Very low frequency oscillations from audio amplifier

[Ionforbes]

Hi all,

A few months ago I built an audio amp and it had RF oscillations; those were quelled, and based on the feedback I got from here I've built a new one  .

This one is also oscillating, but it's very low frequency, like a couple of Hz, and I can't at all figure out where it's from. I've tried various capacitor values/locations, but I don't think this is a compensation issue or anything to do with the sziklai output simply due to how low frequency the oscillations are, and the fact they only start when the supply voltage is increased above 25V or so. They also have a very odd shape. I should point out there are ferrite beads on both of the power transistor's base leads, not shown in the diagram.

Has anyone seen a similar phenomenon or know what could be causing this? It just seems like an unusual way for it to oscillate, at least in my very limited experience

Thanks, Ion

[TimFox]

Is the repetition rate a subharmonic of your power line (50 Hz, I presume)?
I once had pickup of 10 Hz in an audio amplifier under measurement that came from my soldering iron that was idling at a subharmonic of 60 Hz (presumably from a zero-crossing switch in the control) that coupled through the power line (same outlet strip) through the power supply of the amplifier.
This can be checked by setting the 'scope trigger to "Line".

[Gyro]

I would scope what your supply rail is doing. 1000uF isn't a great deal of capacitance. Oscillation frequency that low implies a large time-constant feedback path. Your input is biassed to half of the supply rail via R1 and R2, so if the supply rail moves then so does the input, constrained by the reservoir capacitor and also, the input capacitor C2.

[Ionforbes]

It's powered from a bench supply at the moment (the oscillation peaks max out the current limit) and the frequency is 2.8Hz, so doesn't seem to be caused by line frequencies. I'll try fat a bybass cap at Q12 collector.

[Ionforbes]

Sadly no, not even 4700uF right at the colldector made a difference.

[bdunham7]

I think your problem is back at R5/C3.  Can you explain those values?  Try changing C3 to 10uF and see if your oscillation changes.

[Ionforbes]

R5 is sized to give a gain of 7, C3 is fairly arbitrary, just reuse of the input 22uF. It's getting late here so I'll have to try tomorrow, but why would too high a value cause oscillation?

[Kleinstein]

It's powered from a bench supply at the moment (the oscillation peaks max out the current limit) and the frequency is 2.8Hz, so doesn't seem to be caused by line frequencies. I'll try fat a bybass cap at Q12 collector.

If the current limiting engages the power supply may be a large part of the oscillation. There is the chance to supply may be setting the frequency.
The circuit would a single supply can be a little sensitive to supply variations.

[Conrad Hoffman]

C3 is the only part that could give a time constant like you're seeing. It also sets the point where the feedback is 100% and is normally far larger, like 220 uF or more. As above, I'd change the value and see what happens. FWIW, it makes me uncomfortable that there's no isolation between the current sinks, but I don't know if it matters.

[jimmc]

If it's not the power supply, then could it be thermal?

The dissipation and hence temperature of Q9 will vary with collector (~ output) voltage leading to Vbe change.
Vbe change (Q9) changes base current causing slight change in bias voltage (D1,2) which then feeds back into the input stage.
Probably not a large effect but there is a fair amount of gain about.
Try heatsinking Q9 and see if anything changes.

By the way, the Vceo of a BFW32 is 30v so you're pushing things a bit.

Jim

[magic]

Something dodgy about the output stage, perhaps?

Note that Q10,Q11 have no bias on them other than base currents of the output devices.
What's the output stage quiescent current?

[Doctorandus_P]

I'm a bit guessing here...

Q5 has a low base current.
R6 and R5 form a time constant of half a second with C3

R3 is not an ideal current source, which puts a part of the power supply ripple over D1 and D2, and thus also influences the current sinks Q3, Q7, Q9.
(But you should not have much ripple from that 35V battery... )

I'm guessing that changes in the voltage over D1 & D2 changes the current through Q3 which causes slight shifts in the biassing of Q5, and this somehow creates a feedback loop.

Time constants are determined by RC combinations. (Duh).
If you change the time constant and you see the frequency of your oscillation shift, then you're probably close to the problem.

[StillTrying]

FWIW, it makes me uncomfortable that there's no isolation between the current sinks, but I don't know if it matters.

I was surprised that there's no decoupling on that 1.2V.

[Terry Bites]

Mortorboating?
LF oscillation may occur when LF there is feedback via the smothing caps. I'd be looking at the PSU, speaker wiring and local decoupling arrangements.

[bdunham7]

R5 is sized to give a gain of 7, C3 is fairly arbitrary, just reuse of the input 22uF. It's getting late here so I'll have to try tomorrow, but why would too high a value cause oscillation?

I'm not saying that the problem is that 22uF is too high, in fact it may be just the opposite.  It is a troubleshooting thing.  If subbing in a 10uF cap increases the speed of your oscillation and changes the onset voltage, you know your problem is there. 

Another issue that has been mentioned is that your input section bias is derived by simply dividing the main power supply voltage.   This will give you poor PSRR, and audio amplifiers like this are usually capable of being powered by unregulated raw bridge rectifiier + filter cap supplies because they should have very good PSRR.   So you might want to just look at a schematic or two from an old 70's receiver and see how they set that up with a separate regulated supply for everything but the last stages of the power amp that are within the feedback loop.

With that much amplitude on your oscillation and a gain of only 7, you should be able to pick it up with your scope if you can simultaneously test the output and the bases of Q1, Q5 and Q6.

[Doctorandus_P]

I once read through the two books below, but otherwise did not much with them.
I don't know anymore which I liked best.


Audio Power Amplifier Design Handbook
Douglas Self
ISBN 0 7506 56360


Designing Audio Power Amplifiers
Bob Cordell
ISBN: 978-0-07-164025-1
MHID: 0-07-164025-8
The material in this eBook also appears in the print version of this title: ISBN: 978-0-07-164024-4,
MHID: 0-07-164024-X.

[Conrad Hoffman]

IMO, those are the go-to books on power amp design. They're different and emphasize different things. Both are essential if one is serious about this stuff, but obviously a lot of power amps were built before they were ever written. Some even worked! They don't spend much time on single supply designs, but there are lots of commercial examples to study. Almost all the lower power receivers used single supplies and capacitor coupled outputs back in the day. Once you start seeing current sinks and differential inputs, bipolar supplies become the rule. I'm guessing that if they were going to bother with higher performance circuitry, they'd also want to get rid of output coupling caps, so went with bipolar supplies. Obviously you don't have to, but it makes a certain amount of sense.

One more thought on the above circuit- you'll usually see power applied at the top right, to the outputs, with a resistor isolation the low level part of the circuitry, and a moderate filter cap to ground on the low level circuit side of the resistor. No matter how good the supply is, the output current will tend to modulate it. (I think this was mentioned in one of the posts above.)

[Ionforbes]

Thanks everyone for your replies. I subbed the 22uF with a 100uF and sure enough the oscillation slowed accordingly so this was definitely part of the problem. Then as a sanity check I built another and.....it worked fine, no oscillations! Maybe a defective component? I pulled every transistor from it and they all tested OK 

[bdunham7]

Thanks everyone for your replies. I subbed the 22uF with a 100uF and sure enough the oscillation slowed accordingly so this was definitely part of the problem. Then as a sanity check I built another and.....it worked fine, no oscillations! Maybe a defective component? I pulled every transistor from it and they all tested OK 

If you measure the various transistors' gains, preferably with a curve tracer, I'm betting that you'd see a significant difference somewhere.  You may or may not see it with a simple hFE test.  A common issue in amplifier design is ending up with a design that is intolerant of component variations.  You might try leaving everything else alone and just changing the overall gain on the defective amplifier to see what effect that has.