Is an osciallation like this blowing up a 2N5551 ?

[MathWizard]

I built a roughly 20V 1A CV-CI PSU from an old design that I modded. And for power I've tried 2-3 different transformers. This latest 1 can handle it.

It works, but in the past I've burnt out zener's, and I replaced some of that part with a TL431 ref, and now it just burned that out. Too much power across it. So I'll add a pass BJT to that part and split some other zener's too.

I also added some 5ohm thermistors after the rectifier's, hoping to extend the life if the big caps. And then as usual I cleaned it up, and did a quick run through of resistance checks before powering it back up.

And without a load, I had no voltage on the output. And after a while I found that Q5 the 2N5551 has fried, and also the TL431 (1st time anything fried with this trans tho). The over-temp "shutdown" wasn't connected to the base of Q5.

Q5 converts the voltage you set, to the output drive for the power BJT's. And I guess u could say it closes the loop. IRL I've not tried changing the feedback caps on Q5 or C1/Q2 yet. In the past it's had some osc. under certain loads.



In the sim, see this oscillation here on startup, on the VBE of Q5 2N5551. Would something like that fry it ? The sim does it with or without a cold, 5R NTC off the rekkie's.

[MathWizard]

I'm trying to add the pic of the osc, but the website's lagging so it won;t upload

But the VBE of Q5 goes from 0 to -5V a bunch. A 2n5551, is rated like ~150V VCE and VCB, but still ~5VBE

[moffy]

-5v is the zener reverse breakdown voltage of the BE junction of Q5, not good, the oscillation shows you have an unstable system.

P.S. If you could post a screen capture of your LTSpice circuit as I and I'm sure some others don't have all of your library components.

[MathWizard]

Yeah I'll replace the 7812/5, and some Darlington, might not be stock, or the TL431, and TIP31's. IDK how much the startup matters, in terms of having the voltages end up where they should between 2 voltage rails, so IDK if me adding those inrush limiters matters or not.

[chilternview]

Whatever's driving the 2N5551 into reverse BE breakdown is not right (and probably what is destroying it), first thing to try is disconnecting your current limit circuitry (e.g. the diode that goes to the base of the 2N5551).

If that stops the reverse bias then figure out why your diff pair is pulling that diode down so much. Clamp diodes might help, then work out what's causing the oscillation?

The schematic is only partial, what are the various rails - can't read any of the text apart from the large blue bold stuff due to the scaling used. What's the wire ending in free space for (near a resistor, R12?)

[MathWizard]

Most everything in the sch. should be default parts now. What happens if u try to load a file without all the parts in your lib ? It doesn't load at all ?

I just used the ZZZ and or Broynov ?? lib's.

The main power rail is about over 40V unloaded. And around 20V full load. And it only powers the 4 power trans's, and Q7 , a driver BJT, right after Q5. In the sim tho, that BJT, Q7, it has an awfully high current during some of the start up.

The control rail, is around 44V unloaded, and 43V with ~160mA. The reference voltage it makes, is 5.1V relative to it's GND, and it sits directly attached to the +output lug.

Yeah in the sim, making C1 and C5 bigger, damps most the osc, I've ever seen in the sim. But it's still over my head to properly work out the loop gain and stability, etc.

Well I have lots more 2n5551, so I should reair it again, and see what happens. Maybe remore the inrush limiters. Looking at it yesterday on the scope, trying to find what was wrong, I could see the reference rail take about 100ms to fully charge. And the power rail, was faster, IDK, maybe 70ms.

Some of it might be the 100uF and 47uF caps I added yesterday along the control rail, before the error amp and above the TL431. In the sim, it does osc. a lot more with them, along with the current into the TL431

Ok if I reduce the reference tl431 voltage, the -VBE5 osc does also reduce, and it is about Vref, so...I'll explore that more.

I did briefly try zeners on BE5 today, then I remade the whole circuit to make it presentable on here. I'll try and notice their leakage currents, IDK if they will matter or not.


Ok just adding a 100uF cap on the TL431 output, seems to cure it and VBE doesn't even go negative

[MathWizard]

Well I changed the 470 and 330 resistors for more power, but about the same res. And split some zener's, and upped the rating of a couple of caps for this transformer. I set the ref voltage closer to 4.5V.

I'm glad I didn't power it up tho, it would still kill Q5 if the voltage set pot, is under ~100ohms. And yeah the voltage set diff amp, will make it's output so high as to saturate Q2, and turn off and neg. bias Q5.

When trying to set a few mV, there's lots of little currents going everywhere. Sometimes down the base resistor divider of the current limiter, sometimes back to the reference area.

So for now, I'm just going to add a lower voltage limit with a 470 ohm resistor. Probably add a zener or backwards transistor around Q5


In the sim anyways with AC voltage sources for power, for such a basic PSU, it can have  20uVpp ripple at 10V 1A, s now to see IRL

[MathWizard]

Ok I had to work out a few more kinks, the 7812 was overheating with the min current/max voltage drop. So in the end, I dropped it altogether, and just keep the string of 220R's and my rail is up around 30-40V.

Then I had to rework the overtemp and fan controllers, so that I wouldn't go over the BJT voltage ratings, or anything else. And in a way such that I'd make the minimum changes, and thank god for LTSpice, I found an easy way.

I got everything wired neatly, the NTC's are screwed onto HS. It's all in a wooden box, with mounted V/A meter and knobs, but not the LED's, switches yet, or real lugs/posts.

My heatsink has 4x TIP31C's on it, and a 12V fan. My NTC's are ~12.5k at 70F, and I set the OT to about 3K , which corresponds to when the HS burns my finger.

I just ran 30W into the HS for about 2 hours, and with the fan at about 7-7.5V, it could just keep under the OT.  And the OT LED shuts on and off no problem. It's repeatable, and works to within a few mV, I set it off the 360R voltage of about 0.395V

I left an open collector BJT in the OT circuit. I'll figure out something top do with that, maybe with a couple of options like a cool down time, or a latched output shutdown for if I walk away.

So far I haven't seen any bad noise or oscillations at any outputs I've tried. Actually when the load switch is open, there's some osc. The CV and CI are working pretty good. Except the CI can result in the voltage going to just about the main rail, 43V. I have to look into that, see if that will blow it up.

Other that I'm pleased. Here's the schematic with I think all default parts, so it's not exactly my version. But it works too.


I did blow up 1 cap today, a cheap 100u 25V cap, when I was testing the fan without any cap right on the pass-PNP's collector, bad idea, I was getting 40V spikes. So I added a 1N4002 and 100nF cap, and that did barely anything. And then the 25V cap on the 30V rail blew up, so that would have happened anyways, I just forgot that 1.

So I added a 68u 100V cap, and it looks good on the scope now. Ages ago, I used 100u, like in computer power supplies, so that's not overkill then.

The PSU looks good to a 6.5d DMM on 10G input Z and 100PLC, it's been running 7.007xx-7.009xxV with a current around 0.67A for about 1 hour in my cold basement.