Schottky diode

Interesting. The turn-on delay in a PN diode arises because the E field is initially shielded out by free carriers between the contact and the junction. Those aren't normally present in a Schottky, which is part of the point--instantaneous turn-on. I use 1N5819s in prototype SMPSes all the time, and haven't noticed any voltage overshoot.

How long does the overshoot last? How hard are you hitting the diode?

Is it possible that you got counterfeit 1N5819s?

Cheers

Phil Hobbs

The overshoot lasts ~1/2 of 125MHz period ringing (it is not measurement artifact, believe me). I have a picture of this thing swinging 14V avove rail (ouch!) for ~3ns. The peak current is, say, 3-5A. Counterfeit parts?? This is something I haven't though about. The world we live in... Would anybody pull these tricks with (very cheap) 1N5819? What is the chance of getting counterfeit parts from reputable supplier (Digi-Key)?

On a sunny day (Mon, 23 Nov 2015 07:11:22 -0800 (PST)) it happened Michael wrote in :

Can you do a differential measurement across the diodes (ch1 - ch2) If the pulse is not there, then it is likely PCB. Yes, fake parts are everywhere.

Oh. That's pretty well outside the range of usual uses for a 1N5819!

Well, even with a sinusoid, getting 14V from a 4A current at 125 MHz needs only

L = Vpeak/(dI/dt) = 14/(2*pi*4A*125e6) = 4 nH.

That's an easily believable value for a 1N5819--it's equivalent to about

Well, it doesn't look like they need to be counterfeit to explain the effect.

Cheers

Phil Hobbs

Agree. Somewhat. dI/dt=4A/4ns=10^9 A/s; L=V/(dI/dt)~14nH - ~20mm of 1mm diameter wire. The layout is much better than that. Another argument: MBRX140-TP in identical package does not misbehave. Regards Michael

OK I first need to warn you that most of my ideas are a bit crazy. (So take them with a grain of salt.)

But I would guess a turn on time of some diode capacitance, times whatever the effective resistance is. Perhaps one has less C and or resistance.

George H.

I populated (working) prototype with "counterfeit suspected" parts (1N58190). The ringing (there is always some) is barely distinguishable (only with glasses on) from the one with MBRX140-TP. :o(

Unless I'm misunderstanding, the overshoot lasts half a cycle, so it has to go up and back down in that 4ns. Thus I think you're optimistic by a factor of 2, and 7 nH is an easily believable figure for a 1A diode. (You're really abusing that 1N5819--its abs max rating is 1.5A for repetitive pulses of Regards

Yeah, it's never the cheap simple thing that went wrong.

Cheers

Phil Hobbs

The diodes are similar, and even cost the same. Just replacing them won't answer "why" question. I firmly believe in this rule: If circuit (it) failed once, it will fail again. If it's never failed before it may still fail in the future. :o( Regards Michael

There was an article by Jim Williams where he was looking at (I think) the reverse recovery time of diodes. There was a big variation even (I think) in diodes with the same part number from the same vendor.

George H.

Just out of curiosity, what's the circuit for?

Cheers

Phil Hobbs

TEC driver. Looks like I was pushing 1N5819 too hard, 3A rated parts seem to work - running test as I am typing. I wonder what is the failure mechanism... not overheating for sure (duty cycle is ~1/1000) Regards Michael

Cheers

Phil Hobbs

capacitance?

Maybe a TEC is not a thermo electric cooler? (Why would you want to run it anywhere near that fast??? I mean how much current is flowing through the capacitance of the TEC... I guess at least that doesn't cause any heat..) Weird, George H.