Supply depended oscillation

[K-amps]

Originally posted by Thoru
....

What do you mean with higher rails? Should I just continue to increase the supply voltage? I'm a bit scared of that because I really hate the sight of smoke from such a lot of work. But if you really think it will operate fine at +-30V (which is the final supply voltage), than I'll try

Thanks for your reply,

Remco Poelstra

Absolutely not if they have never been tested with those rails before...

[Thoru]

I've tested everything except with the Q4 modification. It seems I'll have to revert that patch and check for stability again.

Remco

[john_ellis]

Hi thoru

Are you still using two-pole?

The component values I simulated don't give much of a frequency peak in the response, but increasing C6 (only 5 pF) might help.

A 15 pF capacitor might solve your problem.

Another thing to try - normally, Miller type amps are very stable (because of the extra pole appearing at a skyhigh frequency point) but possibly the input stage sees a variable impedance.

Does the oscillation go away if you ground the input .. or what about if the input is open?

If grounding the input stops the osc then try a 100 pF capacitor on the input...

Sorry, haven't looked at your layout to see if any tracks are too close/interfering etc.


cheers
John

[Thoru]

Hi,

Yes, I'm using two-pole and I increased C6 to 330pF to have the fb network only active in a 50kHz bandwith.
The input is grounded most of the time yes. But also with my function generator the oscillation appears.

Thanks,

Remco Poelstra

[john_ellis]

Hi thoru

Simulations indicate possible things to look into.

Should not be a problem, but there is a peak in the differential (input transistor base-base) around 20 MHz (too high?) and a change of slope fro m6 db to 12 db around the unity gain point (may not be so good).

But I noticed that in your circuit diagram one of the power supply line caps is connected to the input (quiet) ground.

Does your layout do this?

Impedance in the earthy side ground and supply line could send signals straight into the input from the rails.

These power rail caps must go to the noisy (power) ground pin, not the input (quiet) pin.

cheers
John

[Thoru]

Hi,

First about the current layout: The grounds are seperated as you can see on the photo I posted a few posts back. Ugly hack, but it works .
While thinking about the peaking in the differential pair I fiddled some with the feedback bandwith. When I changed C6 back to 6p, the oscillation disappeared. This gives the suggestion that it is the actual amplifier not being stable. To check that I chose 226p for C6. The oscillation now was lower in amplitude, what I expected, but it was only slightly de-tuned to 4.1MHz. If it would be the amp being unstabel, I would expect a bigger frequency shift, so it seems that the capacitance of the feedback network only slightly influences the oscillation, while the divider operation of the feedback resistors can kill the oscillation.
So it seems the problem is somewhat solved, although I do not understand where and how the oscillation works.

Regards,

Remco Poelstra