See this and following posts as to why 7kHz + 13kHz is a very good and practical two-tone signal for IMD testing, and probably better than single sine THD (and especially those above 20kHz) and better than 19+20kHz IMD or (199+200kHz for that matter).
For peace of mind we may want to generously extend that to 3x, that is, 21+39Khz and test within a 60kHz bandwidth. Beyond that point any gear should have effective low-pass filters anyway.
good advice, thank you therefore.
Unfortunately, at the time when I was thinking about good amplifier circuits and trying out a few topologies with "Circuitmaker" around 2005 to 2011, I didn't know this yet. I came up on those days with the idea of working at the simulation level with measurement frequencies far outside the audible range and thus found some good topologies and learned that very simple circuits are often superior to the very complicated versions with extremely low THD values concerning sonic character.
Very interesting. Thank you.
I’ve compared power amplifier measurements and don’t see, in this particular case, what better insight would 7/13 kHz test provide. I believe that, as is usual, it is better to make even more measurements and then try to draw conclusions. This can be considered as measurement of OPA828 as a first stage in power amplifier. Half of measured distortion is DAC contribution.
Final measurement is 31 tone signal which produces 16 Vpp at amplifier output.
Attachments
Despite its high GB product of 45 MHz, the opa828 has a straightforward one-pole characteristic, so it is forgiving of board traces and other problems, and operates stably even when used with 100% feedback. It will probably operate stably as long as the basic precautions are followed, such as decoupling of the power supply line close to the IC.
Note that the current consumption is four times that of the TL071.
However, if the original circuit is for TL071, since it may be an old design and the value of the peripheral circuits (especially the impedance of the filter CR at the crossover) may not be suitable for OPA828, it may not be worse than TL071, but may not bring out the full performance of OPA828,
Do not expect too much when rolling only OP-AMPs.
Note that the current consumption is four times that of the TL071.
However, if the original circuit is for TL071, since it may be an old design and the value of the peripheral circuits (especially the impedance of the filter CR at the crossover) may not be suitable for OPA828, it may not be worse than TL071, but may not bring out the full performance of OPA828,
Do not expect too much when rolling only OP-AMPs.
It would be very interesting for me to know how the IM residual distortion curvature looks in this two-tone (7+13Khz) measurement, if one do the same measurement procedure as the magazine "STEREOPHILE" do with the normal THD measurement for getting Fig. 6 under
https://www.stereophile.com/content/bryston-3b-st-power-amplifier-measurements
and fig. 8 under
https://www.stereophile.com/content/technics-su-g700m2-integrated-amplifier-measurements
What would you expect to see?
Anyway, to look at the residual, you'd need to completely notch out the test signal from the DUT output, and IMO that would be complicated. A distortion analyzer would normally use an auto-tuning notch filter to suppress the (only) fundamental. With a two- or multi-tone test signal, you'd need two or more auto-tuning notch filters, and I have not seen that done.
In contrast, it is easy to reduce the test signal in the DUT output by 50-60dB, enough to keep the ADC from adding its own IM products, so that the spectrum shows you IM products from your DUT and not from your measurement rig. However, the resulting signal is still dominated by the test signal, and that's what you'd see on the 'scope:
Anyway, to look at the residual, you'd need to completely notch out the test signal from the DUT output, and IMO that would be complicated. A distortion analyzer would normally use an auto-tuning notch filter to suppress the (only) fundamental. With a two- or multi-tone test signal, you'd need two or more auto-tuning notch filters, and I have not seen that done.
In contrast, it is easy to reduce the test signal in the DUT output by 50-60dB, enough to keep the ADC from adding its own IM products, so that the spectrum shows you IM products from your DUT and not from your measurement rig. However, the resulting signal is still dominated by the test signal, and that's what you'd see on the 'scope: