Hm, not so clear as yesterday after swapping the order of IVY and Legato:

B3 playing -30dB sinus at 1kHz
-> IVY III balanced Vout+ and Vout-
-> V/I via 2x 680Ohm to Legato inputs
-> Legato unbalanced
-> +40dB AMP
-> ESI Juli@ soundcard

rms noise: -108dBV
varying harmonics 60dB bellow signal => distortion patterns are back

Edited by user Saturday, November 8, 2014 5:20:36 AM(UTC)  | Reason: Not specified

My dear Legato. I'm sorry for that. Fortunately it seems it still lives.

And yes, the windmills are my passion... see this.

Using oscilloscope I've tried to measure voltage after the first OPA1632. But, using the probe even with ground tip the images were just oscillations and I was sad.

So.. I've added 82pF capacitance in parallel with feedback resistors.

The distortion patters are reduced now - I'd say it is much better now. Estimated by 10-15dB less. But... another big SMASH by the windmill! The RMS noise without those caps was -112dBV. Now, with the caps it is -101dBV. Aaargh, I need to find another attack pattern!

Here are some stairs created with 82pF caps on both feedbacks resistors, voltage measured between V+ and V-: https://drive.google.com...AcAQZHdVZHA2RE0tVjQ/view

I've made the mistake while I was adding capacitor into feedback loop ... a silly one...

but it helped... a lot. Symmetric solution did not work, neither way.

Low amplitude measurement at IVY unbalanced outputs:
https://drive.google.com...czlpVDg/view?usp=sharing

Work of the single eyed bandit:
https://drive.google.com...aTdHeEE/view?usp=sharing

Schematic of attack against the windmill is attached.

It has nothing to do with stability. :)

The problem is that such caps will only filter to the GBW of the entire circuit -leaving all else to pass through.

If you want a more aggressive filter you have 3 options.

1) Before the IVY which then makes the DAC see a high impedance. This is fine - it is just now you are using the DAC as a voltage source - not a current source.
2) Between the stages as I did (just lift an R and do what you like) This is the best place to apply your tinkering. You might try a 3rd or 4th order filter - but not in the feedback loop.
3) After the IVY - which raises the output impedance.

All three options will add some phase shift.

The circuit you have drawn will actually do quite a lot to harm the SNR :) You have created a very non-linear circuit. :) Such asymmetry will wreak havock on the THD in a fully symmetrical amplifier.

Edited by user Sunday, November 9, 2014 7:51:15 PM(UTC)  | Reason: Not specified

Russ, thanks for this explanation.

I've removed my "tuning" and started again... :-)

In the optimum case the voltage at the current DAC output is 0 Volts. So I've connected oscilloscope probe between GND and Out+ on the DAC.

And I've found correlation between variable distortion patterns seen in the FFT and measurement on the oscilloscope.

It seems that it correlates with something like oscillations in the 4-30MHz range.

Here are the measurements of the "same" low-amplitude signal but with different distortion pattern caused by rotating digital volume knob down and up again.

Any hints what I can try to change to keep these variation constant?

EDIT: check this in case the attached document does not work -- https://drive.google.com...AcAQQ1VzZVNxSGFPVlE/view

Edited by user Monday, November 10, 2014 11:08:25 AM(UTC)  | Reason: Not specified

If it were oscillation you would have seen it when you have the legato between the DAC and the IVY and you clearly did not :)

I have already answered your core question - have you tried my suggestions?