"Is there a version of the AP software that runs on a PC without an AP unit being connected?" Yes for $$$$.
What I would like to see and could help illuminate what is going on is using DiAna to analyz the distortion at several points. If you follow its calibration routine with an APU in the chain it will correct for the insertuion loss of the APU and show the correct harmonic amplitues and phase relationship. It will also recreate the residual so its clear what part of the waveform departs from ideal and how. it can be very informative.
What I would like to see and could help illuminate what is going on is using DiAna to analyz the distortion at several points. If you follow its calibration routine with an APU in the chain it will correct for the insertuion loss of the APU and show the correct harmonic amplitues and phase relationship. It will also recreate the residual so its clear what part of the waveform departs from ideal and how. it can be very informative.
Today I got the scaler. It improves DAC measured performance at full scale (where it is automatic fix gain to 0db, hence a buffer),
but even more improvement at low levels when it can give much better THD numbers.
Anyway, I put the scaler into autorange mode, in between the DACDO300EX to Cosmos ADC in stereo mode (so results could be
even better by +3db, I don't have enough connectors). Here I give two measurements: IMD and THD swept.
In each picture I overlayed two graphs: one with scaler and one without, such that it is possible to compare the scaler influence.
As you can see, it improves. In both cases it reduces the hump and I think little shift it to the right, but maybe it is also a point where
ADC is overloaded. In any the higher DAC level is closer to the place where the scaler is a unity gain, or step above.
I did not include comparison to Nihtila AK5572, but I compared. I can tell, with respect to IMD, AK5572 without scaler is almost like Cosmos+scaler
In THD, AK5572 without scaler is little worse than Cosmos+scaler. I did not check AK5572 with scaler and in both cases I did not incorporated the
APU in THD tests here, so even possible better results could be reached.
but even more improvement at low levels when it can give much better THD numbers.
Anyway, I put the scaler into autorange mode, in between the DACDO300EX to Cosmos ADC in stereo mode (so results could be
even better by +3db, I don't have enough connectors). Here I give two measurements: IMD and THD swept.
In each picture I overlayed two graphs: one with scaler and one without, such that it is possible to compare the scaler influence.
As you can see, it improves. In both cases it reduces the hump and I think little shift it to the right, but maybe it is also a point where
ADC is overloaded. In any the higher DAC level is closer to the place where the scaler is a unity gain, or step above.
I did not include comparison to Nihtila AK5572, but I compared. I can tell, with respect to IMD, AK5572 without scaler is almost like Cosmos+scaler
In THD, AK5572 without scaler is little worse than Cosmos+scaler. I did not check AK5572 with scaler and in both cases I did not incorporated the
APU in THD tests here, so even possible better results could be reached.
Attachments
Very interesting even if I don't have an explanation for the observed behavior. Do we know when the scaler switched or what the respective input level into the ADC was? It is remarkable that there are no clear switching points as with Audio Precision sweeps except for -30 dB which could be a switch point or the hump.
Why would you get lower THD but not lower IMD with the scaler? The only thing I can imagine is that the bins for THD are still large enough to include some noise whereas those for IMD are sufficiently narrow.
Why would it dissappear the hump for -50 to -30 dB and make it smaller from -30 for both THD and IMD? I have no idea at this point!
Why would you get lower THD but not lower IMD with the scaler? The only thing I can imagine is that the bins for THD are still large enough to include some noise whereas those for IMD are sufficiently narrow.
Why would it dissappear the hump for -50 to -30 dB and make it smaller from -30 for both THD and IMD? I have no idea at this point!
For detailed on scaler behavior, please look here:
https://e1dashz.wixsite.com/index/cosmos-scaler
The DO300EX had 5V full scale and the scaler is targeted to 4.5V.
The x-axes is generator level, so you can compute more or less the DAC output and the gain setting from the table of gains in scaler webpage above.
When the graphs are cosmos direct to DAC or via scaler. So, you see that impact of loading more on THD for some reason. The THD swept with scaler is very similar to the same swept with APU Notch.
I think the impact I mentioned of loading the DAC (without scaler) does not enter the play as this loading more effect the harmonics to 19k and 20k, so higher that does not count. I think that the reason no a lot of benefit to IMD with scaler except the clear hump reduction.
The hump is maybe partly shifted because of the amplification I think. Shifted and also much reduced.
https://e1dashz.wixsite.com/index/cosmos-scaler
The DO300EX had 5V full scale and the scaler is targeted to 4.5V.
The x-axes is generator level, so you can compute more or less the DAC output and the gain setting from the table of gains in scaler webpage above.
When the graphs are cosmos direct to DAC or via scaler. So, you see that impact of loading more on THD for some reason. The THD swept with scaler is very similar to the same swept with APU Notch.
I think the impact I mentioned of loading the DAC (without scaler) does not enter the play as this loading more effect the harmonics to 19k and 20k, so higher that does not count. I think that the reason no a lot of benefit to IMD with scaler except the clear hump reduction.
The hump is maybe partly shifted because of the amplification I think. Shifted and also much reduced.
I will check but I am sure it is also loading. You can see the benefit is mostly at low levels, lower than hump interval. This could be also because the ADC has lower performance at low levels where the amplification counts. Near the end of measurement scaler gain is 0 or 7db and the benefit is almost none.
Not sure why the ADC should have poor distortion performance at low level. Usually they are excellent in that range but noise becomes an issue.
Just another idea: in addition to loading the DAC output with ~ 2.5 kOhms and measuring with the Nihtila, maybe you can run them in parallel (and without the extra resistive load)? If you see a difference between resisitive and Cosmos ADC load, this would tell you that the Cosmos ADC injects transients into its load. I haven't seen a schematic for the Cosmos' front end, but I wouldn't be surprised if an op amp with a low impedance feedback network were not able to fully isolate a switched load on its output from its input.
Just another idea: in addition to loading the DAC output with ~ 2.5 kOhms and measuring with the Nihtila, maybe you can run them in parallel (and without the extra resistive load)? If you see a difference between resisitive and Cosmos ADC load, this would tell you that the Cosmos ADC injects transients into its load. I haven't seen a schematic for the Cosmos' front end, but I wouldn't be surprised if an op amp with a low impedance feedback network were not able to fully isolate a switched load on its output from its input.