Lab Power Supply Design / Build

[akis]

quick question, short the oscilloscope at the sensitivity you have chosen, what do you get on the screen?

[MJL21193]

Flat line. A little meat, about half the thickness of the line in the last pic of post #99

[akis]

I have run 1,000 simulations and yes there are oscillations. Very hard to control. It depends where they come from. If however you are talking about ripple then of course there will be a ripple shape at the output, and you'd need to trace it to the source(s). it may be for example you're drawing too much power out of the transformer, or the filter caps are too low and the base ripple level touches the output level creating kinks... Also the quality/stability of the voltage feeding the base is important. If I remember correctly you are feeding the pass transistors from some rudimentary voltage with no regulation.

[MJL21193]

Originally Posted by akis

I have run 1,000 simulations and yes there are oscillations. Very hard to control. It depends where they come from. If however you are talking about ripple then of course there will be a ripple shape at the output, and you'd need to trace it to the source(s). it may be for example you're drawing too much power out of the transformer, or the filter caps are too low and the base ripple level touches the output level creating kinks... Also the quality/stability of the voltage feeding the base is important. If I remember correctly you are feeding the pass transistors from some rudimentary voltage with no regulation.

Honestly, if this is your best assessment of the circuit and the data I just presented, please save it.
You obviously don't get it.

[megajocke]

I'd rather suspect a problem with the voltage regulation loop than your current limiter. It looks like the global loop crossover frequency is as low as about 25 kHz so I'd suspect local oscillation. Perhaps it's the three transistors in the regulation circuit that are acting up in their local loop through the global loop compensation network. It could also be the darlington output stage that dislikes the 10 nF capacitor across the output.

I'd probably redo the voltage loop compensation to go to the base of the "VAS" transistor instead of the LTP input to make sure the regulation circuit itself is well behaved and make the loop crossover frequency a couple of hundred kHz.

It may also be easier to get the thing stable if you use an electrolytic capacitor on the output. A couple of hundred microfarads is often used. As long as the ESR zero (at a couple of tens of kHz) is well below the loop crossover frequency it should be very stable and have low output impedance up to high enough frequencies that it won't matter any more.

BTW, did you set your scope Y amplifier to variable uncalibrated gain on purpose? I've got it's 150 MHz little brother 2445 here. These are pretty nice instruments!

edit: I worry a little that the current sharing of the output stage may not be too stable. It looks like you have your two parallelled output transistors on different (and pretty small) heatsinks and also have small emitter resistors. I fear it could run away to a situation where one transistor does all the work at low output voltages and mid to high current. Do you know the thermal resistance of your heatsinks when mounted in the box that way? The stability can be calculated. If you want to just try it be ware that it could take a pretty long time for the instability to develop.

[MJL21193]

Originally Posted by megajocke

I'd rather suspect a problem with the voltage regulation loop than your current limiter.

It may also be easier to get the thing stable if you use an electrolytic capacitor on the output. A couple of hundred microfarads is often used.

Ok, I tried several things - first I'll re-post the schematic so as we will know what I'm talking about:

LAB SUPPLY SCH.png

I changed C7 to 1nF - slight improvement. I dicked around some with other things then I changed C2 to 1nF(this is big, I know but I wanted to be sure) - bigger improvement, nearly gone. I then added a 220uF electro to the output and viola! Is gone!

So, here is the output with no load at 15VDC:

NPX_988.JPG

And here is the output with a load drawing 1.8A at 15VDC:

NPX_985.JPG

Did someone say ripple?

If I remember correctly you are feeding the pass transistors from some rudimentary voltage with no regulation

[MJL21193]

Originally Posted by megajocke

BTW, did you set your scope Y amplifier to variable uncalibrated gain on purpose? I've got it's 150 MHz little brother 2445 here. These are pretty nice instruments!

TBH I don't know what I have going on with that scope - it's been nearly a year since I used it and I haven't had a chance look it over yet.
Yes, it is a pretty good scope and it didn't cost a fortune - a good deal actually.

Originally Posted by megajocke

I worry a little that the current sharing of the output stage may not be too stable. It looks like you have your two parallelled output transistors on different (and pretty small) heatsinks and also have small emitter resistors. I fear it could run away to a situation where one transistor does all the work at low output voltages and mid to high current. Do you know the thermal resistance of your heatsinks when mounted in the box that way? The stability can be calculated. If you want to just try it be ware that it could take a pretty long time for the instability to develop.

So far, the current sharing seems fine and I have had these sinks hot. With the cover on, it may get a little too warm but I'll keep my eye on that. The case has holes in the bottom under the sinks and the cover has holes directly above. It'll have a bit of a chimney effect, some convection to speed cooling.
If need be, I have space between the sinks and the back panel for a pair of muffin fans. Last resort though - that scopes fan is loud enough without adding another 2.

[MJL21193]

By popular demand, another look. Less blurry pics and a bigger hoop to jump thru
Got to try out the internal dummy load as well.

30VDC output, no load:

NPX_991.JPG

30VDC output, 16 ohm load (~1.8A):

NPX_993.JPG

The subject:

NPX_996.JPG

This is outstanding IMO. I really couldn't have hoped for better.
Ripple is non-existant.
Ultra low noise.
Next to no voltage drop with the load connected.

Sweet!

PS: Thank you Mega for your very effective help - you saved me again!

[MJL21193]

Extras, in case someone out there in TV land thinks my scope doesn't discern ripple.
Output of the doubler supply, NO load on output:

NPX_998.JPG

Note the v/div.

Output from the raw supply (at the main smoothing cap) with a load - 30VDC @ 1.875A:

NPX_999.JPG

I measured voltage drop with the load connected - it went from 30.29V to 30.13V. Not too shabby.

[MJL21193]

Originally Posted by megajocke

Do you really want the weak foldback characteristic given by R21 or is it perhaps tuned to counteract the early effect in the current limit transistor?

Going back to this from page one, I'm beginning to wonder if this is the cause of the instability. Looking at the function, R21 serves 2 purposes: First it (along with R20) provides a dummy load for the output so that it can be adjusted down to zero. Second, it supplies positive feedback to the emitter of Q6. The reason for this is that when in constant current mode (current limiting), as the output voltage falls the additional drive necessary to turn on Q6 harder is supplied via its emitter, through R21. This prevents an increase in output current.
I will have to investigate further.

Another thought: while added to improve stability, what does the 220Uf cap across the output do to the transient response? Is transient response a critical factor for a supply like this? I could rig up something to measure it - a couple of resistor loads and a MOSFET to switch one in parallel with the other, thereby switching the current draw from low to high. This could be driven with a squarewave from my FG. Maybe I'll give it a try.