It's the capacitance, stupid

I got cinnamon buns and Fermat didn't.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Thanks for explaining the diode, I misunderstood it before!

If Q1 is normally saturated on then fighting storage time is an issue, I wonder if he used a mmbt2369 kind of device?

piglet

Sloman has a point. Why design things, when you can look them up in a journal?

You could do something simple like

The D1 capacitance is probably all you need to do the C1 function. R2 can be pretty low. Vh is a clamp, not a source.

This has a linear recharge, which will be ready for another photon a lot faster than exponential.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Ahh, well I do look in journals. Lots are behind paywalls, some are not so good, and then others are too complicated for me. And heck I can come here and sometimes tease ideas out of you all. :^)

Scratch, scratch... Is this a (semi) active quench? V+ is my bias, Vh is below V+(?) so the diode is on mostly.

And why C1?

Linear? I'm missing something big time. V+/ R1 is the current source? and Vh is the bias point of my spad? OK that makes sense. So this is a current source reset, but there is no active quenching. (Is that right?) I guess I could reset with a current source, that would speed it up a bit... (not as much as I'd like) Let's see, a max current of 50 uA (or the spad won't quench) Say 10 V over voltage, what's the C of a diode? 2-3 pF? So I need ~50 pC / 50 uA... I'm still stuck near 1 us.

I will try adding 1pF to ground from the high side of the spad on my circuit.

That should give bigger voltages at not much other cost. (maybe I can skip the comparator and go right inot some logic circuit.)

Thanks for the ideas! George H.

I asked Mario about Q1... his reply. "That is my trade secret." :^) He did say it's now obsolete, but that he has a lifetime supply.

Thanks for the mmbt2369 hint... I may need something like that.

George H.

I think that circuit may have a latchup mode. It works as long as it works, but if it misses once it's latched.

Things that have latch states usually manage to find them.

Designing stuff is more fun than copying circuits from musty old journals. And most of the journal circuits are dreadful.

+V and R1 are basically a current source, well above the Vh clamp voltage. It's self quenching.

C1 stores the charge that gets dumped into R2. R2 should be small if you want speed. As noted, the capacitance of C1, maybe a schottly with a couple of pF, may be enough. Without the equivalent of C1, and a big R1, and by implication a big R2, it will be slow.

I like using dirt-cheap LVDS line receivers as fast comparators.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I thought (possibly wrongly) that latching was intentional and DLY1 is what ends it?

piglet

There are two delay boxes in the circuit. I think that's to make sure everything is 'clear' before the comparator is turned on again.

Oh a rhetorical question.

So the added 1 pF to ground.. that gave me volt level pulses.

Say how about if I recharge through a long string of diodes, and when not recharging they will all be reverse biased and only add ~1pF. Hmm I guess I'd have to wait for each diode to turn on. Any idea how long that is? (Bat41 says extremely fast, but no numbers)

George H.

Small signal diodes like that turn typically turn on in hundreds of picoseconds, much less than nanosecond. Parasitics can be more limiting.

The diodes chapter in Bob Pease's Analog Troubleshooting book also warns of bizarre overshoots.

piglet

A lot of published circuits are inept, but they do tell you what more or less works, and what frequently goes wrong. Sturgeons Law was that 90% of everything is rubbish, but you can learn quite a bit from less than optimal solutions to actual problems.

Who doesn't. But 685 style comparators have more gain, and the logic input that can latch up the input stage within a nanosecond or so can be useful.

SPADs aren't cheap, and spending a bit more on a proper comparator is probably going to be a sensible investment.

--
Bill Sloman, Sydney

It's an async hairball. Things run around in circles and if everything does what you expect it to do, it's OK.

But the latchup state is spad on and current flowing to the left, through Rs etc to +14. That's a static state if the spad doesn't self-quench. The SAP500 is not well specified.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Huh, OK... I sent an email to Mario and posted a link to this thread previously, (when I asked about Q1 for piglet) I'm not sure he'll look, but that's his problem. If I built the thing I was going to just use a higher voltage source.

So what do you think about a diode string?

HV---+-----------+ R some switch(pnp) | | +-|

Thanks piglet, I just wanted to add that, I'm very glad you are following this thread. I'm not very good at transistor level design, and if I start to get near a working circuit, and you think.. "humm maybe a little more C here, or R there." I trust you'll chime in, and offer suggestions.

I guess that goes for any other lurkers too. :^)

George H.

Not a static latchup as long as the schottky blocks reverse bias. With spad on and current flowing to left and schottky diode blocking then the comparator + input sits unchanging at 0.51V (1.8k,330) while other input resets it after DLY1. If spad does not quench but continues conducting then the comparator might oscillate but that is not a static state :)

Circuit does seem critically dependent on Vf of the schottky diode, hence the preset R1?

The schottky diode is rated at 60V reverse but if breaks down and drives the comparator + input above the high tide level (R1+1k+330+1500) then a static latch state exists but HV is not specified and who knows where the 1N6263 really breaks down?

The page after the schematic describes the circuit function in detail.

But yes, I agree, it is a hairball!

piglet

Thats very kind of you George. This spad stuff is beyond my present knowledge so I am floundering - not waving but drowning.

piglet

Hmm? What is nice about Mario's active quench is only side of the spad is a low-C sensitive node. Yours and JL's have both spad terminals cap sensitive. Not sure if that is actually an issue in practice - do you have the comparator and spad all close together on one small PCB?

Mario's de-sensitizing anti-twilight period DLY1 sounds like a good thing too.

I think I was wrong - Q1 must be faster than a garden variety 2N2369 - he may have some RF device in store - one of those natty BFT BFQ BLR things Phil Hobbs and John Larkin write about sometimes?

I was thinking about using a high voltage avalanche transistor transmission line pulse generator to generate a fast quench pulse but to get the transistor to avalanche requires a voltage much higher than the spad needs for quenching so would need attenuating. Getting complex.

Another idea: the fastest switching devices tend to live inside logic chips but have swings at most 5V whereas the spad needs 20 -25V quench step. So what about a 4:1 transformer - could be wideband transmission line to keep risetimes short?

Just adding confusion and noise to the thread!

piglet

It's really pretty simple, conceptually like a spark gap that's triggered by a flash of light, but the flash is single photons. The issue here, as with lots of other electro-optical components, is how poorly the parts are specified for electrical behavior.

I don't know how the quench mechanism behaves electrically, and you can't tell from the data sheet. Or from the academic paper referenced here.

We have a similar issue right now, with some gain-switched diode lasers. The idea is that you blast them with a lot of current (several amps in roughly 1 ns) while they are innocently asleep. They wake up and grumble and output one gigantic optical pulse maybe 100 ps wide. Then you back-bias them hard before they remember that they were supposed to be CW lasers. Problem is, the laser data sheet doesn't mention this mode. Hell, they don't even specify capacitance. Or thermal properties. I suspect this behavior is more accidental than planned.

So we'll get a bunch and test them and probably blow a couple of them up. Luckily, we are working with a brilliant laser physicist who also enjoys blowing things up.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

If the quench is AC coupled, and it ever misses even once, and not quenching fries the spad, things could get expensive.

Async coupled hairballs can be hard to analyze. The designer reasons through how it works but doesn't consider the many ways it might not work.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Sure... it's a solder hairball, but I'll post a pic. (1/2 the bits of the circuit are not hooked up.. at the moment.) (Snap, I'm having severe camera issues... maybe later.)

Right I think I'll need some sort of blanking time to turn off the comparator during the reset.. else the reset will trigger it and I'll have made another oscillator. :^)

OK, Mario does show some snapping looking pulses, but having things slower to start with is fine by me.

Hmm I don't think I need all that much speed. Heck to start with I'd be happy to get it reset in 50 or 100 ns.

As a final note, this whole thing has to be put away for a while. I'm going down to Baltimore for a workshop/ conference thing. I need to prepare for that and then I'll be away most of next week.

A pain, but we've got to sell stuff, if I'm going to get paid for playing with designs.

George H.