Multiple diode question

[yaddatrance]

Not worrying about conservation makes the task easy... I'm assuming that laser-hobbyiest math is at work here (outside of the surplus market, normally big diodes are more expensive than simple current control drivers)...

Diodes pop just by staring at them wrong, but higher power diodes tend to be
more friendly... It's not common to find optical feedback on higher power
diodes. The space between Lasing threshold and Max power on bigger diodes tend to be friendly as well.

1) chart each diode by the current and power output. Mark the current of lasing and when it reaches rated specifications. Remove from the pool any diodes that popped or are sufficiently out of spec.
2) order the diodes in series so that the more efficient diodes are further down in the series...
3) slowly bring up current and make sure that the current never exceeds the lowest spec in the bunch.
4) If the highest output laser is outputting lower than the rated laser output, bring up the current very slowly until the highest output laser reaches rated power.

At this point the system should be reasonably working... It'd be tedious with 6 diodes, but certainly not "hard"... This is the point beyond which you will be venturing into catastrophic failure territory.

5) Crank it up until right before the first diode pops... The trick is that we don't actually know when it'll pop. If you don't mind sacrificing one right off the bat, then take an "passed" diode from the testing earlier and slowly bring up current and measure the power output until the diode pops. Use that as a rough guide for max current and max power. In general I just recommend being gentle...

P.S. You'll only need 1 PBS to conveniently join all 6 optically in the way you were planning... Use a bunch of razor thin bounce mirrors.

[The_Doctor]

How many laser diodes use germanium as a substrate?

Adam

There is that. Steve-o's right though, I overdid the case for diode junctions. While I never tried to blow a germanium diode, and managed not to the few times I used them, they have a gentle onset of conduction that makes it hard to know when enough is enough. Still, most diodes, probably even those, have a wider, safer margin by far than that for front facet damage in an LD. I could still be wrong, so if anyone knows any other kind of diode so ready to puke its sphincter up at the slightest provocation, let me know, if it's cheap I want to have fun with it.

I see a few concerns here, my main one is that although it is possible to place diodes in series, and drive them with constant current, you are making one serious assumption, is that all of your Laser diodes have exactly the same lasing current threshold (point at which they will begin lasing), and that point is what sets the amount of current you can go up from there to achieve maximum output.

Not overlooked, at least not by accident. I don't think it's that significant, in practise. I like your point about using the threshold as a guide to safe max though, that could be VERY useful info if the correlation is good and accurate.

If one diode has a lower threshold (very likely), you will end up with 2 options, #1, turn down the current so that the weakest diode is driving full output, reducing the efficiency (output) of every other diode in the chain since it is not able to put out full power, #2, the current it set at point where most diodes are outputting full power, if one diode's threshold current changes (easy as temperature change in most cases), you would probably burn optical cavity components (mirrors) within the diodes, and make an expensive LED.

That's correct, entirely I think, but I think the margin is slight. In cases where it is not, it's usually an unusual diode in a batch, and should not be planned for. Weak ones will pop early anyway, strong ones will last well, so that's something less to worry about, rather than trying to push the majority into matching them. Given that they can differ, it's enough to buy more than needed and to risk a couple to find the weak end of the scale. If we share data on diode types, we can minimise this waste and cost. I'll add my own findings for 7 ROHM DVD diodes later in this post.

If you look at datasheets, the spec current vs actual lasing current can vary as much as 50%!!! (mainly on smaller diodes), hey why do you think most of them have a photodiode for optical feedback for as opposed to driving them constant currrent?? Unless you are sampling (optically) the output of each of your diodes and running a closed loop current control based on this, I don't think its a terribly good idea.

Not so. If they required a monitor photodiode, they'd have one. The makers didn't drop them because they were lazy or thoughtless. As you say, the low power diodes are far more easily popped, and even those are manageable in ACC mode. The Philips OF4944 is an old gain guided diode, rated 3mW at 655 nm. It can actually put out more than 10 mW, you can get a tad short of this even after a non-AR coated optic! The CC drive current MUST be limited to 76 mA to run for >2000 hours. 77 will kill many in a day, 78 will kill most in a few minutes. I got through several, learning how to handle the anti-static and other risks. (Fortunately retroreflection doesn't harm those.) In short, though they are old, made before the standards rose to what we get now, they could be well overdriven, and despite what data sheets say, most were VERY consistent. Manufacturers are largely being conservative to avoid customer discontent. Where there IS reason for such large variation in spec, it's due to the few diodes that really are off spec. But they're the exception that proves the rule.

Another diode I had a hundred of is a Samsung SLD6351826 index guided MQW diode, rated 5 mW at 635 nm. Those won't overdrive as much (just 7 mW, 5mW after a non-AR-coated lens), but they are tolerant, and have an odd nature, they could be overdriven further without dying, but the power output fell, and would recover when the current was reduced to safe limit. Those could be tuned to best output easily. Curiously, at the SAME drive current that the OF4944 could take as max, 76 mA. Not sure if this is significant though...

One other diode I tried was an Opnext/Hitachi 6526FM (I think it was that number..) I had ten, at £50 a throw. Nasty, I lost 4 before realising that death by retroreflection even existed, I was trying all sorts of stuff to restrict other failure modes before I was sure. I don't know if makers declare this appalling risk, but they certainly didn't then, I had no way to claim anything back. I overdrove them slightly, but they never went far. The last died or was sold in a module before I even had a meter. I sold them as their rated 80 mW. I bet they did more, but I'll never know unless a buyer who got lucky tells me they do. Again, they were all very consistent with their max current. I forget what it was, maybe 136 mA. I could look it up from old notes, if someone with those diodes wants to know. I later got some similar 50 mW diodes, same firm. They WERE highly variable! Unusual. I think they were actually a failed attempt by the maker to make the 80 mW diodes they eventually succeeded in making. I think this because there were other, more expensive diodes at 50 mW from same firm, and because those of the cheaper type that could be overdriven hard were very similar in beam profile and max limits of brightness and drive current to the 80's. So close a match, that it doesn't look like co-incidence. My guess is that's what they aimed for, fell short in so many diodes, that they fell back on the old practise that was used when LD's were newer: sell them underrated so less get complained about. It's the easiest way to recover some of what they spent making them. They can always sell the higher spec when they perfect the process.

Finally, the Rohm RLD65PZB5. These are the ones that will be of most interest. Rated 658 nm, 240 mW pulse 50% DR at 1 MHz or so. I found that in CW they can put out around 180-200 mW at full bore if kept at room temp, or blood heat at highest. I haven't had them long enough to do long term tests, but past experience with other diodes lets me make some predictions based on how they fail. One was poor, mode hopped oddly at strong drive, failed early. The others (7) were consistent, with no obvious mode hops (haven't looked for subtle ones, not important to me). I was getting >200 mW out of them after a non-AR-coated lens rated to let 90% through, so assume around 220 mW raw output. At the start of a week at that current they'd take retroreflection off a Lasercheck's ND filter, but near the end, a flash back off that would pop them, so once I saw that this serious stress pattern was consistent (with most diodes out of the small sampling) I lowered the drive to produce >160 mW after the lens. After ten days, no change is seen since the start, so this is probably in for the long haul. I'm using an old module I made that has some thermistors to do temperature compensation, but it's only partial, designed for the Opnext 80 mW diodes' lower currents. Changes of temp between 4C and 30C didn't kill it. I don't know the exact current yet, it's the output power that is more important right now. (Estimate 290 mA). I think they can safely take more, maybe 310 mA. All this matches closely what others have claimed for them. I think they should do >200 mA out given a good broadband AR coating on an acrylic asphere. I'm in no hurry so I won't be pushing this yet, I'm still happy to watch the current test for a lot longer, and to test some mounting schemes and make designs.

Other issue is that as you add more diodes in series, you increase the forward voltage drop of the chain, this is a problem for modulation, and now the rise and fall time of the power supply is longer, due to a wider voltage swing, and that will slow down modulation.

Why? Most modulation circuits while slower than the diodes they are modulating, are not the limiting factor. Far more important is transient suppression to prevent spikes from causing COD in a tiny fraction of a second. Modulation rise/fall times should be no faster than strictly needed, to avoid compromising the protection too much.

Diodes pop just by staring at them wrong, but higher power diodes tend to be more friendly... It's not common to find optical feedback on higher power diodes.

Nice. It helps to whisper around them too, probably. Agreed though, high power diodes are much more tolerant, and consistent. I've read that diode bars are butch wee beasties so long as you don't zap them with static. I've also read that they can be conencted as a bar, or separated into smaller diodes in some cases. That's interesting because it implies that in one bar, you can run several diodes in parallel without individual current limiters. That would be impossible unless the consistency between each one was extremely good. I guess any high power diode off the same wafer might be consistent enough to allow series chains without agonising over their differences. If one IS different enough to be bothersome, it might be faulty anyway.

2) order the diodes in series so that the more efficient diodes are further down in the series...

Why? If they're in series, the current through all is equal, and apart from their own forward voltage, nothing else can distinguish them from each other electrically. Is it to make an ordered progression for testing visually?

I like the cautious graded approach, but I'd simplify it. Any that pop really early are not widely representative, the spread around a mean is a bit like the bell curve, like that one for showing brightness with wavelengths. Most diodes' limit values are clustered tightly around the central region. When buying one or two, or using unknowns, details are critical, but with more than ten of one type from same batch, it's usually ok to go with declared typical values and not worry about extremes. Just watch the weak end of that small scale.

P.S. You'll only need 1 PBS to conveniently join all 6 optically in the way you were planning... Use a bunch of razor thin bounce mirrors.

Four. Also, they don't have to be razor thin, just front surface leading to a clean edge. One ideally needs a further detail, a corner that is not a right angle, but cut at 30/60 degrees like the corner of a scanner mirror. I worked out that this could make a very tight formation of three beams. A pair of spare scanner mirrors would make an ideal set for each cluster of three, but for cheapness, one for each of two clusters, the other being any plane square front surface mirror. I want to keep some secrets, so I won't spell out the layout of that, but I won't need to, it is possible, and you know what parts go where, roughly, it's just a small puzzle to figure out where exactly. And I've posted entirely too much now anyway.

[steve-o]

Mmmm--
Quite impressive, Doctor.
You sound like you're deep into this stuff..
Are you working for a laser company or starting your own?
What's your take on VCSELs? I've heard that they're more efficient.

Steve

[The_Doctor]

Hoping to go alone. Which means I'm entirely responsible for any failures, including those I fail to see if caused by the manufacturer of the diodes. A hard logic, but it's real, just as they can't expect the customer to blame nature for the maker's failure to manage limits in the elements and compounds they use... So I have to look deep. The only failure risk I can expect to pass on to the customer is death by retroreflection (and absurd deviations from electrical advise and common sense). After all, that was passed on to me. At least I'll be telling the customer if I manage to make and sell stuff, which is more than I got from the makers...

I don't know about VCSEL's, I wish they were more powerful and they'd attract my attention more. I still hope they'll win out though, I think the idea of emission from the surface, not the edge, is a good thing. I'm interested in the NECSEL with its IR and harmonic generation all in one package, but I don't think it's the same thing, exactly. There seem to be more IR wavelengths in lasers than visible, and more tunable ones too, so harmonic generation might remain the main way to get visible. I don't know VCSEL though, so I'll have missed something, probably. What were you hoping for from them?

[steve-o]

Not really hoping to gain anything much.. just thought it was new cutting-edge technology that might be of some use to the "build-em-from-scratch'ers" such as yourself. I really don't know much about the technology except what I've read (these are special apps):

VCSELs with external cavities, knows as VECSELs or semiconductor disk lasers. VECSELs are optically pumped with conventional laser diodes. This arrangement allows a larger area of the device to be pumped and therefore more power can be extracted - as much as 30W.

The wavelength of VCSELs may be tuned, within the gain band of the active region, by adjusting the thickness of the reflector layers.

• Multiple active region devices (aka bipolar cascade VCSELs). Allows for differential quantum efficiency values in excess of 100% through carrier recycling < (I dont know what this means, but it sounds good )

Widely tunable VCSEL with micromechanically (MEMS) movable mirror

[yaddatrance]

Four. Also, they don't have to be razor thin, just front surface leading to a clean edge. One ideally needs a further detail, a corner that is not a right angle, but cut at 30/60 degrees like the corner of a scanner mirror. I worked out that this could make a very tight formation of three beams. A pair of spare scanner mirrors would make an ideal set for each cluster of three, but for cheapness, one for each of two clusters, the other being any plane square front surface mirror. I want to keep some secrets, so I won't spell out the layout of that, but I won't need to, it is possible, and you know what parts go where, roughly, it's just a small puzzle to figure out where exactly. And I've posted entirely too much now anyway.

That would work, there are 4 reds combined in the laser above... I was actually just suggesting thin mirrors for cost, mirrors with custom cuts get quite expensive. A thing to note is that most higher power diodes will have a fast axis so you won't need the cornercut to do a nice tight square when combining a couple using a fine-edged mirrors.

The reason to order them by specifications is just to make sure the actual startup current path and associated ripples have the most headroom. While fast by human standards, there is inherent slew in the start pulse of current through the chain. Not an issue if you're just turning it on once in a while, but if you're modulating it, it's just a free safety margin.

[aijii]

The trick is, is that in the arctos reds, like what yadda has pictured, the pre-mixed beams from the laser diodes are collimated to be quite large - like we are talking at least 1cm accross IIRC.

It is a lot easier to line everything up when you are working in cm, instead of mm or fractions of a mm.

There is a large collimator as you can see, after the polariser cube....

[The_Doctor]

That large beam combinator and subsequent collimator looks to me like a maths expression that while correct, can be reduced further, and should be.

Only linear offsets are helped by larger scales. Any angular change will be multiplied by that final lens, and it's just more expense and loss to consider. It might be worth it to get a very fine beam at aperture at the expense of greater divergence, but I'd not justify it for any other reason. It's easy enough to get high accuracy with small beams, and the parts cost is less. If gathering 3 beams in triangular formation for later combination by PBS, the shape of the mirror and the clean edges are all that matter. If the shape is not right (30&#176;/60&#176; cut corner), it won't get any better by using fat beams, it just gets more expensive. I accept that the special cut might cost too much, which is why I asked about the likely cost of scanner mirrors (as these are already made with that cut). It might be that the beam profile will allow triangular form with simple square mirrors, with enough efficiency. I haven't modelled that much detail though. The ideal is to have one mirror with that cut, as then you know that any round profiled beams will be ideally tesselated, taking all the fudging out of the equation.

EDIT:
Interesting point about that slew through a series chain. Would it be much? The only thing I can think of that could make it significant would be inductance or capacitance. The diodes should themselves be switchable at 1 MHz easily, they're rated for that as it is, 50&#37; duty cycle in the case of the Rohm diodes. It's not very demanding to makea small series circuit able to do this. Incidentally, while I haven't worked it out, I'm still wondering if the switching can be fast enough to allow PWM and TTL to make brightness changes without scanned lines appearing to have dashes in them. With small scan angles, it might be, even with fast scanners. If the PSU and an optoisolator sheild the diode from transients, it might be possible to reduce diode filtering to allow 10 MHz switching. No idea how feasible that is, though. If the slew IS there, it might be helpful, if the scale is small, it could reduce the risk of dashed lines appearing in fast scans of six-diode beams. Deliberate blank periods will be on a lot longer scale, so it won't affect that, hopefully.

[yaddatrance]

The trick is, is that in the arctos reds, like what yadda has pictured, the pre-mixed beams from the laser diodes are collimated to be quite large - like we are talking at least 1cm accross IIRC.

It is a lot easier to line everything up when you are working in cm, instead of mm or fractions of a mm.

There is a large collimator as you can see, after the polariser cube....

Actually that huge collimator is the trick to a better beam profile and low
divergence. The size of the collimator makes alignment easier and isn't
indicative of the beam size...The spotsize out of a tuned arctos is much
smaller than say a laser pointer. The quality of components is what makes it
possible. The beam width out of the red above is 2.1mm with a divergence of
<0.8mrad... (If you're looking for a catch, the only problem is cost)...

Given that it's near impossible to use a fiberfeed with poor spotsize these
photos below should give some indication as to the actual beam profile...





It is certainly possible to reduce the beam width or divergence, but if
you want to improve both, it will require another jump up in optics
complexity.

[allthatwhichis]

OT here,

but I'll need close ups of those collminators, mounts, and fiber cable...

Those are totaly sparate units correct, laser and scanner set up?