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

[tmtomh]

What is difficult for me as a layman (without electronics background) is to reconcile such statements with the real measurements of other reviewers.

The differences are not extreme, but still clearly present. The speaker load that Archimago uses is challenging in the low frequency with up to just under 60 ohms, but around the typical crossover frequency to the tweeter with just over 15 ohms not challenging, up to30-60 ohms are quite realistic. In the range above 2 kHz, a speaker load of about 60 Ohm would lead to significantly larger deviations than in the review.

Looking at just the 4 and 8 ohm FR measurements, the amp is just acceptable to me, but with the speaker load, the frequency response deviation in the frequency range below 10kHz is too much for me. So for me as a layman, the speaker load provides additional information, because I could not have foreseen these deviations in the FR.
View attachment 282654

MEASUREMENTS: Fosi Audio TB10D [Upgraded Version] stereo amplifier (TI TPA3255 Class D). Ethernet is inherently "isolated".

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Speaker load:
View attachment 282655

The "Measurement Of Amplifiers Rig" (MOAR): Standard Tests, Loopback results, and the AMOAR Score.

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

useful for Class-A amps too:
View attachment 282658

REVIEW / MEASUREMENTS: Pass Labs / First Watt SIT-2 stereo Class A amplifier. [And Stereophile steps in with MQA yet again...]

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Thanks for this - very interesting!

Again, I'm not an expert, but there is one point that I think is extremely important here: both of these amps are load-dependent amps, and we don't need the Sony speaker to reveal that.

The first one, the Fosi, is a TPA3255-based amp, and all (or at least the vast majority of) such amps are known to be load-dependent. It is very true, as @Holdt says, that there's no way we can predict the Fosi's frequency response curve into that Sony SS-H1600 (the blue line) by looking only at the Fosi's response with a 4 ohm or 8 ohm resistive load (the green and red lines).

But we don't need the Sony/blue line to tell us that the Fosi amp is load-dependent: the response starts dropping at 3kHz when feeding the amp into a simple resistive 4 ohm load, and it starts peaking at the same point with an 8 ohm load.

Now, I will certainly admit that with the Pass amp, the difference in response at 4 ohms vs 8 ohms is much more subtle. But (a) it is still there; (b) Pass says the amp is designed for 4-16 ohm loads so if that Sony speaker is really presenting 30-60 ohms then it's way outside the operating design of that amp; and (c) Pass's design uses no feedback and he says it uses a special transistor that "behaves like a triode" - in other words he says it basically behaves like a tube amp, and tube amps are known to be load-dependent.

So here is the issue that I'm having trouble understanding: once you know that an amp is load dependent, why would you consider buying or using that amp? Why do you need to know - or why is it useful to know - exactly how its frequency response looks with a particular speaker (like the Sony SS-H1600) when the chances of you ever using that speaker with that amp are virtually nil? In other words, isn't the diverging frequency response into 4 ohms vs 8 ohms sufficient to tell you that this amp's frequency response is not dependable with varying impedance, and so if that is a concern then you should not choose that amp?

I'm not trying to be snarky, rhetorical, or disrespectful here. And of course I have zero problem testing amps with complex loads for curiosity and interest. But I'm genuinely asking - what qualitatively different or new insight do we get by running an amp into a complex load like this? Thanks!

 

[valerianf]

"what qualitatively different or new insight do we get by running an amp into a complex load like this?"
Answer is simple: closest to reality is the qualification test, more accurate is the rating of the amplifiers.

Let us take the analogy with cars.
Would you rate them only with a 0-60mph strait acceleration or will you also test the handling to see if they are able to follow the variation of a bumpy road?
In real life, car handling is the main concern.
Back to the amplifier: it needs to follow the variation of the music with the constraints of a variable load.
Nobody is listening a fix sinus 1khz signal loaded with a fix resistor!
It is only a common test set by engineers to get specifications.
It will never tell you if the amplifier sound quality is good in real life.

Running a complex load is closer to reality and and has more chance to detect the non SOTA amplifier.
Marketing people will look at a test that has 99% chance to satisfy their claims.
Competitive tests will require to test the limit of what the amplifier is capable of.
As a consumer electronics buyer i am looking only at the competitive tests, when I can find them.
ASR tests are competitive tests.
They can always be enhanced to be more selective.

 

[SIY]

Let us take the analogy with cars.

That's always a way to enlightenment.

 

[ctrl]

both of these amps are load-dependent amps, and we don't need the Sony speaker to reveal that....
...TPA3255-based amp, and all (or at least the vast majority of) such amps are known to be load-dependent.

If it is guaranteed that all TPA3255-based amps behave perfectly identically with a challenging speaker load, then the diagnosis of "load dependent behavior" would suffice.
However, if there can be qualitative differences in amp design with TPA3255, then I would be very interested in this information to buy the "best" designed TPA3255-based amp.

Now, I will certainly admit that with the Pass amp, the difference in response at 4 ohms vs 8 ohms is much more subtle. But (a) it is still there; (b) Pass says the amp is designed for 4-16 ohm loads so if that Sony speaker is really presenting 30-60 ohms then it's way outside the operating design of that amp; and (c) Pass's design uses no feedback and he says it uses a special transistor that "behaves like a triode" - in other words he says it basically behaves like a tube amp, and tube amps are known to be load-dependent.

Assuming a layman would know the manufacturer's information and look at a review with the 4 and 8 ohm resistive load measurement.

He/she/it would come to the conclusion "perfect, I'll buy a passive 8 Ohm BR speaker, so I'm exactly between the manufacturer's specifications".

But the expensive passive 8 ohm BR-LS with 93dB sensitivity sounds like crap. He/she/it then reads in a forum that the speaker needs an "impedance linearization" (RLC series circuit correction) to sound good with the amplifier.

With the RLC filter (see blue line) the speaker load and amp FR changes to:

The speaker now sounds a bit better, but still not good, since the load dependency in the low frequency range has not been corrected.

This problem will most likely be ignored by many tube amplifier owners as well, possibly causing severe frequency response ripples in the low frequency and low midrange frequency range.

Loudspeakers with severe impedance peaks (>20 Ohm) in the low frequency range are most likely not correctable with passive filter, because you often need more than 15mH at air coil sizes (very expensive) or the total impedance of the loudspeaker drops below 3 ohms due to the filter.
Any speaker with very roughly more than 10 ohms impedance ripple, used with the Pass Labs amp, will result in significant, audible frequency response ripples (without using EQ/DSP and measuring equipment).

Without measuring the amplifier with a challenging speaker load, a layman cannot estimate this.

This is probably also true for the use of many speakers with tube amplifiers (if no EQ is used).

So here is the issue that I'm having trouble understanding: once you know that an amp is load dependent, why would you consider buying or using that amp?

Because there is not only one load dependency, which is the same for all amplifiers. I have no problem buying a cheap Class D amp with some ripple FR, but the ripple should be as low as possible with 90% of the speakers - I can't estimate that as a layman when only 4 and 8 ohm resistive load is shown.

But I'm genuinely asking - what qualitatively different or new insight do we get by running an amp into a complex load like this? Thanks!

I have already described most of this above.

One aspect that has not yet been addressed concerns "amplifier sound". This might give new insights into the dispute of "subjectivist versus objectivist" by measurements with a challenging speaker load.
Our hearing is not good at perceiving changes in volume level, but it is good at identifying resonances with a small Q-factor.

Source: Sound Reproduction - Floyd Toole

Perhaps measurements with challenging speaker load can give clues whether there are cases where this can explain the difference in sound perception (besides the "imagined differences").

P.S.: I can also confirm that it is analogous with cars, dental floss and toenail clippers - sorry @valerianf

 

[Deleted member 48726]

One aspect that has not yet been addressed concerns "amplifier sound". This might give new insights into the dispute of "subjectivist versus objectivist" by measurements with a challenging speaker load.
Our hearing is not good at perceiving changes in volume level, but it is good at identifying resonances with a small Q-factor.
View attachment 282740
Source: Sound Reproduction - Floyd Toole

Perhaps measurements with challenging speaker load can give clues whether there are cases where this can explain the difference in sound perception (besides the "imagined differences").

@valerianf

This is one of the main reasons I keep these discussions alive. I'm sure this is the reason for people hearing differences in amplifiers. -Often just ridiculed and laughed at when one dares to mention it..

 

[Sokel]

Can please someone explain (pretend that you explain it to a 10yo without any association) what would that do to this speaker?
How would it sound?
Example: "it would sound +2db @50Hz,-2db @350Hz",etc.

 

[pma]

After 5-10 minutes, the human ear would get accustomed to these deviations and the listener would be satisfied, if no direct level equalised comparison is done with the flat FR amp.

 

[tmtomh]

If it is guaranteed that all TPA3255-based amps behave perfectly identically with a challenging speaker load, then the diagnosis of "load dependent behavior" would suffice.
However, if there can be qualitative differences in amp design with TPA3255, then I would be very interested in this information to buy the "best" designed TPA3255-based amp.




Assuming a layman would know the manufacturer's information and look at a review with the 4 and 8 ohm resistive load measurement.

He/she/it would come to the conclusion "perfect, I'll buy a passive 8 Ohm BR speaker, so I'm exactly between the manufacturer's specifications".

But the expensive passive 8 ohm BR-LS with 93dB sensitivity sounds like crap. He/she/it then reads in a forum that the speaker needs an "impedance linearization" (RLC series circuit correction) to sound good with the amplifier.

With the RLC filter (see blue line) the speaker load and amp FR changes to:
View attachment 282734 View attachment 282736
The speaker now sounds a bit better, but still not good, since the load dependency in the low frequency range has not been corrected.

This problem will most likely be ignored by many tube amplifier owners as well, possibly causing severe frequency response ripples in the low frequency and low midrange frequency range.

Loudspeakers with severe impedance peaks (>20 Ohm) in the low frequency range are most likely not correctable with passive filter, because you often need more than 15mH at air coil sizes (very expensive) or the total impedance of the loudspeaker drops below 3 ohms due to the filter.
Any speaker with very roughly more than 10 ohms impedance ripple, used with the Pass Labs amp, will result in significant, audible frequency response ripples (without using EQ/DSP and measuring equipment).

Without measuring the amplifier with a challenging speaker load, a layman cannot estimate this.

This is probably also true for the use of many speakers with tube amplifiers (if no EQ is used).



Because there is not only one load dependency, which is the same for all amplifiers. I have no problem buying a cheap Class D amp with some ripple FR, but the ripple should be as low as possible with 90% of the speakers - I can't estimate that as a layman when only 4 and 8 ohm resistive load is shown.



I have already described most of this above.

One aspect that has not yet been addressed concerns "amplifier sound". This might give new insights into the dispute of "subjectivist versus objectivist" by measurements with a challenging speaker load.
Our hearing is not good at perceiving changes in volume level, but it is good at identifying resonances with a small Q-factor.
View attachment 282753
Source: Sound Reproduction - Floyd Toole

Perhaps measurements with challenging speaker load can give clues whether there are cases where this can explain the difference in sound perception (besides the "imagined differences").

P.S.: I can also confirm that it is analogous with cars, dental floss and toenail clippers - sorry @valerianf

Thanks for the thorough reply. Makes sense, especially the part about a scenario in which one would want a TPA3255-based amp and would want to get the most linear (or least non-linear) model or implementation available.

The one issue I think it's still important to keep in mind, though, is that any particular speaker load is, well, a particular load. I know @Holdt and others have talked about testing amps with a "typical 2-way" load and a "typical 3-way" load, and I certainly have no objection to that in principle. I just wonder how much of the load that speakers present to amps is typical, vs how idiosyncratic each speaker's load is. For example. we know that phase and impedance can vary especially at crossover points, and given the variability of crossover frequencies and crossover designs in speakers - and how sensitive human hearing is in the 1-5kHz range where most speakers have at least one crossover point - I'm not sure how predictive a "typical" simulated speaker load will be of how a load-dependent amp will sound with the actual speakers one might own and plan to use with it (for example, a peak or dip at a 1400Hz crossover is going to sound different than a peak or dip at a 2500Hz crossover). Similarly, if the main anomaly of those Sony speakers Archimago uses is that they have very high impedance at certain frequencies (30-60 ohms is basically a headphone!), I don't know how predictive that is of an amp's nonlinearities, given that a lot of (most?) speakers present challenges to amps in the form of areas where they present very low impedance.

Again, I am not arguing against testing amps with complex loads. But I am questioning how much we can form general conclusions about the particular nature of an amp's load dependency from 1 or 2 sample loads.

 

[IPunchCholla]

It is indeed almost impossible to see from resistive tests how it behaves with the complex loads. The power cube is better. It's still just at a fixed frequency. I keep thinking three equivalent circuits would give me more useful data than the power cube test.
Then I could compare which equivalent circuit resembles my speakers the most and make the assessment from there.

This is a great research question. It is clear, easy to devise a falsifiable thesis, experimental burden isn’t too high. Are there three equivalent circuits that can be used (within predetermined ranges) to simulate a supermajority of speakers? Have you looked into the research at all? (Actual question). Because it does seem like a reasonable inquiry.

My own first guess, based on having built all of one crossover, is that the answer would be no, as I think I did well over 100 variations in design and the impacts of changing the values of inductors and capacitors even small amounts could have very large changes in the overall response. It also important to remember, given what we know of psychoacoustics, that as PMA pointed out, our hearing will adjust to the “wiggly line” in a few minutes, unless we are actively A/B ing against a equalized comparison, so we do have to weight the value of what we might find against the value of possible gains to be had from the research. For me, the value would be small, since I run Advanced Room Correction and also listen in a treated room. For you it might be much higher.

 

[atmasphere]

Has anyone shown that to be an actual problem?

Some class D amplifiers can oscillate when subjected to a capacitive load (ESLs), on account of the output inductor setting up a resonant frequency in parallel with that capacitance. This can result in damage to the amplifier.

The Hypex module of this test/thread is not an example of this problem. It employs enough feedback that it can control the oscillation. However there will be a small sine wave at the resonant frequency, which some might interpret as an oscillation. I explained the mechanism in my prior post.

 

[ctrl]

After 5-10 minutes, the human ear would get accustomed to these deviations and the listener would be satisfied, if no direct level equalised comparison is done with the flat FR amp.

It also important to remember, given what we know of psychoacoustics, that as PMA pointed out, our hearing will adjust to the “wiggly line” in a few minutes, unless we are actively A/B ing against a equalized comparison, so we do have to weight the value of what we might find against the value of possible gains to be had from the research.

Our hearing adapts when the "unbalanced" sound is permanently stimulated. In most cases, however, it's probably different.

Assume a loudspeaker radiates in the frequency range 6-9kHz with high sound power. Then in some songs the sibilants will be just tolerable.

If the amplifier has an additional 0.8dB frequency response hump in the 6-9kHz range because of load dependency. Then there will be annoying sibilants in some songs because the additional hump in this range is the straw that broke the camel's back.

This will never change, because the now too high sound power in this range is not permanently excited.
There will always be parts in songs with aggressive sibilants and this impression will never disappear (for our assumed amp speaker combination).

 

[IAtaman]

I do understand and share the interest to better understand the potential coloration introduced by the interaction between the amp and the varying imp of the speaker load. This graph shared is very interesting for example

But I suspect that graph has a lot more to do with the speaker than the amp. Amp in question seem to have very little load variance, so the graph probably follows the speaker impedance graph, and I suspect if the same amp was connected to another speaker and measured, that graph would follow the impedance graph of that speaker as well. As far as amp measurements are concerned (and given that we are only interested with in spec behavior of the amp) 4ohm and 8ohm resistor measurements tell us what we need to know about the amp, don't they?

If yes, then I think what needs to be done is not to invest time and effort to measure amps with all kinds of different loads but measure the speakers better, is it not?

 

[Sokel]

I do understand and share the interest to better understand the potential coloration introduced by the interaction between the amp and the varying imp of the speaker load. This graph shared is very interesting for example


View attachment 282965

But I suspect that graph has a lot more to do with the speaker than the amp. Amp in question seem to have very little load variance, so the graph probably follows the speaker impedance graph, and I suspect if the same amp was connected to another speaker and measured, that graph would follow the impedance graph of that speaker as well. As far as amp measurements are concerned (and given that we are only interested with in spec behavior of the amp) 4ohm and 8ohm resistor measurements tell us what we need to know about the amp, don't they?

If yes, then I think what needs to be done is not to invest time and effort to measure amps with all kinds of different loads but measure the speakers better, is it not?

It's not the speaker UT,it's the amp measured at 4 Ohm,at 8 Ohm and at the speaker's complex impedance,all at 1V.
I asked the real world listening results example by this but the friends here are probably only interested in theory or keep it to themselves.

 

[IAtaman]

It's not the speaker UT,it's the amp measured at 4 Ohm,at 8 Ohm and at the speaker's complex impedance,all at 1V.

I got that. What I was trying to say is that the variation in level seems to be caused by the speaker, not the amp. This does not show to me that amps that measure almost load invariant still might change the FR with wonky speaker loads, this just shows some speakers have wonky FR.

 

[Sokel]

I got that. What I was trying to say is that the variation in level seems to be caused by the speaker, not the amp. This does not show to me that amps that measure almost load invariant still might change the FR with wonky speaker loads, this just shows some speakers have wonky FR.

What I understand is that wonky speaker loads (all speakers in short) would dictate their own FR to this amp and to every other similar amp that is (as it seems) load variant despite the 4/8 Ohm graph that shows it's not.

 

[charlielaub]

Can please someone explain (pretend that you explain it to a 10yo without any association) what would that do to this speaker?
How would it sound?
Example: "it would sound +2db @50Hz,-2db @350Hz",etc.

The plot above doesn't really make sense. My guess is that both the 4 Ohm and 8 Ohm responses have been normalized to 1kHz by their own output at that frequency. I draw this conclusion because that is about the only explanation I can think of as to why those two curves are essentially identical but the loudspeaker load is all over the place.

The reason that a loudspeaker load causes SPL changes vs frequency is that the amplifier puts out different amounts of power into different loads. The classic example of this is a single ended tube amp with an output impedance that is comparably to the nominal loudspeaker impedance. This appears in series with the loudspeaker impedance and acts as a voltage divider. This creates SPL variation in the loudspeaker's acoustic output just like what is shown in the plot.

If the 4 and 8 Ohm responses are essentially identical, so would the SPL produced by the loudspeaker within and around that range of load impedance.

I did a Google search using the terms "First-Watt SIT-2 output impedance" and there in bold type, right at the very top, is a short paragraph saying the output impedance is 4 Ohms. This proves that the green and red curves are normalized as I guessed.

An amplifier with negligible output impedance (e.g. output impedance << loudspeaker nominal impedance) has none of this variation. That is the whole reason why I find amplifiers with non-negligible output impedance to be pretty stupid. The cause your loudspeaker to sound different, because pretty much all loudspeakers are designed under the assumption of zero amplifier output impedance.

 

[Sokel]

The plot above doesn't really make sense. My guess is that both the 4 Ohm and 8 Ohm responses have been normalized to 1kHz by their own output at that frequency. I draw this conclusion because that is about the only explanation I can think of as to why those two curves are essentially identical but the loudspeaker load is all over the place.

The reason that a loudspeaker load causes SPL changes vs frequency is that the amplifier puts out different amounts of power into different loads. The classic example of this is a single ended tube amp with an output impedance that is comparably to the nominal loudspeaker impedance. This appears in series with the loudspeaker impedance and acts as a voltage divider. This creates SPL variation in the loudspeaker's acoustic output just like what is shown in the plot.

If the 4 and 8 Ohm responses are essentially identical, so would the SPL produced by the loudspeaker within and around that range of load impedance.

I did a Google search using the terms "First-Watt SIT-2 output impedance" and there in bold type, right at the very top, is a short paragraph saying the output impedance is 4 Ohms. This proves that the green and red curves are normalized as I guessed.

An amplifier with negligible output impedance (e.g. output impedance << loudspeaker nominal impedance) has none of this variation. That is the whole reason why I find amplifiers with non-negligible output impedance to be pretty stupid. The cause your loudspeaker to sound different, because pretty much all loudspeakers are designed under the assumption of zero amplifier output impedance.

Archimago's test protocol is well described,here's another measurement ,the description is clear:

 

[Sokel]

And that's his description about the said amp (@charlielaub it seems that the output impedance is as high as 2.7 Ohm! )



So the 4/8 Ohm measurement is clearly misleading here,we would tell people that they are delusional if they described it like this without knowing.

 

[charlielaub]

I had no idea where the plot came from. IMHO as shown it is a bit misleading regarding the levels of the two measurements done into resistive loads.

 

[DonH56]

I do understand and share the interest to better understand the potential coloration introduced by the interaction between the amp and the varying imp of the speaker load. This graph shared is very interesting for example


View attachment 282965

But I suspect that graph has a lot more to do with the speaker than the amp. Amp in question seem to have very little load variance, so the graph probably follows the speaker impedance graph, and I suspect if the same amp was connected to another speaker and measured, that graph would follow the impedance graph of that speaker as well. As far as amp measurements are concerned (and given that we are only interested with in spec behavior of the amp) 4ohm and 8ohm resistor measurements tell us what we need to know about the amp, don't they?

If yes, then I think what needs to be done is not to invest time and effort to measure amps with all kinds of different loads but measure the speakers better, is it not?

An ideal amplifier would not change with the load (speaker). An amplifier with no load variance at its output would not follow the speaker's impedance graph; it would stay flat, the same as for a resistor. This plot shows how the amp's output does vary with a real speaker load, indicating high output impedance (low damping factor). It shows the amplifier is sensitive to the load.