[bigshot]

The chesky discs I’ve heard weren’t that great.

 

[A Jedi]

It could be any disc - the exact disc is not important.

 

[gregorio]

As I said above, any music is way more complex than a 1K sine.

True but any music is way less complex than say white noise. So, you’ve got the same argument against any piece of music vs white noise as you have against a 1kHz sine wave vs a piece of music. If lack of complexity is the argument then a multi-tone (30) test signal is better or white (or pink) noise or a sine sweep.

What I propose is measuring the output of the amp at the speaker terminals as usual but with the speaker connected (measurement leads are in parallel to transducer).

Of what speaker? Obviously everyone would have to have exactly the same speaker because otherwise you could never compare measurements (as Castle mentioned) but in addition:

Is this a say 20w speaker or say a 250w speaker and how would you measure the performance of say a 20w amp on a 250w speaker or a 250w amp with a 20w speaker? Of course you could take such measurements but:

1. The results would be unrepresentative of amp performance if the amp’s gain setting is very significantly outside it’s optimal range.

2. Is the speaker load identical with say an extremely underpowered input signal (a say 20w amp and 250w speaker) as it is with an optimally powered signal?

3. Speakers are manufactured to a tolerance. Two high quality, high tolerance broadcast monitors of the same make, model and version will perform the same within a tolerance of about 1.5dB. In other words, the difference between two identical high quality monitors could be anything from around 0.2dB - 1.5dB and would typically be around about 0.5dB. Consumer speakers, even audiophile ones will typically have greater tolerances, 2dB between the same speakers in some cases but the better ones about 1dB. In addition, the tolerances will almost certainly change/increase somewhat with age. This brings us back to the issue of scale, these differences between identical speakers are fairly insignificant in most cases (except the poor ones), audiophiles rarely mention them but even the best monitors have differences within the thresholds of audibility. So how appropriate are they in a chain when measuring tiny differences 1 or several orders of magnitude below the threshold of audibility? Of course, I’m talking output differences here rather than input load differences and TBH, I don’t know the input load tolerance/differences between two speakers of the same brand/model/version but are they always zero or close enough to zero not to seriously distort the measurements of the tiny, inaudible differences we’re measuring? Of course, we can avoid any question of differences due to manufacturing tolerances or ageing of components within speakers by not using speakers and instead just specifying a simulated load.

G

Reference:

This graph shows the statistical deviation of the frequency response from the averaged responses of over 2500 measured loudspeakers, and it gives a good idea of how closely the performance of individual monitors can be matched across an entire production process, given the will and appropriate techniques. [Courtesy of AES.] - Taken from Sound on Sound/Hugh Robjohns

 

[castleofargh]

I Don't argue just to argue. I'm fine with more measurements of anything, more power to us. But the measurements need to be worth the effort.
Once again half of what I had written is what Gregorio just posted so no point in a long post to say the same again. The case of music vs noise is what I was insisting on the most actually. We have one, 2, 3 tones measurements that are very standardized and give us something to distinguish harmonics from modulated content for example. Once you add more tones, you don't get a progressive evolution and readability. Already at 5 tones we have a hard time identifying what's what, and several measurements already correlate well with noise measurements.
It's a fact that just 2 tones show interactions that one tone won't show. No debate on that. But as you add more, you very very rapidly reach a state where new stuff just can't affect much of what has already been taken for a ride. If it's to get something similar to noise measurements but without the references and knowledge of test with noise, what we end up with is an extra test that tells us even less than the existing ones.
I get that in a context where we have almost no measurement available to the consumer, more is always significant, but even then I'm not sure I would try to go at it with music.

There is one paper on non coherent something I remember(I have it somewhere, it's like 15 or 20 years old), and it did look interesting to me who doesn't understand 1/10th of it. The test allowed music, was crap for precise measurement, but did tend to show "something" when big stuff were going on(like clipping), and so it was hypothesized that it could be a cool measurement to correlate with audibility and subjective impressions for once. How often have I seen that measurement used since? One time and most likely because one of the guys on the paper was also on the one for that novel measurement method. I'd bet he insisted to get it until the others gave up and allowed it. ^_^
I don't know how to do it so I can't even try to find out if it's at all interesting for me/us.




Amps and load, we would realistically need a few different IEMs and headphones(or a circuit mimicking them) to have a range of voltages, impedances, and mix of reactance, inductance and whatever not just for one portion of the audible range. Would we really learn much beside the amp's own impedance? IDK. In some case I'm sure we would, for really terrible gear that probably would also give themselves up in traditional tests. For the rest, most likely the distos of the amp would still be much below what the headphone does to the signal and we're back to wonder how to analyze that?
As you can see I'm super optimistic. But I get why you're asking for that. I tried to measure stuff that way with almost all my gear as a matter of fact. To check that the combo I had was not giving me any big bad surprise. But I don't know what others could gain from that quite anecdotal use and mostly nothing to see? I put that at the level of my own ABXs. informative/reassuring/convincing for me. Near meaningless for others.

Perhaps we can have one of our AI overlord work on this. I'm a solid believer in AI having the potential for better diagnostics than even expert humans. I still do have issues with adding a load that causes by itself more distos than the amp, DAC, Ethernet cable we try to test. Once the signal we're looking for is drowned by louder crap, we're usually out of luck.

 

[A Jedi]

By all of you guy's logic we can deduce that all measurements are useless since the transducer's own distortion swamps everything else. So why bother measuring anything other than speakers?

And I'm not saying I have all the answers, nor am I proposing the end all be all testing procedure.

Want to use white noise? Fine.

All I'm saying is, I have multiple amps who's measurements show differences below the threshold of audibility yet they sound different (on the same speakers/headphones). Therefore the measurements are incomplete. What't the missing variable? The transducer.

 

[gregorio]

By all of you guy's logic we can deduce that all measurements are useless since the transducer's own distortion swamps everything else. So why bother measuring anything other than speakers?

And now we start, you presumably have no reliable evidence or facts to support your argument, so you resort to a fallacious argument. You’re claiming “all of us guy’s logic” must be faulty on the basis of an illogical assertion you yourself have invented and falsely attributed it us? What we’ve actually stated is that all measurements are not useless and are not swamped by the transducers own distortion/noise because we take many measurements without using any transducers. You are the one arguing that measurements be taken with transducers, not us! How does the distortion of transducers swamp the measurements of say a DAC or an amp if there is no transducer in the measurement chain?

All I'm saying is, I have multiple amps who's measurements show differences below the threshold of audibility yet they sound different (on the same speakers/headphones). Therefore the measurements are incomplete. What't the missing variable? The transducer.

What indeed is the missing variable and how should we address this question?

The typical way for audiophiles seems to be; come up with some guess (usually inspired by audiophile myths or marketing) and then defend that guess to the death, regardless of the actual facts, logic or anything else.

The typical way for engineers, scientists or anyone else who values the actual facts and rational reasoning would be a series of logical steps to narrow down the missing variable. So first we’d look carefully at the actual question itself, which in this case contains an apparent impossibility; how is it possible that you’re hearing differences below the threshold of audibility? That’s obviously a huge red flag for any logical or rational reasoning, so a good starting point would be to determine if you can actually hear differences. If in a controlled listening test you can’t, then the missing variable is your imagination (biased perception) and the objective measurements were incomplete because they obviously do not include your personal imagination (and wouldn’t be objective measurements if they did)! If you can actually hear a difference then we need to eliminate other possibilities. For example, are you not human or some sort of super-human and therefore not subject to human audibility thresholds? Or more seriously, are you really talking about “measurements not showing differences below the threshold of audibility” or only one or a few specific measurements? If it’s the latter, then the missing variable is one of the other measurements and therefore “measurements” are NOT incomplete, what’s incomplete is your choice of what measurements you used.

The above is just a couple of example steps but as you can see, there are a whole bunch of potential options to eliminate (some of which are far more likely) BEFORE we could conclude “measurements are incomplete” or the missing variable is “the transducer”. None of the above seems applicable to most audiophiles though because working out this sort of logical/rational flow chart and then actually working through it obviously requires a great deal more time and effort than simply inventing some guess and the apparent impossibility in the question is not a problem/huge red flag because audiophile marketing/mythology came up with fallacious responses to such red flags decades ago (that audiophiles would swallow) and have rolled them out repeatedly ever since. So they don’t even appear to be red flags to audiophiles these days, let alone huge ones.

G

 

[bigshot]

Using music in a listening test is best to judge perceived sound quality, not objective sound fidelity.

 

[A Jedi]

You're missing the whole point. I'll spell it out for you:

Currently measurements on amps are done with a resistive load.

Speakers are not purely resistive - they are reactive.

Since nobody is measuring an amplifier's behavior with a reactive load, the measurements do not reflect reality.

Lab measurements are fine. But the equipment does not operate in a lab.

Feel free to flame me for being a subjectivist (I am not). But telling me that my point is invalid is a sign that you're either not a scientist (no interest in discovery) or you're unable to deal with variables that reflect reality.

If asking for a measurement into a real world load makes me an audiophool so be it.

"Yeah but all transducers are different and there's variance across any batch of "identical" drivers."

Yup that's the way it is. But if you use of handful of drivers consistently (high impedance driver, low impedance driver, planar, etc.) across all the DUTs, you should be fine.

At the very least to ascertain if/how they effect various amps. Maybe across all drivers, all amps will perform as their stated lab measurements. But maybe (likely) they won't. This what I'd like to know.

 

[WoodyLuvr]

You're missing the whole point. I'll spell it out for you:

Currently measurements on amps are done with a resistive load.

Speakers are not purely resistive - they are reactive.

Since nobody is measuring an amplifier's behavior with a reactive load, the measurements do not reflect reality.

Lab measurements are fine. But the equipment does not operate in a lab.

Feel free to flame me for being a subjectivist (I am not). But telling me that my point is invalid is a sign that you're either not a scientist (no interest in discovery) or you're unable to deal with variables that reflect reality.

If asking for a measurement into a real world load makes me an audiophool so be it.

"Yeah but all transducers are different and there's variance across any batch of "identical" drivers."

Yup that's the way it is. But if you use of handful of drivers consistently (high impedance driver, low impedance driver, planar, etc.) across all the DUTs, you should be fine.

At the very least to ascertain if/how they effect various amps. Maybe across all drivers, all amps will perform as their stated lab measurements. But maybe (likely) they won't. This what I'd like to know.

There is a body of evidence, compiled over many decades, which strongly supports the use of test signals and loads and mind you not all of these signal tests are simple pure tones like sine waves but actually very complex test signals covering the entire audio frequency spectrum that are actually far more challenging than any piece of music could ever be. We have a sound understanding on how test "simulated" resistive and reactive loads will affect audio amp performance especially in comparison to reactive loads or even real loads (there is actually a difference between the two, reactive and real loads in the testing world). Reactive loads in themselves are not necessarily any better or worse than resistive loads in most scenarios nor are reactive loads always directly applicable to (performing the same as) real loads... but we do know enough to easily compare and interpret any of these results between any of them therefore drawing rock solid conclusions from them. Why would we therefore over complicate the matter by using real loads loads and music?

A real mind-bender for you: measuring noise produced in an amp circuit need not require the use of a test signal. Mull over that.

Audio Power Amplifiers for Loudspeaker Loads​

 

[Davesrose]

My gosh, this thread keeps going. You're never going to convince someone who has bought into hype (especially when it's literal and they've bought some overpriced "audiophile" ethernet hub). Wireless bluetooth at certain distances, I can understand issues with bandwidth and audio codecs that try to be lossless. But modern ethernet, where 1Gb/s is a normal? I'm doing more streaming of 4K movies with lossless surround sound. Their audio streams (that can be 7.1 THD Atmos or high res concerts) seem to cap around 4MBps (compared to the extra data you need for HDR 4K video). Running an internet test on my wifi even gets 850Mbps (or 106MB). Even cheap ethernet these days has more than enough bandwidth for anything audio.

 

[bigshot]

They aren't actually talking about ethernet cables any more.

 

[Davesrose]

They aren't actually talking about ethernet cables any more.

I thought current dialogue is about perceived noise, from noise on an ethernet cable? They're exclusive: digital mediums can have more noise in the signal but still carry more than enough data. Hearing noise in your transducer: for many applications that means you're listening too loud or have your input settings in your sound control too high.

 

[bigshot]

I think they're talking about the best way to measure amps for speaker systems now.

 

[Davesrose]

I think they're talking about the best way to measure amps for speaker systems now.

Well at least in the last page it all still started with the source. Sorry, don't think anyone has tried to drive a 25watt speaker with a 250watt amp. Yes, after the digital source, there can be a subject about proper amping. But all is irrelevant to the topic of ethernet, and should we now get into all the invariable aspects of sound reproduction?

 

[gregorio]

You're missing the whole point. I'll spell it out for you:

No, I understand your point but I disagree with it, due to the facts of scale and due to history. I’ve dealt with scale but you don’t seem to get the point

40 years ago Bob Carver famously built a relatively cheap ($400) amp, lined it up with a couple of the most expensive (and significantly different) audiophile amps on the market, did a double blind test with “golden eared” Stereophile reviewers and no one could tell the difference. Carver stated he used old measuring techniques on the outputs of the audiophile amps to match the output of his amp. So it clearly was possible to use measurements to characterise the performance of amps to inaudible levels. You think maybe that measurements and measuring equipment and techniques have become worse over the last 40 years?

Around 30 years ago we started to digitally model amps, by about 20 years ago we were modelling guitar amps and cabs and their interaction (including their non-linear behaviour) so well that even guitarists were impressed and today, in carefully controlled tests, guess what … How is this possible if we can’t measure what we’re trying to digitally model?

Lab measurements are fine. But the equipment does not operate in a lab.

So you’re saying that consumer audio equipment/amps know when they’re in a lab and change their performance when they’re in an audiophile’s sitting room? What about in a recording studio, which are often similar to lab conditions?

Feel free to flame me for being a subjectivist (I am not). But telling me that my point is invalid is a sign that you're either not a scientist (no interest in discovery) or you're unable to deal with variables that reflect reality.

But making up false assertions and falsely attributing them to “us guys” is scientific is it? Carver, digital modelling software and numerous other bits of history and science/engineering proves that science IS able to deal with “variables that reflect reality”.

But if you use of handful of drivers consistently (high impedance driver, low impedance driver, planar, etc.) across all the DUTs, you should be fine.

And how do you spread that “handful of drivers” all over the world amongst those thousands of people, engineers, etc, who objectively measure amp performance? Or are you saying that the relatively large variations between drivers of the same model don’t affect the measurement of relatively minuscule audio properties? As an example, most DACs have a freq response variation throughout the 20Hz - 20kHz range of about 0.4dB. How are you going to measure those differences when using different drivers (of the same model) that have FR variations of anything from about 0.5dB - 2dB? Exactly the same DAC could measure anywhere from a 0.0dB FR variation to a 2.4dB variation depending on the production line tolerances of the same driver. You’re not actually measuring the FR of the DAC, the majority of what you’re measuring is the difference between drivers of the same model. And the scale of this problem is far larger with many other DAC performance metrics, say jitter distortion or self-noise.

Incidentally, to everyone else: My understanding is that all this relates to the OP due to the suggestion we should measure the (supposed) noise/distortion of Ethernet switches and cables (or of anything else) under “real world” listening conditions, EG. After the amp and speakers/HPs.

G