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

[DonH56]

One thing I have always wanted to see in measurements (but rarely have) is an impedance plot showing the actual (effective) inductance and capacitance over frequency and not just the magnitude and phase. I used to do that with a VNA but no longer have access (and the last one I had access to would not reach down to the audio range).

 

[theREALdotnet]

I used to do that with a VNA but no longer have access (and the last one I had access to would not reach down to the audio range).

I’d love the NanoVNA, such a nifty and cheap device, unfortunately it only goes down to about 50kHz.

 

[SIY]

One thing I have always wanted to see in measurements (but rarely have) is an impedance plot showing the actual (effective) inductance and capacitance over frequency and not just the magnitude and phase. I used to do that with a VNA but no longer have access (and the last one I had access to would not reach down to the audio range).

It seems to me that this is derivable from the mag/phase data and a little Excel manipulation? Personally, I'd like to see the return of Nyquist/Heyser impedance plots.

 

[theREALdotnet]

It seems to me that this is derivable from the mag/phase data and a little Excel manipulation? Personally, I'd like to see the return of Nyquist/Heyser impedance plots.

Any linear network is completely characterised by magnitude and phase response, but, and I go out on a limb here, I think it’s impossible to pin down a network from a given magnitude and phase response, because there are probably an infinite number of possible networks to fit the bill.

 

[Hayabusa]

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.

View attachment 273767

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.

I have measured the NC400 with Final model 15 electrostatic as load an did not see this behavour.
isn't the start and the stop of the burst the problem? It contains much higher harmonics because of the discontinuity?

 

[SIY]

Any linear network is completely characterised by magnitude and phase response, but, and I go out on a limb here, I think it’s impossible to pin down a network from a given magnitude and phase response, because there are probably an infinite number of possible networks to fit the bill.

Thus my meh attitude about “equivalent capacitance” or whatever. Networks are just models, and there’s certainly more than one way to model a load. Even more when you add in non-classical elements like Warburg. One of the things I do in research is construct impedance models for electrodes (EIS) and there’s always spirited debate on the physicality of those equivalent circuits…

 

[DonH56]

It seems to me that this is derivable from the mag/phase data and a little Excel manipulation? Personally, I'd like to see the return of Nyquist/Heyser impedance plots.

It is, but I'd have to manually enter the data from the curves as most reviews do not provide tabulated data. Nyquist plots (or the 3-D Heyser plots) would provide it in a different way, natch, as would a Smith Chart that I've probably used more. But I'm lazy.

In this day of Web links for everything (else), a link to the raw data would be nice to include in the measurements.

Any linear network is completely characterised by magnitude and phase response, but, and I go out on a limb here, I think it’s impossible to pin down a network from a given magnitude and phase response, because there are probably an infinite number of possible networks to fit the bill.

But for the load an amp sees, we (or maybe just I) only care about the terminal impedance, not what's inside the black box. All I want is the driving-point impedance.

And the whole premise of linearity is open for question, of course, given various non-linear aspects of most any impedance over various driving and environmental conditions. I choose to ignore those for the most part, at least to first order and assuming nothing is overdriven.

 

[theREALdotnet]

But for the load an amp sees, we (or maybe just I) only care about the terminal impedance, not what's inside the black box. All I want is the driving-point impedance.

Yes, exactly. If you don’t want to (or cannot for any reason) test the amp with your actual speaker and have to resort to a replacement “model” network, then you better be prepared to make a network with knobs so you can adjust it depending on frequency. Or be prepared to make a very complex network.

The Smith chart of a speaker does’t provide much help in designing a simulated load. But one thing I think it would show is that speakers don’t tend towards zero impedance at high frequencies.

 

[atmasphere]

In a class D self-oscillating amplifier you have only two issues that affect how the amp behaves with a capacitive load. I covered this earlier!
The first is the resonance that occurs between the output filter choke and the capacitive load.
The second is the feedback.
Self-oscillating class D amps apply so much feedback that the phase margin of the circuit is exceeded and the amp goes into oscillation. The feedback network is designed to only allow one oscillation frequency, which will vary around the center point as the amplifier makes power. That frequency is of course the switching frequency of the circuit. In case its not obvious, this allows a lot of feedback (+35dB!) with excellent stability. The old NCore module uses north of 50dB of feedback with some clever techniques to prevent bad behavior when the amp is clipped (which stops switching), allowing only one frequency solution when the switching restarts.

When you run that kind of feedback you don't get the weird stuff claimed on this thread; any resonances are controlled by the feedback. You can get interactions as I previously outlined; its to the credit of the circuit that its able to handle it, especially then the circumstances as pointed out can never be real-world.

Such an amplifier works fine on ESLs when capacitance load really is a thing. Those class D amps that are not self-oscillating can get into trouble on such loads. But not this one!
Can we move on now?

 

[DonH56]

Yes, exactly. If you don’t want to (or cannot for any reason) test the amp with your actual speaker and have to resort to a replacement “model” network, then you better be prepared to make a network with knobs so you can adjust it depending on frequency. Or be prepared to make a very complex network.

Power Cube... But, I was thinking more of measuring the impedance, not modeling it. I agree that emulation and modeling can get tricky using real-world components, but we can usually get pretty close with a reasonable number of components. Especially if you give up sweeps or have a controllable impedance load; then, set the impedance to match the speaker at the test frequency, and run the test. This is only truly valid for single-frequency testing, natch, and arguably not even then as harmonics (e.g. when clipping) or oscillation/ringing frequencies may not see the proper impedance.

The Smith chart of a speaker does’t provide much help in designing a simulated load. But one thing I think it would show is that speakers don’t tend towards zero impedance at high frequencies.

I was kidding about the Smith Chart; I do not believe I have ever tried to use it for an audio circuit. Especially for speakers, what would your reference impedance be? Regular old H/Z/Y and Nyquist plots work well enough for me for audio circuits.

 

[Head_Unit]

If the manufacturer specify the hiss of its active speakers

Ya know, why the @#$% would a manufacturer ever ever EVER release an active speaker where you can hear the hiss?!? I do not get that, it is just ridiculous. Some interesting monitor candidate I was reading a review of mentioned hiss, very disappointing, no way for me.

 

[Head_Unit]

Consumption concerns are better suited for some 'ecology' chapter of the forum, not mixed with sound evaluation.

Changing the world to Class D from AB is hardly going to stop the greenhouse warming crisis, yet I think don't efficiency is irrelevant if you are trying to get a lot of power out of the wall, particularly as I suspect few people have dedicated circuits for their audio, and more particularly in the case of multichannel listening. Nor do I see why all non-sound concerns should be shunted off elsewhere, since in real life things like aesthetics DO matter to almost all (despite denial from some...beauty is in the eye of the beholder ha ha)

 

[Overseas]

I do not stare at my amp.
If one does, SINAD is not going to help.
Sound/ music and looks/ paintings are 2 different arts.

 

[pma]

It should belong to this thread as well. I was doing tests with 4.7ohm//33uF load, which at 1kHz represents 3.3ohm impedance magnitude and -45° phase. Please take into account that only at 1kHz. Parallel 4.7ohm//33uF combination equals to serial combination of 2.41ohm + 67.75uF at 1kHz. The amp was doing well with this load, at 1kHz, and the result was posted here.
It says that the 33uF capacitor did not send the amp to oscillations and the high order feedback loop was working well.

So I have decided to re-measure the amp with the original load of this thread, 4.7ohm//2.2uF, hoping that some miracle might have happened to make it stable. But it did not. The result was same bad as in this thread, and the new measurements are posted here.

So, it is confirmed, without any doubt, by independent measurements with an audio analyzer, then with a fast digital oscilloscope, and finally by ears and speaker, that the NC252MP is unable to work with the 2.2uF load properly. The high order feedback loop goes to oscillations near 70kHz, there are audible by-products near 6-8kHz, the noise in audio band rises from original tens of microvolts to audible millivolts and the distortion accelerates of 60 dB higher than normally. So there are areas of capacitive load range where the NC252MP is able to work properly and another areas where it is not able to work properly. Neither usual reviews nor datasheets would give you this info.

 

[hoppy IPA]

It should belong to this thread as well. I was doing tests with 4.7ohm//33uF load, which at 1kHz represents 3.3ohm impedance magnitude and -45° phase. Please take into account that only at 1kHz. Parallel 4.7ohm//33uF combination equals to serial combination of 2.41ohm + 67.75uF at 1kHz. The amp was doing well with this load, at 1kHz, and the result was posted here.
It says that the 33uF capacitor did not send the amp to oscillations and the high order feedback loop was working well.

So I have decided to re-measure the amp with the original load of this thread, 4.7ohm//2.2uF, hoping that some miracle might have happened to make it stable. But it did not. The result was same bad as in this thread, and the new measurements are posted here.

So, it is confirmed, without any doubt, by independent measurements with an audio analyzer, then with a fast digital oscilloscope, and finally by ears and speaker, that the NC252MP is unable to work with the 2.2uF load properly. The high order feedback loop goes to oscillations near 70kHz, there are audible by-products near 6-8kHz, the noise in audio band rises from original tens of microvolts to audible millivolts and the distortion accelerates of 60 dB higher than normally. So there are areas of capacitive load range where the NC252MP is able to work properly and another areas where it is not able to work properly. Neither usual reviews nor datasheets would give you this info.

can you please list some speakers that would fail this test you made up?

 

[pma]

can you please list some speakers that would fail this test you made up?

I have already done it several times, so again -


It is a design flaw. Good amplifier must handle 2uF in parallel with 4ohm. Period.

 

[hoppy IPA]

I have already done it several times, so again -

View attachment 292267 View attachment 292266
It is a design flaw. Good amplifier must handle 2uF in parallel with 4ohm. Period.

so $20k/pair speakers with a $600 amp yeah? literally nobody is doing this. i surely wouldn't pair a 150w amp with speakers rated to 500w.

 

[Deleted member 48726]

so $20k/pair speakers with a $600 amp yeah? literally nobody is doing this. i surely wouldn't pair a 150w amp with speakers rated to 500w.

I find the price argument silly. What does price have to do with it?

I would rather question the probability for other speakers than these to drive the amp to oscillation. I don't recall other impedance / phase graphs that are like this. And if these are just about the only ones that does it, then why should we care about this particular issue if it only exists with these particular speakers?

But I'm still adcocating for complex load testing of all amps as it reveals some behaviour that explains that you may not be able to pair any speaker to any amp and get good results.

 

[SIY]

so $20k/pair speakers with a $600 amp yeah? literally nobody is doing this. i surely wouldn't pair a 150w amp with speakers rated to 500w.

More appropriately, we're talking about an extremely tiny niche of speakers, none of which will be driven at those powers at those frequencies. Assuming the measurements are correct (and I have a few doubts), an amp that does fine with 99.99% of speakers is certainly a good amp. I have three different Hypex (or Hypex licensed nCore) amps here and they do very well driving both dynamic and electrostatic speakers.

 

[DonH56]

so $20k/pair speakers with a $600 amp yeah? literally nobody is doing this. i surely wouldn't pair a 150w amp with speakers rated to 500w.

Literally I am doing that. So I guess that is one... Price and rated speaker power handling are but two of many factors in determining what amplifier and speakers to buy and pair.