NC252MP (class D) vs. A250W4R (classAB) burst measurements into 4ohm//2.2uF load

[pma]

NC252MP (class D) vs. A250W4R (classAB) burst measurements into 4ohm//2.2uF load

As I have been doing, quite intensively, wide range of power amplifier measurements, I have decided to compare 2 different power amplifier design philosophies, class AB and class D. The amplifiers chosen have equal rated power, 2 x 250W at 4ohm load. The main difference, except for class of operation, is that NC252MP uses a common power supply for both channels, however A250W4R are in fact 2 independent amplifiers in one case, completely separated, with 2 independent linear power supplies (2 transformers and two rectifiers/filters). So each channel of A250W4R "does not care" if the other one is loaded or not.

The bursts were generated with DAC at 96kHz sampling rate. It can be seen in the class AB measurements, in the class D amplifier this fact is masked by its high ultrasound noise.

The amplifiers under test are described here and here.

1. Complex load used

The load used for the test is 4ohm resistor in parallel with 2.2uF polypropylene capacitor, KEMET automotive type with max. allowed dv/dt = 200V/us. Below please see measured impedance of this load:

At 20kHz, we have 2.762ohm impedance magnitude, -40° phase and 1.07ohm EPDR (equivalent peak dissipation resistor).

2. Measurements with 1kHz burst
(Maximum allowed distortion is 1%)

A250W4R maximum 1kHz burst power is 2 x 308W/4ohm, one or two channels driven (no difference)

NC252MP max. 1kHz burst power is 2 x 234W/4ohm (distortion 0.4%)


2. Measurements with 10kHz burst

A250W4R max. 10kHz burst power is 2 x 299W/4ohm (load 4ohm//2.2uF)

NC252MP max. 10kHz burst power is 2 x 227W/4ohm (load 4ohm//2.2uF)


3. Measurements with 20kHz burst

A250W4R max. 20kHz burst power is 2 x 298W/4ohm (load 4ohm//2.2uF). With respect to EPDR, max. burst power into this load is 2 x 1112W/1.07ohm.

NC252MP cannot handle the same input burst amplitude as at 1kHz or 10kHz. Distortion rises to 6.4% (H3). The shape is not sinusoidal.

Reduced amplitude for NC252MP 20kHz burst

With burst amplitude reduced to get 20.5Vrms output from NC252MP, the amplifier works "better", but still unacceptable.


4. Conclusion

The "beefy" class AB amplifier has a bit higher distortion with traditional measurements at 1kHz (with 20kHz BW), but it handles much better the complex load with capacitive component of -40°/20kHz. It is also more powerful because of two separated power supplies, though the rated power is equal.
4ohm//2.2uF load is not good for NC252MP. It activates oscillations near to 80kHz, as an interaction of NC252MP output filter with the 2.2uF capacitor. Care with complex load with such amplifiers.

There was a different windowing of 10kHz/20kHz measured spectra for A250W4R and NC252MP, for the reason that A250W measurements are older and taken with Picoscope 6 SW (older version). So the spectral resolution with Picoscope 6 was lower at lower frequencies, thus the different shape below 50kHz.

The longer I test test the Hypex NC252MP, the more my initial passion for this class D amplifier is fading away and the more my scepticism to class D is growing.
Next step would be capturing music files from the output of both amplifier loaded with 4ohm//2.2uF and the ABX test. Output power will be up to 2 x 200W/4ohm.

------------------------------------------

Appendix:

NC252MP comparison of THD+N vs. power at 1kHz into 4ohm resistor load and 4ohm resistor in parallel with 2.2uF capacitor load

This is the proof that the NC252MP is unusable with 4ohm//2.2uF load (simulation of the worst-case elstat. speakers)

 

[MaxwellsEq]

Your work on this is fascinating. It may also be a turning point in identifying what measurements are sufficient when testing power amplifiers. Thank you.

 

[Matias]

So we are condemning the module for failing a 20 kHz burst of 250W. How is this anything close to real music listening? On average real music has -40 dB at 20 kHz...

 

[pma]

So we are condemning the module for failing a 20 kHz burst of 250W. How is this anything close to real music listening? On average real music has -40 dB at 20 kHz...

I am afraid you do not understand the point. The module is unable to handle the load used. Even if you lower the amplitude, it does not work. Check this plot, you can see all kinds of distortion in the burst.

With zero input signal, the NC252MP is happily oscillating at 66kHz, 1.7Vrms, with 4ohm//2.2uF load. Do you think it is fine? Me, I do not think so.

Years ago I tested NC400, with R//C load. Same behaviour. It is a design issue and it is inevitable for the NC concept. The additional capacitance changes FB behaviour and makes it oscillating.

 

[Matias]

I see, thanks for clarifying. Yes, that is indeed undesired, but still above audibility, so no big deal imo.

 

[Sokel]

Ok,to make it easier to understand,what that 2.2uF load would represent in real world conditions?

 

[pma]

Ok,to make it easier to understand,what that 2.2uF load would represent in real world conditions?

Some of the elstat speakers, quite exactly.

 

[Schollaudio]

Looks like we need a double blind listening test too.

 

[jhenderson0107]

The vast majority of speakers have negligible output above 40 kHz. So this is much ado about nothing. It is not a problem in actual practice.

 

[Sokel]

The vast majority of speakers have negligible output above 40 kHz. So this is much ado about nothing. It is not a problem in actual practice.

Yes but here in ASR engineering excellence is a serious top subject too and it's nice to follow the progression as we do with dacs which came a time where audibility issues are far but engineering is always very important.

 

[Kachda]

What’s the cost difference between the two?

 

[MCH]

What’s the cost difference between the two?

The cost, no idea, but the weight i bet ca. 20kg difference

 

[jhenderson0107]

Yes but here in ASR engineering excellence is a serious top subject too and it's nice to follow the progression as we do with dacs which came a time where audibility issues are far but engineering is always very important.p

Perhaps a 60 kHz oscillation is interesting from a scientific or purist perspective. But it’s is irrelevant from an engineering (practical) standpoint. Since the typical speaker load cannot respond to a 60 kHz stimulus, it is not relevant to the analysis of an audio amplifier. So, why measure it and portray this as a “defect” compared to another amplifier topology?

 

[Matias]

Here is another test I propose: 2 test signals, one full scale at say 80 Hz and another at -40dB 20 kHz. Output this to burst around 200W into the load. Measure the remaining artifacts in the audible range, not only in THD % but also total watts sent to the drivers. This should be interesting.

 

[pma]

Some examples of speakers that would be a pain for NC amplifiers

Sound-Lab A-1 electrostatic loudspeaker Measurements

Sidebar 6: Measurements

www.stereophile.com

This is more difficult than my dummy 4R//2.2uF load

Another one:

MartinLogan Montis loudspeaker Measurements

Sidebar 3: Measurements I used DRA Labs' MLSSA system and a calibrated DPA 4006 microphone to measure the MartinLogan Montis's frequency response in the farfield, and an Earthworks QTC-40 mike for the nearfield and spatially averaged room responses. A note on these measurements is in order:

www.stereophile.com

At 10kHz, a capacitance of 10uF.

 

[pma]

I hope that at least someone will understand that the assumption "if SINAD is very good, everything is good" is not true and is not based on facts, at least in case of power amplifiers. My recommendation would be - never make a choice of power amplifier based on 5W/1kHz SINAD charts!!

 

[jhenderson0107]

I hope that at least someone will understand that the assumption "if SINAD is very good, everything is good" is not true and is not based on facts, at least in case of power amplifiers. My recommendation would be - never make a choice of power amplifier based on 5W/1kHz SINAD charts!!

I agree - one should never make a choice of power amplifier based on 5W/1kHz SINAD charts.

Since the distortion in the electronics chain is ~1000 times lower than the speaker/room chain, there's no need to fret over streamer/preamp/dac/power amplifier SINAD charts!! Instead choose devices for your electronics chain which meet your budget, mechanical constraints, ergonomic preferences, output power requirements, and aesthetic since nearly all perform adequately and well beyond audible thresholds.

 

[amirm]

There was a different windowing of 10kHz/20kHz measured spectra for A250W4R and NC252MP, for the reason that A250W measurements are older and taken with Picoscope 6 SW (older version). So the spectral resolution with Picoscope 6 was lower at lower frequencies, thus the different shape below 50kHz.

Why is this in the "conclusion" section instead of intro before the graphs? What does it mean? You don't have both amps side by side to test the same way?

 

[amirm]

I hope that at least someone will understand that the assumption "if SINAD is very good, everything is good" is not true and is not based on facts, at least in case of power amplifiers.

No one has said that. What is said by me is that SINAD is a predictor of the performance of the product at the extremes. A superb SINAD predicts, but does not guarantee, that the rest of the performance would be excellent. If it were guaranteed, I would not have run the test of the tests. If this was the motivation for you to do these corner tests, then it was for not.

You also haven't got enough evidence here to matter. You have tested only one class D amplifier. That does not mean all class D amps are behaving this way. They may but you better do your homework before making sweeping statements.

 

[amirm]

With zero input signal, the NC252MP is happily oscillating at 66kHz, 1.7Vrms, with 4ohm//2.2uF load. Do you think it is fine? Me, I do not think so.

The whole amp is oscillating at its switching frequency which is part of its functionality. So finding oscillations is like saying day is brighter than night. We don't worry about the switching frequency oscillations in exchange for superb performance elsewhere.