PSU protection

[grumpydoc]

A while ago I acquired 4 HY3020 power supplies (0-30V 20A). These are fairly straightforward linear designs originally by ? and cloned by many far east vendors.

They all basically had the same fault - blown series pass transistors. The original design had 6x 2N3055, in rebuilding them I used 6x2N3773 since I wasn't convinced that the thermal design was up to continuous operation at 20A and the 3773 has a lower junction->case thermal resistance (meaning that the thermals just work). I considered something more modern such as a MJ15015 but the 3773 won on price (though I notice that it might not have such a margin as Farnell now list the 15015 at £2.54 ea).

All seemed well testing the refurbished supplies and I sold a couple with no problems, I've just got around to selling the last two so and the moment I have one left to go.

But  the buyer of the third has just been in touch to say that it has died - I suspect when I get it back I'll be replacing the pass transistors again.

Before selling the PSU I ran it at full output for a few hours to make sure that it was OK so I'm not sure that just raw load killed it, they said they had it up to about 15A

So, I find myself wondering whether the design has any inherent flaws or it's just that they connected a "difficult" load such as a big DC motor (lots of back EMF) or an inverter (possibly large caps on the DC rail+big switched inductances). I've asked them but they haven't come back to me yet.

Can anyone see any inherent dodgyness in the design?

There is a 1N5402 across the output which will clamp any reverse polarity but that's about it for protection - if there's a large capacitive load the pass transistors could end up reverse-biased and there isn't really anything to clamp a largeish positive EMF. I'm considering adding a further 1N540x backwards across the CE junctions of the pass transistors to clamp and reverse-biasing voltage (should just dump it back through the bridge into the transformer) and perhaps a 50V mov across the output to just generally clamp anything presented to the output to a low value - any thoughts on whether those measures are  worth it?

[SeanB]

That most likely will work. I am just not happy with the parallel power diodes though, they really are not good, as one will hog the current ( generally the top ones as they are the hottest) and will be horribly overrun. Preferably a proper 35A or higher bridge rectifier on a heatsink in place of them. I have seen rectifier diodes that got hot enough to desolder themselves and fall out of the board, even with long leads.

[grumpydoc]

I am just not happy with the parallel power diodes though, they really are not good, as one will hog the current ( generally the top ones as they are the hottest) and will be horribly overrun. Preferably a proper 35A or higher bridge rectifier on a heatsink in place of them. I have seen rectifier diodes that got hot enough to desolder themselves and fall out of the board, even with long leads.

If money were no object I would replace the bridge with an LT4320 and 4 MOSFETs but there is a practical limit to how much redesign of the power supply is sensible.

Actually, in practice, the parallel diodes seem to work OK - the 1N5408 is supposed to be resistant to thermal runaway and I've had them loaded up to 20A for 2-3 hours without the bridge destroying itself so although it's not ideal I don't think it's a terribly weak point of the design.

[pipe]

Perhaps a little bit off-topic, but as a newbie I'm curious about the design here. It seems to me that they actually generate a negative voltage in reference to the ground point in the schematics, and then use that negative voltage as a 'fixed reference point'. Of course, since everything is relative and the ouputs are floating, it doesn't make much of a difference, but is this a common way to design these power supplies? Or am I reading this the wrong way?

[grumpydoc]

Perhaps a little bit off-topic, but as a newbie I'm curious about the design here. It seems to me that they actually generate a negative voltage in reference to the ground point in the schematics, and then use that negative voltage as a 'fixed reference point'.

One reason is that means you can tie one end of the current shunt at 0v and use a single ended voltmeter to display current rather than needing a differential input.

Moving back to the protection for the pass transistors - I'd never really noticed the diode at the output so I pulled the one I still have out and took a look. The diode is somewhat hard to get at but I don't think the output transistors fell off the cliff - I think they were pushed, one 1N5408 completely wrecked suggesting that the output just might have been abused a little.

[SeanB]

Rather much to expect a diode that optimistically is rated at 3A ( chinese rating) to survive something that smoked a 20A pass element.

[grumpydoc]

Rather much to expect a diode that optimistically is rated at 3A ( chinese rating) to survive something that smoked a 20A pass element.

The Fairchild data-sheet for the 1N540x says 3A so it should be real amps not Chinese ones. The surge rating is 200A. Whatever hit that diode physically blew it apart so it had some energy behind it - I'm not surprised it took the pass transistors out as well.

I did debate putting something beefier in but ultimately I think I'll stick to protection against reasonable scenarios, not unreasonable ones.

[grumpydoc]

Well, the "faulty" PSU turned up yesterday and I've had it on the bench this afternoon.

Not a single problem as far as I can see - it's been doing 12V @ 7A half the afternoon without breaking a sweat and was fine at full output. Oh, and it's not blowing fuses or tripping breakers here either.

Not sure what to do now. Do I give a refund knowing that ebay is likely to side with the buyer anyway? Can I trust ebay to recognise that B2B transaction aren't covered by the distance selling regs. Do I stand my ground?

Hmmm, difficult life is, as Yoda might say.

Probably a bit more testing before I message them back.

[Monkeh]

Not a single problem as far as I can see - it's been doing 12V @ 7A half the afternoon without breaking a sweat and was fine at full output. Oh, and it's not blowing fuses or tripping breakers here either.

What's your line voltage?

[grumpydoc]

What's your line voltage?

According to my UT61E just shy of 240V

They didn't say whether it took the current trip or the RCD trip. It passed simple PAT testing but I only have the most basic of testers - a Seaward Primetest 50.

I can sort-of see the HY3020 taking out a twitchy current trip. They have a pretty meaty toroid (easily > 1kVA but I don't know the exact spec) and a commensurately large inrush. They will blow anything smaller than a 10A fuse even with no load connected. However the fuses haven't blown eithe in the plug top or the unit itself.

But they also claimed "no output" which I don't see.

[SeanB]

Perhaps a little bit off-topic, but as a newbie I'm curious about the design here. It seems to me that they actually generate a negative voltage in reference to the ground point in the schematics, and then use that negative voltage as a 'fixed reference point'. Of course, since everything is relative and the ouputs are floating, it doesn't make much of a difference, but is this a common way to design these power supplies? Or am I reading this the wrong way?

Negative supply rails are there to allow the opamps can stay within their common mode input range, allowing you to use a cheap multiple sourced common opamp that will work well. It also allows the output to all the way to 0V while still having the opamps controlling the output voltage. Otherwise they would have to use a more expensive rail to rail opamp and many are not able to operate off power rails higher than 6V.

As to Grumpydoc and the return I do not know, possibly PEBKAC or they tried to run something that needed more power than the unit can provide. I know that Fairchild semi made good 540x devices, but there are also a lot that are to put it politely rather optimistic in the ratings. Might have the case and the thick leads but I have seen quite a few with steel leads and that fail at around 3A, while I have seen genuine ones run 5A with no problems. Does look like the blown one suffered from a large overcurrent though, probably a motor running drawing 20A and the reversing switch was flipped with it running at full voltage. If it is tripping the mains probably a good addition would be a PTC inrush limiter of around 5R.

[grumpydoc]

As to Grumpydoc and the return I do not know, possibly PEBKAC or they tried to run something that needed more power than the unit can provide. I know that Fairchild semi made good 540x devices, but there are also a lot that are to put it politely rather optimistic in the ratings. Might have the case and the thick leads but I have seen quite a few with steel leads and that fail at around 3A, while I have seen genuine ones run 5A with no problems. Does look like the blown one suffered from a large overcurrent though, probably a motor running drawing 20A and the reversing switch was flipped with it running at full voltage. If it is tripping the mains probably a good addition would be a PTC inrush limiter of around 5R.

Just to make sure there's no confusion the blown 1N540x is not out of the returned unit but out of the last of the four which I finally got around to repairing last week. Any that I've put in were sourced from Farnell and I ordered the Fairchild part. Obviously the originals could have been from anywhere.

They claim that the first time they used the PSU it was OK but then started tripping their breakers and when they did get it powered up there was no output. As they deal in solar power systems and inverters and whatnot I assumed they had a clue and the PSU really was dead - but, unless I accidentally installed magic self-repairing 2N3773's I am now wondering about user error as a possibility.

Tomorrow morning I'm off to Luton but I'll run it a bit longer before deciding how to approach them.

To be honest if they were a bit closer I might just offer to go over and see what they've been doing with it but they're more than 3 hours drive and I have a day job as well.

[Monkeh]

Solar people? Taking bets they have a rooftop covered in panels and an excessive line voltage..

[grumpydoc]

Solar people? Taking bets they have a rooftop covered in panels and an excessive line voltage..

Maybe - they didn't have when Google stopped by for photos but the industrial unit that they now occupy was in use by someone else at the time.

Even if the line voltage was a bit higher it probably wouldn't have a huge impact.

Think I'll try to get some more information out of them first.

[Monkeh]

Solar people? Taking bets they have a rooftop covered in panels and an excessive line voltage..

Maybe - they didn't have when Google stopped by for photos but the industrial unit that they now occupy was in use by someone else at the time.

Even if the line voltage was a bit higher it probably wouldn't have a huge impact.

10V can make a difference when it comes to inrush. Easily enough to have it tripping 16A or 20A B-curve MCBs. 20A ones will go quite often with 1500VA site transformers (I speak from experience).

An industrial unit will probably have a pretty low source impedance, too.

[NiHaoMike]

A MPPT controller can cause the output to swing wildly as it runs into a "sharp" current limit (with little voltage drop prior to that) as opposed to a gradual voltage droop followed by a "soft" current limit. When I was testing some MPPT controllers (though for a stationary bicycle, not solar panels) for my senior design project, the lab supplies would wildly chatter with a MPPT load.

[grumpydoc]

10V can make a difference when it comes to inrush. Easily enough to have it tripping 16A or 20A B-curve MCBs. 20A ones will go quite often with 1500VA site transformers (I speak from experience).

An industrial unit will probably have a pretty low source impedance, too.

As I said I can easily see these triggering a current trip - obviously as I run the "business" (i.e hobby of fixing things with occasional sales) out of the garage I probably have more forgiving breakers.

However they claimed that even when powered up there was no output which currently I'm finding a bit harder to swallow.

Don't forget that all these PSUs have is a current limit so if they are asked for a bit too much then they just collapse the voltage until the current is ? 20A

For the time being all I can really do is ask what they tried to drive with the PSU and exactly what happened and work with them to see if I can figure out the problem.

It doesn't help that I'm also miffed that it has sustained new damage whilst with them - it wasn't perfect but was pretty good and now has new dents in the case.

[grumpydoc]

If it is tripping the mains probably a good addition would be a PTC inrush limiter of around 5R.

I presume you mean NTC (although I know that PTCs are also used).

I don't have much experience choosing such a beast. I'll need something rated at least 6A, I think - does this http://uk.farnell.com/epcos/b57238s509m/thermistor-ntc-5r-6-4a/dp/1704471 look about right.

How hot would it get at operating temp?

[SeanB]

It says in the short form 170C, and they reach this at full current then drop a little as they lower in resistance. Mount in a location on a tag strip such that is is in free air away from anything it can cook. I did a lifetime buy of 3 and 4 position tag strips a while ago, still have around 100 left. You use a single position as the ground and the mounting, and then use the other 2 to solder the device and the power leads.

I have placed them inside a lamp holder as inrush protection, where they were fine as the lamp in any case was running at 100C or more in the base anyway. Made the lamp last longer, or at least I never changed a lamp in that desk again, as 6 months later she dropped it on the floor and threw the broken pieces away ( so I was told, probably took it home to use) and only when I went around to check inventory ( on an asset register so I check annually) it was not there with the story.

[grumpydoc]

It says in the short form 170C, and they reach this at full current then drop a little as they lower in resistance. Mount in a location on a tag strip such that is is in free air away from anything it can cook. I did a lifetime buy of 3 and 4 position tag strips a while ago, still have around 100 left. You use a single position as the ground and the mounting, and then use the other 2 to solder the device and the power leads.

Yes, hmm. They have it on a 16A "B" breaker so we need to keep inrush < 48A - your 5ohm suggestion was spot on 

Farnell do an EPCOS 5ohm NTC for inrush suppression rated at 6.4A. The data-sheet says it has standard R/T curve 1207, a max power dissipation of 3.9W and a 20mW/K thermal resistance in air.

I found a document with all the standard curves here, very useful.

That says for curve 1207 at 170^oC the multiplier compared with the resistance at 25^oC is 0.03670 or 0.1835ohms.

If we assume that the PSU takes 5A fully loaded that makes the effective impedance of the transformer primary 48ohms (I'm assuming a realistic 240V in the UK rather than notional 230V), adding the NTC resistance at 170^oC gives 48.1835 ohms or 4.98A  - I²R gives 4.55W

That's over the max power, and at 20mW/K gives a 227.5^oC rise (not forgetting that's over ambient which could easily be 40^oC inside the case).

So that one will hit equilibrium somewhere well north of 175^oC and probably fry itself at full output, or try to desolder itself or both.

Back to the drawing board, possible the 4ohm 9A one will stay the right side of 175 degrees although the inrush might still be a bit high. I think I need to find something with a higher power dissipation and/or a higher beta - perhaps closer to 4000K

Ah, there's a higher power EPCOS one on Mouser - S364 series - 5ohm, 8.5A but they give the graphs rather than reference the standard curves. Will stare at the datasheet tomorrow.

However I'm still looking at something running close to 175^oC so I'm going to have to crimp it or use high temperature solder, plus high temp sleeving and find something suitable to mount it on and somewhere with at least a couple of cm clearance on all sides which is going to be a little tricky.

Well, I've offered it as a solution but I'm not sure I'll pull it off if they want me to go down that route.

[Monkeh]

The customer might find it easier to simply change the breaker. A C16 or maybe C20 if the circuit allows should solve the issue.

[grumpydoc]

The customer might find it easier to simply change the breaker. A C16 or maybe C20 if the circuit allows should solve the issue.

I've suggested that as well.

[grumpydoc]

The customer might find it easier to simply change the breaker. A C16 or maybe C20 if the circuit allows should solve the issue.

I've suggested that as well.

Looks like they're going for the "change the breakers" option.

In the end I don't think that the NTC idea would have been practical as I don't think there was a location within the PSU that I would have been totally comfortable with installing something that was going to be over 150^oC in operation.

Two 2.5ohm NTC's in series would have helped with the temp a bit but  it would still have been pretty toasty.

However a random through occurred - would using a zero-crossing SSR also be a way of limiting the inrush?

[richard.cs]

However a random through occurred - would using a zero-crossing SSR also be a way of limiting the inrush?

Not on a transformer load, you would want to switch on the peak not on the zero crossing and I'm not aware of any off the shelf stuff that does that. A 5 Ohm wirewound resistor shorted by a relay after a few mains cycles would be easier - there are various designs around that use a 48V relay run off of a capacitivly derived d.c. supply and switch it on with a single npn after an RC delay. You find them in microwave ovens quite often since those transformers are designed in a way that gives them very high inrush currents.

[grumpydoc]

Not on a transformer load....

OK, yes - that makes sense now that I think about it, you can't change current through an inductor instantaneously so even if you "threw the switch" at the peak of the waveform no current will flow.

Which brings me to two questions:

First off what is the mechanism for the inrush? Is it as simple as the resevoir caps charging on the secondary side?

Secondly what exactly happens when you "throw the switch" to an inductive load at the zero volt point of an AC waveform. With everything running current lags voltage through an inductor so peak current corresponds to zero volts¹ but if I start at 0V and ramp up with a sine wave what happens? Unfortunately my maths falters when we come to AC circuit analysys (ideally I'd go back and get my calculus brushed up - should have done a numerate subject as my first degree mutter muble).

[1] Another concept I'm struggling to get a mental model of - current with a small voltage behind it I'm OK with but current flow at 0V?