4 microsecond high power pulses through LED

[mikeselectricstuff]

Bear in mind different manufacturers may well use different phosphors - even shining a UV LED light at a random bunch of white LEDs produces significantly different amounts and colours of fluorescence, so may be worth trying some different types.
RGB might be interesting but combining all colours homgeneously might be tricky. Sequencing the colours might make for some interesting shots as you will be able to tell what order things happen in. RGB is certainly nowhere near as efficient than white, and the spectrum will be spikier.

I wonder if there might be some mileage in blue laser + phosphor.
Maybe even blue laser remotely, firing at a local phosphor so you could get it really close for macro shots?
Lasers will probably be less tolerant of overdriving though due to optical damage. 

[mribble]

Mike, I've thought about RGB LEDs and even just separate strips of them.   While like you I initially thought the color spectrum would be worse, I've read it's actually better.  I'm sure if you had a certain type of object it would be worse, but most photographers say they are generally better.  Another interesting thing in this area are that high end cell phones actually use 2 different white leds (one cooler and one warmer) and can control the temperature of the flash.  Anyways, with all those cool options I still think for this initial version I'm just going with the fast simple option.

I did the photodiode measurement tests and I do see some ringing at this new 30V level.  It's not as bad as it was on the current limiting resistor, but it is certainly there.  This was with no probes attached so now I know there is some ringing all the time.  At this point I think I'll just try a better pcb layout and see if fixes the ringing in the led.  See attached photos of the light output.

[mribble]

I think I found most of the source of my ringing.  I was using a wire 5W 1 ohm wire wound resistor to limit LED current for these pluses.  When I removed that the small amount of ringing on the light output is gone.  I didn't have a good place to measure voltage directly in the circuit anymore so I don't know if a little remained (I'm guessing it did and am working on a better PCB layout).

Anyways I plan to go to a metal oxide 3W resistor since those are supposed to have much less inductance.  Does anyone have a suggestion for a better approach?

Also is there a rule of thumb about how much you can overdrive a resistor if it's just for a few microseconds?

[alsetalokin4017]

I was going to mention that your 5W resistor was likely wirewound and thus had significant inductance which would contribute to the ringing.
Why not try an actual non-inductive resistor here? There are several types you might try. I like the Ohmite precision Ayrton-Perry types:
https://www.digikey.com/catalog/en/partgroup/wn-series/6251

[mribble]

Thanks for the link.  I'll look into these and other non-inductive resistors.  The issue I see with those is they are expensive and I need 4 per unit.  On my next version I'll test the light output with these and a cheaper metal oxide resistor to see if they make a difference.

I still am wondering if I could use something like a 2W resistor for 20V at 5A if it is just on for 4 microseconds.

[kony]

Carbon composition resistors handle peak transients very well (used commonly in input protection due to that property), you don't really care that much about tad higher noise and worse accuracy/tempco.

[T3sl4co1l]

Carbon comp is an anachronism.

Carbon, metal and metal oxide film types are all excellent.  Wirewounds have more inductance, unless otherwise stated.

Regardless of type, expect on the order of 1-2 nH per mm of component length.  This works for wires just as well as resistors, capacitors and so on!

Ringing can also be dampened by adding an R+C across the switch, C >= 2 * Coss and R = sqrt(L/C), where Coss is the switch output capacitance (MOSFET Coss parameter; note that this is voltage dependent, so pick a voltage typical of the ringing), and L is the stray inductance (if you don't have a number, estimate it from the ringing and Coss).  The resistor doesn't need to be big here, since it's only consuming energy peaks, not continuous duty.

Tim

[mikeselectricstuff]

Multiple surface mount resistors in parallel would probably give the lowest inductance

[mribble]

Carbon comp is an anachronism.

Carbon, metal and metal oxide film types are all excellent.  Wirewounds have more inductance, unless otherwise stated.

Regardless of type, expect on the order of 1-2 nH per mm of component length.  This works for wires just as well as resistors, capacitors and so on!

Ringing can also be dampened by adding an R+C across the switch, C >= 2 * Coss and R = sqrt(L/C), where Coss is the switch output capacitance (MOSFET Coss parameter; note that this is voltage dependent, so pick a voltage typical of the ringing), and L is the stray inductance (if you don't have a number, estimate it from the ringing and Coss).  The resistor doesn't need to be big here, since it's only consuming energy peaks, not continuous duty.

Thanks for the info.  A little ringing doesn't matter to me and it should be greatly reduced with my new pcb and when I stop using wire wound resistors so I think I'll try skipping the RC across the switch.

That said, you mention resistors don't need to be big since it's just for peak.  Are there any rules of thumb.  Let's say I have 5A and 20V for 5 us.  Do I need a 1/2 watt, 1W, 2W, or 5W resistor? 

I'm happy to use surface mount resistors, but again I need to have an idea of what wattage I need to handle for these pluses.

[T3sl4co1l]

Thanks for the info.  A little ringing doesn't matter to me and it should be greatly reduced with my new pcb and when I stop using wire wound resistors so I think I'll try skipping the RC across the switch.

That said, you mention resistors don't need to be big since it's just for peak.  Are there any rules of thumb.  Let's say I have 5A and 20V for 5 us.  Do I need a 1/2 watt, 1W, 2W, or 5W resistor? 

I'm happy to use surface mount resistors, but again I need to have an idea of what wattage I need to handle for these pluses.

What's pulse rep rate / duty cycle?

5A * 20V * 5us = 500uJ, about a flea fart.  An SMT chip resistor might poof, but I don't think the average 1/2W resistor would even mind.

Strictly speaking, you need to shop for pulse-rated resistors, and find where on their power(time) curve 5us and 100W lies.

Informally, resistors usually go about as t^1/3 below the dominant (thermal) time constant.  1/2 exponent would make more sense (thermal diffusion), or proportional (fixed energy), but, eh, for whatever reason, that seems to be the most common pattern.  Most power resistors are rated for a 5x or 10x overload for 5 seconds.  Combining these two properties, we might expect a 6-decade faster pulse to have 2-decade more power.  So a 100W pulse could be handled by a 1W resistor (no overload condition needed) or a 0.2W resistor (5x overload 5s).  Resistors quite this small, I would worry that their time constants will be much shorter, so erring on the side of 1W would be better.

This assumes a pulse rep rate much longer than the longest thermal time constant, i.e., so that the resistor temperature returns to ambient before the next hit.

For a one-shot flash, I'm guessing this will be sufficient?  If you're instead doing (or "also considering", since it's trivial once you've built the hardware) a stroboscope, then the temp may not return to zero between pulses, and you have to worry about peak loading on top of the baseline temperature.

It seems likely that, for resistors over 1W, you'll have no problem with peak loading, so you can use average power just as well.  For example, 2W resistors would get you a duty cycle up to 2% (a pulse rep rate of 5kHz, and more if you reduce the pulse width / intensity).

Tim

[tggzzz]

That said, you mention resistors don't need to be big since it's just for peak.  Are there any rules of thumb.  Let's say I have 5A and 20V for 5 us.  Do I need a 1/2 watt, 1W, 2W, or 5W resistor? 

I'm happy to use surface mount resistors, but again I need to have an idea of what wattage I need to handle for these pluses.

Read the data sheet, looking for pulse power. If not listed, find a data sheet that does list it!

[mribble]

You guys are awesome!  I tend to sort resistors by lowest costs so I guess it isn't surprising the ones I've looked at don't list pulse ratings.  I did a little looking at some more name brand SMD resistors (like ones from Bourns) and they have pulse ratings.  There are cheap 2 watt resistors from Bourns that handle 600W for 100uS.  They are thick film and have much shorter leads so should be low inductance from what I've read.  I think I'm set.

Tim, since you asked my duty cycles are low.  The worst case will be 5 us on then off for 100us (repeat 5 times so total on time is 25 us), then off for  second.  Most of the time it will just be a single 1 us pulse and then off for minutes.

[mribble]

Thanks for all the great feedback.  I've incorporated all the great suggestions into a new pcb.  I'm attaching the schematic, board files (zip because svg files aren't allowed), and the gerber files.  If you have any more feedback based on this info let me know.