Shame on Topping for holding performance back for a measly $2 more in production costs and go from 'typical digital' (mechanical/metalic) sound to 'analog' (warmer more neutral) sound by the touch of a swapable op-amp.
Would have loved to see the measurement differences in response and IM products 19-20kHz to see if electrically things changed in an audible way...
Just teasing you Thorsten.
I will show measurements when I get my QA-404. Sadly right now my DSO has a FFT that is NG below -90dB but it goes out to several 100MHz... ;-) Measured the same on that, of course.
But the post asked audible changes, so I commented on what I heard.
Some extra notes.
The D10S runs the analog stage on appx. +/-5.6V created from a switcher running at ~ 1MHz with fairly vestigial filtering and switching ripple is observable on the Op-Amp supply lines with a decent fast 'scope.
The LME49720 has no specification of 1MHz PSRR, but at 100kHz it is ~54dB which suggests 34dB PSRR at ~ 1MHz.
The LM6172 has ~ 50dB PSRR at 1MHz.
The LME49720 has ~ 1.3uV output noise in the actual circuit in the D10S, the LM6172 2.1uV output noise, relative to 2V signal levels, so ~124dB to ~120dB SNR.
The LM6172 is not specified for high impedance THD, but being designed to drive 100 Ohm, there is a lot of bias current in the output stage, especially because the "single stage current feedback" nature of the design cannot rely on extra NFB to reduce THD.
The -106dB H3 for a gain of 6dB with 1V PP into 100Ohm rapidly drop into higher impedances. The output stage distortion changes pretty much linear with impedance, so for 2k Loads we can estimate -132dB H3 and -140dB H2. That compares favourably to the LME49720. IME, once the output stage runs into high impedances, the H3 element in distortion drops rapidly, leaving H2 dominant.
To conclude, technically speaking, the LM76172 has slightly higher noise (~ 4dB) than the LME49720, much greater PSRR at the switching frequency of the Power Supply feeding this Op-Amp and comparable to lower Harmonic distortion.
The LM6172 is based on a "high bias, fast silicon" model with a single stage of gain, relatively low levels of loop feedback (< 80dB 10Hz OLG) at lower (audio) frequencies and 100MHz GBWP with 2,500V/uS slew rate.
The LME49720 is based on "low bias, low current noise, slow silicon" model with two stages of gain, very high levels of loop feedback (< 140dB 10Hz OLG) at lower (audio) frequencies and 55MHz GBWP with 20V/uS slew rate.
I think for low noise use with clean, low noise (especially low RF noise) supply rails the LME494720 is a better choice (the discontinued LM6182 would address the noise if it can be used).
For use with high frequency noise on inputs and rails the LM6172 would appear superior, in technical terms. If RF IMD is reduced, this could (usually does) correlate with a sound that has less "grain, grit, edginess, brightness" to use the negative (what it does less wrong) form of statement.
Thor