Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Variable psu output vulnerability?

Status
Not open for further replies.

throbscottle

Well-Known Member
I'm in the last 1% of my hybrid psu design, and think I have found out a problem that must affect many variable bench psu's.

So, take an adjustable psu with an emitter follower output, connect a capacitor across the output, turn up the voltage, quickly turn down the voltage. Suddenly the emitter of the output transistor is now at a much higher voltage than it's base, possibly exceeding its vbe and killing it.

So I think that's why I keep blowing output transistors anyway. Does the theory hold water? Does it affect all variable psu's? Or is it a nasty trick I have uncovered because the pre-regulator lowers the collector voltage when the output is turned down?
 
Does the theory hold water?
Yes.
The B-E junction is much like a 5 volt Zener diode. In one direction you get 0.65 volts approx. and in the other direction the diode conducts at about 5V. The wattage rating is small and you can kill the B-E junction by reverse voltage.

Place a diode across the B-E of the transistor in the direction that reverse voltage will flow through the new diode. 1N4001 1A 50V or something. The diode will take the discharge current from the output capacitors.
 
Nice to have got something right for a change! The solution creates another problem however in that the energy from the output caps now flows through the control transistor. which easily could put it outside it's SOA if the voltage was cranked up fairly high, then pulled down sharply by the current limit (or even just adjusted quickly). Though I suppose I could use a beefier transistor, which is annoying because I've just bought some BCX38C's (60v / 800mA / 1W / E-Line / darlington) for the job, which I thought were a bit over powered, now it looks the other way...

So far my 2 alternative ways around that new problem are:
1) put in another transistor, PNP, with it's emitter on the psu's output, base at the output tranny's base, and collector feeding back to the control amp so that the output is kept high for as long as it takes for the cap to discharge.
2) put the diode in, but make it a a low voltage (4.3v) zener, which relieves the output tranny but also drops some voltage from the control tranny
In either case a diode goes back across C-E of the output transistor in case the pre-reg voltage drops below the output as well.
 
Thinking about it, if the current limit pulled it down, the voltage on the cap would go down quickly anyway because of the load. It must be when I've adjusted it with no load that it happens.
 
if the current limit pulled it down
No.
Current limit stops the flow of current. It, on its own, does not pull anything low. The load or short pulls the output low. So I think you are OK.
I can not see your schematic so......I am guessing. based on many hundreds of power supplies I know.
put the diode in, but make it a a low voltage (4.3v) zener,
Will not work. The diode might short out the B-E. Zeners pass current in both directions. You could use 2 or 3 diodes from B-E to get 1.2V or 1.8V.
 
I would look at the way 7805 and others are protected against this situation, they simple have a diode from the output to the input terminal, which dumps the charge from the output caps back to the input caps.
 
Sorry I threw in a red herring there, didn't mean to! My current limit works by turning down the control voltage until the current output from the pre-regulator matches a value you set, but it also has a foldback current limit to protect the transformer, right after the main cap. So I actually meant pull down the control voltage.

I'm not posting the schematic, it has too many distractions in it! When it's all working happily I'll put it up as an article. Stick with guessing.... ;)

Good point about zeners. It's easy to forget they conduct both ways! I soon found out the voltage reduction isn't worth bothering with anyway.

Ok if I put in the diode as you say, but with a resistor in series. The cap on the output is 330uF (actually fairly arbitrary to be honest, it could probably be much smaller since the PSU only generates 2.5mV of ripple/noise with this one) - if I put a 220R resistor in series with the diode it just barely keeps in the control transistor's SOA. Conversely, the output transistor is now exposed to the voltage on the output cap for some tens of ms - I guess it will survive that?
 
I am not sure why you want to put that resistor in there? A schematic would really help see the big picture.
 
Well OK, but it's still a work in progress so not everything might make sense, and not all the values are final. You can see why I wanted to get you to imagine a standard type of supply....

I've added the proposed diode/resistor to the channel at the top (B) but not the bottom (A) so you can see both versions.

The rectified & smoothed output from the mains transformer can be up to 55V, (50v nominal, can go as low as 45v) the pre-regulator should be outputting between 3 and 5v more than the final output (it was 5 but I intend to re-calculate to get 3).

I chose to use a sziklai pair for the output as an experiment, and I've stuck with it because I like it - it's effectively an NPN emitter follower with almost as much gain as a darlington. It's the small driver transistor that suffers the vulnerability in question, and what happens as far as I can work out is that it can fail so all pins are shorted due to it's Vbe being exceeded, which then gives a path for excessive base current to flow in the PNP (or exceeds it's Vbe because the input voltage has been lowered, take your pick), completely shorting it also, and also destroys the control transistor into the bargain (which fails open collector) because it is exposed to too much current. I'm running out of suitably rated old transistors, I don't want to start blowing up new ones!

When the voltage is turned down, the darlington control transistor conducts bringing down the output. However if there's no load, or a very light load, it will try to pull the voltage all the way down to zero until the output catches up. Now suppose the output has been turned up to 40v, or the high 30's at least, and then I decide to turn it down quickly. So now with the diode arrangement suggested by ron, there is a capacitor with 40v sitting on it, discharging through a diode and through the fully on control transistor. Obviously quite a lot of current can flow there, which will definitely destroy the transistor, hence the resistor is there to limit it. If you look at the datasheet for the BCX38C, and use the handy calculator here, you will see that with 220 ohms this keeps the transistor just barely inside its safe operating area as the voltage falls.

I put the diode back across the output transistor into the schematic ages ago, but like a dummy I haven't implemented it yet.

The strange looking 2 zener (because they are the values I happen to have, no other reason) and resistor combo is there in case the control transistor fails O/C, it will keep the voltage from the constant current source down to a little over 40. The tracking design would work just as well with a resistor providing the base current, but the CC source gives some ripple rejection.
 

Attachments

  • Schematic_Design__Hybrid_sch_non-Linux-generated_files.pdf
    196 KB · Views: 136
Large capacitors across the output of a power supply can cause several kinds of problems, and you have found one. Another has to do with invalidation any kind of current limiting. If you have the current limiting set to, say, 0.5 A, but have a 10,000 uF output capacitor, you can have amperes of current for a few milliseconds, more than enough to turn a small wiring error into smoke.

ak
 
Been playing with it in LTSpice. The current with just the diode only goes to 40mA, but if I connect a much larger cap across the output, (suppose for a minute I connect one externally for some reason) the current goes into 100's of mA. Putting a resistor in series with the diode in this circumstance completely defeats it's purpose when a larger cap is connected, so I'll forget about that.

Found another solution. I can connect another TIP42C (I'm calling it the discharge transistor) with it's emitter on the output, base to the control transistor's collector, and its collector to ground. As soon as the output's Vbe is enough to turn on the new transistor, it starts to discharge the output cap. Testing it in LTSpice, it works fine as it is, but with a very large cap on the output (100,000uF) it runs into problems. Back to the diode then.

Ok so what I now find is that if I put a resistor in series with the output transistor driver's base, so the CC source is going to the control transistor, connect the diode as it should be (no resistor), hey presto, the resistor dissipates most of the energy from the capacitor, the control transistor's collector voltage drops to about a volt, and the output voltage isn't significantly affected by the resistor being there. The resistor has a negligible effect on output voltage.

Both schemes work well for big caps, simulation showing some oscillation in the control transistor current with smaller ones.
 
AK, I wish I'd known before I destroyed quite a lot of transistors with this thing. Ain't nothing like learning the hard way...
 
Another has to do with invalidation any kind of current limiting. If you have the current limiting set to, say, 0.5 A, but have a 10,000 uF output capacitor, you can have amperes of current for a few milliseconds, more than enough to turn a small wiring error into smoke.

ak

I never thought of that before! :eek:

What is the most basic trick to overcome that, AK? Not reducing the caps, I hope...
 
Decoupling caps are called that because they decouple the effects of transien current demands from steady state current supplies. That's great, right up until it isn't. A big cap, or a group of three cap technologies in parallel, are a great way to get a very low power supply output impedance. But there is that 20 A transient side effect. Kubeek is right, the only way to get around it is to have a wide bandwidth control loopin the egulator, and smaller output caps for high freq management only.

This problem was the very first thing I did with my new EE knowledge.

ak
 
I'm going on the principle that one day I'm going use the psu to charge very large caps for something, and then there's a power cut. It's the most destructive scenario I can think of.

Aside from that, is the 330uF I have sitting on there too big? (It's the first one I found in my junk box with adequate voltage rating of what seemed a suitable value.)

I have been thinking I should put a few smaller caps in parallel there instead, maybe totalling < 100uF, benefit from the lower ESR without the disadvantage of a big cap in there.

As it stands there is 2.5mV of ripple at 52KHz, which looks like a line of little spikes sticking up on the 'scope. Obviously adding more capacitance isn't going to improve it, so it's the ESR and inductance that need to come down.
 
Last edited:
When the voltage is turned down, the darlington control transistor conducts bringing down the output. However if there's no load, or a very light load, it will try to pull the voltage all the way down to zero until the output catches up. Now suppose the output has been turned up to 40v, or the high 30's at least, and then I decide to turn it down quickly. So now with the diode arrangement suggested by ron, there is a capacitor with 40v sitting on it, discharging through a diode and through the fully on control transistor. Obviously quite a lot of current can flow there, which will definitely destroy the transistor, hence the resistor is there to limit it. If you look at the datasheet for the BCX38C, and use the handy calculator here, you will see that with 220 ohms this keeps the transistor just barely inside its safe operating area as the voltage falls.
Looking again at that schematic, why don´t you put a resistor between the control transistor and the power pair to limit the current, and drop R124?
 
I wish to congratulate you on a very complete and challenging project.

Could you substitute the compound output transistor with a Mosfet?
You can find Mosfets with Vgs(max) of 20 volts, which would alleviate most if not all of your failures.
 
Hi Kubeek
After I discovered that the arrangement in the schematic doesn't really work very well, I started playing with other ideas in LTSpice, and I was experimenting with exactly the scheme you describe last night. Nice! The voltage on the control transistor's collector goes right down. The resistor just needs to be adequately rated.
 
I wish to congratulate you on a very complete and challenging project.

Could you substitute the compound output transistor with a Mosfet?
You can find Mosfets with Vgs(max) of 20 volts, which would alleviate most if not all of your failures.

I think it would work with pretty much anything up there to be honest.
I wouldn't call it "very complete" though. Still got to change some component values, couple of transistors to choose, small modifications like the one discussed here - bits and pieces.
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top