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

IAtaman · Apr 3, 2023

fpitas said:
I'm not sure. But now I'm curious how that would sound!

I bet it won't sound warm.

fpitas · Apr 3, 2023

IAtaman said:
I bet it won't sound warm.

It might play chill music really well, though.

SIY · Apr 3, 2023

valerianf said:
1) Amp test on resistive load and (simulated?) speaker load.
2) Test of the hiss by placing the hear close to the speaker ( a decibel meter would enable quantifying the result).

Only think that is missing is a thermal measurement with the amp at full load.

The load was an actual loudspeaker, in this case, a sealed box 2-way minimonitor.

I've tried measuring hiss, but at those low levels, room noise overwhelms it. Not the case for some active speakers where the hiss was loud enough to rise above background noise.

tmtomh · Apr 3, 2023

fpitas said:
I tend to believe you. Still, it's an interesting topic that perhaps deserves some attention, since our job here is to obsess over measurements

Indiscriminate skepticism is no more useful than blind faith.

I don't mean to sound harsh - it's just that without a heuristic approach, anything and everything can seem equally plausible and we can potentially ask endless questions without ever figuring out which questions are most important to ask given the limited time and resources we have, and the knowledge that's already been established.

tmtomh · Apr 3, 2023

pma said:
As you know, it is an Oxymoron. Everyone (almost) tests with resistors only and when listening, no speaker is a resistor only.

And music isn't a 20kHz square wave, and humans can't hear ultrasonic frequencies, and we already have measurements of IMD in the Class D amps you obsessively criticize, but that hasn't stopped you, has it?

jimk1963 · Apr 4, 2023

amirm said:
And no one listens to continues sine waves which you used in your testing. And the load you created doesn't match any real speaker whatsoever. You sure you want to keep going this argument?

Remember what I said at the outset. We do this test and then what? What reported audible problem are we catching with it?

Beating-the-dead-horse-dept:

I (re-)submit that the most complete way to test an amplifier is to include a rigorous load-pull, same as we do with RF amps. One doesn't need to re-run every parametric test, but every amp should have its output (in particular) exercised through a wide range of impedance to look for (a) oscillations, (b) aliased noise and/or spurious folded back into the audio band, and (c) any other odd behavior such as a strained SMPS, component whine, etc. When amplifiers exhibit strange artifacts with load variation, it's generally fixable on the design end with some TLC... often but not always involving the power supply, i.e., additional bypassing, etc. It's one indicator of the amplifier's up-front design quality.

Taking an extreme, if @amirm load-pulled my mythical $25K Mark Levinson amp and saw any artifacts of any kind whatsoever, I'd lose my mind. I would be paying $25K for a perfect amp - one that is nearly invariant to load conditions, able to light up a city. Or, I just had a ridiculous amount of cash to burn and wanted to impress my friends. Ahem. A more practical example was posted by @amp, where one amplifier exhibited artifacts under heavy capacitive load while a second amp exhibited virtually none. That test doesn't necessarily tell us of a "problem!", it's just another design quality indicator. And besides, the amazing Class D designers have already achieved such incredible, unbelievable performance, they need a new challenge!!

Maybe one way to quantify load-pull behavior is via a 2-D plot of total in-band total noise/spurious vs. reactance, with reactance on the X-axis. Use a fixed resistance and an assumed frequency (say 4 ohms and 1 kHz for example). Imagine the x-axis starting at -jXX, heading to 0, then up to +jXX. Plot the total in-band noise as reactance is varied; maybe it looks like a bowl - higher at the two extremes, lowest at the purely resistive point in the middle. Maybe lesser amps have a steeper bowl, better ones a more shallow/flat bowl. Dunno, just spit-balling, needs more thought.

damonhill · Apr 4, 2023

amirm said:
Remember what I said at the outset. We do this test and then what? What reported audible problem are we catching with it?

whilst I don't agree with @pma pma test as implemented here, I would have thought the usefulness of testing with loads representative of real speakers would be obvious.

Two different amps. Black traces are the simulated speaker load. Rest are resistive loads.

damonhill · Apr 4, 2023

jimk1963 said:
Beating-the-dead-horse-dept:

I (re-)submit that the most complete way to test an amplifier is to include a rigorous load-pull, same as we do with RF amps. One doesn't need to re-run every parametric test,

For sure. You don't need to test every conceivable load to see if an amp is load invariant or if it will be potentially audibly affected by different speakers.

charlielaub · Apr 4, 2023

@pma I'm coming upon this thread only now, but it is not too old so I thought I would ask:

What is with this crazy load you invented? It's rather pathological. As shown in post #1, the total reactive load above 20k is approaching 1 Ohm! That is below the rated load of the Hypex. How does this relate to a real loudspeaker? (sorry if I missed that in the thread).

@pma Why did you decide to try this load, but no other loads that might more realistically characterize a transducer, most of which are inductive at high frequencies?

For others who are citing equivalent circuit models with large capacitance in them (like whomever linked to this page) seems to fail to see that Rc (e.g. at least a few Ohms) and Lc are both in series with it, so that is not at all like what PMA is using as a load. The last plot at the bottom of the page shows that the impedance magnitude is increasing (inductive) at high frequencies, NOT capacitive.

OP is a real head-scratcher for me...

amirm · Apr 4, 2023

damonhill said:
whilst I don't agree with @pma pma test as implemented here, I would have thought the usefulness of testing with loads representative of real speakers would be obvious.

Two different amps. Black traces are the simulated speaker load. Rest are resistive loads.

View attachment 276921
View attachment 276922

You already see the problem with different resistive loads in the second graph. You don't need the simulated one to know the output impedance is too high. I run such a test on all class D amps and point out when there is an issue like that:

A speaker simulator would show a wiggly response but the point is already made without it.

tmtomh · Apr 4, 2023

amirm said:
You already see the problem with different resistive loads in the second graph. You don't need the simulated one to know the output impedance is too high. I run such a test on all class D amps and point out when there is an issue like that:

A speaker simulator would show a wiggly response but the point is already made without it.

Thanks Amir. I think this is the third or fourth different way you've tried to explain and show that this "simulated real complex speaker load" test does not reveal anything that your existing tests don't already reveal. I don't know why so many folks here are unable or unwilling to grasp this basic information that you keep trying to provide.

Deleted member 48726 · Apr 4, 2023

IAtaman said:
Well, the original premise was interesting in that the amplifier load was presented as a rare but potentially realistic case for a speaker. The more the topic was detailed the more apparent it became that this is not the case. In that regard, the measurements are not really more interesting than connecting your amp to the fridge to see what happens in my opinion.

That's a huge jump to conclusion. We haven't seen a real complex load test. And even if one does it, does that bear precedence to all other amps?

Deleted member 48726 · Apr 4, 2023

tmtomh said:
Indiscriminate skepticism is no more useful than blind faith.

I don't mean to sound harsh - it's just that without a heuristic approach, anything and everything can seem equally plausible and we can potentially ask endless questions without ever figuring out which questions are most important to ask given the limited time and resources we have, and the knowledge that's already been established.

I read this as a rephrasing of turning the blind eye to what potential could be more knowledge. -Which should be every ones priority. Science is also to break new grounds. Maybe it's succesful, maybe it's not. But it's saddening to not try.

Deleted member 48726 · Apr 4, 2023

charlielaub said:
For others who are citing equivalent circuit models with large capacitance in them (like whomever linked to this page) seems to fail to see that Rc (e.g. at least a few Ohms) and Lc are both in series with it, so that is not at all like what PMA is using as a load. The last plot at the bottom of the page shows that the impedance magnitude is increasing (inductive) at high frequencies, NOT capacitive.

I did links to this. -As a way of means to argue for a better load circuit and to question that of OP's. Not to defend OP's test circuit. Quite the contrary. I think that is obvious if one don't just skims the thread but reads the posts.

damonhill · Apr 4, 2023

amirm said:
You already see the problem with different resistive loads in the second graph. You don't need the simulated one to know the output impedance is too high. I run such a test on all class D amps and point out when there is an issue like that:

A speaker simulator would show a wiggly response but the point is already made without it.

Sorry Amir but your understanding is not correct and demonstrated by the plots I already posted.

This plot shows that the various resistive loads (coloured libes) present a relatively flat frequency response in band, yet the reactive load is all over the place.

This plot shows flat response on resistive and reactive loads.

You also have the secondary issue of reactive loads creating excessive heat dissipation in class AB amps or bus pumping in class d.

damonhill · Apr 4, 2023

tmtomh said:
Thanks Amir. I think this is the third or fourth different way you've tried to explain and show that this "simulated real complex speaker load" test does not reveal anything that your existing tests don't already reveal. ( don't know why so many folks here are unable or unwilling to grasp this basic information that you keep trying to provide.

Amirs understanding is wrong.

Deleted member 48726 · Apr 4, 2023

amirm said:
You already see the problem with different resistive loads in the second graph. You don't need the simulated one to know the output impedance is too high. I run such a test on all class D amps and point out when there is an issue like that:

A speaker simulator would show a wiggly response but the point is already made without it.

That "wiggly response" I think shows a lot more useful info than the flat line of the resistive tests. It reveals how and where the sound profile is affected.
Below is from a post from @damonhill (black trace is speaker load). (NO SOURCE @damonhill ??)