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

[damonhill]

What is the source of your data?
Or, at least the output impedance of the amps?

Sorry if I missed where this was already given.

Stereophile. Just a couple of random examples.

 

[tmtomh]

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 ??)

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

I'm not sure. But now I'm curious how that would sound! I bet it won't sound warm.

audiosciencereview.com

View attachment 276951

It only reveals what it will sound like if your actual speaker happens to present the same complex load as the simulated load does. If your speaker is even a little different, then the peaks and valleys of the "wiggle" response shift in frequency and the sound is altered.

And that's part of @amirm 's point: you don't look at an amp's measurements to see exactly how the amp will sound if you connect it to your speakers. You look at measurements to see whether or not the amp's frequency response is linear regardless of load, or whether it has load-dependent nonlinearities.

If it's load-dependent then you probably avoid that amp (unless it's a budget amp and your particular use-case does not require dependably ruler-flat response, like for example in a secondary or background music system).

If you want a particular nonlinear response curve from an amp, well, that's why the good lord invented EQ and DSP.

Amirs understanding is wrong.

Yeah, that's totally what's going on here.

 

[Deleted member 48726]

It only reveals what it will sound like if your actual speaker happens to present the same complex load as the simulated load does. If your speaker is even a little different, then the peaks and valleys of the "wiggle" response shift in frequency and the sound is altered.

And that's part of @amirm 's point: you don't look at an amp's measurements to see exactly how the amp will sound if you connect it to your speakers. You look at measurements to see whether or not the amp's frequency response is linear regardless of load, or whether it has load-dependent nonlinearities.

If it's load-dependent then you probably avoid that amp (unless it's a budget amp and your particular use-case does not require dependably ruler-flat response, like for example in a secondary or background music system).

Of course it changes from load to load. Why are you making this to an argument or feel the need to explain that? That's the whole point of having some sort of representative load circuits for complex loading and the whole turning point of the thread.

Please explain to me how I'm able to deduct that this amp is going to have audible changes with speakers, from the resistive traces alone?

 

[Deleted member 48726]

Stereophile. Just a couple of random examples.

Please link the sources.

 

[valerianf]

It is difficult to find a detailed measurement on the web.
Let us look at these measurement from Stereophile:

https://www.stereophile.com/content/yamaha-mx-d1-digital-power-amplifier-measurements

From a serious manufacturer (Yamaha) we can clearly see that the frequency response is varying with the load up to showing waving in the FRC.
My guess is that the trace that shows the waving variations is the one with the simulated speaker load.
Thus the big question is about the THD+N that may be degraded.
I did not find any THD+N made on a simulated speaker load.

If we want to go to the bottom of this discussion someone needs to make the two THD+N measurements (i.e. a resistive 4 ohm load and a complex 4 ohm load) and then compare the results.
May be there will be a difference if PMA findings are happening again.

 

[tmtomh]

Of course it changes from load to load. Why are you making this to an argument or feel the need to explain that? That's the whole point of having some sort of representative load circuits for complex loading and the whole turning point of the thread.

Please explain to me how I'm able to deduct that this amp is going to have audible changes with speakers, from the resistive traces alone?

View attachment 276959

@amirm already explained it: he provided an example of an amp whose frequency response changed when he tested it with an 8 ohm load instead of a 4 ohm load. (To be clear, that's far from the only amp he's tested that is load-dependent.)

The point he has tried to explain over and over again is that there's nothing magical about a "complex" load - just as there's nothing magical about "complex musical signals" as opposed to test tones. Amir includes a 32-tone test in his reviews to simulate "complex/real musical signals," but he has repeatedly stated - and the 32-tone measurements routinely show - that this test doesn't show anything different than the other tests do, and he has repeatedly noted that he includes this test in his reviews only because some folks have hassled him about it so much - he doesn't think it's necessary (Amir, correct me if I'm mistaken). The gear that has low noise and distortion with test tones shows the same thing with the 32-tone test. And the gear that has high noise, or power supply harmonics leaking into the signal, or rising distortion in the higher frequencies, shows the exact same behavior when he runs the 32-tone test.

Similarly, the point he's trying to explain in this thread is that an amp whose response is load-dependent will show frequency response nonlinearities with resistive loads and with "complex"/"simulated real speaker" loads. The traces will look different than each other, of course, because the complex load's impedance will vary. But the problem - the load-dependence - is clear as day when Amir shows 4 ohm vs 8 ohm frequency response (or occasionally 2 ohm) for amps that have load dependency.

The only logical conclusion one can draw from your unending pursuit of this issue and your constant insistence that we are missing something and being close-minded and "sad" by not agreeing with you, is that you believe there are amps that are load-independent with simple resistive loads but could or would reveal themselves to be load-dependent if only they were tested with more complex loads.

But Amir has already done that on a representative sample of amps and it hasn't shown anything different than the resistive load tests have. You just don't want to believe that - you seem to think that we can't rule it out unless every amp is tested with such loads. At some point the burden is on you to establish some plausible case for the probability of that outcome. "You never know until you try" is not, in fact, an adequate summary of the scientific method.

Finally, I have to say that this thread is a great example of why I am so critical of @pma's approach to discussing the issues and how he draws conclusions from the tests he runs: I can't think of any other member here whose posts generate such deep, extended amounts of confusion and muddying of the waters.

 

[jimk1963]

The only logical conclusion one can draw from your unending pursuit of this issue and your constant insistence that we are missing something and being close-minded and "sad" by not agreeing with you, is that you believe there are amps that are load-independent with simple resistive loads but could or would reveal themselves to be load-dependent if only they were tested with more complex loads.

This is not the conclusion drawn. You are stating that by testing an amp with two resistive loads, which show load variance, the amp is therefore fully characterized and let’s call it a day. This is a fundamental misunderstanding of how reactive elements affect an active circuit, in a different manner than resistive loads. For example, using @amp ‘s extreme load, by your logic I could replace the magnitude of its impedance at frequency X with an equivalent resistance of the same value, and get the same result. That’s not accurate. Reactive loads impart phase shift, and phase shift in any system with feedback alters behavior. It’s instructive to understand this load sensitivity, no different than with any other amplifier design.

 

[amirm]

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.

Only for that sample load which can't possibly represent more than one speaker. My measurement on the other hand shows that across all loads there will be variations. Showing it for this and that complex load will only be random examples, not instructive of anything.

 

[amirm]

This is not the conclusion drawn. You are stating that by testing an amp with two resistive loads, which show load variance, the amp is therefore fully characterized and let’s call it a day. This is a fundamental misunderstanding of how reactive elements affect an active circuit, in a different manner than resistive loads. For example, using @amp ‘s extreme load, by your logic I could replace the magnitude of its impedance at frequency X with an equivalent resistance of the same value, and get the same result. That’s not accurate. Reactive loads impart phase shift, and phase shift in any system with feedback alters behavior. It’s instructive to understand this load sensitivity, no different than with any other amplifier design.

Once more.... PMA's load doesn't represent any real speaker. And any sample complex load you pick, will only show one speaker's response, assuming it is perfectly emulating one which most likely it is not.

My test however clearly and nicely shows load dependency. Actual impact will be speaker dependent which you can measure acoustically for your setup. Or better yet, buy an amp that shows no dependency.

 

[jimk1963]

Once more.... PMA's load doesn't represent any real speaker. And any sample complex load you pick, will only show one speaker's response, assuming it is perfectly emulating one which most likely it is not.

My test however clearly and nicely shows load dependency. Actual impact will be speaker dependent which you can measure acoustically for your setup. Or better yet, buy an amp that shows no dependency.

Your last line sums it up - except you won’t know that without running a load-pull test beyond the resistance line. You provided an earlier example showing peaking at 8 ohms resistance. You’re implying that if no peaking had occurred with 8 ohms resistance, the amp is therefore mostly load invariant. But in reality, that same amp might show peaking with a reactive load of similar magnitude impedance. Further, as noted over and over, it’s not only the swept frequency response that’s important - it’s also other nonlinear effects that might fold noise back into the audio spectrum. Especially with Class D’s architecture. It’s not so difficult to acknowledge that one will know more about an amp’s overall behavior by doing basic load-pull testing. For the life of me, I can’t understand why this is even controversial.

 

[IAtaman]

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?

I am not talking about whether complex load tests would reveal relevant amp characteristics - people with a lot more experience than I do are talking about that, I am simply stating a cap parallel with a resistor is not a realstic load for an audio amp.

 

[amirm]

You’re implying that if no peaking had occurred with 8 ohms resistance, the amp is therefore mostly load invariant. But in reality, that same amp might show peaking with a reactive load of similar magnitude impedance.

Well, show that and then we can talk. Right now the facts are clear: if the class D obligatory filter is in the feedback loop in which case, frequency response is independent of the load. Otherwise it is load dependent. My tests clearly show this effect and as such, are highly diagnostic without inviting arguments. After all, 8 ohm resistive is a valid reference as much as 4 ohm is.

I have also added reactive load stress testing for amplifiers. That item if we keep going, will cost me nearly $15,000. So wishes are not free.

 

[damonhill]

Only for that sample load which can't possibly represent more than one speaker. My measurement on the other hand shows that across all loads there will be variations. Showing it for this and that complex load will only be random examples, not instructive of anything.

It doesn't matter that it only represents one reactive load. Your measurent showed there was a variation in FR out of band. it does not show there will be variations across all loads at all, or within the relevant in band region.

My example showed an amp that was fine with resistive loads, yet went bananas with a reactive load. So it is very instructive where your example isn't.

 

[damonhill]

You’re implying that if no peaking had occurred with 8 ohms resistance, the amp is therefore mostly load invariant. But in reality, that same amp might show peaking with a reactive load of similar magnitude impedance.

This is absolutely correct.

 

[amirm]

It’s not so difficult to acknowledge that one will know more about an amp’s overall behavior by doing basic load-pull testing. For the life of me, I can’t understand why this is even controversial.

It is controversial because of the impact it will have on the industry. Let's make the case simpler. What if I wake up tomorrow and say all amps need to handle 1 ohm load and if they don't, I am going to ding them. You think this is proper? People should avoid buying an amp that can't handle 1 ohm load? How about amplifiers costing more money to be able to drive a 1 ohm load? Not good either.

So far, vast majority of what I have pushed the industry to do has come at basically no cost. What you and OP are asking for is not going to be that way. So better come back with strong backing of what is being asked. A random dummy load is not remotely useful test in this regard.

 

[damonhill]

My tests clearly show this effect and as such, are highly diagnostic without inviting arguments.

They don't. I have shown clearly above that an amp can appeR to be load invariant with resistive loads but seriously affected by reactive loads.

 

[amirm]

It doesn't matter that it only represents one reactive load. Your measurent showed there was a variation in FR out of band. it does not show there will be variations across all loads at all, or within the relevant in band region.

My example showed an amp that was fine with resistive loads, yet went bananas with a reactive load. So it is very instructive where your example isn't.

View attachment 276980

Nope. It is you that is not understanding the topic or that measurement. First, show the link to the reviews on stereophile and not just the graphs. Here is the link: https://www.stereophile.com/content/primaluna-dialogue-seven-power-amplifier-measurements

Right there, JA explains what is going on which is what I explained:

"In 8 ohm impedance feeding an 8 ohm load transfers more power to the load than would a greater or smaller impedance. However, there is a price to be paid: because loudspeaker impedances vary considerably with frequency, high output impedances result in significant modification of the amplifier's frequency response, due to the Ohm's Law interaction between the amplifier and loudspeaker impedances. This is illustrated by the gray trace in fig.1, which shows the DiaLogue Seven's frequency response from the 4 ohm tap in ultralinear mode into Stereophile's standard simulated loudspeaker."

As I already explained, the response changes due to high impedance of the amplifier, in this case 8 ohm. This clearly shows up in response differences of 4 vs 8 ohm. Ohm's law works there and as JA says, it works exactly the same in a varying impedance load. At any point in that varying load, you can use the simple Ohm's law to compute the impact. No new insight is gained other than application of Ohm's law to resistive vs reactive load. This is why i said the impact on your actual speaker will be different. Best to avoid these amps with high impedance. But if you can't avoid it, then use acoustic measurements to deal with it. Nothing is gained by looking at at that simulated line. The only insight was 8 ohm impedance of the amplifier. Period.

 

[damonhill]

My test however clearly and nicely shows load dependency.

It doesn't. It doesnt show any massive problem in band.

 

[amirm]

It doesn't. It doesnt show any massive problem in band.

Again, you don't understand. Class D amplifiers with load dependency have variable output impedance that is frequency dependent. My resistive tests shows this very clearly. You can compute the impedance vs frequency from my measurements. Once there, the exact same thing will apply to a reactive load. To the extent the impedance of a class D amp is very low at low frequencies, there is no impact with resistive load or reactive one at lower frequencies. So there is nothing to see "in band." The issue is limited to high frequencies which once again, my test shows.

 

[jimk1963]

It is controversial because of the impact it will have on the industry. Let's make the case simpler. What if I wake up tomorrow and say all amps need to handle 1 ohm load and if they don't, I am going to ding them. You think this is proper? People should avoid buying an amp that can't handle 1 ohm load? How about amplifiers costing more money to be able to drive a 1 ohm load? Not good either.

So far, vast majority of what I have pushed the industry to do has come at basically no cost. What you and OP are asking for is not going to be that way. So better come back with strong backing of what is being asked. A random dummy load is not remotely useful test in this regard.

I’m not “asking” for anything. You’ve gotten very defensive over your present test methods, which was not my intent at all. I’m simply pointing out what any amplifier circuit designer knows very well - namely, amplifiers (and other active circuits) can and will do strange things with reactive loads. This is very, very well known in the art. At RF, we can easily and quickly check a very wide range of impedances using line stretchers. I have zero idea how one might go about it at baseband, which I stated early on but got no reply from you. (Plenty of replies from others, who aren’t the ones doing this great ASR testing.)

If there is a way to easily, accurately, repeatably do a load pull check in your audio amp testing, then great. Like that load pull box someone mentioned. If there’s no easy way to do this, or you’re not inclined to for any other reason, that’s perfectly fine too. This is free information to all of us, please don’t think it is unappreciated for one second. It’s awesome, valuable info! My experience-based dialogue here was meant to stimulate, not generate hate. Apologies if it went off the rails too far. Please, carry on!