I have been using these for years. The basic design is good but it seems to be limited in distortion in ways that don't make sense. It should be limited by the performance of the opamps but the floor is much higher. Also its dominated by second harmonic. I have several and I wanted to get one tweaked up for a friend.
I finally had the opportunity to start really digging in. I have an ultra low distortion source (all harmonics are -120 dB or better) (Shibasoku AG16B) and have set up an ADC (AKM5394A demo board) with distortion residuals at almost the same levels.
Now equipped and with the beast on the bench I found that the input board seems to pass the source with the harmonics at the -117 or better level. The filter board was the first bad guy. The distortion is essentially zero up to U8. The problem seems to be some nonlinearity in the fet switch (A1-U7 LF13201) Replacing the bipolar opamp with a JFet one (A1-U8 OPA604) reduced the current and removed most of the distortion at that stage.
Still high second harmonic on the monitor output.
The notch board has the active notch filter and a variable gain amp on its output. 3 LME49720's in A2-U1,U2 and U3 reduced the harmonics at the test points to - 117 dB but there is still a high second harmonic on the output. close inspection shows two dipped tantalum caps in the gain amp, implemented in a way that could create a lot of distortion. Replacing A2-C18 and A2-C19 with back to back electrolytics dropped the second harmonic over 10 dB.
After these changes the unit will need recalibration, something very quick and easy if you have suitable resources.
The generator is now the limiter. That will be the next installment.