Complex impedance load amp FR influence - WiiM Amp review by Erin

[Yorkshire Mouth]

It is quite clear these kinds of deviations are generally audible in A-B tests.

Cheers. Do you have a link for that?

 

[Rick Sykora]

While I get that ideally you want to test with a complex load, using simple resistive loads, Amir came to the same conclusion about load dependency of this amp as Erin did. @pma has been doing complex load testing and this topic has been litigated at length in other threads) For that matter, @amirm started to do more complex loads and seems to have abandoned.

So I get the feature, where is the major benefit of more complex load testing? If it is worthwhile, then let's do the science, document it and then all can test to a common reference. 

 

[dominikz]

Cheers. Do you have a link for that?

Sure, here's one (link to full AudioXpress article):

Peaks and dips are a major manifestation of frequency response anomalies. Peaks in frequency response are caused by resonances and can be characterized by a central frequency, and a Q that is associated with the height and width of the resonance. Toole and Olive have investigated the audibility of resonances (4).

Figure 4: Detection thresholds for high, medium, and low Q resonances from reference 1.

Figure 4 shows the detection threshold for resonances of various Qs in the presence of typical program music. You see that very narrow resonances (high Q) must be about 10dB above the average level to be heard, whereas very broad resonances need only be 1 to 2dB higher to be detected.

You can also find this information in dr. Toole's amazing book "Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms".

However, it is also easy to test this for yourself - e.g. try toggling a +1dB boost at say 1kHz (Q=1 or lower) with PEQ when listening to a spectrally dense recording (or pink noise). You can even process a file and do an ABX vs an unprocessed file in foobar2000.

 

[dominikz]

So I get the feature, where is the major benefit of more complex load testing?

IMHO I don't really see a benefit of testing amplifiers with complex loads, especially since there are no standards defining them.

Instead, I'd propose to do amplifier output impedance measurements (ideally including phase).
From that and the unloaded ("open circuit") frequency response you can predict the exact frequency response for any known load (complex or otherwise).
But more importantly, it allows you to easily estimate the scale of the FR deviation - for this I really like the Benchmark spreadsheet calculator (link).

 

[Yorkshire Mouth]

Sure, here's one (link to full AudioXpress article):


You can also find this information in dr. Toole's amazing book "Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms".

However, it is also easy to test this for yourself - e.g. try toggling a +1dB boost at say 1kHz (Q=1 or lower) with PEQ when listening to a spectrally dense recording (or pink noise). You can even process a file and do an ABX vs an unprocessed file in foobar2000.

Specifically, the science there says for broad resonances the difference needs to be 1 dB to 2dB. So 1 dB absolute minimum.

Have I read that right?

 

[dominikz]

Specifically, the science there says for broad resonances the difference needs to be 1 dB to 2dB. So 1 dB absolute minimum.

Have I read that right?

Personally I wouldn't interpret it like that, at least not without reading the original research.
The way I read it is: how audible a resonance is depends on both its magnitude and bandwidth, and broad resonances require less magnitude to be audible. Whether 1dB is the absolute minimum or not is not explicitly stated. E.g. might be if they used even wider filters they'd find 0,5dB differences were audible - I really can't say.

But it is indeed important to keep in mind that when we're talking about thresholds of audibility we're definitely talking about small differences, where many people would probably require training to identify differences consistently, and which may sometimes not be audible at all without instant switching and appropriate audio content.

EDIT: I just did a quick test and I'm reasonably sure I could consistently ABX a +0,5dB high-shelf filter at 1kHz (Q=0,7) with pink noise as stimulus (but don't trust me until I post ABX log )

 

[Yorkshire Mouth]

The more I read the more I learn.

Apparently, the limits to what people can ‘easily hear’ are 17.4k for teenagers, 15k at 40, 12k at 50.

I’m 58. So a lot of the errors we see in these measurements will be inaudible to me. If we take (as you have) a difference of 0.5 dB being where we can notice differences, and we accept not just an upper limit of what we can hear, but a degree of roll off before there, I think we’re looking at very small audible differences at worst.

 

[hex168]

Playing with EQ, I can easily hear a 1 dB broad Q change. However, that is different from having a preference. For a 1 dB change in the midrange I did not have a consistent preference across varying program material. (Note: not blind.)

 

[dominikz]

The more I read the more I learn.

Apparently, the limits to what people can ‘easily hear’ are 17.4k for teenagers, 15k at 40, 12k at 50.

I’m 58. So a lot of the errors we see in these measurements will be inaudible to me. If we take (as you have) a difference of 0.5 dB being where we can notice differences, and we accept not just an upper limit of what we can hear, but a degree of roll off before there, I think we’re looking at very small audible differences at worst.

I agree.
If you read a few of my recent posts you will see I find the load dependence of e.g. WiiM Amp acceptable - especially given the feature set, usability and the rest of the measurements which are quite OK. EDIT: But I would be happier if it was made to be load-independent. There's no pleasing some people.

In short - just because something is (barely) audible under optimal conditions it doesn't mean it is really a practical concern.

Which makes it even more interesting that a lot of people in the audiophile hobby are concerned about 'issues' that have never even been formally demonstrated to be audible.

 

[voodooless]

Playing with EQ, I can easily hear a 1 dB broad Q change. However, that is different from having a preference. For a 1 dB change in the midrange I did not have a consistent preference across varying program material. (Note: not blind.)

If it were a tube amp, people would scream about lifted veils

 

[IAtaman]

Playing with EQ, I can easily hear a 1 dB broad Q change. However, that is different from having a preference. For a 1 dB change in the midrange I did not have a consistent preference across varying program material. (Note: not blind.)

I think visually not impaired would be the appropriate term.

 

[voodooless]

I think visually not impaired would be the appropriate term.

No! I’m visually impaired, not blind. They are not the same thing!

 

[Rick Sykora]

I agree.
If you read a few of my recent posts you will see I find the load dependence of e.g. WiiM Amp acceptable - especially given the feature set, usability and the rest of the measurements which are quite OK. EDIT: But I would be happier if it was made to be load-independent. There's no pleasing some people.

In short - just because something is (barely) audible under optimal conditions it doesn't mean it is really a practical concern.

Which makes it even more interesting that a lot of people in the audiophile hobby are concerned about 'issues' that have never even been formally demonstrated to be audible.

I do like Erin’s representation graphically, but to some of your point, I think audio enthusiasts really want to know how well an amplifier performs with their speaker. Lacking that, Amir‘s gut check approach seems simpler. I know some are questioning how he could possibly recommend an amp with a clear load dependency, but as long as he conditions it with “for its class”, I can decide whether I want to spend more for something better.

 

[levimax]

It is interesting that some consider an amp with a SINAD of 60 dB to be unacceptable but give a pass to an amp with frequency dependancy that is going to be far more audible in some cases. An amp like this has the possibility to sound more colored than most quality amps made during the last 60+ years. Nice features and power but if it can't get FR right it really makes no sense to me.

 

[Deleted member 48726]

I do like Erin’s representation graphically, but to some of your point, I think audio enthusiasts really want to know how well an amplifier performs with their speaker. Lacking that, Amir‘s gut check approach seems simpler. I know some are questioning how he could possibly recommend an amp with a clear load dependency, but as long as he conditions it with “for its class”, I can decide whether I want to spend more for something better.

"For its class" -What class is that?

I don't generally see AVR's getting this designation excuse when having relatively poor (but inaudible) THD+N or lack of sustained power even if it does a hundred things more than a power amplifier.

Why is the class, price or the number of features a remedying factor for the main thing an amplifier should do; amplify a signal transparently?

 

[antcollinet]

I don't generally see AVR's getting this designation excuse when having relatively poor (but inaudible) THD+N

Then you've not been looking. I think it is common in Amir's reviews.

Marantz Cinema 40 AVR Review

This is a review and detailed measurements of the Marantz Cinema 40 9.4 channel AV Receiver (AVR). It is on kind loan from a member and costs US $3,500. As is typical in this class, you get a very large box. At least the look is unique and to my liking. The port hole display while limiting...

www.audiosciencereview.com

Giving the benefit of doubt and going with this value of SINAD (ratio of noise+distortion), we get a reasonable ranking for an AV product:

Or:

Anthem AVM70 Review (AV Processor)

This is a review and detailed measurements of the Anthem AVM70. It was sent to me by the company for review and costs US $3799. The AVM70 has a very modern and attractive front panel and display: Company takes full use of the large display as you see with nice icons. Importantly, the...

www.audiosciencereview.com

Dynamic range is good for the class:

Just the fact there is a separate Sinad chart for AVR's shows they are judged to their own standard.

 

[dominikz]

It is interesting that some consider an amp with a SINAD of 60 dB to be unacceptable but give a pass to an amp with frequency dependancy that is going to be far more audible in some cases. An amp like this has the possibility to sound more colored than most quality amps made during the last 60+ years.

Here's my perspective: I currently use an integrated amp that is both load-dependent (almost 1dB variance with my speakers) and has SINAD of only about 70 (this is the amp).

This concerned me for a while so I've tried FR compensation for the specific load variance seen with my speakers, I've done listening tests of the compensated vs uncompensated response, and I've also tested amps that are both load-invariant and have better SINAD. Some of it described in this post.
EDIT: This online ABX test I prepared is IMO also relevant in this context, because it shows the vast majority of participants (almost 350 in the most recent results) couldn't reliably differentiate two DACs where one has -0,3dB/+0,7dB FR variation compared to the other.

The result of my tests were that the audible impact was IMO totally negligible (but not zero!), since the in-room response is affected more by opening a window (see this post) or moving to a different seat, or by using a different EQ target/approach/DRC SW (see this thread), or by using a different loudspeaker. Also, the fact that a difference is perceivable doesn't imply that there's a preference for one or the other - and I usually have no preference if the audible difference is very small.

This is of course not to say that we should generally accept poorly-measuring gear - we should definitely push the industry to produce better-measuring stuff - but we should also try to avoid falling into the trap of thinking every measurable improvement translates into an audible improvement. Gear with less-than-stellar measurements can still produce 'transparent' sound.

I can only speak for myself, but as long as I find a device to be close enough to transparent I'll decide whether or not to buy it based on things like functionality, simplicity, how well it integrates with a TV (and other tech I have), ease of use, how much additional boxes I'd need to buy to make it work, form-factor, warranty, price, etc...

This is why I like various AVRs and the WiiM Amp more than I like separates - even though separates often measure better.

All that being said, I'm also very surprised that so many amps are load-dependent these days when it is known how to make them load-independent. Seems like such low-hanging fruit that I hope the rise of importance of objective measurements (due to ASR and similar sites) will push manufacturers to fix it in the future!

 

[ctrl]

[off topic]

Specifically, the science there says for broad resonances the difference needs to be 1 dB to 2dB. So 1 dB absolute minimum.

Have I read that right?

Sorry, but no.
Generally, it is said that 1dB represents the threshold of perception for a difference in volume/loudness (i.e., the entire frequency spectrum changed by 1 dB).
For resonances with a small Q factor, as has been mentioned before in this thread, the thresholds of perception are much lower.

EDIT: This online ABX test I prepared is IMO also relevant in this context, because it shows the vast majority of participants (almost 350 in the most recent results) couldn't reliably differentiate two DACs where one has -0,3dB/+0,7dB FR variation compared to the other.

There still seem to be strong doubts about how audible resonances with low Q are with real music. Especially in the presence range of 1-4kHz, were the measured load-dependent amp shows +0.6dB FR deviation with a complex dummy load.

Therefore, here is another ABX test with a short sample of real music (see attachment). The original was changed using Audacity with a maximum 0.4dB peak in the 2-4kHz range:



The changes were verified with the software DeltaWave:



It seems that the overall volume of the modified sample has been reduced by about 0.05dB, but this has no impact. The peak is approximately 0.4dB (or 0.45dB if one adds the loudness decline) at 3kHz.

From the ABX results you can see that a low Q 0.4dB resonance peak around 3kHz can "easily" be detected (made a second ABX test because 8/10 alone is not convincing, made only two full ABX tests without training, one ABX test I had to cancel because it was right after the first one and my concentration slipped):

Please don't misunderstand me, by "easy" perceptible, I don't mean that this is the case with every piece of music at every point. As mentioned in the first ABX example in post#39, it's more of an additional aggressiveness, brightness, and sharpness at certain points that is noticeable.

Some of you will say, the difference is so small, it's subjectively hardly perceptible. Good for you, personally, it drives me crazy when passages with additional aggressiveness, brightness, and sharpness are audible in many pieces of music, even if the music is only playing in the background

 

[dominikz]

There still seem to be strong doubts about how audible resonances with low Q are with real music. Especially in the presence range of 1-4kHz, were the measured load-dependent amp shows +0.6dB FR deviation with a complex dummy load.

Therefore, here is another ABX test with a short sample of real music. The original was changed using Audacity with a maximum 0.4dB peak in the 2-4kHz range:



The changes were verified with the software DeltaWave:



It seems that the overall volume of the modified sample has been reduced by about 0.05dB, but this has no impact. The peak is approximately 0.4dB (or 0.45dB if one adds the loudness decline) at 3kHz.

From the ABX results you can see that a low Q 0.4dB resonance peak around 3kHz can "easily" be detected (made a second ABX test because 8/10 alone is not convincing, made only two full ABX tests without training, one ABX test I had to cancel because it was right after the first one and my concentration slipped):

Please don't misunderstand me, by "easy" perceptible, I don't mean that this is the case with every piece of music at every point. As mentioned in the first ABX example in post#39, it's more of an additional aggressiveness, brightness, and sharpness at certain points that is noticeable.

Some of you will say, the difference is so small, it's subjectively hardly perceptible. Good for you, personally, it drives me crazy when passages with additional aggressiveness, brightness, and sharpness are audible in many pieces of music, even if the music is only playing in the background

Thanks for sharing!
Actually I don't doubt these results at all - I'm not surprised that people are able to hear low-level, low-Q resonances. As far as I understand that is pretty much established to be audible by previous research (and even tracks with my own experience using PEQ).
That being said, I wonder whether the majority of people who are able to identify such low-level FR deviations in ABX would also consistently indicate preference for flat FR vs such low-level resonances in blind tests with diverse audio content - perhaps there's already some research on that?

A note regarding the ABX I mentioned in the section you quoted: the +0,7dB resonance was at around 17,5kHz, which probably explains why the vast majority of people were struggling to hear it. I put it as a reference only because sometimes (but not always) the most severe FR deviations of load dependent amps can end up very high in spectrum.

 

[Yorkshire Mouth]

[off topic]


Sorry, but no.
Generally, it is said that 1dB represents the threshold of perception for a difference in volume/loudness (i.e., the entire frequency spectrum changed by 1 dB).
For resonances with a small Q factor, as has been mentioned before in this thread, the thresholds of perception are much lower.



There still seem to be strong doubts about how audible resonances with low Q are with real music. Especially in the presence range of 1-4kHz, were the measured load-dependent amp shows +0.6dB FR deviation with a complex dummy load.

Therefore, here is another ABX test with a short sample of real music (see attachment). The original was changed using Audacity with a maximum 0.4dB peak in the 2-4kHz range:

View attachment 351572

The changes were verified with the software DeltaWave:

View attachment 351575

It seems that the overall volume of the modified sample has been reduced by about 0.05dB, but this has no impact. The peak is approximately 0.4dB (or 0.45dB if one adds the loudness decline) at 3kHz.

From the ABX results you can see that a low Q 0.4dB resonance peak around 3kHz can "easily" be detected (made a second ABX test because 8/10 alone is not convincing, made only two full ABX tests without training, one ABX test I had to cancel because it was right after the first one and my concentration slipped):

View attachment 351577 View attachment 351578

Please don't misunderstand me, by "easy" perceptible, I don't mean that this is the case with every piece of music at every point. As mentioned in the first ABX example in post#39, it's more of an additional aggressiveness, brightness, and sharpness at certain points that is noticeable.

Some of you will say, the difference is so small, it's subjectively hardly perceptible. Good for you, personally, it drives me crazy when passages with additional aggressiveness, brightness, and sharpness are audible in many pieces of music, even if the music is only playing in the background

You start with “Sorry but no”, but then go in to discuss how there are doubts, etc.

I appreciate it’s a complex area, and we can all oversimplify occasionally - especially with the necessity of internet brevity, but I still stress that we’re talking about relatively small* differences, which in any case only occur when this extremely low cost device is pushed.

* Small in the context of the entire chain, but just other amps.