Oscillating Power Supply....WTH?

[EchoWars]

Battling on old Philips 210WPC amp...my buddy blew it up, and it required an output bjt or two, a few drivers and resistors, and the pre-drivers to put it right. It also uses an opamp (4558) at the input for a subsonic filter. I replaced the 4558 with a NE5532, gave the 5532 a decent bypass, and fired it up.

Once running, and while looking at the signal to the main amp with a scope, I saw some nasty oscillation (about 3MHz or so, amplitude of about 1V P-P). Figured the layout was just not right for a 5532, so put the 4558 back in. STILL oscillating. What's more, I see that ALL the voltage supplies in the amp show this oscillation riding on the rails (several zener regulated +/-15V, a standby +12V, and another +14V for the front LED's).

OK...so I completely remove the opamp from the circuit (socketed now), thinking there is a broken trace or some such contributing to the problem. Even with the opamp completely out of the amp, the rails show the oscillation.

I note that...with a scope probe on the chassis, I see nothing odd. However, at the 'common' of the big filter caps, I see oscillation. Very odd, as they are tied to the chassis a few inches away. An ohmeter reading shows a short, as it should. It's a good hefty wire (about 12ga) and connection is well soldered.

For the hell of it, I take a .01µf stacked film cap and 'short' the common of the filter cap to the chassis. Oscillation disappears. I don't know why, but I also took the .01µf cap and 'shorted' across one of the diodes of the big diode bridge for the main high-current supply. Oscillation almost disappears. I tried adding a cap across all of the bridge (I guess these are called 'snubber caps' when used like this??), and the oscillation is totally vanished.

The amp uses a quite large transformer with a single high-current winding common to both channels, but has a separate diode bridge for each channel. Doesn't matter which bridge I add the snubber caps to (if that's what they are called), but once added oscillation is completely gone.

Amp seems to run fine now, but I can't help but wonder how/why the oscillation started...was it always there? Has something gone bad that is not-so-obvious starting the oscillation? Why in the world would an unregulated supply begin oscillating in the first place? Do bridge diodes really create that much high-frequency trash?

[EchoWars]

Turns out that adding the caps to the bridge only took care of the oscillation at idle. Once a signal is added, the output waveform gets the 'fuzzies' on the negative half of the wave.

The driver boards are connected with a female plastic connector that pushes onto male pins on the chassis. By adding .01µf caps to the ground wires running to the driver board, almost all the oscillation is gone.

However, I noticed that the most improvement came when adding the .01µf cap to the ground for the Zobel filter. I suspect that this filter, which consists of what looks like a cheap cap and a wirewound resistor, is the problem now. When the speaker switch is set to disconnect any load, there is no oscillation at all. With the dummy load connected, there is the small bit of oscillation on the negative peak of the waveform. Since the oscillation, small though it may be, disappears with the load disconnected, I'd like to replace the filter with some better parts and see what happens.

Below is a pic of how it is set up now. I thought it odd that the filter is on the far side of the inductor/resistor combination. Is that why the values are so large (33 ohm and 1µf..), or is there any reason why I can't use a more standard value of like 5-10 ohms and a 0.1µf cap?

[lucpes]

If you can, I'd put the zobel 'before' the L||R thingie, using a good quality 0.1uf cap & 10R3W. Also try to 'bypass' the power rails with .1uf as close to the output devices as possible and I'd add a tinyl value cap (68pf or so) across the amp's input (signal+ to signal ground.)

[EchoWars]

Thank you very much for the reply.

It won't be easy to move the filter, but it can be done. I certainly don't look forward to pulling those driver boards out yet again, but I got no choice.

The amp input has a 100pf cap already between input and ground. Signal at the amp input stage looks quite nice, and even at the output it looks very good until power out hits about 100W or so. Then the fuzzies start to show up on the negative peaks.

I'll see if I can't get the parts I need.

If anyone ever tries to give you a Philips amp, politely decline.

[maylar]

Interesting problem. I can only comment with generic thoughts that may or may not help.

You need gain and feedback for an oscillation to sustain itself. Zener-transistor regulated power supplies don't have that, so the source of oscillation must be somewhere in the amp.

My first technician's job was at a power supply company. When measuring the output ripple it was common to see high frequency hiss on a 'scope. The chief engineer said it wasn't "real", and proved it this way: short out the 'scope probe and touch it to the circuit. If the noise remains it's an anomoly in the measurement system, not part of the circuit. We usually put the 'scope on bandwith limit to ignore such nonsense.

[Eva]

These 3Mhz thing dependent on output load and instantaneous voltage look much like output stage oscillations due to unstability [feedback applied at frequencies where loop phase shift is 180º or greater]

You should study the amplifier circuit. Dry small/mediul electrolytics or faulty compensation capacitors may cause this problem. Fake, low quality or slower than original transistors may also cause it. Finally, it may be a design flaw of the original amplifier and in this case you should think about increasing closed loop gain or increasing compensation in order to make it stable

Drawing an schematic and posting it may help to identify what components could cause the unstability

You work would be finished when the amplifier could drive a 1uF capacitor direcly connected to the output, entering into clipping and without apreciable oscillations

You should not let the amplifier oscillate since this oscillation markedly increases THD and IMD and there is some risk of blowing output devices due to cross-conduction

[EchoWars]

maylar...
I certainly considered this, but how can I not trust what I see with the scope? In any case, it appears that adding a slew of .01µf bypass caps from the chassis to several of the internal ground connections has improved the situation markedly.

Eva...
Both channels exhibit exactly the same behavior, so I think a failed cap is unlikely. I believe this is a design issue...

The 3MHz oscillation is gone. What I have now I call 'oscillation' for lack of a better word. It actually is more of 'mild instability' on the negative half of the wave, slightly blurring the waveform and becoming more pronounced as the wave approaches its negative peak. The positive half of the waveform looks just fine. Moreover, the phenomenon I describe does not present itself until...

1) A load is connected, and
2) Power output approaches 100W

With no load the wave is clean as can be. This is why I feel it may lie with the poorly chosen wirewound resistor for the Zobel filter. Any other thoughts on moving the filter before the R||L output coil? Or the size of the filter components?

I'd scan the schematic, but sections of it are shaded, and may be tough to read. Still, I'll see what I can do.

[traderbam]

it required an output bjt or two, a few drivers and resistors, and the pre-drivers to put it right.

What exactly did you change?

[EchoWars]

Wow...

Well, this thing was fairly smoked. Q210 (driver) was shorted, as well as Q212 IIRC. Anyway, since I had no idea what these '355NI' devices are (although they were labled Toshiba), I moved Q211 to where Q212 was, and replaced the drivers with 2SD555's.

The pre-drivers were also toast, and replaced with 2SA1535's and 2SC3944A's.

All of the .47 ohm emitter resistors were open, and were replaced, as well the 10 ohm 2W resistors for the drivers.

Bias transistor was shorted, and replaced with a 2SD756A.

D206 in the protection leg was shorted, and replaced with a 1N4148.

[EchoWars]

Yes, I know that's tough to read...

Here's a zip that is a lot easier on the eyes.