catch diode problem

[trevwhite]

Hi. I have the following simple circuit that switches on a resistive load that is approx. 6A.



I am using a diode rated at 200V, 1A as the catch diode. This is a S1DL. I measure the current through the diode and it never reaches more than 2Amps for less than 5us. This is well within specification. I actually used a current probe and a 1N4007 to measure the current the diode would see in this situation. 

The problem I am getting is that sometimes the diodes are blowing up. It appears also that the diodes are blowing up when the device is being switched on but I can not confirm this at present so it might not be an absolute fact. Also when I say blowing up I do mean it. The cases are cracked or completely in bits and they stink of smoke.

Also when this board went into initial production run the production company switched over the diode and use a brand I have never heard of before. It is supposed to be a like for like but I can not be assured of how good that particular brand of diode is.

I really need to understand if there is any way on switch on that the diode could be stressed? Is it possible from being switched off that the diode might conduct initially when the circuit is switched on and this shoot through current could be killing the diode? I understand that diodes have a recovery time but I believe that is when they are conducting and then have a reverse polarity connected across them but that does not seem to be this situation. The diode starts off in a non conducting state.

Any help would be really appreciated.

Trev

[acbern]

If you have an inductive load the current flowing through the load, as a rule of thumb, initially flows through the diode. Thus the diode should generally be rated with the same current to be on the safe side. Now you can qualify the load decay current and relate that to the allowable pulse current (with tolerances of course). However, as you see, your measurement may be wrong (bandwidth of probe?) or maybe also the repetition rate of the on-off cycles does not match, otherwise your diode would have survived (assuming your peak current integral is within SOA of diode as you seem to have analyzed; I have not checked).

[trevwhite]

Hi and thanks for the response. The strange thing is I am being told this is happening when the system is switched on. In that scenario the diode should not conduct. It should withstand the 33V as its rated for reverse voltage of 200V.

Also it does in the catch scenario appear to work well within tolerance as the diode is rated for pulse currents of 30A for 8.3mS I think. I am doing 2A for <5us.

This is why I am trying to think is there a way it could conduct in reverse polarity initially and blow up. It really does not make sense to me. I am hoping it is down to a shite diode product and I replace it with branded and life goes on but I want to be able to reproduce whatever the problem.

I am currently switching various voltages on and off across the diode in reverse polarity to see if I can create some failures.

[AndersJ]

Miscellaneous thoughts:
- A resistive load does not need a diode, unless cables are long and inductive
- If the diode is installed backwards it will blow on turn on
- If the diode is not what you think it is, but a zener < 33V, it will blow on turn on

[trevwhite]

I appreciate the misc. thoughts. Good to get some other ideas.

My first assumption was that they had plugged in the unit the wrong way round but that has since been proven not the case.
The diode is installed the correct way. A batch of boards were made and a hand full have blown up. On some it can be seen the diode is the correct way around.
It is definitely not a zener, its a like for like S1DL made by a Spanish company called Fagor Electronica.

There is a 3m cable running to the board which powers it up. But the diode is still not being used in this way. The diode circuitry only sees 33V. The cable to the heater is literally 15cm long. I put the diode in just for good practice in case the cable was longer.

[AndersJ]

Major transients on incoming power?
Automotive load dump?

[trevwhite]

No. The power to this board is turned on from a FET on a cable a few meters away.

Anyone got any information on how age affects diodes?

[Seekonk]

Your diode is not very fast, it will keep conducting after turn off for a while.  I wonder if you drive circuit is not stable at power up and there is high frequency oscillation for a short period.  One test would be to put a faster diode in and see if the problem goes away.

[trevwhite]

Hi Seekonk

I am wondering if maybe something turns it on at power up. I am going to investigate that. I it not a normal reverse voltage situation because I have done tests now to satisfy it is something more complicated.

I can not currently repeat the cause of failure so replacing the diode is a little bit of a close my eyes and hope thing. I really want to prove what the problem is first. I think it must be something more complex to do with start up as you suggested. I am going to move on to testing that and see what I can find.

Thanks all

Trev

[Seekonk]

I would also be looking at the circuit that turns on the power to see if that is stable.  All FET will initially turn on when power is first applied due to miller effect.  If something causes that to continually repeat for a short period it would kill the diode.

[Circlotron]

I'd stick a 1N5819 40V 1A schottky diode there instead. That would be right at home in a cct like this.

[Andy Watson]

Are you turning the heater on and off, or are you pulse-width-modulating the drive? !N4007s are not good at high frequency.

[trevwhite]

The heater is turn on or off, no PWM. They are saying to me the system appears to fail when the system is turned on so I think it must be a start up condition.

The max voltage the circuit will see is 39V so the 1N5819 is a little close to the wind. If I am not able to recreate the situation I will have to take some course of action. Maybe fitting a fast diode might just change things and we will have to keep and eye on it. It would be really good to be able to recreate the fault condition.

[Andy Watson]

How does the power supply to the mosfet drive rise/fall? If there is a significant period where the gate voltage is such that mosfet enters its linear mode it could be oscillating - this will dump a lot of power into the diode. Is there a stopper resistor in series with the gate?

[trevwhite]

The rise time is <2us of the FET switching on. Even so when the FET switches on the load is not in the diode. It is only when the FET switches off if there is a spike that the diode starts to work

[Andy Watson]

I was thinking along the lines of what happens at power-up and power-down, when (as Seekonk was suggesting) the drive conditions might be undefined. What provides the power for the heaters - presumably it is rectified (?) does it have significant smoothing, i.e. power storage capacity?
Is there anything that could drag the main supply negative ?

[Siwastaja]

Is there proper local bypass caps on the DC link (over both the diode and FET), i.e., a combination of an elcap and some ceramics? Your 2m cabling to the system has quite a lot of inductance, so any current surge could induce short overvoltage peaks. I'd expect them to kill the FET first, but the diode may go as well.

[trevwhite]

The heater is a cold resistance of approx. 6R. It has always switched on really nicely without any surge current spikes.

There isnt any capacitors across the supply and ground for the 33V line. Maybe there is some high frequency latching? Maybe its a simple as 1nf cap? but I just cant get a system to fail here so that I can detect whether I have fixed it. That is the frustrating thing about it.

[David Hess]

I am using a diode rated at 200V, 1A as the catch diode. This is a S1DL. I measure the current through the diode and it never reaches more than 2Amps for less than 5us. This is well within specification. I actually used a current probe and a 1N4007 to measure the current the diode would see in this situation. 

How was the current through the diode measured?  Was this the current through the diode at turn-off?

Why is there a resistor in series with the source of the power MOSFET?  What type of resistor is it?

I like the idea that the power MOSFET is breaking into oscillation during turn-on and turn-off.  A storage oscilloscope should have no problem capturing this if it is happening by measuring the gate-source and drain-source voltages.

[CJay]

It is definitely not a zener, its a like for like S1DL made by a Spanish company called Fagor Electronica.

I suspect the diode is underrated but others will be able to tell you better.

My input to this though is that Fagor have been around a number of years and are pretty good, not the best quality but far from the worst.

[trevwhite]

Thanks for clearing up Fagor, that is good to know they are okay.

I do not believe the diode is underrated. I measured the current through it and the worst I saw was 2Amps for < 5us. Considering it can handle 30A for 8.3mS it surprises me it might be failing. I would be open to changing it if I could prove it was underrated but all my measurements have failed to show a situation that directs me to that conclusion.

The switching looks good on the scope. I am thinking more in the direction of some RF getting in. There is no capacitance across the input supply to the board. There are in other places but not directly at the inputs of the power leads and they are actually something like 1.5m each in length.

I wish I could just get one to blow up in front of me. Least then I would have something to work from.

[CJay]

Thanks for clearing up Fagor, that is good to know they are okay.

I do not believe the diode is underrated. I measured the current through it and the worst I saw was 2Amps for < 5us. Considering it can handle 30A for 8.3mS it surprises me it might be failing. I would be open to changing it if I could prove it was underrated but all my measurements have failed to show a situation that directs me to that conclusion.

The switching looks good on the scope. I am thinking more in the direction of some RF getting in. There is no capacitance across the input supply to the board. There are in other places but not directly at the inputs of the power leads and they are actually something like 1.5m each in length.

I wish I could just get one to blow up in front of me. Least then I would have something to work from.

I suspect you've got more of an inductive load than you think, don't forget the supply wires at 3M in length will also contribute to that inductance to push the 'spike' well over the PIV rating of the diode.

Have any of the original spec diodes failed or is it just the Fagor ones?

[trevwhite]

It seems they never used the original diode I specified. They stated they were sent the Fagor one. I am am not 100% sure it is a Fagor part but the markings on the diode suggest it is.

I did start to see a little bit of voltage spike on switch off but we are talking <0.5us and I am not sure how real that is. It did not show up as any current through the diode anyway. I use a A6312 current probe for measurement.

The switch on is really nice. The current raise is over 2us and its just smooth up to constant current. No over spike, nothing. That measurement is with the A6312 connected around one of the wires to the heater.

Anyone got some input on RF noise. As said I have no capacitors directly connected across the input supply? Could this cause some problems with 3m lines?

[Kleinstein]

My guess would be RF oscillations of the MOSFET in some rare cases, e.g. during power down of the driving circuit. Not having decoupling on the supply and possibly having more than one such channel can contribute to this. So a series resistor at the gate is a good idea but also have a look at the driving circuit if this is the part oscillating the resistor is not enough.

The diode looks like large enough for normal slow operation (e.g. less than 100 Hz), but it could be a problem with high frequency oscillations, as the diode will than see µs pulse of current and reverse voltage to fast, before it can recover. So a faster diode like UF4003 would have better chances to survive. But it's better to  avoid fast switching.

[Siwastaja]

There is no capacitance across the input supply to the board. There are in other places but not directly at the inputs of the power leads and they are actually something like 1.5m each in length.

Then this cannot work at all, except with pure luck. Input power bypassing with caps is an absolute must in any electronic product using semiconductors, no matter how simple case of switching it appears.

Both sides are as important; the inductance caused by 1.5m leads on the input side is as problematic as it would be on the output side. In the output side, you deal with it using the freewheeling diode; in the input side, you need bypass caps to make the loop smaller, to decrease the inductance.

If we assume 2uH of inductance on the input leads (approximated using https://www.eeweb.com/toolbox/wire-inductance/ , pair of lines, 1mm dia, 10mm spacing), resistance-limited current of 33V/6ohm = 5.5A, then we are storing quite a lot of energy in that inductance. The voltage WILL overshoot quite a bit. I can't come up with a simple way of saying how much exactly, but you could easily simulate it in LTspice.

Assuming that the culprit is the current, you measured diode current, but you didn't measure the DC link voltage, did you?

I'm not saying this is the definite answer, but this is the most prominent culprit to me. Supply bypassing is an absolute must when switching anything with MOSFETs, unless you really slow down the switching.