poweramplifier trace elliott GP11: can output MOSFETs whistle ?

[gasoil]

Hello all,

I'm trying to repair my old bass amplifier, the power amplifier works as intended when loaded by 10Ohm at more than 100Wrms, exepting that a quite loud whistle can be heard when the signal is 8kHz or more. And the whistling freq heards like it follows the signal freq.

1st I've think about the transformer (but if it was the culprit, the whistle should be 50Hz, right ?), but tighten the screws that "close the EI" does not change anything. Then I was thinking about the power resistor, after all it looks like a coil inside. But there again, it wasn't the guilty.

I've extended the cabling to use the amp PCB outside the chassis, and now I'm sure the whistle come from the PCB... but there is no coil nor transformer on it ! The only components that can see high currents are the power mosfets in TO3 case.

So my questions are:
 - is it possible for power mosfets to whistle ?
 - if a power mosfet whistle, is it a bad sign regarding mosfet EOL or other ? (if the whistle is at the same freq that the signal, I guess it won't harm that much at 180Wrm when the load will be a speaker !)
  - any advice to localise the whistler ? the board is something like 200mmx50mm, and touching some parts (especially mosfet and their rad) do nothing on the noise.

I can post a picture of the PCB/amplifier and/or its schematic if it helps. Not sure about the schem copyright, but google with trace elliott GP11 will easily find it.

Last: this is my 1st post there, and i'm not a "forum guy", in fact this is my 2nd forum ever... Sorry for any awkwardness.

Thank you

[Circlotron]

  - any advice to localise the whistler ?

Get a 300mm length of plastic tubing about the same diameter as a pencil and put one end (carefully!) in your ear and move the other end about the board.

Capacitors can make a noise, particularly multilayer ceramics.

https://www.murata.com/en-eu/support/faqs/capacitor/ceramiccapacitor/char/0020

[gasoil]

Thank you for your advice, i'll try to be soft while putting the pipe in my hear !

Here are pictures of the pcb. You can see what kind of components there's on it: poweramp from the 80s i guess, no CMS.

The close range one show all there is to see excepting the output stages (2 mosfets push-pull in //), this is the channel I'm debbuging right now, the other part of the board if for the 2nd channel which is powered but no signal in (same amp, unless one push-pull only and HPF in input). Since there's is no whistle when there's no load plugged, I guess the noise can't come from the 2nd channel.

In your opinion, what would be the usual suspect ?

I'll try to localize the whistle with the pipe asap.

Thanks

[gasoil]

Hard to be sure with the "pipe experience", I can't hear anything while "checking" the board components with the pipe. I can hear the whistle again while the pipe is over the rad... I'm not able to point one mos with the pipe, sound is more or less equal whatever the mos. I can't be sure it is a mos since the pipe is then at a sufficient distance from the board so I can hear the whistle again even if i'm not over the rad (not sure to be clear).

By gently pressing any mos (even the ones that does not output power), the whistle disapear and come back as soon as I'm releasing the pressure. Thus I guess one mos make vibrate the rad that amplifies the noise. I've tighten the mos on their rad, the whistle is still there... Maybe weaker, hard to tell. But I'm not able to make it lower by pressing  mosfet anymore...

More, I found they become quickly hot: a K-probe on TO3 show more than 60° in less than a mn for something like 75Wrms (100Vpp in 10Ohm, pure R), is it OK for a double push-pull (2SK135 and 2SJ50). But I don't have any experience about this so your opinion is welcome.

[coppercone2]

capacitor, disc cap, i just eliminated 4 on a hp supply that were loud as fuck

[temperance]

The famous Trace Elliot oscillators. A friend of mine has some of those. Those amps are a copy of a very old Hitachi application note.

What I did to stop them from oscillating:
I added some 1µF film capacitors on between +Vcc and GND and -Vcc and GND onto the amp board close to the output stages.


Extending the cables to use the amp outside the chassis will for sure result in oscillation. The wiring between the power supply capacitors should be as short as possible and at least 1.5mm².

Touching the MOSFET's can stop the amp from oscillating as well.

[shakalnokturn]

Are you scoping the output while load testing?
If you're getting parasitic oscillations (often close to saturation) that's a problem to solve because they can cause the output stage to overheat.

If the noise you're hearing is only at the audio input frequency I'm not sure it's a problem, every amplifier above 30W I've tested into a resistive load has whined.
The whole power stage sings, the resistive load sings too.
The noisiest components will usually be the inductors and wound resistors because the magnetic effect is quite important there but the whole power stage will sing more or less due to the thermal effects of the varying current.
You can even try feeding it music while load testing you should hear the power stage sing in the 1-8kHz range.

If your transistors are getting too hot check they don't need re-tightening it would be a shame to destroy those rarities.

[gasoil]

capacitor, disc cap, i just eliminated 4 on a hp supply that were loud as fuck

I'll try that, thank you.

[gasoil]

The famous Trace Elliot oscillators. A friend of mine has some of those. Those amps are a copy of a very old Hitachi application note.

If you could retrieve the ref it would be great !

Extending the cables to use the amp outside the chassis will for sure result in oscillation. The wiring between the power supply capacitors should be as short as possible and at least 1.5mm².

OK so i'm the ususal suspect... There's 200nF (or maybe 400nF) on each power rails, but the wire are probably 5x longer than usual.

Touching the MOSFET's can stop the amp from oscillating as well.

And that's what i was experiencing before I've tighten the mos on their rad...

Thank you for all, i'll try to put 1uF on each power rails input and see how it runs. If no more whistle, i'll try with shorter power wires without the cap and let's hope it solves it...

[gasoil]

Are you scoping the output while load testing?
If you're getting parasitic oscillations (often close to saturation) that's a problem to solve because they can cause the output stage to overheat.

Yes 'im scoping with a micsig this very same forum make me buy to restart my elec hobby more seriously!
And the output loaded by 10Ohm at 150Wrms is beautiful, 44dB THD when my GBF is at 45dB, and precision is not where the micsig is really good at. So the power stage looks like it works as expected, only it does'nt need any speaker to sing !

If the noise you're hearing is only at the audio input frequency I'm not sure it's a problem, every amplifier above 30W I've tested into a resistive load has whined.
The whole power stage sings, the resistive load sings too.
The noisiest components will usually be the inductors and wound resistors because the magnetic effect is quite important there but the whole power stage will sing more or less due to the thermal effects of the varying current.
You can even try feeding it music while load testing you should hear the power stage sing in the 1-8kHz range.

In my case it sounds like the power stage only sings, if i fut away my resistive load the whistle stay near the board. But there is no inductor nor resistor bigger than 1/4W on the board. It really sounds like it is the rad that sings. I've not tested with music, it should be fun, but the whistle starts at some kHz (can't remember) and follow the GBF freq past 12kHz, after I can't hear anything . And the scope show a crystal clear sinus up to 70kHz. Probably sufficient for a bass amplifier !

If your transistors are getting too hot check they don't need re-tightening it would be a shame to destroy those rarities.

I've re-tightening them already, and more than 60°C in less than 1mn at about 50% full power is it too much ? I'm a rookie.

Thank you

[floobydust]

Given the age of the board, outright replace all the (five?) electrolytic capacitors - especially the Philips axial part, they don't last 40 years.
The output Zobel network capacitor is usually the noisiest due to the high voltage AC it experiences. The green one at the loudpspeaker output.

[temperance]

I had those schematics. But they went into the old paper bin long ago. But after two minutes searching the web one can find something like this:

http://www.cma4ch.org/chemo/ftp/Hitachi-power-Mos-Fet-application-note.pdf

Fig8-3 is the schematic for the trace amp.

Replacing the capacitors is not going to help much. The Trace Elliot PCB layout is pretty bad and requires some care to get it to work without the the amp going into oscillation. I don't remember exactly how I modified this amplifier about 25 years ago. I think I cut the traces to the diff pairs and installed 4.7...10R resistors in series with the power supply lines and then added 100nF + 10µF capacitors to decouple the diff pair from the output stage.

Its a bad idea to switch an oscillating amplifier on as it might kill the output stage. A new set 2SJ50/2SK135 will cost around €200...500 today.

[gasoil]

Sorry, I'm still around but really buzy.

I hope to have some times this week-end to try yours suggestions.

[Gyro]

Sorry, I'm still around but really buzy.

You go very nicely with your whistly amp then! 

[gasoil]

http://www.cma4ch.org/chemo/ftp/Hitachi-power-Mos-Fet-application-note.pdf

Fig8-3 is the schematic for the trace amp.

Is it usual for a 2nd stage to be differential ? What are the benefits ? +6dB easy ? avoiding capacitive coupling with 1st stage ?
The doc is talking about class A vs class B for the voltage stage as if class A <=> diff... Not clear for me.

The potentiometer is there to set the mias current at 10mA, so I should read 1V (100Ohm x 10mA) between the high voltage rail and the emitter of the current load ? (the components labels are not readable sorry) 

Replacing the capacitors is not going to help much. The Trace Elliot PCB layout is pretty bad and requires some care to get it to work without the the amp going into oscillation.

By looking in detail at the PCB, there is a strange routing: each channel has its own power rails, but the cap are "interleaved": the + of one channel goes through the cap to - of the other one ! Not a problem on nominal configuration, but since in my case the other channel is not powered, it is as if there's not cap at all ! I want to change that, but maybe I'm missing someting ?

Before to try it again, I would like to recap it: but I have difference between value I can see on the components, and the one in the BOM:
the coupling input cap is at 1uF instead of 1.5uF, I guess the bigger the lower freq pass ? And I have 10uF instead of 47uF in the feedback loop. Should I stay with 10uF ? Are the ceramic cap aging -> good idea to change them too ?

And a last question for my info: it looks like there's 3 red stripes on each trans of the input diff pair. Does it mean they are appaired ? The voltage amplifier is a diff pair too, but no stripes on its trans, does it means appairing on the voltage stage is less critical ?

Thank you.