Differences & Similarities of Red and Blu Diodes

[IgorT]

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...

This is exactly what i've been thinking!

See, the way i did it, was, i ramped up the current to 640mA, wondered how the hell the diode is not dead yet, took it out of the testing apparatus and mounted the diode in a new (cold) heatsink, and started assembling the laser...

By the time the laser was done, the diode was very cold. And when i turned the laser ON, the diode died so rapidly, that i never even saw the optical pulse that killed it, just the result..

Only after repeating this two or three times, did it occur to me, that maybe i should try ramping up the current to 640mA several times during testing. I was hoping this would eliminate diodes that can't survive those currents, without having to waste the time to build a laser, just to see the diode die every time...

I was thinking, that perhaps, the diode can only get to those currents and powers once, and gets damaged enough during the first test, that the next time it dies. So with the next one, i ramped it up twice. It survived! In a laser, again dead at first powerup...


What i did not test, because it only occured to me afterwards was just what you said above..

Thing is, on my diode analyzer i don't always have to start from zero.. I could simply turn the diode OFF at 640mA let it cool off and turn it back ON again. This way i would find that one diode, that could survive a rapid powerup, without wasting the time to build a laser around it first (this was when i still thought such a diode must exist - because of Billg's 510mW LOC)...

I could also test what would happen if i turned it OFF WITHOUT allowing it to cool off before turning it back on again!
This would definitelly be an interesting thing to test, and i will, as soon as i have time to "waste".

When testing this, i should perhaps even reduce the heatsinking, to make sure it doesn't cool off too much during the time it's OFF..
Or, as you said, turn it off for a very short time, using a computer interface or an MCU.


However, even if i do try this and the diode survives a rapid powerup when hot, this still leaves AT LEAST two possibilities:
1. Thermal stresses were avoided, since the diode was still hot from before
2. Optical flux did not go far above the previously measured point, since the diode was powered up hot, with the efficiency still reduced from before.

But in all likellyhood, it's the combination of both (or more?) factors..

----------------------------​

Then again, the whole reason i was pushing diodes to these absurd currents was Billg's freak 500mW+ LOC...
I was trying to find one diode that could survive the sudden powerup to these currents and powers...

But after thinking about what he told me - he accidentally selected a fixed 3.3V voltage regulator, instead an adjustable one, that could be configured for constant current - it suddenly occured to me that i can't really replicate his results with a constant current driven laser and a rapid powerup!

I was trying to replicate his results under completelly different circumstances!


What i did not realize until he told me he used a constant voltage source was, that when he turns his laser ON with a cold diode, the diode's Vf is higher than it will be, once it's hot! And in his laser, this means that the current starts lower than at the 780-820mA he measured!

The diode then starts warming up from the current, it's Vf drops, and more current starts flowing through it since the voltage is constant...
The power increases and due to more current the diode also produces more heat, warms up to a higher temperature and it's Vf drops further, allowing the current to climb further - over and over again...

This is a good scenario for a thermal runaway, but he seems to have given it just enough heatsinking, so that the current then stops climbing at around 800mA..

Basically, his constant voltage source behaves almost like a slow powerup!

- A constant current source will push the diode to the full current immediatelly, and then keep it there.
- A constant voltage source will allow the current to change with temperature - the hotter the diode, the more current will flow!

This is why we don't use constant voltage to power diodes... It's actually dangerous. But in the opposite way than with constant current.

At a constant current, a cold diode will produce more power - if a diode was supercooled, it could die at a current it can normally survive!
But under normal circumstances, constant current is the safest way to power diodes - as a diode warms up, it's Vf drops, and the driver will reduce the output voltage, to keep the current constant. Additionally the efficiency drops, and the diode drops in power, meaning the optical flux is lower, and the stresses actually reduced, when the diode warms up in a constant current setup!

At a constant voltage, a cold diode will produce less power, as it will allow less current to flow through, due to a higher Vf! But it will start warming up, the Vf will drop and the current will increase. The more it warms up, the higher the current will become!


In Billg's case, it is heat that gradually raises the current from something the diode can evidently survive even with a sudden powerup, to the current where it produces those absurd amounts of power!

Since i did not know just how his accident happened and what the driver was like before i started killing diodes, i was trying to replicate his experiment, but under completelly wrong circumstances...

Billg, by accidentally selecting a fixed constant voltage source, already introduced some sort of a "gentle" powerup into his laser. Altho the first part of the powerup is still rapid in his case, but evidently low enough for the diode to survive, and only after that does the diode climb to it's final current and power.


This means, that a constant current driver with a gentle power-up would be even better, as it would get to the final current even more gently, than the constant voltage source - the entire slope would be gradual - and it would also keep the current constant afterwards, eliminating the possibility of a thermal runaway....



I believe the constant voltage source explains HOW and WHY he was able to make a 500mW+ laser, and it shows, that using a constant current source with a slow power-up will be even "nicer" to the diode, increasing the chances of succesful replication of his results. :yh:

P.S. It does not mean that CV is good for diodes however! I explained what can happen in the worst case (thermal runaway)...

 

[IgorT]

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.

That's good info, but i also see something else in it..


You said the reason the reds show that knee is the result of the material's sensitivity to heat. Good to know! This explains a alot.

You also mentioned that when pulsed, reds don't show a knee... This is something i've been wondering about for a while (altho mostly about BluRays and their kinks), so thanks for the clarification (do you perhaps know if the same is true for 405nm's and their kinks?)...

You go on to say that they prefer being pulsed, to CW at the same power (which we know as a fact), and don't have problems with the resulting thermal gradients..

This last part makes it seem you're saying that the thermal gradients are worse when pulsed...

But are you sure about that?


I'm asking, because when i try to imagine it, i don't see it the same way.. (But of course i'm no expert, and could be imagining it all wrong..)

For one thing, the total amount of heat created is lower during pulsing.. If the duty cycle is 50%, the diode only produces half as much heat on average, as in CW. In fact, if the knee does not appear, the diode produces less than half as much heat on average, than in CW, due to the MUCH higher efficiency...


But that's not the main factor here, we're talking about changes in heat while pulsing, compared to those during a CW power-up...

Were you thinking about pulsing a diode to the same power (in this case 400mW), or pulsing it with the same current (in this case 640mA)?

Let's say we are pulsing it to the same power...
- In this case, since it's pulsed and won't display a knee, it will require MUCH less current to reach the same power..
While it is being pulsed rapidly, it is producing much less heat (on average AND during the ON state), and for two reasons:
1. It needs less current to reach the same power, due to the absence of a knee.
2. It produces less heat at this lower current, than it would at the same current in CW, due to the higher efficiency in the absence of a knee.

Now let's say we are pulsing it with the same current...
- In this case, again, since it's pulsed and won't display the knee, it will create more power than at the same current in CW...
And again it is producing much less heat, since more of the power going IN is coming OUT...

Additionally, as already mentioned, in both cases, if the duty cycle of the pulses is 50%, the diode produces only half as much heat on average - less than half actually, in both cases, due to the higher efficiency in the absence of the "knee"...


So if i try to imagine it, i only see the first pulses being hard on the diode when it comes to thermal stresses, and even these first pulses less hard than the same powerup to CW. With every next pulse, the diode is warmed up more and more from before, until it reaches a certain temperature at which it stabilizes..


The pulses surelly aren't nice to it, but after the diode reaches it's operating temperature, during the ON state it just warms up a bit from the average and during the OFF state it just cools off a bit from the average temperature... Basically, it's just going up and down a bit around a certain average, when pulsed...

The way i see it, neither is nearly as harsh to the diode as a rapid powerup from a cold diode producing no heat at all, to a sudden onset of 2W of heat dissipation, and the resulting thermal stresses (at least when i try to imagine it).


We know that pulsing is nicer to diodes. Are you sure it's harsher, when it comes to thermal gradients?

I mean, i could be imagining it all wrong, i'm just thinking out loud again. :angel:



EDIT:

In response to PullBangDead:
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.

Heh, looks like you already covered what i was trying to say above..

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.

Spiking was among the first things on my list of suspects.. In the laser, where the diodes were dying, i was using a FlexDrive, and i added a huge cap on the diode, to further soften the blow, but i know that the Flex can misbehave in some rare situations (which is the reason i always add caps to them), so i tested this particular one, just in case there was something wrong with it.

But it wasn't the driver... The powerup was normal with or without the additional caps...


Besides, i think i already figured out why Billg's diode keeps surviving, while i couldn't replicate his result even after killing multiple of them.
His constant voltage source already acts as a sort of "slow power-up" driver from a certain step onwards..

So what i'll do is, i will take one of my SEPIC drivers, and add the slow start circuitry to them. Then i can go through several diodes again, and i'm sure i'll find one that will survive 400mW... In fact i think i'll find that all of them survive it at least for a little while, if the powerup is gentle enough, just like when i was ramping up the current by hand.


Especially since i "tested" a LOC at 600mA+ once before... Well, it wasn't really supposed to be a test, i was trying to kill it before hammering it out of the module (cos someone jammed it in sideways), and i thought to myself, why not also get some info out of it before killing it...

I simply hooked it up directly to my Lab PSU, at 0.6A with only a capacitor in between.
I set the current limit to 0.6A and then slowly increased the voltage, until the current limiting engaged. I expected it to die right away..

Surprisingly, that diode survived 600mA+ for 24h straight! Of course it wasn't cycled, which explains why it lasted so long, and the way i started it up (setting the current limit and slowly increasing the voltage) was also a slow power-up!

So i'm thinking that once i have time to do these slow start-up experiments, i'll find that many LOCs can be forced to produce 400mW for a while, if only they are treated gently enough...


After that i can also test how much difference the slow start makes at 300mW, which practically all LOCs and LCCs can survive even with a rapid powerup, by cycling them side by side, half with a normal driver, and half with a slow start-up driver...

Then repeat the same tests on PHRs...



EDIT 2: Hmm, i was thinking some more...
I could easily replicate Billg's results using my PSU again. I could set it to 3.3V, and additionally limit the max current.
Then power up, and see how the current behaves - where it starts and how fast it climbs with heat...

This should replicate his constant voltage setup quite nicelly, with an added safety of the current limit...

 

[IgorT]

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?

Do you have a datasheet or part number of your diode?

I would be interested in recordings of the diode death at powerup...