Console PSU keeps blowing diode bridge. What to do?

[Ilya]

At some point we should take a step back and ponder why we are essentially redesigning a product years after it has been put into service. Were these PS always a hot mess, or has something degraded? In my experience caps fail or they don't. If you suspect the caps, such faults will usually reveal in a visual inspection. Look for swelling or leakage. Sometimes if they are left powered off for very long periods of time, they may need to be re-formed in, but I'm talking sitting for years not days.   

That's exactly what I was thinking.

I don't have the PSU at hands right now, but I remember looking at the caps and I'm pretty sure they looked ok. I'm also quite confident that when I check them with a dmm I'll see a normal behaviour - that is initial 0R resistance that is slowly rising up to infinity. Otherwise (if they were failed shorted) the PSU won't work at all. So I'm not really sure how to check them if they are still any good.

When I had that beast at home, I started it slowly with variac and had no problems whatsoever.

Replacing all caps is quite expensive - they run up to 70$ each, and there're 4 of them. So this brute force approach is reserved as a last resort.

I've also found a 200A diode bridge that I can get locally (around 60$), as well as semikron diodes mentioned by analogguru (approximately 50$ each). But installing these looks like a bandaid. So I really need to find a way to check the caps reliably. I'm planning a trip to the studio tomorrow, will see what I find.

 

[JohnRoberts]

I do not expect the caps to be bad and a 200A diode bridge seems like over kill.

If you study the data sheet for most diodes, they will tolerate momentary overloads 10x nominal ratings or so.

The failure mechanism for diodes is overheating (literally melting the silicon). The transient high current to charge up the caps the first few cycles causes a lot of heat momentarily.

Is there any way to help the diodes get heat out more quickly like thermal epoxy or a heat sink?

A silly idea, but a long extension cord will reduce the initial peak current pulses. 

JR

 

[Ilya]

John,

The diodes (or the bridge in its previous incarnation  ) are mounted on the massive heatsink with a fan. I doubt that making it bigger will make any difference.

While  inspecting the failed shottkies I found one to be black from burning, and another had a crack and recess right in the middle with something white poking through. Before that, the bridges failed with their legs welded off.

Extension cord from the stepdown is approximately 5m long.

It's curious that most of the time the positive part of the bridge fails (marked on the pic)

 

[Kingston]

I have seen those exact parts fail in a PSU where I completely forgot to rate the diodes high enough. It worked for some years constantly at double the current rating of the diodes (!!), then one day the two diodes fizzed out and "pop, goes the weasel!" (mains fuse).

Pretty sure it's an over current situation. In my experience not all caps can be visually inspected for failure. Some of the modern ones are sturdy enough so they'll basically implode somewhere deep within, nothing you could see from the outside.

I would just go ahead and order some caps already.

[edit]

the above advise related to the fact the PSU even was unloaded during a failure. I would also disconnect each of the separate regulator sections and measure the start up current surge with each of them separately. Or set up adequately rated fuses to each separate regulator section if you don't have the measurement gear to do this.

 

[Ilya]

I would also disconnect each of the separate regulator sections and measure the start up current surge with each of them separately. Or set up adequately rated fuses to each separate regulator section if you don't have the measurement gear to do this.

My Fluke can measure up to 10A which seems not sufficient in this situation. I think of measuring voltage over, say, 10R resistor connected between the caps and the regs. What do you think? Maybe the scope would be even more convenient.

Also, would you suggest going for 200A bridge?

 

[jplebre]

In my experience not all caps can be visually inspected for failure.

the power supply in question had no visible damage, neither on the caps (no swelling, pop, droopage etc.) neither on the discharge resistor.

 

[Kingston]

My Fluke can measure up to 10A which seems not sufficient in this situation. I think of measuring voltage over, say, 10R resistor connected between the caps and the regs. What do you think? Maybe the scope would be even more convenient. Also, would you suggest going for 200A bridge?

I thought you mentioned the raw DC voltage was fed to a bunch of 3A regulators. You could certainly measure one section at a time in isolation even with the fluke.

That would be the next step if the caps turn out fine.

Don't know about 200A bridges. Seems overkill when clearly the problem is elsewhere.

 

[Ilya]

I thought you mentioned the raw DC voltage was fed to a bunch of 3A regulators. You could certainly measure one section at a time in isolation even with the fluke.

Not sure. The schematic calls for 24A going into each reg rail. Am I correct suggesting that the initial spike would be even higher than that?

Meanwhile, I've adjusted some values on the delay circuit to make it work from 20VAC.

 

[JohnRoberts]

Since you are getting different advice good luck.

I am speculating that the diode failure at turn on is related to the initial current surge from charging that much capacitance from 0V to 24V at first turn on.  This peak current will be limited by all the resistances in series, including transformer winding impedance, line cords etc. 

How much current does the power supply actually deliver in steady state?

You could reduce the peak cap current by adding say  0.1 or 0.2 ohm resistors in series with diode bridge. This would reduce ripple sightly and drop unregulated voltage slightly.

70,000 uF is a lot of capacitance. Does it really need that much? How much ripple voltage is on the unregulated rail in normal use.

JR

 

[analogguru]

70,000 uF is a lot of capacitance. Does it really need that much?

Per textbook:  2200µF/1A  30A x 2200µ = 66.000µ

I don´t understand for what this 30A are needed.... I thought this is a digital console (therefore mostly running on 5V) ?

 

[JohnRoberts]

70,000 uF is a lot of capacitance. Does it really need that much?

Per textbook:  2200µF/1A  30A x 2200µ = 66.000µ

I don´t understand for what this 30A are needed.... I thought this is a digital console (therefore mostly running on 5V) ?

Per JR 70,000 uF at 100Hz refresh rate (50Hz mains) is roughly 0.14V of ripple per amp draw.  30A = 4.3V p-p ripple

30A at 24V is lot of console... which is why I ask.

JR

 

[Ilya]

I am speculating that the diode failure at turn on is related to the initial current surge from charging that much capacitance from 0V to 24V at first turn on.  This peak current will be limited by all the resistances in series, including transformer winding impedance, line cords etc.

I agree. Actually, before this problem occured, I tried different places to apply power to the system. Mainly before the stepdown and after it. I never managed to power up the console without tripping the wall fuse when applied power before the stepdown. This forced me to leave the stepdown constantly powered and this also devoids me of the benefits of its winding inductance.

How much current does the power supply actually deliver in steady state?

I've never measured this, but the PSU is specced for 24A per rail.

You could reduce the peak cap current by adding say  0.1 or 0.2 ohm resistors in series with diode bridge. This would reduce ripple sightly and drop unregulated voltage slightly.

This is what I'm going to do, but I doubt I can find 0.1R resistor powerful enough for that. I'm planning on using a 10-20R relay shorted resistor instead in the DC path. If you suggest moving it into AC path before the bridge, I'll do that.

70,000 uF is a lot of capacitance. Does it really need that much? How much ripple voltage is on the unregulated rail in normal use.

Well, in max config this thing has 24 channels and master section. Regs in the master section suggest that it consumes about 2-3A, so it leaves approx. 0.9A per channel. Never measured ripple. When I manage to get that beast started I'll do the measurements.

30A at 24V is lot of console... which is why I ask.

This is a large scale analogue desk with digital control. It "eats" accordingly.

I have 4 resistors rated at 22R, 10W. My initial plan was to put 2 of them in parallel, thus making 11R, 20W composite resistor. My calculations suggest that at 11R the current will be 1.8A, and dissipated power will be 36W which is too much. If I make another composite 11R 20W resistor and add it in seriesto the first one, I'll get 22R 20W. This will cut power dissipation in half to 18W which seems acceptable. Am I correct here?

I'm planning on leaving these resistors in circuit for 2-3 sec. Is it ok?

 

[JohnRoberts]

If PS is really putting out 36A a .1ohm R in series with bridge is 3.6V drop to DC, more to rectified AC.

Relay is looking good again.

JR

 

[bruce0]

It ran fine for years correct?

Is it possible you have some sort o a sneak circuit (perhaps intermittent) that is increasing current draw and in combination with the reservoir is just too much current? 

Would a bad regulator among your fleet of regulators, acting badly cause this problem?

Others have said it, but it ran fine for years, so trying to limit current at startup now seems like working on the wrong problem.

Is there a schematic?

 

[Ilya]

It ran fine for years correct?

Correct

Would a bad regulator among your fleet of regulators, acting badly cause this problem?

I doubt it because in this case I'd very likely had other problems with power down the line. All regulators seem to be working fine.

Is there a schematic?

Yes, it's a very basic linear PSU with a bunch of 3A pos and neg regs. Nothing fancy. Just uberpowerful.

 

[bruce0]

How about an (intermittent) short or open on one of the regulators (or supporting caps)... lost in the wash of all the other regulators during use, smoothed out by your great reservoir, but enough to cause current to be high during startup.

 

[JohnRoberts]

To melt a diode the obvious vector is cap charging current.  Have we inspected a voltage breakdown on the diode that first causes it to short circuit, then the huge AC current from driving that much cap with AC brings the heat. 

An intermittent cap short fits the symptoms, but not my experience with cap failures. When caps draw huge current they usually stay broken.  Likewise IC do not usually fail intermittently and self heal. While i concede no experience with that size caps.

Has anything else changed?  A torroid instead of an EI power transformers could have lower winding impedance.

JR

 

[Ilya]

Here's an update to the situation.

I've found 3 protecting diodes on the regs that failed shorted. They were between the in and out terminals, so they probably couldn't be the case of the bridge failure.

I've checked the filter caps again and they look fine. The top plastic cover is flat, although I can feel that it's moving up and down like membrane when pressed. The caps are Mallory extra reliable series 105C. I can't find any suitable replacement caps of the like grade unfortunately.

I've unscrewed all caps and checked them with the dmm. They all measure fine.

Didn't manage to install the relay. I spent all the time disassembling this monster and to replace those diodes all regs had to be desoldered and unscrewed.

I'm thinking of putting it on the variac, but this sort or defeats the purpose of slow start circuit. Maybe I'd have to discharge the caps completely before real test drive.

[edits]
Nothing changed in key components. Toroid was there since its birth.

 

[JohnRoberts]

I would try to look at voltage on transformer side of diodes with scope.  Diodes should only see some 50 odd volts across them reverse polarity but look for spikes or external interference.

If the replacement bridge was not the same as original build that could be a difference.

Sorry I'm guessing at this point. Shorted diodes across regulators could be secondary failure, but probably not. I don't think I've ever seen that happen either. Diodes are there to protect the regulators. Depending on the regulator, a short across it could be fatal.

JR

 

[Ilya]

I'll check with the scope, but I'm really not so willing to turn this thing on without any protection. Each failure is expensive and takes quite a lot of time to fix. Besides, I hate soldering 8awg wires

Will it be as informative if I check the voltage with slow start engaged?

I believe that shorted diodes are a legacy from the last failure, when 3 regs went to the better place. I think that these diodes were right there with those regs.