What can kill an SCR...

I just got a unit (a netMCA spinoff) which was basically short circuit at its 12V power input, 0.1 to 0.2 Ohm. The third part I unsoldered turned out to be it. An SCR, a C106. Its job is to short the input power (past the fuse) if any of the other 4 similar protections for the lower voltages (5, 3.3, 2.5 and 1.5) got triggered. Once I replaced the shorted SCR everything worked, no other victims No artifacts on the lower voltages, these never trigger except during development when I short something with the probe or something (I am quite impatient so they have come to rescue more than once).

Any idea what could have caused this SCR to get short? After nearly

5 years of service that one. I have never had this and we have netMCA-s (standard, this one is repurposed somewhat but not in a way relevant to what is discussed) and they typically run 24/7 for years, the first ones for over a decade now.

I can't think of another reason than someone tampering with the unit (I had two units sabotaged at this customer some 2 years ago, but it was mechanical, someone had put some aspirin or something on the heaters (260C) and they got corroded - with the remnants of the stuff on them for me to find). So foul play is my first thought. But could it be something else? Part defect showing after 5 years...? I am just wondering, whatever the conclusion my course of action will remain the same (give them the repaired unit, that is; anything else could be reputation damaging).

------------------------------------------------------ Dimiter Popoff, TGI

formatting link

Reply to
Dimiter_Popoff
Loading thread data ...

The usual reason is inadequate gate drive, which reduces the dI/dt capability.

When an SCR fires, the conducting region starts at the gate contact and spreads out across the die. In a crowbar application, there's a race between the expansion of the conducting area and the inductive rising current waveform: Will the SCR have enough conducting area that it doesn't melt?

The often-seen zener + resistor + SCR crowbar circuit is a disaster for this--the SCR cannibalizes its own gate drive as it turns on.

You really want nice beefy gate drive with a sharp trigger edge to get the most out of your crowbar SCR.

Back when linear supplies were more common, there used to be dedicated crowbar controllers for this job, such as the MC3423,

formatting link
. There's a discussion of this problem on pages 5 and 6.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

The gate drive of the dead SCR is not via a zener,it is a 2907, emitter to +13V, collector to the SCR gate via some 100 Ohm; OK, instead of torturing you and myself with explanations here it is:

formatting link
The dead SCR is SR205; I cannot say if any of the rest have been triggered at all, no reason for that I can think of for this to happen on its own. I looked at the voltages, clean. The delayed/soft start - as it has to be. No signs on the case of the C106 from being fried (hence it was my third choice to unsolder, a .... zener that is not a zener, one of these overvoltage things came first, then an electrolythic cap (pathetic attempt, it was visibly good). Then the 12V the unit gets are not so powerful ( a stepdown from 18V in a separate case with other stuff), I really have no idea what may have happened. Even including tampering, what do you do to cause that?! I would expect the stepdown to die, not the shorting SCR... Then except for a fuse there is a current limiting coin thingie, "recoverable fuse" or whatever they call them, it's been a while since I did it, its resistance becomes pretty huge above a current threshold.

Reply to
Dimiter_Popoff

The 2907 is powered off the same supply that the SCR is trying to crowbar, with no local reservoir cap. That's not the ideal, but since the rail is 12V, it might well work fine. Provided nothing steals the

2907's base drive, the 12V rail has to come down to 2-3 volts before it'll turn off. There won't be much energy left in the caps at that point.

Your average switching regulator has cycle-by-cycle current limiting, so it won't fight the SCR very hard. If the SCR died the way that I suspect, it's the reservoir caps that are the killers.

Probably a polyswitch. Those typically take a few hundred milliseconds to switch even with a big overload.

Of course, it could just be a bad SCR, or else the dI/dt or I**2t capacity could be marginal.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

Of course it does work, hey. I know what I am doing, you know :). The +12 has at least 1000uF (on the board where the SCR is, which consumes about 0.5A off the 12V).

My average stepdown is very similar to these you see on the posted circuit. I rarely use off the shelf stepdowns, they invariably have had stability surprises for me etc. so I prefer to have things under control so the first board iteration has things work as I expect. Not much current limiting though the MOSFET resistance provides some; the stepdown in question had not died, not sure it was the one powering the board when the SCR was killed... If they have toyed with it - I know they have opened the case with the stepdown for no valid reason (from my point of view). Who knows what voltage the+12 input has seen.

Oddly I still can't remember what it was called. Though I know at which shop to order them.... (it is a local one, comet.bg) Most likely exactly what you say, I think something like that starts to come out of the fog... :)

Can't imagine the SCR would become bad after nearly 5 years of service (that unit was delivered 2018 and it has had its online support on most of the time so I know it has been in use a lot). Could not the SCR be triggered by dV/dt? (Not likely with >1000uF on the rail of course). Most likely they have powered the unit off something else, probably at much higher voltage than 12. I can imagine some of the 2907-s turning on because of that - and the power being applied just long enough to kill the SCR by overcurrent without the case showing signs of it.. (that polyswitch thingie?) (??? - not an easy postmortem anyway :).

Reply to
Dimiter_Popoff

Your customers are obviously very creative folks. ;)

The polyswitch could easily protect the box and still allow the SCR to turn to lava.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

Hah, some of them are indeed. Most are normal folks just using their netMCA-s to measure spectra; but I have had so far 3 customers trying to clone it (mission impossible *for me* without documentation). These do use their units to measure with but apparently they are trying to be the fourth ones...

This seems to be the likeliest scenario, thanks for the insight, I do appreciate it.

Reply to
Dimiter_Popoff

The SCR collapses the supply that the gate drive comes from, so that shouldn't be a big problem, I wouldn't think.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.