Hi all, I've designed a high-voltage discrete opamp which is supposed to take a negative DC voltage and simply double it. So if Vin is -23V then Vout should be -46V. The circuit is powered from a +30V supply and a -120V supply, each from their own winding. The circuit is fine in LTSpice, but has been giving me nothing but trouble in real life.

At first, the circuit was drawing way too much current. We fixed that with help from this forum by removing two bias diodes and replacing them with a short...this connection is now Rbias on the attached schematic. Now what is happening is the circuit outputs the positive rail - .7V and dominates the input voltage, forcing it positive (that's bad)! I measure the same 29.3V on top of R2, R3 as well.

I'm out of my element here using BJTs as amplifiers (I understand them as switches) and what's making this more confusing is that I can't get the circuit to not work in the simulator. Any advice on how to approach the circuit to find out what is going awry or does anything stick out as being wrong? Thank you!
< Schematic thumb

 

Post your circuit.

 

Hi Marley, it's there as a thumbnail at the end, just hard to see!

Mod: Enlarged the image.E

 

Which electronics software makes a schematic with such low contrast that it cannot be seen??
I added contrast:

 

That would be an old version of Eagle, lol! Noted, I'll export in black next time.

 

I like seeing young ladies dressed in pink, not schematics.

 

Could this be a latch-up issue? The +30V supply rises much slower than the -120V supply. I recall reading somewhere about an across the rails bypass cap (C1) causing problems.

 

Check your ground connection on the feedback. If Vout = 29.3V then the base of T2 should be around 14V.
How much voltage do you have across R1? If the LTP emitters are at 29.3V then how can you have enough voltage across the darlington to switch it on? Have you got all the transistors the right way round?

 

Check your ground connection on the feedback.

Good call on this. There's 2 of these circuits and found a mistake where the feedback connection was floating. I fixed it on one but was measuring the other. After that however, the circuit behaves the same.

The voltage across R1 was .7V...basically everywhere I measure is ~29.3V! I've also checked both the transistor orientation and made sure I didn't put a 6517 where a 20 should be and vice versa. No luck there either.

The voltage across R1 and your comment about turning on the Darlington make sense. In the sim, R1 has ~1.2V across it which would be enough to turn it on. I only used the Darlington connection because in the sim, it seemed like the more gain I had there in the VAS, the less current I would need in the output stage.

 

Put a 2.2k resistor across base-emitter of T4, just to make sure that it isn't being turned on by leakage current, either through T3 or across the pcb.
It would seem very difficult to get 29.3V on the output. The most you should possibly get is 28.6V, because there will always by 2x Vbe, one in the darlington and one in the emitter follower on the output. If you get 29.3V on the output, then something in the area of T3/T4/T5 is short.
You should now get 14.7V on the base of T2.
I suggest that you print out a copy of the circuit, and write on the voltage at every transistor pin. You can then see which ones look right and which ones don't.
Remembering this form the first time you posted it - what actually is the load on the output? Will it always be negative?

 

Would you mind posting the LTSpice's .asc file?

 

Put a 2.2k resistor across base-emitter of T4, just to make sure that it isn't being turned on by leakage current, either through T3 or across the pcb.

Tried this and same results. I also checked briefly for shorts but not thoroughly because I forgot which were in question...none so far. I'll report back in a bit.

Remembering this form the first time you posted it - what actually is the load on the output? Will it always be negative?

Ya, you are to thank for helping with this monster the first time! Thank you! You know, I wonder if what is happening now is related to the output stage drawing way too much current before? I still have a hard time with why LTSpice showed such a massive discrepancy.

The load is just the feedback resistors and a reversed bias diode to keep out voltages below the output...it's essentially a clipper.

 

Would you mind posting the LTSpice's .asc file?

Here you go. The little circuit in the box on the left generates a non-static negative DC input. I use that to show that the opamp will follow the input signal, but 2x. This file still has the bias diodes in that I referred to in OP.

 

To get the output you are observing, the output stage must be switched on (or shorted). In that case the darlington must be on (or shorted), and T1 must be on (or shorted). All four of these transistors (2 in the darlington) must have at least 0.55V across base and emitter, and that can't be true for the output voltage you are measuring. So something is wrong somewhere.

 

To get the output you are observing, the output stage must be switched on (or shorted). In that case the darlington must be on (or shorted), and T1 must be on (or shorted). All four of these transistors (2 in the darlington) must have at least 0.55V across base and emitter, and that can't be true for the output voltage you are measuring. So something is wrong somewhere.

So odd. I poked around on the transistors in question and found no shorts anywhere. I even checked all the package pinouts in the software to be sure I didn't make a mistake there and nothing.

If this was your circuit, how would you approach this? Replace all the transistors? Maybe start removing a stage at a time (output, VAS, diff input) to try to find where things are going wrong? Is matching very important with this circuit? Maybe get rid of the positive supply to simplify things? Since I originally had the problem with this circuit I got two books on the subject but its not clicking yet unfortunately.

 

1) I'd measure the voltage across all the Vbe junctions
2) I'd print out the circuit diagram and write all the voltages at all the transistor pins on it.
Matching hardly important at all. T1 and T2 are the most critical, but if they were unmatched, there may be an output error of 100mV at the maximum.
Do you have 680 ohm resistor in the collector of T2 by any chance?

Unfortunately, you can't remove stages, as they are all needed to complete the feedback loop.

 

Thank you, Ian0!! I'll do this right now. No resistor in T2's collector.

 

I doubt this helps but ya...
EDIT: T4's base is 29.35V

 

How are those 47k 2W resistors? Warm? (maybe they are open circuit, but I think the next suggestion is more likely)
What is the impedance of your reference signal? How has it managed to get to 28.4V?
What appears to have happened is that during startup the output swung positive (if the two supplies start up at different speeds then it can wander all over the place until things get stable). Current from the feedback circuit has gone through D2, and pulled the reference up to 28.4V, where it has got stuck.
The amplifier is now doing its damn best to make the output equal to twice the reference = 56V.

What connects to the output of this amplifier? And where does the other end of whatever connects to the output connect to?
Please draw the circuit that gives you NEG.VDC.IN

You might be tempted to try it without the protection diodes, but don't! You could try back-to-back 5.6V zeners (in anti-series, not anti-parallel)

 

looking at your voltage measurements in #19:

The voltage across R2 is 28.6 - -120 -> 148.6 volts
so the current through R2 is 148.6/47.5e3 -> about 3 mA

Where is this current coming from? T1 or T2?

Looking at the base voltages on T1 and T2:
T1 Vbe 28.46-28.6 -> -0.14 -> T1 is off
T2 Vbe 28.6-28.6 -> 0 -> T2? strange, base-emitter short?

The voltage across R1 is 29.35-29.08 -> 0.27 volts
and the current through R1 is 0.27/681 -> 396 uA
So the 3 mA for R2 isn't via R1.

The 0.27 volts across R1 is driving T3 & T4 but 0.27 isn't enough to
turn on even a single Si BE junction. So there's more mystery there.