Microsecond LED Flash Ringing Issue

[mribble]

A few weeks ago I got some great help on this forum in this thread: https://www.eevblog.com/forum/beginners/4-microsecond-high-power-pulses-through-led/msg1416524/#msg1416524.  A quick refresher is that I designed a test circuit that was supposed to create a microsecond burst of light from an LED.

I fixed several things in my circuit and had a new PCB with things scaled up to where I think it will create enough light for my needs.  I'm now running it at 160V with 4 strings of 6 LEDs that I overdrive (measured 8 amp through each led so total current of the system is 32A).

The problem I'm seeing is that the LED doesn't turn off when I ask it to.  The issue seems to be a ringing problem, but the odd thing is the ringing doesn't decrease over time.  Instead it happens for a long time and then stops.

The basic induced voltage equation is V=L (di/dt).  I estimate my circuit is 20 nH, and measured the rise time and current at 32A in 0.1us.  This gives me an induced voltage of 6.4V.  I measured 6.8V in the ringing circuit so I think that confirms it.  I would like to estimate the resonance frequency in the same way, but I can't get the equations to work out.  I thought it would f=1/(2*pi*sqrt(L*C)).  With a C of 470uF I calculate a frequency of 117Mhz, but I'm only measuring 10Mhz on the scope so that seems wrong to me.

I have also verified that the ringing isn't contained to the critical cap/led/mosfet loop.  If I go all the way back to the pin on the micro that is driving the mosfet driver that pin also shows the ringing.  Maybe this is why the ringing continues since the micro is actually driving the mosfet on/off?

I would also like to simulate the ringing issue in Spice LT.  I have the boost converter and a trigger event simulated.  That works as expected by creating a nice square pulse, but when I added in the 20 uH inductor I still didn't see any ringing like I expected.  Did I add it to the wrong place or something?

What I'd like to do is verify/understand the issue is ringing and then figure out a way I can make this circuit work.  I don't really care if there is some ringing if it's short or doesn't output much light from the LED.  I just want to be able to output around a 1 us burst of light from these LEDs.  Right now this ringing is causing the LED to go on/off for a long time (20-100 us) after the original 4 us where it should be on.

I have 2 guesses on what might help with this.  One is to add a resistor that consumes a few percent of the current across the LEDs.  This would help dampen the ringing.  Another is to add a resistor between the mosfet driver and the mosfet to slow how fast I turn on/off the LEDs.  Are those my best options or is there something else I should consider?

I've attached my schematic (pdf), pcb layout (svg in zip), spice LT file (in zip), and 2 pictures of the voltage across the 1 ohm resistors that limit current through the LEDs (one shows the 4 us pulse followed by ringing and the other is zoomed in on x axis so you can see the sinusoidal shape of the ringing).

[capt bullshot]

My guess:
You've got either a software related problem - check the uC output without the LEDs
or a massive GND bouncing problem that leads to this kind of oscillation. Can you show us the layout?

Edit: I've found the layout file now.
The MOSFET should have a gate series resistor, even if it is as low as 3.3 Ohm. The layout is a bit hard to read, it seems the line from the uC to the gate driver IC makes a long loop around the LED array, and the ground plane appears to be splitted - these might be the the cause for the oscillations

[mribble]

I've attached the gerber files so if you have a gerber viewer that will make it easier to read the layout.

The software is really simple right now so I doubt I have a bug in that, but anything is possible.  I will rule that out soon.

I will also add in a resistor on the mosfet gate as you suggested since that's an easy thing to do on my pcb.  I'm using this mosfet (https://www.mouser.com/ds/2/196/irfs4227pbf-1227657.pdf) which has a gate capacitance of 100nF.  100nF at 12V stores 7u Joules.  To charge that gate capacitance with a 3 ohm resistor would limit it to 4A would take about 0.15 us which is more than quick enough for me.  I'll find a resistor in the 3 ohms or bigger range and try it.

[mribble]

The smallest resistor I had was 10 ohms so I tried that in series with the big mosfet's gate.  That solved my issue.  I also see this didn't really have an impact on the rise/fall time of the mosfet.  The rise time is around 70ns.  This is smaller than I calculated before, but that makes sense since the mosfet's get doesn't need to get to 12V.  It starts opening at 4V and is fully open just after 6V.  I can probably put an even bigger resistor in there if I wanted to get rid of the rest of the ringing.  The remaining ringing is a quickly decaying ring that lasts for about 0.3 us.  I see a tiny amount of ringing on the LED output, but it seems acceptable.

So I can live with what I have now, but anyone has suggestions on how to make the light output a little crisper let me know.  I can experiment with a bigger resistor for the gate on the mosfet.

[StillTrying]

2 pictures of the voltage across the 1 ohm resistors that limit current through the LEDs (one shows the 4 us pulse followed by ringing

Even if your 12V supply is very floating, connecting the scope's ground clip to the switching mosfet's drain can cause strange effects, have you tried just looking at the voltage across the source-drain, much of the ringing might not be real.

anyone has suggestions on how to make the light output a little crisper let me know.

I've produced much crisper <1uS light pulses, but with only 50 - 600mA through a single red LED, considering your much higher volts and Amps and white LEDs, I think your light pulses are quite good, are you sure you're not at the limits of the photodiode.
https://www.eevblog.com/forum/beginners/4-microsecond-high-power-pulses-through-led/msg1414944/#msg1414944

[mribble]

Even if your 12V supply is very floating, connecting the scope's ground clip to the switching mosfet's drain can cause strange effects, have you tried just looking at the voltage across the source-drain, much of the ringing might not be real.

I've produced much crisper <1uS pulses, but with only 50 - 600mA through a single red LED, considering your much higher volts and Amps and white LEDs, I think your light pulses are quite good, are you sure you're not at the limits of the photodiode.
https://www.eevblog.com/forum/beginners/4-microsecond-high-power-pulses-through-led/msg1414944/#msg1414944

I was using a battery for the power supply so it was floating.  I verified the ringing also by checking the LED output so I'm sure it was real.

I have no idea if I'm reaching the limits of the LEDs.  Not sure how to check that.  I think the link you included was to the wrong page since I don't see a post from you there.

Thanks for the suggestions!

[StillTrying]

"I think the link you included was to the wrong page"

It's your 4uS light pulse, with ~250ns rise and fall times I think it's quite good and will be difficult to improve on.

The fastest on and off light switching edges I've been able to produce and view using the common BPX and SHF photodiodes is 76ns, but that was just a single red LED.

[mribble]

Oh, I thought you were trying to link to a project.  My falling edge with this higher power version isn't quite as good as the previous single LED version, but it's close.  So I'll say it's pretty good.

One other thing for people here.  I tried increasing the voltage from 160 to 200V and it didn't affect brightness in my test rig.  Do you think I'm just reaching the max brightness from these LEDs (it's acceptable, but I'd be happier with more), or do you think some other component like the cap would be limiting me?

[StillTrying]

"I tried increasing the voltage from 160 to 200V and it didn't affect brightness in my test rig.  Do you think I'm just reaching the max brightness from these LEDs"

What is your test rig. You'll have to devise an experiment somehow to see if increasing the current is worth it in increased light output. Perhaps you could photograph the flashes reflected off a white paper with a camera on a fixed manual exposure.

For the single red LED I used the 1 transistor circuit there: (the opamp is not needed unless it's driving a cable)
https://www.eevblog.com/forum/chat/20w-halogen-bulb-viewed-by-a-photodiode/msg1317933/#msg1317933

Its output is linear up to about 2.5Vpp, so I fixed the single LED and photo diode together in a small tube so they couldn't move. The ~10mVpp output at rated LED current gave me >100X the light headroom to measure the extra light, I seem to remember around 30X the current gave around 25X the light as expected.

"do you think some other component like the cap would be limiting me?"

Dunno, is the cap a good one! If the current holds up the voltage must be holding up. In theory, after the 4us 32A pulse the voltage on your cap should only be down by ~0.3V, so I think it's 470uF is more than enough.
I'd have tended to use more than one cap of different values to supply 32A pulses.

You've probably already see this one.
https://petermobbs.wordpress.com/2015/02/06/experiments-with-led-based-flash-gun-for-high-speed-photography/

[mribble]

My test rig is exactly what you suggested.

I'm going to try some other caps.  Currently I'm using this one: https://www.mouser.com/ProductDetail/EPCOS-TDK/B43501C2477M?qs=sGAEpiMZZMvwFf0viD3Y3XPvHoPNDFjbo3YJj9bNF5E%3d

I can't actually find any low ESR small caps on mouser or digikey.  There are a few crazy expensive ones, but I'm looking to keep my total cap cost under $5.  Do you have any suggestions?  Considering I could probably get buy with just a few 10 uF caps is there something I should be looking for instead of electrolytic?

[StillTrying]

The 470uF's specs look good to me, for 32A I'd probably use something like 220u + 100u + a 0.1 to 0.47 poly for good luck.

With the scope probe on X10 it should be quite safe the watch the 160V between the mosfet source and top of the LEDs to see how much it drops during the pulse time.

[David Hess]

It's your 4uS light pulse, with ~250ns rise and fall times I think it's quite good and will be difficult to improve on.

The fastest on and off light switching edges I've been able to produce and view using the common BPX and SHF photodiodes is 76ns, but that was just a single red LED.

It is easy to get much faster and cleaner than this with a different topology; switch the MOSFET at its source instead of its gate using a two transistor cascode configuration.

Another interesting way which is more suitable for shorter pulses is to dump a charge line into the LEDs.  This could be done with a discrete charge line for longer pulse duration.

[mribble]

David, I tried googling for the terms you used like "switch mosfet at its source", "two transistor cascode", and "charge line circuit".  I learned about some things that didn't seem related to what I'm doing like cascode amplifiers, but nothing that seems to apply to my case.  Could you explain your suggestions some more?

[David Hess]

David, I tried googling for the terms you used like "switch mosfet at its source", "two transistor cascode", and "charge line circuit".  I learned about some things that didn't seem related to what I'm doing like cascode amplifiers, but nothing that seems to apply to my case.  Could you explain your suggestions some more?

This PDF shows the general idea but any two transistor types can be used.  In high performance applications, it used to be common to use a fast low voltage MOSFET or low voltage bipolar transistor to switch the emitter of a slow high voltage bipolar transistor yielding a fast high voltage device but this is less common now although some parts are specifically designed for it.  Any two transistor types can be used and in your case, a pair of MOSFETs are suitable.  GaN depletion mode power MOSFETs are often used this way with the GaN part being used for the upper transistor.

The advantage is that the upper MOSFET isolates the swing in drain voltage from the lower MOSFET so less charge is transferred through the lower MOSFET's reverse transfer or gate-to-drain capacitance.  The lower MOSFET can now also be a smaller lower voltage device for lower capacitance anyway.

This type of circuit is not only faster but has less coupling between the control signal and the output so the output waveform can be cleaner.  In the past it was often used to generate fast clean edges into a transmission line for calibration purposes.

Differential amplifiers can be seen as a twisted form of this with one emitter/source driving the other emitter/source.  It can also be used to make a switch using a pair of diodes but this has no amplification.

The charge line idea is suitable for high power short duration pulses.  Instead of discharging a capacitor into the LED string which would result in an exponential decay of current, a transmission line replaces the capacitor so it "empties" for a duration of 2 times the transmission line length producing a constant current through a ballast resistor for a short duration.  The switch only has to turn on, usually regeneratively like with an avalanche transistor or SCR, and then the pulse ends once the charge line is depleted.

[mribble]

Thanks David.  That helped.  I'll consider these, but my current circuit is working fairly well now.  I suspect the quality of mosfets today reduces the need for some of this.  I do want to support longer pulses at lower voltages so the charge line isn't ideal.

[StillTrying]

Are you producing enough light to take a short range flash photo?

[mribble]

Yes.  I've done testing and it's enough light for at least a 6x4 inch area at iso 800 F11 with a 2 microsecond burst.  I'm hoping to get that better by optimizing the circuit and optics. 

If you're interested in this, I just posted a video about my current circuit and how I'm starting to work on the optics. 

[StillTrying]

Looks good.
For such a LED panel, rather than a lens, I've often thought a medium length horn type reflector built from glossy white card would work best at close range to the subject. Something vaguely like this.

[Twoflower]

Nice!

How about placing the LEDs horizontal shining on a 45° metallic reflector. The reflector should have a rough surface to scatter the light a bit. Such surfaces are often used for film/photographic reflectors.

[mribble]

stilltrying, that won't work very well.  I have tried those in the past and the corners cause uneven light distribution.  I also tried printing a parabolic reflector and with my 3d printer and covered it with reflective tape.  This didn't have the problem with hot spots, but it didn't really affect the intensity of the light through the lens.

Twoflower, that is a reasonable idea, but I think it will add cost/complexity.  I'm trying to keep it simple.

Going forward I'm going to focus on the lens.