Amplifier "sample test reviews" using Paul's Multitone Analyzer software

[pkane]

Looking forward to X-axis cal option in [W] .

Give this version a try, Pavel: https://www.audiosciencereview.com/...loopback-analyzer-software.27844/post-1478232

 

[pma]

Give this version a try, Pavel: https://www.audiosciencereview.com/...loopback-analyzer-software.27844/post-1478232

Thank you I did, but got a very very powerful amplifier suddenly! (joke) .



In the top right corner box there is a correct max. power of some 282W, however the plot says 1300W ...
Log scale on X-axis for watts should be used, scale setting check box does not do it. But Paul, I am not pushing on you, for DIY purposes the SW is OK, for semi-professional I am continuing in REW use.

[V] sweep seems to have X-axis calibrated in Vpeak rather than Vrms.

 

[pma]

Cause and effect is pretty clear from those, thanks Pavel! Jives with the earlier plots that imply it is mostly distortion, and I agree looks like supply drop at low frequencies.

More on the subject Don, as it is interesting, at least to me. Level step is now 0.4dB, and THD, N and THD+N plots vs. power posted. As one can see, it is useful to have them separated, because THD+N (SINAD) vs. power plots sometimes do confuse the real issue. I will also post these findings into my NC252MP thread.

 

[DonH56]

More on the subject Don, as it is interesting, at least to me. Level step is now 0.4dB, and THD, N and THD+N plots vs. power posted. As one can see, it is useful to have them separated, because THD+N (SINAD) vs. power plots sometimes do confuse the real issue. I will also post these findings into my NC252MP thread.

View attachment 263286

View attachment 263287

View attachment 263288

View attachment 263289

View attachment 263290

Yes, interesting indeed. My observations, which I have no doubt you have already made:

• THD is nearly flat from a few watts up to clipping, and roughly equal to the noise around 30 W. Classic behavior would have THD falling at lower power, but for these curves I see dBr, so assume referenced to the power level on the x-axis and not full power?
• Noise is relatively independent of frequency, but THD is higher at lower frequency, especially at 10 W and above. Supports that performance is limited by THD and not noise more so at lower frequency. I suspect the dip around 130 W in the 5 kHz curve is due to a happy coupling or resonance cancelling the distortion.
• 1 kHz and 200 Hz are pretty close until >100 W or so, then 200 Hz rises closer to the 22 Hz curve. That seems to explain the rise ("hump") in distortion below 800 Hz seen before.
• By eye, the 20 Hz clipping waveform appears more asymmetrical than the 1 kHz clipping waveform, with a bit more flattening or drooping of the peak past that first little corner at the bottom of the waveform, and is worse on the bottom. That would seem to support your conclusion that the power supply is giving up and sagging.
  • If so, a longer-term test at higher frequency might also show dropping power/increasing distortion over time.
  • A two-tone test using 20 Hz and 1 kHz might be interesting, or even a 100 Hz square wave, but probably not show anything more useful.

Really cool stuff, thank you very much for digging more into this!

 

[pma]

Noise is relatively independent of frequency, but THD is higher at lower frequency, especially at 10 W and above. Supports that performance is limited by THD and not noise more so at lower frequency. I suspect the dip around 130 W in the 5 kHz curve is due to a happy coupling or resonance cancelling the distortion.

Class D behaviour is different from Class A or Class AB, we (probably both of us) are more used to live with linear amplifiers and test them. NCore has quite complicated feedback loop too handle class D issues like dead time etc., please see here and here. It might be interesting to make more detailed investigation of Purifi as well.

 

[DonH56]

Class D behaviour is different from Class A or Class AB, we (probably both of us) are more used to live with linear amplifiers and test them. NCore has quite complicated feedback loop too handle class D issues like dead time etc., please see here and here. It might be interesting to make more detailed investigation of Purifi as well.

Agreed, my last class D design (for RF) was many years ago, and audio before that, so I've not kept up. Thanks for the links, more required reading, when I have some spare time.

 

[pkane]

Thank you I did, but got a very very powerful amplifier suddenly! (joke) .

View attachment 263284

In the top right corner box there is a correct max. power of some 282W, however the plot says 1300W ...
Log scale on X-axis for watts should be used, scale setting check box does not do it. But Paul, I am not pushing on you, for DIY purposes the SW is OK, for semi-professional I am continuing in REW use.

View attachment 263299

[V] sweep seems to have X-axis calibrated in Vpeak rather than Vrms.
View attachment 263302

An honest mistake The X axis was using the generator level instead of the input. Will be fixed in the next beta version:

https://www.audiosciencereview.com/...loopback-analyzer-software.27844/post-1479545

 

[pma]

An honest mistake The X axis was using the generator level instead of the input. Will be fixed in the next beta version:

https://www.audiosciencereview.com/...loopback-analyzer-software.27844/post-1479545

Paul, it is much better, however still we have some minor issues.

1. Below 1W, the X-axis power scale does not show decimal point values, it is rounded and truncated to 0, 1.
2. I had to change load input in the chart from 4ohm to 2ohm to get proper power value with 4ohm load. It seems that the SW shows half the actual power. But the THD and TD+N readings are correct - important.

REW plot:

 

[pma]

Class D behaviour is different from Class A or Class AB, we (probably both of us) are more used to live with linear amplifiers and test them. NCore has quite complicated feedback loop too handle class D issues like dead time etc., please see here and here. It might be interesting to make more detailed investigation of Purifi as well.

Agreed, my last class D design (for RF) was many years ago, and audio before that, so I've not kept up. Thanks for the links, more required reading, when I have some spare time.

The same set (THD+N, THD, N vs. power) measured with a linear class AB amplifier, MOSFET output stage with error correction. The plots are more predictable and clipping behaviour is less violent than in case of the class D amp.