Ashdown MAG300 oscillation

[glebert]

The output on this amp is going crazy. Probing the power amp module with no signal it is clean up until I get to the Vbe multiplier section at TR15. The oscillation starts at any point of contact with TR15, including the collectors of TR7, TR12, and then through the whole output section. I replaced TR15 but it didn't change anything.

Schematic can be found here: https://music-electronics-forum.com/...ad.php?t=38478 Note this is a MAG300 but the PSU/power amp board is a MAG250.

Any thoughts? I got this amp for free and am mostly just using it as a learning opportunity.

Thanks,
Greg

[nickb]

Do you see oscillation on the speaker out?
What if you turn the master to zero?
Does it still happen if you have so speaker connected?
What are the characteristics of the probe you are using?

[Helmholtz]

C1 (47µ) O.K.?

[J M Fahey]

The output on this amp is going crazy. Probing the power amp module with no signal it is clean up until I get to the Vbe multiplier section at TR15. The oscillation starts at any point of contact with TR15, including the collectors of TR7, TR12, and then through the whole output section. I replaced TR15 but it didn't change anything.

It-is-not-a-tube-amp

You have no "separate" gain blocks ("tubes"), the whole power amp is a single gain block, a giant Op Amp.

What is the actual problem you are trying to solve?

[glebert]

Do you see oscillation on the speaker out?

Yes, full rail, kind of a square wave with an RC decay down from the peaks

What if you turn the master to zero?

Don't think it makes a difference but I will check

Does it still happen if you have so speaker connected?

Yes

What are the characteristics of the probe you are using?

I have used my Tek digital scope (TDS420A) with P6100 probes and also my old tube Telequipment scope with regular wire leads and it is similar with both. I can't catch the behavior very well on either since it is too big for the digital scope and too long a period for the analog to make a solid trace.

[glebert]

C1 (47µ) O.K.?

Interesting. I am not seeing any appreciable DC across C1. Stay tuned...

[glebert]

Replaced C1 and the rail to rail oscillation has gone away. Now I have about 53V DC and a 5vpk-pk 120Hz triangularish ripple on the output. Time to start looking at the output devices I guess. At least I can use a scope or meter to see what is happening, before everything was all over the place.

So what is the hint that this particular capacitor was the place to look? I guess the feedback being necessary for stability this is a very sensitive point, as Juan pointed this out being a system.

[nickb]

Replaced C1 and the rail to rail oscillation has gone away. Now I have about 53V DC and a 5vpk-pk 120Hz triangularish ripple on the output. Time to start looking at the output devices I guess. At least I can use a scope or meter to see what is happening, before everything was all over the place.

So what is the hint that this particular capacitor was the place to look? I guess the feedback being necessary for stability this is a very sensitive point, as Juan pointed this out being a system.

The cap is in the feedback path. If it develops a higher than designed-for internal resistance then you will have too much feedback at higher frequencies. If you also have an extra 180 degrees of phase shift through the amp and feedback (you always do at some frequency) then the negative feedback turns positive it oscillates. It's not an unusual problem.

IN my questions were I was trying to find out whether the oscillation was in the power amp and if it was how your tests might be affecting it.

[J M Fahey]

Painting with a VERY broad brush, 4AM here and going to sleep after a loooonnnggg day, even with a "non broken" ampyou have 2 ways to increase or introduce instablity.
1) apply too much feedback, as Nick said above.
2) since there are many tiny "parasite" capacitances all over the place, sometimes applying a probe to certain points , may add enough capacitance to make amkp oscillate.
A probe, even if shielded, is some kind of antenna.
The shield/braid itself is an antenna with one end connected to ground, and just the act of touching certain points may trigger oscillation.

FWIW a tried and true way to kill a cheap digital multimeter is to measure power tube plate voltage

In a surprising amount of cases that is enough to make power tube oscillate and NUKE the multimeter.

It´s so widespread that you´ll notice some schematics add the warning:
* "don´t measure at plates" or its cousin "only measure voltage at OT centertap" which means the same.

[Helmholtz]

Replaced C1 and the rail to rail oscillation has gone away

Glad it worked and stopped the oscillation.

The idea behind was that E-caps have a limited lifetime and their impedance increases over time. C1 AC grounds R5 which is the lower feedback divider resistor. When the impedance of C1 increases, more signal gets fed back and too much feedback often causes instability.

Not necessary, but as there is little DC across C1, you may consider to use a non-polarized (bipolar) E-cap for better reliability.

[glebert]

Not necessary, but as there is little DC across C1, you may consider to use a non-polarized (bipolar) E-cap for better reliability.

Should have been more clear, there was no DC with the old cap, now there is 16V. The old cap was shorting. Regarding effect of probes, this amp was definitely oscillating without being probed. When I first got it and hooked it up to a junky old speaker I had lying around (maybe a 20W) it fried it almost instantly. And yes I had measured for DC on the output, but not AC.

[Helmholtz]

Should have been more clear, there was no DC with the old cap, now there is 16V. The old cap was shorting. Regarding effect of probes, this amp was definitely oscillating without being probed. When I first got it and hooked it up to a junky old speaker I had lying around (maybe a 20W) it fried it almost instantly. And yes I had measured for DC on the output, but not AC.

Wow, E-cap shorting is rather unusual. Somehow spoils my explanation of feedback induced instability. But would shift DC levels of course. With some volts DC across there is no need for a bipolar cap.
But I doubt that 16V are normal condition.

[g1]

But I doubt that 16V are normal condition.

Agree, that 16V must be 'feedback' of the DC on the output.

[g1]

I'll add this schematic as it has a few bits not shown on the other.

[Helmholtz]

Z1 zener voltage correct?