Tangent Young-Jung power supply

[tangent]

Now I'd start poking around the CCS area. It's beginning to look like it might be misbehaving for its own reasons, not just indicating problems in the surrounding bits.

Q1 might not be a transistor any more, for instance; it's rated for 50 V, but a small inductive spike on top of that when turning it off for the first time could have pushed it over the edge. Use a diode tester to see if the EB and CB junctions still look like diodes. If there's any doubt, remove Q1, then power it up again to see if D2 lights. If it does, Q1 was probably fried, and explains everything...no CCS, no regulator.

My second best guess is that your use of nonstandard components has something to do with it. You didn't get "creative" with the R3 or R4 values, did you? You've checked for shorts or opens with all those free-range components?

[schubert]

I don't have a diode tester - I assume you mean something that will test whether current will go only one way through a diode. Could I rig one using a 9V battery, LED and a resistor?

Anyway, I just went ahead and removed Q1 and powered up. Sure enough, D2 lit up, albeit dimly. Large voltage drop over the LM317 resulted. I replaced the Q1 with a new transistor, but it didn't do any good - I'm back where I started.

Originally Posted by tangent

My second best guess is that your use of nonstandard components has something to do with it. You didn't get "creative" with the R3 or R4 values, did you? You've checked for shorts or opens with all those free-range components?

I definitely try to avoid being creative, especially with a circuit like this. I had already thought of the possibility that I somehow got the wrong resistors in R3 and R4. I pulled R3 and checked, and both R3 and R4 are spot on. I've also checked all the rest of the resistors, and they're OK too. I've been over the rest of the circuit too, and there doesn't seem to be any connectivity problems.

The only parts besides the hodgepodge of resistors (which are all 1% metal film) that are not right off the parts list are C9 and D2. I ordered the quantities specified on the list w/o checking the schematic (having missed a forum warning about discrepancies) and ended up with 4 C5's and not enough 100uF for C8-C11 (somehow I missed C11 when ordering). So, C9 is an axial electrolytic 100uF cap from Radio Shack. Somehow I doubt that's the problem, though.

So that leaves D2. Could the wrong LED be causing the problem? To be truthful I don't remember what I did when I ordered, but I do remember putting an ordinary diode in that spot. (The recommended part is currently non-stocked at Digikey and has a 5000 piece minimum - and it's listed as "obsolete" at Mouser, though a "638" replacement version is available). Upon rechecking (and before I ever powered up) I realized I needed an LED there, so I removed the diode and put in a green LED from RS. Before putting it in I hooked it up with a 9V battery and a resistor and measured the voltage drop. It was very close to 2V, so I thought it would be OK. (Yes, I suppose this does qualify as "getting creative" ) Have I made a raw newbie mistake? In a way I sure hope so, because the relief of getting this thing working would far outweigh the embarrassment!

[schubert]

I replaced the D2 LED (and C9) with the recommended parts, and (no surprise, I suppose) there was no change - still ~1.56 volts over the LM317. The LM317 is healthy - I took it out and tested in the following circuit:
yhlmccs.gif
The voltage drop over the resistor was 1.24V, about what was expected.
I also tested all the transistors for conductivity using a 9V battery, 1K resistor and an LED. All tested identical to unused transistors right out of the package - that is, the PNP versions conducted as long as the negative was connected to the base, and the D44 was opposite. I lifted R6 before testing Q1, and removed the D44 to test it.
Here are the voltages:
R3: .617V (+adj to out-) (.617mA)
R4: .513V (+adj to D44-) (.622mA)
I think these may be normal - they add up to the current expected with a 1K resistor in the adj-out loop of the LM317.
D2: 1.241V
R5: 38.2V (1.91mA)
R6: .513V (+TP5 to Q1-) (2.06 mA)
Q1: 7mV (+E to C-)
Q1: 726mV (+E to B-)
Q1: 723mV (+C to B-) (yeah, they don't add up - the numbers fluctuate a little)
D44: 1121mV (+C to E-; 539mV over CB, 603mV over BE)
output: 38.5V (+TP6 to gnd-)
R8 is still disconnected.
I'm tearing my hair out over this one. Anybody have any bright ideas about what to try next?

[tangent]

At what points have you replaced the op-amp? Right now, it may be dead due to the high output voltage, but if you replace it and the rest of the regulator is fine, as seems to be the case, it shouldn't die because it's going to maintain proper regulation.

[schubert]

These voltages are all with the op amp removed. I was concerned because I wasn't getting the ~2.3V drop across the LM317 I was expecting (though I do get ~2.1V from the input pin to the adj pin). I have a new op amp - should I go ahead and put it in? I haven't done so yet because I didn't want to fry another one. Besides, I still have 38.5V across the op amp rails.
Without the op amp, is the 1.1V drop over the D44H11 OK? I thought it should be more.
Also, D2 is still not lit - I assume it should be.

[tangent]

I'd investigate that D2 failure first.

Also, when you do decide to put the op-amp back on, I'd at least temporarily wire the transformer secondaries in parallel to drop the output voltage to something that can't harm the op-amp if the regulator continues to not be a regulator. Only revert to your current series configuration once you've got the circuit behaving sanely.

[schubert]

Well, I rewired the transformer for parallel and tried again - same result. Now I'm getting around 20.4V unregulated, the same 1.5V drop over the LM317, and 0.8V over the pass transistor rather than 1.1V. D2 still does not light, though it's physically fine.

Then I put a new AD825 in, and it made no difference except that D2 lights for about a second and then shuts off (as it did at first with the 70053 transformer). There's now only 20V or so across its rails, so I'm pretty sure I haven't fried it. I also replaced Q1 (again) and the D44H11 to no effect.

A question about the LM317 - should there not always be a 1.25V drop across R3 (LM317 adj and out pins)? Isn't that what an LM317 is supposed to do? I'm only getting around 0.6V, but if I take the same LM317 out of this circuit and put it in a simpler one, I get the 1.25V drop just fine.

I'm trying to gain some understanding of this circuit. As far as I can see the output voltage is determined by controlling the current at the base of Q2. With no op amp present, D2 doesn't light because the current that would usually light it is going through Q1. There's no reason for it not to, since the base is at a lower voltage than the emitter. I'm guessing there's also some EC current, since there's minimal voltage drop over Q2.

When the op amp is added, now there's current from D3, which heads to the Q2 base pin and controls Q2's conductivity. This also opposes current from Q1, only temporarily in my case, since current is forced through D2 for about a second. Ideally this should be steady-state (more or less) and D2 should remain lit, but in my case the current from D3 is being stopped and current from Q1 is leaving Q2 open, and I get a very small voltage drop, which is unaffected by VSET.

I need to figure out why current is getting through Q1 when it should be going through D2 (at least when the op amp is installed). The Q1 I just took out had an hFE of around 350 (the data sheet said 250 min) and the one I put in was around 650, so I think they're both OK.

Does this analysis sound OK, or am I totally off track? As you can tell I'm obviously no electrical engineer, but I'm trying to learn as much as I can. Thanks to tangent for putting up with my ignorance, and to anyone else who might be able to help.

[tangent]

Originally Posted by schubert

Then I put a new AD825 in, and it made no difference except that D2 lights for about a second and then shuts off

I know of nothing in this circuit that could do this other than the protection circuitry in the op-amp itself. Everything else should act more or less instantly. Protection circuits, though, cut in after a short time and stay clamped, until either power or the badness goes away, depending on how they're designed.

There's now only 20V or so across its rails, so I'm pretty sure I haven't fried it.

Yes, if nothing else, this exercise has removed one source of FUD.

A question about the LM317 - should there not always be a 1.25V drop across R3 (LM317 adj and out pins)? Isn't that what an LM317 is supposed to do? I'm only getting around 0.6V, but if I take the same LM317 out of this circuit and put it in a simpler one, I get the 1.25V drop just fine.

Yes, you understand it correctly, but I can't offer any enlightenment about its behavior.

With no op amp present, D2 doesn't light because the current that would usually light it is going through Q1.

No. If Q1 is still a PNP transistor, there's something on the order of 1 uA coming out of its base, and going down to ground through D2's ballast resistor. (The exact value is Ic / hFE.) Current flows exactly opposite to the way you propose.

the base is at a lower voltage than the emitter

I think you're assigning FET-like behaviors to a bipolar here. Bipolars are current-controlled, not voltage-controlled. It doesn't matter whether the emitter is above the base or below it. What matters is the current going into or out of the base, and what's going on around the transistor. I wrote up how this CCS works in the op-amp linear regulators article.

Remind me: with Q1 out, D2 lights? What about when only Q2 is out of the circuit?

Is D3 still a zener?

Does removing Q3 and Q4 help?

Have you checked for PCB trace breaks or manufacturing errors?

[schubert]

I read the linear regulator article, though complete understanding is yet a little way off.

Originally Posted by tangent

Remind me: with Q1 out, D2 lights? What about when only Q2 is out of the circuit?

Yes, D2 lights dimly - not nearly as brightly as it does (temporarily) with the op amp in. I thought about powering up without Q2, but I wasn't sure it was safe or advisable. I'll try it.

Originally Posted by tangent

Is D3 still a zener?

No clue. I know what a zener is and what it does, but not how to test it. I'll read the data sheet and try to figure out a way to see if it's still functioning. To tell the truth I didn't realize D3 was a zener - is it reversed biased when the unit is functioning normally?

Originally Posted by tangent

Does removing Q3 and Q4 help?

Haven't tried that either, mostly because the basic problem is there when Q3 and Q4 are out of the circuit (that is, with no op amp and R8 lifted).

Originally Posted by tangent

Have you checked for PCB trace breaks or manufacturing errors?

I've checked part of the board, but I haven't gone through it systematically since I've been focussing on the parts. I'll do it. It's the only thing left that makes sense, since the parts seem to be OK.

I'm also going to lift one end of R4 and see if I get 1.25V over R3.

Originally Posted by tangent

If Q1 is still a PNP transistor, there's something on the order of 1 uA coming out of its base, and going down to ground through D2's ballast resistor.

That's what I meant to say - current is draining out of the base of Q1 and to ground through R5. With no op amp, this current has to come from R6, doesn't it? I didn't think it could come from the base of Q2. If it's coming through R6, then it's bypassing D2, which then doesn't light.

[tangent]

Originally Posted by schubert

Yes, D2 lights dimly - not nearly as brightly as it does (temporarily) with the op amp in.

Something to investigate, then. There's apparently some sneak path changing the current through the LED. Normally there should be only one thing controlling that, R5, which is constant. If somehow the other path were through the op-amp output, it could be a surge of current that makes the LED very bright until the op-amp output shuts down due to overcurrent.

I've stared at the schematic and no obvious path for this pops out at me. Every theory I come up with makes no sense, if you assume no shorts on the PCB and the components are all functioning.

I know what a zener is and what it does, but not how to test it.

Just check the voltage drop across it.

In fact, why not check the V drops everywhere? Print out a copy of the schematic, write down all the voltages you see relative to TP3 on every straight line on the schematic. For instance, between Q1 and Q2, you have several choices of where you can take the measurement: Q2 base, Q1 collector, D3 cathode, and + side of C8. Check at the easiest one to get to in each case, so you reduce the chance of slipping with a probe and blowing something else up, invalidating the test. In this particular case, it's probably the D3 cathode.

Then, go through and calculate all the voltage drops across each pair of nodes.

I'd confine myself to just the second page of the schematic. Everything from TP4 back seems to be behaving sensibly.

If you study the result and nothing pops out at you as a cause, scan it in and post it here.

D3...is it reversed biased when the unit is functioning normally?

Yes. 6.8 V.

current is draining out of the base of Q1 and to ground through R5.

Yes, but again, it should only be a tiny amount, in the single-digit microamp range, not enough to affect anything materially.

With no op amp, this current has to come from R6, doesn't it? I didn't think it could come from the base of Q2. If it's coming through R6, then it's bypassing D2, which then doesn't light.

Current goes down both paths. It's not either/or. There is no switch between R6 and D2, causing current to take one path or the other. The current through R5 is almost entirely that through D2, plus a tiny bit of leakage through Q2's base. What current passing through R6 doesn't leak through Q2's base goes out Q2's collector.

[rds]

When can we expect to see more boards available?

[tangent]

I'm holding off on finalizing the v1.2 design until I get comments from one other prototype builder, at least. I'd also like to get schubert's issue resolved before going to production, too, in case there's a wider problem to be dealt with, rather than just some one-off issue. On top of all this, I have other projects I have to get to first.

All of this together means final boards are no less than a month away, probably farther.

[grenert]

One more report of a successful YJPS v1.1 build. Mine is a rock-steady 12V. I used a 15+15V transformer with parallel secondaries. Part selection for the resistors was:
R10=499R
R11=348R

Unloaded, I had 22mV RMS ripple after the filter caps. I have no way to measure ripple in the later stages.

Thanks to Tangent for another great design!

[schubert]

Congratulations grenert. I hope to be able to post the same message soon, though my success will be through a very circuitous route!

[schubert]

Originally Posted by tangent

Current goes down both paths. It's not either/or. There is no switch between R6 and D2, causing current to take one path or the other. The current through R5 is almost entirely that through D2, plus a tiny bit of leakage through Q2's base. What current passing through R6 doesn't leak through Q2's base goes out Q2's collector.

I realize that it will head both directions - I was mistaken about how much current Q1 would let pass through its base. Without the op amp I had 1.91mA through R5 and 2.06mA through R6, so I figured Q1 was a wide open gate (Which I guess it can be, E to C) and D2 wasn't getting enough current to light (only 1.24 volts across it). Since removing Q1 and thus restoring that microamp current was enough to light it dimly, that lends support to the idea of a rogue path to ground, probably through R5 - otherwise where is R5's current coming from?

With the op amp there is only 0.9mA through R5 and almost exactly 1mA through R6 - still indicative of a leak somewhere.

Originally Posted by tangent

Just check the voltage drop across it.

I already had checked the voltage across D3 (18V), but I'm not sure it will reveal much since the op amp has pretty much shut down and there's no current through D3 (no voltage drop over R7). I checked the resistance of D3 and it behaved the same as an identical uninstalled part - around 470 ohms in one direction and unmeasureable in the other.

Originally Posted by tangent

In fact, why not check the V drops everywhere? Print out a copy of the schematic, write down all the voltages you see relative to TP3 on every straight line on the schematic.

I did this with the op amp removed and posted it earlier (as a list), and I've redone some with the op amp back in. The only thing that stands out that may be significant is that the 2.3V drop I should be seeing TP4-TP5 is being scrupulously maintained over TP4-TP6 instead. I noticed the same thing with the other transformer setups as well, except with the higher voltages it was around 2.6V instead of 2.3V. It didn't click with me until I put the opamp back in and noticed that even though the voltages over R3 and R4 changed by around -30% or so, TP4-TP6 remained rock steady at 2.3 V (1.5V TP4-TP5 and 0.8V TP5-TP6 in each case). LM317 must be confused somehow about the resistance between its out and adj pins - but not nearly as confused as I am! Any ideas?

I'd like to find that 1.25V drop the LM317 is supposed to maintain - it must just be in a different place than I think it should be (that is, across R3). I thought maybe it could be that 1.24V drop across D2...but with the op amp it's 0.977V.

There doesn't seem to be any trace breaks that I can find. I'll start looking for sneak paths - this will be a lot of trial and error.