- a dark blue line for the 8x's primary axis (as above - optical power (Po) in all graphs)
- a pink line for the LOC's secondary axis (showing either Vf, Efficiency, El. Power (Pe), or Heat)
- a light blue line for the 8x's secondary axis (as above - depending on the graph)


- Plot #1 is a normal PIV plot of an 8x and a LOC
It shows all the electrical characteristics of each diode. It has a meaning for each of the two individually, but for comparing them it is (almost) completelly useless.. All it shows is that the Vf of an 8x is much higher than that of a LOC, and that the LOC needs WAY more current to even start lasing..

Since the two diodes are plotted by the input current, the two plots stand very far appart. The 8x seems MUCH more efficient than the LOC, since it produces so much more power at the same input currents.
But it's much higher Vf shows that in reality, the electrical power into the 8x diode is a lot higher at the same currents than with the LOC...


Next i made a new plot, where the optical output power of each diode is plotted against it's input power..


- Plot #2 is a Po/Pe & Efficiency plot
It shows Optical Output Power plotted against Electrical Input Power for both diodes, as well as their ACTUAL efficiency (a percentage showing how much of the electrical input power is converted into optical output power)..

The plot still consists of the same data as the first one, except that here i multiplied the input current (I) and forward voltage (Vf) data for each diode, to get the electrical input power (Pe), and plotted their output powers (Po) against the results. Instead of the current, the X axis shows the electrical power into the diode.. Since i now had both the input and the output powers, i could also calculate actual efficiency of each diode and plot that onto the secondary Y axis...

Because they are plotted against el. input power, the two plots now stand much closer together, and for a while, the diodes exhibit significant similarities, and at the same time the actual differences start becoming apparent...

The new plot shows that the open can needs more electrical power to start lasing, but the difference is not even nearly as big as the first plot seemed to suggest.. The plot also shows, that for a while the slope of the LOC is steeper than that of the 8x.. The open can almost wants to exceed the 8x in efficiency (if this was a low efficiency 8x, the open can would exceed itfor a while).

But then it starts falling into the knee all reds are limited by. The power of the LOC starts increasing less and less, the further i go - every next increase in input power results in a lower increase in output power and a higher increase in heat dissipation. This is even more noticable in it's efficiency plot, which starts dropping rapidly after the peak.

BluRays on the other hand don't have these knees, and simply continue upwards in a straight line. Every increase in input power results in the same increase in the output power. Well, almost the same. The power line looks almost straight, but the efficiency line does show a drop.
In the case of BluRay diodes however, this efficiency drop is not the result of a knee or anything similar, but rather the result of the heat that developed during plotting.
If the temperature of the 8x diode was kept constant throughout the testing, the power really would form a straight line, and the efficiency plot would be horizontal after the peak...


Some interesting things i've noticed so far:
- With reds (i plotted LOCs and LCCs) the Vf climbs linearly as the current is increased - the diode's forward voltage increases by the same amount with each step of current increase - the Vf plot is a straight line... But their output power line starts curving down more and more.

- With BluRays it is the other way around! The Vf line starts curving down more and more as the current is increased, while their output power keeps climbing linearly.. Each step of current increase brings the same amount of power increase - their efficiency is only affected by their own heat..

The two diodes are behaving in the opposite way!
This might also provide a small clue as to where the knee comes from with reds, and why BluRays can be overdriven to such bizzare levels..


Since the electrical input power is a factor of current and voltage (P = U x I), and with reds, voltage increases linearly with current, this means that the electrical power going into the diode is increasing exponentially, as the current is raised! If i were to plot input power against input current, the plot would form an upwards bending curve..

But with BluRays, every next current increase results in a smaller voltage increase. It looks like the Vf increase is dropping just enough with each step, so that the input power increase stays the same with every step in current!


To check this, i made (yet) another plot..


- Plot #3 shows El. Input Power (Pin) & Opt. Output Power (Pout) plotted against Current...

As predicted, i get a straight line for the 8x and an upwards bending curve for the LOC..

This means, that as the current into a BluRay diode is increased, the input AND output powers both increase linearly with every step!
But with a Red the input power increase grows exponentially, while the output power increase drops exponentially with every step, until it gets to a point, where the output power no longer increases, and a current increase only results in more heat created...

The third graph also shows something else. The optical output power is plotted against the left Y axis in mW. The electrical input power is plotted against the right Y axis in W. In this case i put both the left and the right Y axis into the same scale - they both end at 2.4W...
Because of this, if you draw a vertical line from the diodes el. input power to it's optical output power, the length of that line directly corresponds to the amount of heat created at that current (= difference between input and output power).. I drew such a line at the highest current each diode was pushed to... With the red you can see it outputs 0.4W, when it's being fed 2.4W, meaning it is creating 2W of heat...

(This is also why in this plot i wrote Pin for Pe and Pout for Po, because the difference between the power going in and power coming out is Thermal Power, or HEAT, since it's impossible for less power to come out than goes in...)



And since i was thinking about heat now, i made a fourth plot...


- Plot #4 shows Output Power & Heat vs. Current
Here i wanted to show how heat grows exponentially with reds, and to show this, i had to plot it against current.. If i plotted it against input power, the curvature would not be this obvious, since input power already grows exponentially with reds.

It also shows that with a BluRay the heat grows linearly with input current (as mentioned before, the slight but noticable curvature is a result of the diode not being kept at a constant temperature)...

This graph is not that much different from the previous one, the bottom two lines are the same as before, but the top two now show the calculated heat - the difference between the input and the output power i was talking about while describing the previous plot...



For the last plot, i again wanted to show the two diodes in a way they could be compared better, so i repeated the previous plot, but this time i used input power - Pe (or Pin) - for the X axis...


- Plot #5 shows Output Power & Heat vs. El. Input Power
This plot is again much better suited to compare different diodes, because the data is plotted against input power on the X axis...

Now the lines are suddenly so close together, that the two diodes almost seem to display more similarities than differences.. But only until you realize, that it's the outer two lines that represent the LOC, and the inner two lines that represent the 8x! And the LOC lines are curving appart, while the 8x's lines are straight (again)..



While i have the feeling i've been repeating myself for a while now, i want to show something interesting on this last plot:

- Let's start with the LOC... If you look at the plot, it takes 1.3W of el. input power, to reach 300mW of output power. And at this same point the LOC produces 1W of heat (just go up from 1.3W first to the red line and then to the pink one).

- The 8x on the other hand reaches 300mW at 1.15W of input power, and produces 0.875W of heat, judging by the graph (i used a higher resolution graph to get to these numbers, cos i don't have exact numbers for that specific power recorded)....

So far, while there is a difference, it is relativelly low, but it's clear that the 8x is quite a bit more efficient.


But let's say i want 400mW!

- To reach 400mW with the LOC, i have to feed it 2.4W of electrical power, and at that point it produces 2W of heat! I only reached 33.3% more power than before, but the diode is producing TWICE AS MUCH HEAT as it did at 300mW!

- Now let's look at the 8x... It reaches 400mW at 1.56W input power, and produces only 1.16W of heat! Again a 33.3% power increase, but the heat doesn't seem to have increased all that much!

Let's calculate how much the heat increased (or does anyone want to guess?).. 1.16 - 0.875 = 0.285W => 0.285 / 0.875 * 100 = 32.6% more heat...

I would have guessed 33.3%, because as i mentioned multiple times in this post, every aspect of the 8x except it's Vf behaves linearly.. The reason the calculation was off is that i didn't have measured numbers for 300 and 400mW, but had to read the values off the plot. Considering this, the error is quite small...

I could now also try to calculate how much more input power it took to go from 300 to 400mW with the 8x, but again i'd be more accurate if i simply said 33.3% more input power, or rather one third more...



So with an 8x, to increase the power by one third, the input power has to be increased by one third, and the amount of heat produced will increase by one third.

With a LOC, it takes 84.6% (2.4 - 1.3 = 1.1 => 1.1 / 1.3 x 100 = 84.6%) more input power, to reach one third more output power, and the poor diode creates TWICE as much heat!



"Conclusion":
It's no wonder reds can only be pushed so far.... You're probably wondering since the very first graph, how i was even able to plot a LOC to 640mA! The only reason i was able to do it is that i was increasing the current slowly, by hand, on my Diode Analyzer.... In a laser, where this current comes on suddenly at power-up, the diode unforunatelly dies instantly.

I tried it several times, on the Diode Analyzer i could bring them to 640mA every single time, in a laser they would pop instantly. I repeated this over and over again, until i had a mixture of six or seven dead LOC's and LCC's...

The only reason i even tried going so far is that i saw a post showing a 510mW LOC laser - the result of a freak accident.. I was trying to replicate the result, but no luck... I did gather a lot of interesting data however...




One thing i forgot to mention... This particular LOC seems to exhibit something resembling a kink, most noticable in the first two plots. Especialy in the second one, the efficiency line also displays a sudden drop. Unfortunatelly i did not plot the other open cans i killed - i just tried to see if they can survive 620-640mA..
I did plot an LCC tho, and it does not show anything like that.. It's plot is a smooth curve. I don't know if this was a weird LOC or if others have it, i'll have to test a couple more. But i think it's an exception...


In any case these plots make it a little easier to understand why BluRays can be overdriven to such bizzare levels, but much more than that, it shows why reds can't!

One thing's for sure, 405nm's are weird diodes! But that's nothing new, i've been saying it for over a year now..
But these comparisons did give me a slightly better insight into their behavior.

I just hope i'm not the only one obsessed enough to find all this interesting... :yabbem:



P.S. I've been thinking a lot about the fact, that the reds survived (or rather didn't die instantly) when i raised the current slowly, even if i tried repeating it, but died every single time, when i put them on a FlexDrive set to the same current...

If a driver was made to raise the current VERY slowly at powerup, one could make a 400mW+ red that would light up at least a few times.. Of course it would ultimatelly still die quite rapidly, because the amounts of electrical power going in and heat being created are beyond bizzare....

But still, a very slow powerup allowed me to bring reds to 400mW several times, while an instant powerup killed them on the spot!
If it makes such a big difference here, it would surelly also make a difference at the currents they can survive..

It is well known, that it's the power-up stresses that torture diodes the most, and they often die just when you turn them ON..


I can't help but wonder how much longer diodes could live if they were powered by a driver that would raise the current so slowly, you could actually see the power gently climbing to max..


I'm thinking of performing another test on the Cycler, killing several cheap diodes (PHRs and reds), half of them powered by a regular driver and the other half powered by a very gentle slow power-up driver, to see HOW MUCH of a difference it could make at "normal" currents...

 

[pullbangdead]

Ah, the magic of materials science.

And even more so, the magic of GaN. That's the difference, GaN/InGaN instead of the AlInGaP/GaAs used in red lasers. How does GaN do it? There are still a lot of mysteries about it, actually. Basically, it's an amazing solid, with weird physics underneath.

Just compare some of the qualities of GaN, and it's one of those things that seems like it should never work, and 20 years ago people had pretty much completely written it off as impossible. Heck, just last year computer modeling people were telling us that green InGaN laser diodes might be close to impossible, because the detrimental effects were just too large. But look where we are now, multiple companies reporting working green laser diodes.

For instance, if you look at crystal dislocations, basically defects in the crystal that can impact performance: In Si nowadays with how good we are at growing Si crystals, you can get less than 1 dislocation per cm^2. But then you look at GaN, and the best in the word are still something on the order of like 10^6 or 10^7 per cm^2. So it's 6 orders of magnitude worse in defects than a silicon crystal, and yet still performs better than the best red laser diodes available!

Red laser diodes are a much older technology, they've been mostly perfected in many ways. And yet the new kid on the block, the GaN laser that has existed really for little more than 10 years, kills other materials systems for performance in many ways.

How does it do that? Answer that, and you'll do VERY well for yourself in a science career.

But really, look how much more ceiling there is! We're at least 6 orders of magnitude from having crystals the same quality as Si crystals, that's 6 orders of magnitude of room for improvement!

Of course InGaN/GaN isn't all rosy, there are some real difficulties with it that you don't have to deal with in other materials systems. But the potentials are amazing, and half the ideas out there haven't even been tried yet.

But yeah, it's crazy stuff, I have a lot of viewgraphs from Shuji's class, and some other classes, showing a lot of the major points you hit, and showing a lot of the mysteries still surrounding GaN. Plenty of research still to be done.

 

[jayrob]

Really nice work Igor... :thanks:

About the idea of a driver that will power up slowly. As an example of what you have in mind, what kind of power up time are you thinking? Maybe 5 seconds to reach the set max? Or what?...

 

[bryce007]

Wow thanks for doing this! (admits to not having finished reading it yet), but it's good so far!

 

[Dusty]

Awsome info there Igor, that was a very interesting read...:yh: you are doing the forum a great service by carrying out experiments such as these...:beer:

 

[IgorT]

About the idea of a driver that will power up slowly. As an example of what you have in mind, what kind of power up time are you thinking? Maybe 5 seconds to reach the set max? Or what?...

I'm not sure how gentle the slope has to be for the diode to survive absurd powers, or to live longer at normal ones... Obviously the less steep, the better, but i think from a certain point on it doesn't really make a difference if the powerup was even slower..

And i think that point to be below a second... How much it is would be hard to figure out preciselly. But if the powerup was slow enough for the power climb to be noticable to the human eye - and it doesn't even have to be a second for that - i believe it would be slow enough for the diode to survive..


I can adjust the powerup speed to almost whatever i want by changing the value of one component on the add-on circuit, so i could experiment to find the right value...


I was thinking about what it is, that kills the poor LOCs and LCCs, when they are powered up rapidly at these currents, when they can "survive" them with a slow powerup.


Normally, there are several factors when it comes to diode degradation and their ultimate demise:
- First, there is the optical flux at the die... Under normal operating conditions, it the diode's own output power, that gradually "eats away" at the end facets... But in this case, every single diode was clearly capable of handling the ~400mW of optical flux produced at 640mA at least for a little while (I should probably mention, that LOCs are rated for 150mW CW / 400mW Pulsed)...

- Then there is the temperature, which makes diodes age faster, but again the diodes i tested were clearly capable of dissipating the ~2W of heat into the heatsink efficiently enough not to overheat (good thermal design!).


So the LOCs and LCCs were both capable of withstanding ~400mW of optical flux WHILE producing and dissipating 2W of heat!

The thing they could not survive was getting to this point rapidly. The power-up stresses alone or in combination with the optical flux were too much for them to handle..

The most obvious power-up stress is the thermal one. Not the 2W of heat created and flowing through diodes while they are already running, but the sudden "acceleration" from zero to 2W of heat suddenly developing at one spot of the diode in only milliseconds...

Because it comes on so rapidly, the diode doesn't have the time to heat up evenly. And it consists of several parts made of different materials, each with their own thermal expansion coefficient. If the "hot" side of the die suddenly jumps to producing 2W of heat, while the rest of the diode is still cold, this produces serious stresses for the diode - different parts expanding at different rates means tensions are created between them. And if it happens too rapidly, it can't be good...

I'm guessing that's what kills them. Or at least weakens them enough for the optical flux to finish them off....


But if the diode is brought to it's final temperature slowly, it's parts can heat up evenly enough, that it doesn't "crack".. The thing that kept them alive when powered up slowly, was the fact that they were already pre-warmed!

So the startup has to be slow enough to allow this, and it would be most beneficial for the severelly overdriven diodes, which produce large amounts of heat...





P.S. A thought occurs... What if the BluRay dies are attached to the pedestal with the hot side of the die, just like the high power common anode diodes (so that the heat doesn't have to flow through the die before leaving the diode through the pedestal), but unlike the common anode diodes, insulated from the pedestal electrically, but not thermally?

Fact is, they are electrically insulated from the pedestal, since they are neither case negative (like reds) nor case positive (like high power multimode diodes - IR pumps and similar).. The die could be oriented either way and it would definitelly make more sense to attach it to the pedestal with the hot side first.

This would also explain why they are floating pin... They have to work in the same metal sled as case negative reds and IRs.. If they are mounted with the "hot" side towards the pedestal, but would not be insulated from it, it would be impossible to power them, since it wold create a short over the sled.

The different die orientation could be another part of the reason they can be pushed to such high levels..


PullBangDead: Do you know anything about this perhaps? I just remember something i read about common anode diodes once, and it made me think now. I hope i didn't misunderstand it..

Need to go check some datasheets..

 

[ElektroFreak]

I'm not sure how gentle the slope has to be for the diode to survive absurd powers, or to live longer at normal ones... Obviously the less steep, the better, but i think from a certain point on it doesn't really make a difference if the powerup was even slower..

And i think that point to be below a second... How much it is would be hard to figure out preciselly. But if the powerup was slow enough for the power climb to be noticable to the human eye - and it doesn't even have to be a second for that - i believe it would be slow enough for the diode to survive..


I can adjust the powerup speed to almost whatever i want by changing the value of one component on the add-on circuit, so i could experiment to find the right value...


I was thinking about what it is, that kills the poor LOCs and LCCs, when they are powered up rapidly at these currents, when they can "survive" them with a slow powerup.


Normally, there are several factors when it comes to diode degradation and their ultimate demise:
- First, there is the optical flux at the die... Under normal operating conditions, it the diode's own output power, that gradually "eats away" at the end facets... But in this case, every single diode was clearly capable of handling the ~400mW of optical flux produced at 640mA at least for a little while (I should probably mention, that LOCs are rated for 150mW CW / 400mW Pulsed)...

- Then there is the temperature, which makes diodes age faster, but again the diodes i tested were clearly capable of dissipating the ~2W of heat into the heatsink efficiently enough not to cook themselves.


So the LOCs and LCCs were both capable of withstanding ~400mW of optical flux WHILE producing and dissipating 2W of heat!

The thing they could not survive was getting to this point rapidly. The power-up stresses either alone or in combination with the first two factors were too much for them to handle..

The most obvious power-up stress is the thermal one. Not the 2W of heat creation and flow through them, but the sudden "acceleration" from zero to 2W of heat suddenly developing at one spot of the diode in only milliseconds...

Because it comes on so suddenly, the diode doesn't have the time to heat up evenly. And it consists of several parts made of different materials, each with their own thermal expansion coefficient. If the "hot" side of the die suddenly jumps to producing 2W of heat, while the rest of the diode is still cold, this produces serious stresses for the diode...


Different parts expanding at different rates means tensions are created between them. And if it happens too rapidly, it can't be good...

I'm guessing that's what kills them. Or at least weakens them enough for the optical flux to finish them off....


But if the diode is brought to it's final temperature slowly, it's parts can heat up evenly enough, that it doesn't "crack".. The thing that kept them alive when powered up slowly, was the fact that they were already pre-warmed!


So the startup has to be slow enough to allow this, and it would be most beneficial for the severelly overdriven diodes, which produce large amounts of heat...





P.S. A thought occurs... What if the BluRay dies are attached to the pedestal with the hot side of the die, just like the high power common anode diodes (so that the heat doesn't have to flow through the die before leaving the diode through the pedestal), but unlike the common anode diodes, insulated from the pedestal electrically, but not thermally?

Fact is, they are electrically insulated from the pedestal, since they are neither case negative (like reds) nor case positive (like high power multimode diodes - IR pumps and similar).. The die could be oriented either way and it would definitelly make more sense to attach it to the pedestal with the hot side first.

This would also explain why they are floating pin... They have to work in the same metal sled as case negative reds and IRs.. If they are mounted with the "hot" side towards the pedestal, but would not be insulated from it, it would be impossible to power them, since it wold create a short over the sled.

The different die orientation could be another part of the reason they can be pushed to such high levels..


PullBangDead: Do you know anything about this perhaps? I just remember something i read about common anode diodes once, and it made me think now.
I hope i didn't misunderstand it..

Need to go check some datasheets..

VERY nice work, and a very interesting read. I do have one comment about your theory on the effects of rapid thermal change at hot-spots on the diode. I have my reservations about how much stress that actually causes. Certainly temperature changes stress any material, electronic or not, but I use diodes that are modulated anywhere from the low AF range into the mid-to high RF range. Modulation cycles also cycle the temperature, especially at low frequencies, but this does not seem to affect lifetime as long as the modulation signal never exceeds the maximum rated current. One would think that mixing rapid temperature changes with normal gradual facet damage would lead to shortened lifetimes.

Also, I'm curious as to what kind of specialized diode-testing equipment you use (from your sig)?

 

[suiraM]

Thermal effects kill the diode, as evidenced by the time scale.

I am not a physicist, but it seems likely to me that you are blowing the junction apart.

You are effectively sending a thermal gradient shock wave through the lattice. At and near the surface of the wavefront, the distribution of kinetic energy among the atoms in the lattice is such that we no longer need to look to the absolute far tail end to find a lot of ballistic ones. Now, ballistic atoms are limited by free path length, but the path length can get pretty long at the surface of the wavefront. And, more to the point, the free path is anisotropic and asymmetric. Sure, we can look at optical effects, but the round trip time of a photon in a laser diode cavity is just shy 6.7ps, wheras the thermal lag time is most likely four to six orders of magnitude slower, definitely three or more at best (i.e. cooling on the top side of the chip, everything laid out for optimal heat transfer, etc.). Unless your driver is spiking at turn-on (you've checked it on the 'scope, right?), there is almost no chance of a non-thermal cause of death.

If you want to test this theory, then try this with a cheaper diode:

Measure the steady-state forward voltage at the current you want to arrive at, then make a precise shunt that will clamp the forward voltage at that level. Then drive the diode with a constant current source that delivers the current you want. You will still damage the diode, but I don't think you will blow up the junction, since you will now effectively be clamping the magnitude of the thermal gradient. If the junction still blows up, it should be to stresses outside the junction proper, probably causing a cleavage of the crystal that you can see after a postmorten can-ectomy. Those cleavage surfaces will probably be damn sharp, so be careful if you do try it.

Last diode I blew up, it seemed like the end mirror had failed, but there was a hint of thin lines radiating out from the central point of the failure, almost like an aurora that's been wrapped around a tube, or certain dual medium shock waves. Is that how the projections from the murdered diodes look, as well?

You've done great work so far, which is how I even have anything to speculate about.

I hope you'll view the above in that light and not take it as anything but curiosity and appreciation of your efforts. I'm in no way implying that my hypothesis is intuitive (or even correct), nor that the work you have done is in any way lacking. I'm happy that you've done the empirics and shared the results. Curiosity just doesn't end, that's all; always want to know more.

 

[maxh]

suiraM: Very interesting. I was going to suggest that IgorT test his hypothesis by ramping up the current on an LOC to 640mA, then rapidly switch it off and back again. If the diode survives, thermal stress would indeed seem to be the culprit, and if it dies, then back to the drawing board. But now I wonder if the cooling rate upon switching off is great enough that in order to perform the experiment in a meaningful way, the switching would have to be done by a microcontroller instead of a slow human hand? Hmmm... just random thoughts...

 

[pullbangdead]

Thermal effects kill the diode, as evidenced by the time scale.

I am not a physicist, but it seems likely to me that you are blowing the junction apart.

You are effectively sending a thermal gradient shock wave through the lattice. At and near the surface of the wavefront, the distribution of kinetic energy among the atoms in the lattice is such that we no longer need to look to the absolute far tail end to find a lot of ballistic ones. Now, ballistic atoms are limited by free path length, but the path length can get pretty long at the surface of the wavefront. And, more to the point, the free path is anisotropic and asymmetric. Sure, we can look at optical effects, but the round trip time of a photon in a laser diode cavity is just shy 6.7ps, wheras the thermal lag time is most likely four to six orders of magnitude slower, definitely three or more at best (i.e. cooling on the top side of the chip, everything laid out for optimal heat transfer, etc.). Unless your driver is spiking at turn-on (you've checked it on the 'scope, right?), there is almost no chance of a non-thermal cause of death.

Whaaaaa? :thinking:

Free path of the atoms...in a solid? Ballistic atoms in a solid? Heck, ballistic anything in a laser diode? Color me...skeptical.

Maybe we're confusing some terms here...But the definitions of those things, at least the definitions that I know, just don't fit with what you're saying.

Any clarifications here?

------------------------------------------------

As far as the mode of death, I agree it's likely thermal, but I am skeptical that it's a thermal gradient during turn-on that is doing it. Remember, pulsed in kHz-MHz is BETTER for laser diodes than CW. They like being pulsed that fast, and have no problems with any thermal gradients.

In fact, the roll over in output that we see in the red diodes is purely because of heat. If you plot the same diode running pulsed, the L-I curve will stay linear and won't roll over, at least not until much higher. The decrease in slope of the L-I curve is purely heat. In fact, if you know the physical parameters of the diode, you can calculate that rollover ahead of time, it's a well-understood phenomenon. You don't see the same rollover in the violet diodes because GaN is much less bothered by temperature difference. Amazing stuff, that GaN.

 

[IgorT]

I have my reservations about how much stress that actually causes. Certainly temperature changes stress any material, electronic or not, but I use diodes that are modulated anywhere from the low AF range into the mid-to high RF range. Modulation cycles also cycle the temperature, especially at low frequencies, but this does not seem to affect lifetime as long as the modulation signal never exceeds the maximum rated current. One would think that mixing rapid temperature changes with normal gradual facet damage would lead to shortened lifetimes.

Well, to be honest, in this case i was mostly just thinking out loud, trying to imagine what the cause of death could be...

There are of course other possibilities. For example, when i was raising the current slowly on my diode analyzer, the diode was warm by the time it reached 640mA. The increased temperature resulted in a lower efficiency and because of that, a lower optical flux.

When the diode is then suddenly powered up cold to that same current, it starts out at a higher efficiency, the optical flux is higher than it was with a warm diode, and in theory at least, the extra output power could be enough to push it over the limit... Fact is, that a hot diode can sometimes survive a higher current than a cold one, just for this reason.

Problem is, i have no way of knowing how the diode behaves in these first miliseconds, when it is cold and suddenly powered up - no way of measuring just how high that initial power peak really is, so this is again just another guess.


What do you think the reason is?

Also, I'm curious as to what kind of specialized diode-testing equipment you use (from your sig)?

Oh..

That would be my "Diode Analyzer", which i use to make PIV plots of all diodes before using them, in order to sort them by efficiency (and weed out the weaklings).
Also the Cycler circuit, which i use to torture diodes in order to find out how long they can survive certain powers (the diode is turned ON for a certain amount of time and OFF for a certain amount of time, while it's beam is blasting against a photo transistor toggling a timer or a counter, to measure how long it survived automatically).

Just now for example i am preparing to kill two 8x diodes in an attempt to find a power where they are relativelly "reliable" as a hobby laser pointer, or at least figure out how long they can survive the powers we are already setting them to.

Specialized, because i made these devices specifically for the purpose of testing and experimenting with laser diodes. :yh:

 

[ElektroFreak]

What do you think the reason is?

I couldn't really speculate on that myself, as I have a limited understanding of these things compared to PBD and yourself. From my perspective, things get so random when you're pushing electronic devices past their ratings that it's really hard for me to tell.. I just thought I'd share my feelings and experiences on the temperature issue.



Oh..

That would be my "Diode Analyzer", which i use to make PIV plots of all diodes before using them, in order to sort them by efficiency (and weed out the weaklings).
Also the Cycler circuit, which i use to torture diodes in order to find out how long they can survive certain powers (the diode is turned ON for a certain amount of time and OFF for a certain amount of time, while it's beam is blasting against a photo transistor toggling a timer or a counter, to measure how long it survived automatically).

Just now for example i am preparing to kill two 8x diodes in an attempt to find a power where they are relativelly "reliable" as a hobby laser pointer, or at least figure out how long they can survive the powers we are already setting them to.

Specialized, because i made these devices specifically for the purpose of testing and experimenting with laser diodes. :yh:

Very interesting.. they sound like very useful devices.

 

[Deleted member 8382]

actually we should setup and IgorT Laser FAQ, if we put together all his explanations I'm sure it beats the SAM's one o.0

Thanks for these great graphs and info!

 

[suiraM]

Whaaaaa? :thinking:

Free path of the atoms...in a solid? Ballistic atoms in a solid? Heck, ballistic anything in a laser diode? Color me...skeptical.

Okay, will you be happier if we move on to dealing with ballistic phonons and the mean free path of excitons and so forth? A crystal lattice is, like any non-hyperdense matter, mostly empty space. The atoms move back and forth a bit, due to the kinetic energy (heat) they possess. When you start pushing the heat up, this allows migration of impurities through a solid, which is one of the ways to refine e.g. silicon: zone melting. It's also the basis for annealing, and a lot of other processes.

As far as the mode of death, I agree it's likely thermal, but I am skeptical that it's a thermal gradient during turn-on that is doing it. Remember, pulsed in kHz-MHz is BETTER for laser diodes than CW. They like being pulsed that fast, and have no problems with any thermal gradients.

Which is what I was saying.

The thermal lag is orders of magnitude slower than the duration of a pulse when you run it within spec. The pulses generate a lot of heat power (Watts) for a short period of time, so they don't generate a lot of heat energy (Joules). This prevents the gradient from being a problem, as the limited energy readily diffuses through the bulk and averages out to a low heat flux that leaves the chip at an acceptable average temperature. When you switch it on without switching it off, however, the heat energy is much greater, and the thermal lag becomes a real problem, allowing the gradient to build to a higher level, too high to move out of the junction in time.

Correspondingly, if you start low, and increase slowly enough to circumvent the thermal lag problem, you can run the chip CW at its rated pulse power with no problem. If you add TEC to the picture, and ramp that slowly enough to avoid generating a significant gradient, you can exceed the pulse power rating in CW operation, up to a limit determined by the facets and the minimum gradient enforced by the bulk thermal resistivity of the junction. You will still see a shorter lifespan, but you can get more out of a red than you might think, if it is run the right way.

I hope I've explained things more clearly this time around.

 

[suiraM]

Problem is, i have no way of knowing how the diode behaves in these first miliseconds, when it is cold and suddenly powered up - no way of measuring just how high that initial power peak really is, so this is again just another guess.

I can ship you a Hamamatsu photodiode, or even a 3000 frames per second binning imager, if you have the EE know-how to use it to find out. If you lack the appropriate ND filters, I could pack a few of those, too. Would that help?

 

[pullbangdead]

Okay, will you be happier if we move on to dealing with ballistic phonons and the mean free path of excitons and so forth? A crystal lattice is, like any non-hyperdense matter, mostly empty space. The atoms move back and forth a bit, due to the kinetic energy (heat) they possess. When you start pushing the heat up, this allows migration of impurities through a solid, which is one of the ways to refine e.g. silicon: zone melting. It's also the basis for annealing, and a lot of other processes.

Ballistic phonons does make a lot more sense than ballistic atoms, on several levels. And yeah, I get phonons and diffusion and zone refining and all of those kinds of things. Got that stuff covered.