I am currently building a HV supply (+/-5kV)
As usual, the debugging process has been littered with various mishaps, like probe slips, etc.
Nothing exceptional, and I had installed spark-gaps and others measures in case things went wrong (which they did, big time).
Despite all my precautions, the regulation was lost, resulting in voltages well over 5kV appearing in unexpected places, causing arcing everywhere.
Since the bypass caps are 10nF, the discharges were severe, and when I tried to restore the circuit, I noticed anomalies: some diodes used for function selection were damaged, low-impedance as expected, but anomalies remained.
In the end, I resorted to test them all thoroughly, and I noticed that one of the 1N4148 had gone completely open: no leakage or conduction in any direct way.
Basically, it was completely open, which is exceptional for a diode having a whiskerless construction. It is a Telefunken device, completely ordinary, and it showed no external damage.
Such diodes normally fail short, or disintegrate completely, which was not the case here.
I lost quite an amount of time trying to diagnose what was wrong, because I had never seen such a behaviour before.
The moral is that anything can happen and will happen: never disregard even the the most unlikely possibilities. I lost a significant amount of time trying to figure out what was wrong, because I excluded the open possibility.
As usual, the debugging process has been littered with various mishaps, like probe slips, etc.
Nothing exceptional, and I had installed spark-gaps and others measures in case things went wrong (which they did, big time).
Despite all my precautions, the regulation was lost, resulting in voltages well over 5kV appearing in unexpected places, causing arcing everywhere.
Since the bypass caps are 10nF, the discharges were severe, and when I tried to restore the circuit, I noticed anomalies: some diodes used for function selection were damaged, low-impedance as expected, but anomalies remained.
In the end, I resorted to test them all thoroughly, and I noticed that one of the 1N4148 had gone completely open: no leakage or conduction in any direct way.
Basically, it was completely open, which is exceptional for a diode having a whiskerless construction. It is a Telefunken device, completely ordinary, and it showed no external damage.
Such diodes normally fail short, or disintegrate completely, which was not the case here.
I lost quite an amount of time trying to diagnose what was wrong, because I had never seen such a behaviour before.
The moral is that anything can happen and will happen: never disregard even the the most unlikely possibilities. I lost a significant amount of time trying to figure out what was wrong, because I excluded the open possibility.
Completely open? No leakage? That's impressive. Are you sure there was a diode in the package? 🙂
Tom
Tom
1N4148 is a pretty small diode, so you probably blew its innards to smithereens. Ain't nuthin' in there left to conduct anything no mo! What's a little 100V diode like that doing in an HV circuit? Was it part of a pass element control circuit?
Hopefully you have a way of ramping the power supply up slowly for initial testing? HV probes, ground sticks, etc.? Never liked having to discharge big, oil filled cap banks used for 5MW pulse forming. Not that its what you are doing here, but HV with some fair energy storage is still dangerous.
Last edited:
So where did it GO? When the innards blow themselves to smithereens, pieces of glass go flying. It’s not like an air cavity package where you can fuse a bond wire open. Usually the whole thing becomes molten inside and turns into a zero ohm resistor when it hardens and cools (complete with black band indicating the value).
Yeah, completely open. Maybe the crystal was initially absent, but I doubt it very much: there seems to be "something" in the glass, between the terminals and anyway all diodes undergo a minimal test of functionality by the manufacturer.
If I can find it the bin, I may make a breakdown voltage test
The supply is lab type, and it has various low-voltage logic control functions. Under normal circumstances, the diode never saw more than 12V
It is a relatively low-power supply, +/-5kV 4mA, but yes there are minimal precautions, like the spark-gaps, series resistors, etc.
I managed to recover the corpse and carry out further tests: I used my zener tester at a 200µA current, first on 100V max, then on 1000V.
On 100V, it remained completely open, but at 1000V something happened, and kind of repaired the diode: the voltage dropped to ~0.5V, thus one Vbe (the test was made in the forward direction).
I checked in the opposite direction, at 100V, and it was open as it should, but when I reversed the polarity, always at 100V, it reverted to the open condition.
A kind of bistable operation, but it didn't last: after the application of 1000V, it conducted again, but the reverse breakdown was unstable, drifting between 75 and 100V, which is too low for a 1N4148.
Thus, crystal was isolated from one of the electrodes, probably because the shock had volatilized a thin layer of metallization, and somewhere between 100 and 1000V a conducting bridge was created, restoring normal operation, but in a degraded mode due to the extreme stress it had to endure.
It is worth noting anyway, because I had always assumed that the failure mode for this type of construction was a dead short, or at the very least low-impedance (unless the case is physically destroyed, obviously).
Under extreme conditions, this is not necessarily the case, though it is extremely rare: it is the first such instance I encountered in decades of electronics
On 100V, it remained completely open, but at 1000V something happened, and kind of repaired the diode: the voltage dropped to ~0.5V, thus one Vbe (the test was made in the forward direction).
I checked in the opposite direction, at 100V, and it was open as it should, but when I reversed the polarity, always at 100V, it reverted to the open condition.
A kind of bistable operation, but it didn't last: after the application of 1000V, it conducted again, but the reverse breakdown was unstable, drifting between 75 and 100V, which is too low for a 1N4148.
Thus, crystal was isolated from one of the electrodes, probably because the shock had volatilized a thin layer of metallization, and somewhere between 100 and 1000V a conducting bridge was created, restoring normal operation, but in a degraded mode due to the extreme stress it had to endure.
It is worth noting anyway, because I had always assumed that the failure mode for this type of construction was a dead short, or at the very least low-impedance (unless the case is physically destroyed, obviously).
Under extreme conditions, this is not necessarily the case, though it is extremely rare: it is the first such instance I encountered in decades of electronics
It seems you have a lot of spare time 🙂
Open diodes are not that special although shorted ones occur more. A few weeks ago I had an intermittent leaky diode in a Graetz bridge. Also not common. At first having the issue and causing the bridge to become warm and then functioning normal.
Open diodes are not that special although shorted ones occur more. A few weeks ago I had an intermittent leaky diode in a Graetz bridge. Also not common. At first having the issue and causing the bridge to become warm and then functioning normal.
Last edited:
In general, that is true: devices having a bonding wire or plastic-encapsulated do fail open, but the glass case, whiskerless construction seems to never fail open, unless the glass is broken.
For me, that is a first anyway.
I checked the other day a few 1N4148 datasheets and noticed in an older datasheet from Philips from 1999 max voltage is 75V, 2004 NXP datasheet states 100V continues and repetitive, others mention 100V either just as repetitive, and some as non-repetitive only, so there a few different bids on it, I didn't see any Telefunken datasheet while doing just a quick look around, how long time ago was it when they were around manufacturing these?
It has happened at times I go in 'paranoid mode' and measure resistors, bjt's, diodes to make sure I didn't misread and/or just trying to catch a faulty component.
It has happened at times I go in 'paranoid mode' and measure resistors, bjt's, diodes to make sure I didn't misread and/or just trying to catch a faulty component.
:)
They date back from the late eighties or early nineties, I don't quite remember. Telefunken is now long gone. In general, the measured breakdown for a 1N4148 in good condition is around 125~150V, independent from the origin
Maybe the 1n4148 was missed in testing at the factory ?
Its surprising what sometimes gets through, I once found a one legged polyester capacitor in a bag of caps.
I would just replace it, that will rectify the problem !
5KV flying around is lethal. Anything gets in the way will be vaporized.
Its surprising what sometimes gets through, I once found a one legged polyester capacitor in a bag of caps.
I would just replace it, that will rectify the problem !
5KV flying around is lethal. Anything gets in the way will be vaporized.