New problems with old Gainclone...

[KSTR]

Looks like with the high impedance level, you lowered the frequency of a parasitic pole (node impdance + input/stray capacitance at the inverting input). And now, if still present in your circuit, the 220pF is probably major part of the problem, adding deliberately to this capacitance (that's why I said it is dangerous), the zero introduced by the 1k is way to high, thus the 220pF can be seen as going more or less "directly" to GND.

Also, you might have now significant parasitic coupling from the output/supply to the input(s) and more nasty stuff of sorts... all due to the too high impedance level.

For a thorough discussion on these matters, see
http://eportal.apexmicrotech.com/mainsite/pdf/an19u.pdf
and
http://www.ti.com/litv/pdf/sboa015

Some lower supply voltage is easily derived from the rails, then you could add a buffer, built ie on a strip of veroboard, and you get rid of all the problems, turning down the impedances to the lower levels again...

- Klaus

[sek]

Klaus,

thanks for your reply, very nice to have someone knowledgable looking at it.

Looks like with the high impedance level, you lowered the frequency of a parasitic pole (node impdance + input/stray capacitance at the inverting input).

In order to check this possibility, I'll take another look at my grounding scheme.

At the moment, my supply has a power ground star and a signal ground star. Both are sitting close to each other on the PSU board and meet via a short and thick copper trace.

The power ground is split, sinking (and sourcing) current for the positive supply capacitor and it's negative supply counter part via separate links.

So the above PCB has three connections to the PSU star: From the "left" supply capacitor, the "right" supply capacitor, and the signal input (see attachment below).

I thought this would reduce hum and buzz by providing separate current return paths of low impedance for every part of the circuit.

But while this seems to work properly (no hum or noise audible), maybe this has introduced some stray capacitance or inductance I didn't think of! According to the high frequency and low amplitude of the oscillation, we're talking about Nanohenrys and Picofarads.

Does anyone have any idea about my grounding scheme's possible inadequacies?

the 220pF is probably major part of the problem

The 220pF capacitor went out of the circuit as my first measure.

the zero introduced by the 1k is way to high

There's also no 1kOhm resistor in the circuit any more.

The part denominated as R_in has been tried down low to a short (0 Ohms), the problem persists no matter if I shift R_in in front of C_in.

For a thorough discussion on these matters, see http://eportal.apexmicrotech.com/mainsite/pdf/an19u.pdf and http://www.ti.com/litv/pdf/sboa015

Thanks for the links, I'll give both documents a read as soon as I can.

I am familiar with bode plot discussion, the influence of poles and zeroes on transfer functions, basic filter theory and signal processing in general, etc.

I just didn't expect so many difficulties with a tried and proven circuit.

you could add a buffer

As said above, this will happen in the future. I just can't accept that the circuit shouldn't work without it. I still expect an analytical (or maybe lucky) solution to be easier, quicker and cheaper than four improvised buffers (that would have to be thrown out again later anyway).


Thanks for any help again
Sebastian.

[digi01]

i don't understand your power supply.
have you use CT transformer?I wonder if the psu in working order?have you checked the terminal voltage?

I just can't accept that the circuit shouldn't work without it. I still expect an analytical (or maybe lucky) solution to be easier, quicker and cheaper than four improvised buffers (that would have to be thrown out again later anyway).

i appreciate your patience i think the igc should work without buffer.

p.s.
do you have a photo of your amp?

zang

[KSTR]

Seba,

I think your PSU layout is just fine (from what one can tell without fotos)...
... but this has nothing to do with the stray capacitance I mentioned.

Analytical solution?... go non-inverting, which requires nothing more that to rip up and patch you amp-PCBs a little, not a problem IMHO. You even have a lot of space left in the center area of the pcb, drill a few holes and stuff the parts you need there, rewire the new input to the pinheader etc. At least that's what I would do (you won't believe how much this type of "PCB revisioning" is used in the industry -- "new PCB? Naw! Let's hire some students who will happily modify this batch of, er, 2000 boards here. Nobody will see it anyway").

- Klaus