It's the capacitance, stupid

ote:

te:

burned.

number of papers from his group. There are plenty of other publications.

ed the input-latch feature of that style of (fast) comparator.

r wheel.

It's an interesting idea. The quenching pulse is fairly narrow, and a 4:1 t ransmission-line transformer would have a high side impedance of 200R (mayb e even 14x75R or 300R) which is might be usefully high.

You can do better.

Nexperia (use3d to be Philips, then NXP) has a transistor that might work a s direct quencher.

Newark has 476 in stock at $0.44 each in small quantities. It's good for 40 V, so you'd have to cascode a couple to get to the SPAD operating voltage.

The datasheet is a bit vague on the device capacitances - you only the "fee dback capacitance" which is low at 0.3pF.

The Spice model

.MODEL BF840 NPN(IS=4.948E-16 ISE=1.08E-16 ISC=2.118E-16 XTI=3 BF =105.8 BR=2.665 IKF=0.08 IKR=0.05 XTB=1.5 VAF=125.5 VAR=5.34 VJE=0.5956 VJC=0.3305 RE=0.447 RC=1.9 RB=19.38 RBM=2.549 IRB=

is a bit more informative.

A string of three or four in cascode might look like less capacitance than a single junction, if you were lucky.

There is some critical current (probably a little different for each part.) When you get above that current it can take longer to self quench.. you'll see an initial pulse and then other smaller pulses... very much like a ~20V zener near the knee. I haven't pushed it much beyond that point.

Sounds fun! (I wish I had more spad's that I could blow up.)

George H.

Probably not. But an avalanche depends on having enough of an electric field across the gap to accelerate the charge carriers to an energy where they can generate more charge carriers (both positve and negative) as they cross the gap.

Drop the voltage across the gap below the level that generates extra charge carriers, and the avalanche turns off.

Win and I have both posted links to papers that spell it out (though not all that recently). The gap is usually tiny and transit times of the order of picoseconds.

The catch is the the sites inside the gap that got bashed hard enough to generate pairs of free charge carriers stay excited for a few nanoseconds and can generate after-pulses (which is talked about in the paper to which George Herrold prosted a link).

It's the famous "sweeping out stored charge" which you've run into in step recovery diodes.

Step recovery diodes have doping profile that sharpens up the trailing edge of the cloud of stored charge as it get swept out.

You ought to have remember this - you've played with step-recovery diodes.

Being willing to risk blowing things up is a virtue, if you can afford to lose parts. Enjoying blowing things up is a trifle psychopathic.

Fun.

If you get lasers with epi layers on the facets (such as VCSELs), the facets are a lot more durable. In the first few atomic layers at the facet, the electrons feel less restoring force than in the bulk, so the bandgap is effectively less. If the facet is made of the same stuff as the active region, it becomes absorbing at the laser wavelength, and so heats up a lot when the output power is large. A higher-bandgap epi mostly fixes the problem. Facet damage can happen in a few nanoseconds, but with a 100-ps pulse you can probably hit it harder without huge problems.

This time last year I was purposely blowing up the diode lasers in Uber's Fuji lidars. They didn't die in vain though--their sad demise demonstrated that Uber was not infringing Waymo's patent, even under the doctrine of equivalents. ;)

Waymo's contention was that a resistor was equivalent to a diode in the circuit under discussion, so I tested this by swapping Fuji's resistor with Waymo's diode, which sure did make the laser brighter while it and the drive FET lasted (a few seconds). Whee!

Cheers

Phil Hobbs

BF840 might do.

Avalanche transistors aren't closely specified in avalanche. Using them is tricky, and the best parts seem to come from obsolete fabs in Russia, which will wear out eventually.

Mario's circuit stretches the quench well past the anvalanche killing period to prevent after-pulsing. The avalanche transit time is picoseconds, so the inevitable nanosecond of overkill makes a "miss" very improbable.

So here's some 'scope shots. I decreased the quench R to about 100k, chan. 2 (top trace) is the high side spadvoltage (capacitively coupled through 2.2pF so gain is not calibrated) Chan 1. (bottom trace) is the signal from sense resistor. the first is normal,

Then with some longer quenches

And then more voltage... sometimes it takes longer than the 'scope screen time to quench.

George H.

Although not shown in Marios diagram I was expecting the HV DC supply to be current limited at below the self sustaining level (50uA ??) so that eventually once the 100nF bypass dries out the spad will self quench if it somehow fails to pulse quench. Should only be a few tens of microseconds? Short enough not to fry?

piglet

I don't think he has any current limit on the HV. It's the first thing I asked about when I wrote to him. (maybe a month or two ago.) I was describing my current limit circuit and wondering if others did something like that. He seemed to have no idea what I was talking about. But maybe he was just playing dumb too... I don't know.

George H. Oh here's the spec sheet for the sapd.

(if the link doesn't work search for "SAP500 photodiode")

it can take 200 uA CW... I set my current limit at 100 uA. at 50 uA it doesn't self quench.. but will happily break down and not destroy itself.

Looks like making a self-quenching fast spad detector with HV current source of 30-40uA with 1-2pF stray capacitance is non-trivial. Sticking with active quench ideas I had two ideas:

Updates Mario's scheme but replaces his boutique Q1 with a logic line driver and tapped coil.

Looks very elegant but throws away sensitivity, have fun kicking holes in it :)

piglet

Thanks John, I guess quenching a spad has similarities to quenching a xenon arc as 1960s/70s "thyristor" photo flash guns did. Except dealing with tens of nanoseconds instead of tens of microseconds so can't use an SCR :(

piglet

May need kilovolts on +V to get fast recharge even tho' only 200/300V on Vh ?

piglet

Or a current source. Or compromise a bit and get not exactly linear recharge.

A series inductor would help, too.

Oh boy, Thanks piglet. I've copied those into a note book and will stare at them on the road today. (add some timing diagrams.) I never think of using inductors/ transformers... (the third rail of RCL's for me. :^)

I don't know the 74abt541. It looks snappy. Silly question, why does it list the OH (output high?) current as -32 mA (negative current) and OL as positive?

George H.

I guess simply because the current is flowing out of the chip instead of into and they picked that convention?

Faster buffers are around but that one works on 5V and I am still fond of 5V :)

piglet