Wandering bias on an amp.

I'm repairing a Sansu AU-417 amplifier for a local guy.

When it arrived in my hands, the amp wouldn't come out of protection. On the left channel there were a few burnt up resistors and a shorted output transistor - from what I've read the bias pots on these things go bad and overbias the output stage, and sure enough, the bias pot wiper measured intermittently.

Not wanting to risk counterfeits, I ordered OnSemi replacements for the transistors - MJE15034/5 drivers instead of the original 2SB537/2SD382, and MJL1302A/3281A output transistors instead of 2SA1105/2SC2580, all parts chosen to match the Ft and hFE of the originals as close as possible. Also replaced the bias transistor and the output protection transistor with the same part number and hFE bin. For no other reason than "everything should match", I changed all the same parts on the right channel.

Anyway, the amp behaves fine and does a great job playing audio, but the bias on the left channel wanders around. Not by much, but it is a bit unsettling:
https://www.youtube.com/watch?v=YMXjBTrey0o

Yet the right channel is rock solid:
https://www.youtube.com/watch?v=ptc0TjzIVGs

This measurement is a voltage measurement across both emitter resistors, Sansui specifies 20mV in the service manual.

Looking at the outputs on the amp with an oscilloscope I don't see any oscillation. Changing the output loading on the amp doesn't influence the bias current, so I don't think it's compensation/stability related. Bias pots are new, every part related to the bias circuitry measures OK, though I haven't measured the VA stage transistors which drive current through the bias transistor. Unplugging the audio input of the amp module doesn't change anything, so it's not audio rumble or anything influencing the bias.

And annoyingly enough, if I run the bare amp module off a +-25V lab supply which would let me easily probe the card circuitry, the bias is solid. I haven't reached the point of pulling every single part off the board and individually measuring it, but it may come to that.

Any suggestions?
 
EFs do have tendency for parasitic oscillations so that would be where my first suspicion would fall - especially given it goes away when merely changing the power supply.

I looked up the spec sheets of your original transistors but the ones I found didn't give much information about Fts, other than a typical. Looks as though your replacement transistors are most likely wider bandwidth parts, though not by a lot.

I'd dab on a couple of RC networks onto the wandering channel's output emitters.
 
How's the bias generated for the Vbe multiplier? Current source? Could that be unstable?

Perhaps measuring the bias current with an o'scope (differential probes or CH1 - CH2) would shed some light on what's going on.

I wonder if the board suffered damage when things fried.

Tom
Here's the schematic. The bias generator is tugged on in each direction by a sort-of current source, hard for me to eyeball it since I'm not 100% familiar with the operation of these things.

bC574io.png


Blown parts were TR707 (top output), R44 (it's emitter resistor), TR12/R36 driver and emitter resistor, R39/R32/TR09 protection circuit. TR05/TR07 got hot (board is discolored underneath) but test fine, and the board is similarly discolored on the right channel.

With R44 blown, output would swing full negative. FET01b would be off, FET01a would be on, so TR02 conducts and not TR01, turning on TR05 and pulling the bias cell circuit to the positive rail. There might be some current through TR03 (can't really tell by eyeballing it) which would partly turn on TR07, driving some current through those circuits.

I'll pull the resistors around TR05/TR07 and measure. Sadly I don't have a DMM that measures hFE or a curve tracer to measure those parts themselves, I suppose I could jig something up with a dual bench supply, couple of DMMs and some resistors.
 
Update: all resistors around TR03 through TR07 measure fine in-circuit with the board hauled out and unpowered. C12/C13 measure fine on a LCR meter.

Maybe one or more of the pF sized ceramics is damaged, but I think that'd show up as instability instead of wobble. I'll start soldering wires to nets of interest and probing things tomorrow, and see about measuring TR05/TR07's hFE.
 
r23 and r25 are fusibles, notorious troublemakers in Sansui, and the corrosive glue really eating componentleads
Surprisingly those resistors are fine, from what I've read it's the first thing that fails on most of these amps.

Thankfully Sansui didn't use much glue on these cards, bit of corrosion on a few component leads but nothing catastrophic. Got all that glue cleaned off now.

Gonna attempt to gear up a +-50V supply and continue with bench testing. Everything looks great at +-25V, which is the limits of the supply on my desk.
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
If it is oscillation (and it could well be) then it often changes dramatically if you touch or even go near components around the problem area. It might also be worth soldering wires to the measurement point and trapping those firmly into your DVM terminals.

Another angle could be that the DC offset is altering and causing the apparent fluctuation(you mention differing loadings... how about with no load ?). Could there be another issue that caused the original failure in the first place such as a problem with the offset trim pot, a dry somewhere (VAS) or an intermittently failing VAS transistor or its current source above. These will run hot and so could be likely suspects.
 

anatech

Administrator
Paid Member
2004-06-06 8:31 pm
Georgetown, On
Hi gmarsh,
I would replace the transistors that got hot, same for resistors. They can still measure good but be bad. I've seen junctions open up intermittently with heat. The important thing to accept is that those parts were operated outside their SOA and are therefore highly suspect.

Matching the transistors in diff pair arrangements always bring improvements to the operation of the circuit. It's a real pain to match them, but the result of this is very satisfying.

To match transistors, it is critical that the device temperatures should be exactly the same. A small difference in temperature will change the Hfe. You have to use tweezers to position the transistor for test. Make up a diff pair on some bread board and put the test transistors in there, tied together and covered to isolate them. Measure the two collector voltages (meter goes between them). The closer matches will have the least voltage between them. I use 100R resistors for collector loads, about 10 VDC, and a CCS to generate the tail current. 1 ~ 10 K resistors connect the two bases to common ground. You need a negative voltage for the tail current source.

Good luck with the repair. You're almost done.

-Chris

Edit: Mooly pointed out something I assumed was true. Your speakers or loads must be disconnected from the amplifier if doing bias current adjustments. Of course that would mean that you have a voltage instability, and my advice would still remain the same.
 
You say powering the card on the bench and it is rock steady? So perhaps instead of the problem being on the card, it is in the delivery of power supply to it. Bad connection somewhere, bad solder joint at the other end of the supply wire, etc. And don't ignore common/ground.

never hurts to check, but I tend to doubt measuring part values will find the wander. If a part being off value was causing issues, I'd expect them to be constant.
 
So I stayed late at work, series-ed up a couple bench supplies and made +-50V. Here's the findings.

The amp has a separate ground for the output zobel. With this not hooked up, the amp oscillates, putting out about 2Vpp at ~6MHz. I'm not quite sure that a zobel should be required for stability with no load present, so I should probably recompensate the amp.

Bias wanders around as before. Measuring across R23 through R25 seems to show that the current running through that stage is wandering around too, making me wonder if there's some sort of strange effect going on there. I even replaced R23/R25 with fresh parts (the old ones were looking a bit brown, even though they measured 150 ohms) and replaced C8/C9/C10 in case any of those were overheated and changed value... and absolutely no difference.

Never got a chance to measure TR05/TR07's hFE.

What bugs me is the other stage is built identically to this one and works perfectly fine. I've compared every part value between the two and can't find any difference, and measured every part I can and found nothing different. Pretty much all that's left right now is throwing more semiconductors at the thing.
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
Oscillating with no Zobel is perfectly normal for many amps. Measuring the hFE won't reveal much, the possible problem with transistors shows under use when working voltage appears across them.

Gather clues. Does the reading fluctuate as you touch any parts in the output stage. That could be a sign of oscillation.

Give that offset trim preset a few turns just to see if changing offset reflects in a small change in bias. Do that with no load. Prove that it has (or has no) effect.
 

anatech

Administrator
Paid Member
2004-06-06 8:31 pm
Georgetown, On
Hi Mooly,
Measuring the hFE won't reveal much, the possible problem with transistors shows under use when working voltage appears across them.
I suggested this to improve the amplifier. The match would not create a wandering bias if they were mismatched.

Hi gmarsh,
Was there a load connected when you ran your measurements? Otherwise Mooly's comments are true. The Zobel must be connected properly, or the amplifier will oscillate. Also, be certain the the heat sink is grounded as well, or the amp may oscillate as well if it's floating.

-Chris
 
Hi Mooly,

I suggested this to improve the amplifier. The match would not create a wandering bias if they were mismatched.

Hi gmarsh,
Was there a load connected when you ran your measurements? Otherwise Mooly's comments are true. The Zobel must be connected properly, or the amplifier will oscillate. Also, be certain the the heat sink is grounded as well, or the amp may oscillate as well if it's floating.

-Chris
No load connected, but zobel hooked up during testing. The heatsink is isolated from the board ground on the module itself, but they're electrically connected once installed in the amp chassis.

All signs point to some sort of low frequency instability in the VA stage, causing the current driven through the VBE multiplier to wander around. I can't measure anything on the output of the amp so it doesn't appear to be caused by global NFB, and the global NFB seems to be doing a great job keeping whatever is happening in the VA stage from showing up on the output.

I was thinking hFE testing would indicate a transistor that's damaged/out of spec. Transistors are cheap, I'll just replace everything in that stage and see what happens.

Is the quasi-differential VA stage they have going on here a common thing, or is it one of those odd 70s-80s hifi experiments that shouldn't have happened? I'm used to seeing either a single differential amp and VA transistor in simpler amps, or everything doubled up like in a Leach amp, but I haven't seen this kind of design before.
 

anatech

Administrator
Paid Member
2004-06-06 8:31 pm
Georgetown, On
Hi gmarsh,
I think it's more likely to be a noisy bias control, or possibly some HF osc. Whenever you run the amplifier, ground the heat sink to the amplifier board ground. That will prevent some oscillation for you.

This input stage is not uncommon. It can work quite well. As for replacing the transistors, try to match the diff pair, don't touch FET01, you can't get them easily anymore.

-Chris