Critic's Corner

Discuss a review. Provide constructive feedback. Talk to the industry.

Return to Critic's Corner

Message Sort: Post Order or Asylum Reverse Threaded

JA on music waveform

32.217.133.178
Posted on October 22, 2019 at 16:39:16
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 In this month's letters section, JA responds to a reader who points out that we'll never understand the technical/sound divide of an amplifier when tests only use steady-state signals. In his response, JA said the music signal can be de-constructed into sinewaves of diff. frequencies and amplitudes. And therefore derive a test-sound quality standard.

Is this true ? Or tech-babble ? I would think we need better tests to understand the audio signal as it's processed by a (very complex) audio component...

 

Hide full thread outline!
    ...
Tech-babble is for manufacturers :), posted on October 22, 2019 at 18:03:04
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 Any repetitive time-domain musical waveform can be deconstructed as suggested by JA (Fourier analysis). But such an input is the complex thing, not the audio component.
If an amplifier has a non-straight transfer characteristic then when a sinusoid is passed through it harmonics are generated (the shape of that transfer characteristic determines the amplitudes of the harmonics).
If you pass two sinusoids through the same amplifier you get the harmonics of each sinusoid plus the intermodulation products between them (determined by the transfer characteristic).
If you now pass a very complex signal made up of a large number of sinsusoids then you get the harmonics of each and the intermodulation products of each pair of tones and end up with a huge amount of output distortion signals. Though the input signal of the last case might be more representative of a music signal (though not really because they not repetitive) but you haven't really learned anything new that the 1-tone and 2-tone tests told you. In fact the big multi-tone output could be so hard to interpret you might not learn anything.

People have proposed mutli-tone testing for audio that might show up in a google search. In certain communication systems testing multi-tones are used but they are looking for intermodulation products showing up at a specific frequency.

Regards
13doW

 

Yes, it is called Fourier analysis. (NT), posted on October 22, 2019 at 18:17:22
Kal Rubinson
Reviewer

Posts: 12435
Location: New York
Joined: June 5, 2002

 

RE: Tech-babble is for manufacturers :), posted on October 22, 2019 at 19:31:54
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Thanks for you response. In my view, Fourier must have failed because of the wide-gap between tests and perceived sound quality.

Although sometimes, high distortion could mean better sound, even with SS circuits. One example is the great darTZeel amplifier.

 

RE: Tech-babble is for manufacturers :), posted on October 22, 2019 at 20:05:55
Davey
 Because subjective evaluation doesn't always agree with objective evaluation, Fourier analysis is a failure?
I love the logic of audiophiles.

Dave.

 

RE: Tech-babble is for manufacturers :), posted on October 22, 2019 at 20:37:05
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 The logic of audiophiles is that tests don't describe sound quality.

The reader's (sincere) letter in 'phile expressed this. And JA himself said new tests are coming that might show more...

 

RE: Tech-babble is for manufacturers :), posted on October 22, 2019 at 20:38:37
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 Tests tell us exactly how well an amplifier works i.e does it make the input larger without adding anything or taking anything away. How people perceive the final sound is far from exactly defined.

I am sure Fourier is rotating in his grave at a constant angular velocity.

13DoW

 

RE: JA on music waveform, posted on October 23, 2019 at 02:45:29
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Well, yes. And, no.

Music is composed of complex waveforms, all of which can be broken down to a summation and multiplication of individual sine waves. Joseph Fourier described that a couple centuries ago, as has been stated here by a few already.

That's not the rub in all of this. Or, rubs.

First thing is that simple harmonic distortion measurement doesn't really characterize the linearity of an device. Yeah, it's a definitely a measure, just as a physician taking your body temperature is a measure of health.

Complex waveforms, such as sound, create lots of distortion products of various mathematical order called intermodulation distortion. (Best to look that up on your own...) As you add more of those sine wave components as described by Fourier, the number of distortion products exponentially increases. Depending on the linearity of the device, the power in these IMD products can approach half the power of the desired signals. Researchers and mathematicians published papers on this in the mid 20th century. (Again, look for these yourself.) These IMD products fall all over the audio band - their number can be so great that they appear as noise in a spectrum analyzer plot unless you look very closely.

Since these are such a significant distortion mechanism, you'd think that measuring them would be a good idea. In fact, some test gear like the latest Audio Precision models can do exactly that. For a price, of course... This kind of characterization is performed all the time on communication systems that use complex modulation like multi-channel QAM and OFDM.

This kind of testing gives more insight into the linearity characteristic of a device than simple harmonic distortion does, and stresses the device under test a lot more. What you inevitably find is that the linearity is a function of frequency. Because of the math of it all, worse linearity at the higher end of the frequency band generates more IMD at the lower of the band.

In other words, less linearity in the treble region can and often does muddy up the bass.

But, there's still more. To be continued...

 

RE: JA on music waveform, posted on October 23, 2019 at 03:09:53
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Typical multi-tone testing is performed by feeding some number of tones, like 30 or so, across the audio band at just the right frequencies, and then looking at the distortion products that fall between these tones. As you might imagine, creating a set of low distortion tones and then being able to monitor them is no small feat. That's a big part of why this testing hasn't been performed much until somewhat recently and isn't yet that popular.

Over at ASR, you can see lots of examples of this testing as performed on an Audio Precision test instrument.

Now, here's the second rub. Both good news and bad news.

In some ways, things are not as bad as they appear in the plots you might see at ASR. The standard test uses tones of equal amplitude across the spectrum. But, with sound, the spectrum just isn't flat. Think of the plots you can find all over the place of the musical spectrum. The energy is hardly flat from 20 Hz to 20 KHz.

So, a better test might be to apply spectral shaping of the tone amplitudes that roughly approximates some more realistic sonic spectrum. That way you'd see how the distortion falls when reproducing sound with the device under test.

Here's another thing. The auditory system is more sensitive at some frequencies than others. That's where those Fletcher-Munson curves, weighting filters, and loudness controls come in. To make multi-tone tests even more realistic, it would make sense to apply a spectral mask based on the Fletcher Munson curves. That way distortion products that fall into a frequency range where the ear is more sensitive would get greater import.

But, equal amplitude multi-tone testing certainly has its place. If you want to be sure that all the units coming off a production line perform the same, running this test really can tell you a lot. That's kind of a separate question issue from characterizing how a device might sound, though.

More yet to come... (Time to run, now!)

 

RE: JA on music waveform, posted on October 23, 2019 at 03:45:13
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Now, here's a harder one.

One of the ways that the spectrum analyzers make it so that you can peer at distortion products low in amplitude is by using a form of signal averaging. The idea is pretty basic. Distortion products are presumed to be constant in nature while noise is considered to be random. So, if you scan or sweep the frequency spectrum multiple times - the more, the merrier - tones will stay constant in level, while the randomness of the noise will average to some kind of lower mean level. Voila! The tones pop out of the noise.

This one presents more rubs.

First, while this is pretty good for simple harmonic testing, complex waveforms change in amplitude over time - kinda fundamental. The patterns repeat, but it's over a long period depending on the test waveform. (Find a time based display of the audio spectrum for an example of this.) The peak amplitude can and does change by a pretty good amount over time. (Look up the idea of "crest factor.") As you might imagine, the peaks can create very different operating conditions for a device under test than the more average signal levels do. The resultant distortion changes, too.

So, if you average the spectrum to get a better glimpse into the noise, these peaks and their effects get averaged out. Oops. That's because the "noise" isn't truly random.

In addition, it's quite common for devices to have a time dependent distortion characteristic. One simple example of this is thermal effects that occur over time. But, there's others as well. Think about how a power supply voltage might droop when certain signal levels are reached and how that might affect the device performance. Or, noise from rectifier diodes when the current demand increases. The list is long. Even the phase noise of the effective conversion clock in a DAC will change over time as the various things go on from sample to sample.

This isn't just a problem for audio. In communications systems, there is a metric called MER. (Another thing to look up.) Basically, this is kind of a combined score of the signal to noise ratio of a signal along with other degradations like distortion. You can trace the Bit Error Rate performance through the MER. Sorta. That's because usually the MER is measured by averaging a large number of symbols (think of that idea as similar to sweeps of a spectrum analyzer) over time. A typical number might be 1024 symbols. Well, guess what! Of those 1024 symbols, one or more can get completely obliterated by some phenomenon and it barely shows in the resultant MER. That's because the MER value is the average of the performance over time. Losing one symbol does very little to the average. But, every bit in that symbol is demolished and bad. So, the resultant bit error rate might be 1 in 1024 (example) or 1 in 10^3. Not really very good.

The same thing probably is true in audio. Imagine averaging over 32 sweeps. That gets the noise level right down there. But, what if one of those sweeps contains some awful screech or noise at some frequency? That barely affects the measurement. Think you might hear that?

Now, think back to that complex waveform that approximates sound. The peaks occur every so often, but repeatedly. In order to capture these peaks (or nulls, to be fair), you need to continue to sweep over a long time. Doesn't that actually increase the averaging, too?

The answer to this is to not use averaging. Instead use "peak hold" or "max" or whatever your test gear calls it. With an FFT (another term to investigate) based analyzer, that means capturing the peak value in each FFT "bin" and saving that. You keep the largest value for that bin over a number of sweeps. That captures all the bad things that happen over time. Yeah, you can't measure as low "into the noise" but you do get what the system does, worst case.

What audio test have you ever seen that does that?

OK, just one more post...

 

RE: JA on music waveform, posted on October 23, 2019 at 03:50:49
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 One last observation.

Another flaw in the typical testing regime is that the device under test is tested by itself. That makes sense in a lot of ways, because you isolate its performance from everything else.

But, in an actual audio system, that's not how it all works. Instead, the device is subject to all sorts of noise on the AC power mains, not just from the noise everybody talks about but from common mode currents that come from other parts of the audio system that are conducted through the various cables including the power. (This time, look up the collected works of Henry Ott and Ralph Morrison. Warning - they cover way, way more than audio topics.)

How is this performance captured?

In my mind, the problem overall is that there is an established test regime that was developed a half century ago. Technology has advanced such that we can make harmonic distortion measurements way beyond what was imagined way back when. But, getting more resolution doesn't necessarily provide more insight. It all has to be placed in context and with respect to what you really need to know. If the technology is such that you can determine that your aircraft is flying at 31027.734921 feet above the ground, that isn't as, ahh, interesting, than knowing whether it is 10 feet above the runway or six inches.

OK, that's enough from me. A year's worth of comments - or more. I hope anybody not interested in this just skipped over it and didn't get too annoyed.

 

RE: JA on music waveform, posted on October 23, 2019 at 07:55:44
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Thank you for your fantastic response. I wish this happened all the time.

It will probably take me all day to digest this, but I will try...

 

"Better Tests", posted on October 23, 2019 at 08:06:37
Inmate51
 In the pro audio world, including sound reinforcement, reproduction, recording, and design/development, there have been more useful test signals for decades, as well as a few newer ones in the past few years.

For some reason, the home consumer hi-fi world - in general - has always been slow to adopt or innovate in this area.

 

RE: Tech-babble is for manufacturers :), posted on October 23, 2019 at 08:16:59
Davey
 I can't argue that. But that's not what you said previously.

Fourier analysis encompasses a MUCH bigger patch of real estate than how it might characterize audio component performance.

Dave.

 

RE: JA on music waveform, posted on October 23, 2019 at 08:28:10
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Here's hoping you don't get indigestion...

 

Thanks, posted on October 23, 2019 at 08:37:26
E-Stat
Audiophile

Posts: 37580
Joined: May 12, 2000
Contributor
  Since: April 5, 2002
 Lots of reasons why simple uncorrelated sine wave testing on gear tells you so little.

 

RE: Thanks, posted on October 23, 2019 at 09:48:01
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I *swore* to myself that I would keep quiet after posting all that I did - I used way more than my allotted public bandwidth in this forum.

But, I have to ask.

More than once, you've talked about uncorrelated simple sine wave testing being bad.

Just what the heck does that mean?

If you have a single test tone, just what is it correlated with?

If you have multiple test tones, they are likely never correlated, unless they are at the same fundamental frequency or are harmonic to each other.

In a much more general engineering sense, (as > https://jackschaedler.github.io/circles-sines-signals/dotproduct2.html), it's really hard to, ahh, correlate test tones to musical instruments. Instruments have harmonic structures that are not necessarily correlated in the engineering sense - the seventh harmonic may be inverted, for example.

In addition, some people - Nelson Pass as one example - have found that some people prefer uncorrelated even order harmonic structures in amplifiers to be pleasing to listen to. Some times. At a recent Burning Amp gathering, he handed out gadgets to visitors that created that condition for them to try.

So, please explain.

 

RE: Thanks, posted on October 23, 2019 at 09:55:14
E-Stat
Audiophile

Posts: 37580
Joined: May 12, 2000
Contributor
  Since: April 5, 2002
 If you have a single test tone, just what is it correlated with?

Nothing at all! Therein lies the challenge.

Unless you listen to single frequency sine waves (I don't!), those tests do not replicate what audio gear is intended to reproduce - complex, harmonically rich waveforms, aka music.

 

RE: Thanks, posted on October 23, 2019 at 10:38:07
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Well, that was my first point.

And, that's what multi-tone testing is supposed to address. My second point was that it's not quite so easy as pushing the "run multi-tone test" button on the test gear. Or that the traditional approach to testing is based on what needs to be tested, rather than what could be tested a half century ago.

My third point was that there's more to this than just the non-linearities of the devices. There's lots of other distortion mechanisms as well.

I suspect that it'll be a couple generations before anything beyond the traditional testing used today catches on. For one thing, it's not likely that a paper will emerge tomorrow that shows that 7th order IMD in the band of 2.13 to 3.37 KHz (or whatever) is the key to best sound reproduction, based on actual brain scans of people listening to a wide range of music. And, even if it did, there will be people who will just trash it, no matter how solid the proof is. In fact, I'd wager that the six people who (mistakenly) took the time to read what I wrote, based on testing and characterizing systems even more complex than audio reproduction systems, probably didn't even get through the first post without thinking about what an idiot I am.

 

Sure, posted on October 23, 2019 at 10:47:57
E-Stat
Audiophile

Posts: 37580
Joined: May 12, 2000
Contributor
  Since: April 5, 2002
 And, that's what multi-tone testing is supposed to address.

Fifty years after I started reading Julian Hirsch and actually believing such tests were meaningful, we're still no further in audio reviewing publications.

Just show 'em the THD plot and be done with it. Who cares if it is useless information? :)

 

"Throw Out The...", posted on October 23, 2019 at 11:08:02
Doug Schneider
Reviewer

Posts: 881
Location: North America
Joined: April 16, 2005
 This isn't a direct response to your comment -- more of a response to comments below.

Many anti-measurement audiophiles like to throw the baby out with the bath water when measurements don't tell the whole story. And here's the thing -- they don't tell it. But they do tell part of the story.

One can argue that more complex tests could/perhaps-should be done, but that doesn't mean current testing is not relevant. It is...

Several years ago, we had one "boutique" speaker manufacturer whose speakers always showed anomalies in their measurements say something to the effect: "Their simple measurements don't reveal what our more complex measurements show." Unfortunately, they never showed their more complex measurements... ever.

That said, at the time another manufacturer charged forth and said that if a component can't pass those measurements, it's not going to look better when you go deeper. That, I'm afraid, is 100% true.

Simple sine-waves can tell a lot -- like, if it can pass a simple sine wave without various distortions. If a component can't, do you think it's going to perform better with a more complex music signal? Outside of amplifiers, look at all those non-oversampling DACs with their ladder-stepped outputs with a simple sine wave input -- do you think those problems are going to go away with a more complex signal? I can't find it now, but I remember a video where Harbeth's Alan Shaw showed a turntable playing a 1kHz signal -- it couldn't, really.

That's not to say more measurements wouldn't be helpful. We are actually in the process of assessing our measurements and expanding them. On the loudspeaker side, it involves a significant investment in test equipment to do different tests that aren't right now being done. On the electronics side, same thing. Some of the latter was brought on by a recent trip to Anthem's development center where we produced a series of videos linked below. In those videos, Marc Bonneville shows some significant and telling tests. But outside of that, in the lab, they conduct many tests when they're developing their electronics that show the performance under various challenging conditions. Some of those could be quite telling if brought properly into an audio-measurement suite.

All in all, however, a good set of measurements like we're seeing now, and have seen for years, can tell you a lot -- if you know what to look for.

Doug Schneider
SoundStage!

 

If you want to be precise about it, Fourier analysis applies to LTI systems only, posted on October 23, 2019 at 11:52:39
Dave_K
 LTI = Linear, time-invariant

An LTI system has a linear transfer function whose parameters don't vary with time. A lot of audio processing can be modeled as an LTI system, but not all. For example, most analog and digital filters are LTI. But quantization is not.

One property of an LTI system is that frequencies out = frequencies in. The transfer function of an LTI system can only change the magnitude and phase of each frequency component, it can't produce new frequency components that weren't there in the input.

Obviously, an audio amplifier is not truly a linear system. Frequencies out = frequencies in + distortion products + noise + power supply harmonics + any other junk like spurious tones. It's not strictly time invariant either due to thermal variation, but that occurs on a long time scale relative to audio signals.

Practically speaking, if an amplifier has low noise and distortion, you can treat it like an LTI system for the purpose of spectral analysis of large signals. But technically speaking, you can't assume it behaves as an LTI system when you're analyzing its distortion characteristics.

That doesn't mean that Fast Fourier Transforms aren't useful. As I'm sure you've seen before, people use FFTs all the time to characterize the relative levels of harmonics using sine wave inputs. But strictly speaking, the principle of time and frequency domain equivalence is not valid when analyzing the performance of an audio amplifier.

Note there are various methods for spectral analysis of non-linear systems, but none are as general as Fourier analysis. I'd have to brush up my knowledge of the various techniques to say this with confidence, but IIRC non-linear spectral analysis methods depend on being able to decompose signals into non-linear basis functions that are generally application specific, i.e. they would be different for an amplifier stage using MOSFETs vs. BJTs vs. triodes etc.

 

Good posts (nt), posted on October 23, 2019 at 11:59:12
Dave_K
 nt

 

RE: Tech-babble is for manufacturers :), posted on October 23, 2019 at 12:53:15
highendfan
Audiophile

Posts: 159
Location: Ontario
Joined: March 6, 2009
 THD < 1%. Granted not the best out there, however I am sure most speakers are at least 1 to 3 % off dead accurate. I suppose a square wave output response would tell the tale. If it sounds good, it is good. IMO

 

RE: If you want to be precise about it, Fourier analysis applies to LTI systems only, posted on October 23, 2019 at 13:20:35
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Actually, I think thermal affects are more insidious than you suggest.

(Example: https://www.electronicdesign.com/analog/op-amp-audio-realizing-high-performance-buffers-part-i)

Despite what everybody wishes, almost all amplifier circuits have memory issues due to various dielectric absorption issues, imperfect magnetics, and less than ideal thermal variants. Bob Cordell discusses many of these topics in his audio amplifier book.

I'm not at all arguing with you - just offering additional supporting evidence.

 

Fourier notwithstanding -- I suspect a good case could be made for square waves..., posted on October 23, 2019 at 13:36:50
mhardy6647
Audiophile

Posts: 16010
Location: New England
Joined: October 12, 1999
Contributor
  Since: October 23, 2016

if nothing else, the reproduction of square waves of various "fundamental" frequencies says something about the phase linearity of an amplifier and also of its transient response.

Just my not-particularly-hard earned observation; no more & no less.


all the best,
mrh

 

As amplifier age do they lose their memory?, posted on October 23, 2019 at 16:37:36
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 :)

 

RE: "Throw Out The...", posted on October 23, 2019 at 16:38:29
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 To Doug:

Your tests of the darTZeel amp in 2005 showed high distortion, the 'summary' even used those exact words. Yet this was one of the most transparent amps of all time, great reviews from Stereophile to back it up.

This means, with some amps, that the test system is inverting the expected readout. At least when the (perceived) noise and distortion is factored in. Do you agree ?

 

RE: "Throw Out The...", posted on October 23, 2019 at 16:49:29
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 It means that test system is telling you exactly how good an amplifier it is but that many people will perceive some distortion as preferable to no distortion - or they will prefer a large amount of second harmonic to a very small amount of higher-order distortion.

13DoW

 

RE: "Throw Out The...", posted on October 23, 2019 at 17:39:14
Doug Schneider
Reviewer

Posts: 881
Location: North America
Joined: April 16, 2005
 >>>>Yet this was one of the most transparent amps of all time, great reviews from Stereophile to back it up.

Is this a consensus around the world? Honestly, I've never heard anyone else say it's one of the most transparent amps of all time. Not saying it is or isn't -- just never heard that.

>>>>This means, with some amps, that the test system is inverting the expected readout. At least when the (perceived) noise and distortion is factored in. Do you agree ?

I'm not sure I can agree or disagree when I can't figure out what you're trying to say.

Doug Schneider
SoundStage!

 

RE: As amplifier age do they lose their memory?, posted on October 23, 2019 at 17:46:01
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I forget.

 

RE: "Throw Out The...", posted on October 23, 2019 at 18:03:11
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Based on what I heard and the blow-out reviews from 'phile (where JM and WP wanted to create a 'super-amp' category for it).

And Jeff Fritz's very positive review in '05. The high-distortion numbers were opposite of what (I) and these reviewers heard. For a SS amp, not tubed -the normal culprit of high-distortion readouts.

 

RE: "Throw Out The...", posted on October 23, 2019 at 18:44:02
Doug Schneider
Reviewer

Posts: 881
Location: North America
Joined: April 16, 2005
 I think you're grasping at something with one amp with one set of measurements and then extrapolating a lot from that. Once again, I've never heard anyone say that it's the most transparent amp out there. I'm not saying you're right or wrong -- but seems to be a lot of conclusion jumping there.

That said, as far as I recall, Jeff's review was somewhat positive, but we took a lot of heat for the negative parts of it, which Jeff just wrote about a couple of months ago. I've linked his article below.

Doug
SoundStage!

 

RE: "Throw Out The...", posted on October 23, 2019 at 18:51:16
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 The reason for the 'heat' was the awareness (of the greatness) of this amp. Read John Marks and Wes Phillips reviews in Stereophile.

Great amp, high distortion, that's all I'm saying...

 

RE: "Throw Out The...", posted on October 23, 2019 at 20:35:51
Ivan303
Audiophile

Posts: 48887
Location: Cadiere d'azur FRANCE - Santa Fe, NM
Joined: February 26, 2001
 "While its distortion is not as low as is usually found in modern solid-state designs, perhaps of greater importance is the fact that that distortion comprises low-order harmonics, and that the distortion doesn't change its harmonic character with frequency"


First they came for the dumb-asses
And I did not speak out
Because I was not a dumb-ass

 

Stability tendencies too! (nt), posted on October 23, 2019 at 21:25:55
Steve O
Audiophile

Posts: 12359
Location: SE MI
Joined: September 6, 2001

 

RE: "Throw Out The...", posted on October 24, 2019 at 05:58:27
Doug Schneider
Reviewer

Posts: 881
Location: North America
Joined: April 16, 2005
 I always say that the measurements are what they are. If you like it, that's OK.

Doug Schneider

 

Agree about square waves, posted on October 24, 2019 at 07:46:39
E-Stat
Audiophile

Posts: 37580
Joined: May 12, 2000
Contributor
  Since: April 5, 2002
 I believe they show why I find switchers challenged at the top. Courtesy of Stereophile, here's a great example of a $30k Ncore based design.

Dan D'Agostino calls the result "fuzz":



Even when filtered, the 10k one looks decidedly "loopy" unlike what you find with good class A/AB amps.




 

RE: "Throw Out The...", posted on October 24, 2019 at 07:54:11
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Soundstage/NRC measurements are the best. Their summary of this amp was "high distortion".

If harmonic-character is all that counts, then why is the industry drilling low-distortion numbers into our heads since the 1970s ? The test equipment is clearly out-of-sync with perceived sound quality. A WIDE gap in the darTZeel's case...

 

the switching amps do have peculiar properties (measurements) vis-a-vis Class A or AB (nt), posted on October 24, 2019 at 08:09:43
mhardy6647
Audiophile

Posts: 16010
Location: New England
Joined: October 12, 1999
Contributor
  Since: October 23, 2016
 enn tee


all the best,
mrh

 

RE: JA on music waveform, posted on October 24, 2019 at 10:02:28
John Atkinson
Reviewer

Posts: 4045
Location: New York
Joined: November 24, 2003
 >it's quite common for devices to have a time dependent distortion
>characteristic. One simple example of this is thermal effects that occur
>over time. But, there's others as well. . .The list is long.

This is a good point. The assumption made when using steady-state
measurements of audio products is that there is no relationship between
the signal and the devices's transfer function, ie, the transfer function
doesn't vary with signal level, signal content, or time. Most of the time
this assumption is correct, and when it isn't, such as when an amplifier's
output devices change their operating parameters with temperature, this is
addressed by the preconditioning I perform when testing amplifiers for
Stereophile.

Something related to this possible dependence of an amplifier's transfer
function on the properties of the signal it amplifies is that I believe an
amplifier tends to be preferred when the distortion "signature" is
low-order - the second and third harmonics only - and doesn't vary either
with frequency or output power. As long as the distortion isn't sufficiently
high that it produces significant levels of intermodulation distortion, such
an amplifier will sound "musical."

Someone in this thread mentioned multi-tone testing. I actually created
four types of multitone test signal back in the early 1990s to test lossy
codecs and included them on Stereophile's Test CD 3. They worked well
for revealing codec errors but were not revealing with amplifiers, perhaps
because the files were 16-bit PCM and therefore had too high a noise floor.


John Atkinson
Technical Editor, Stereophile

 

RE: JA on music waveform, posted on October 24, 2019 at 10:42:44
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Tech-babble from the man himself. What "pre-conditioning" do you do ?

If Inmate51 (above) is correct "there have been more useful tests for decades". And if Dave K (above) is right, "Fourier analysis applies to LTI systems only".

 

RE: JA on music waveform, posted on October 24, 2019 at 10:43:31
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I think, if you investigate more closely, you might find that thermal effects, which are the result of temperature variations, even happen over a very short time. A good search term to research is "thermal tails". There's also a discussion of this in Cordell's book with regard to higher dissipation devices. So, while soaking a device at elevated temperatures may expose some flaws, it doesn't necessarily expose others.

There's other factors that add to this kind of behavior. A good search term for this is "settling time" with regard to opamps and similar devices.

I'm not going to say that a specific settling time anomaly can predict how a device sounds - there's been few studies of this to my knowledge. But, it is just one measure of signal imperfection that has a real effect. Conceptually, having signals lag after the event isn't much different than unwanted acoustic reflection inside the listening room.

As for multi-tone testing and the like, various other industries have used their own version of this for decades. Analog telephony used noise power ratio testing as a way to investigate linearity and noise. How long ago was analog telephony obsoleted? Cable television has used multi-tone testing to separate odd order distortion products from even order distortions, both of which certainly limit system performance - far, far more than simple harmonic distortion does.

In both of the above cases, the use of simple harmonic testing just wasn't representative of the actual use of the system nor did it really stress the system very much. So, they moved onto more complex and expensive test methods.

So, just why hasn't more sophisticated testing caught on? I'm not in a position to answer - I'll bet you are, though! You've tried advocating for something better yourself.

One of the conceptual problems I think most observers of audio performance have is a lack of practical, if not theoretical, idea of how IMD products manifest themselves. The math has been there since, well, forever - Brockbank and Wass published a paper on the subject back in 1945.

For those interested, I've attached a link to a proposal from the mid 70's for a test regime. I'm not advocating for using this approach, but there's a lot of good references in the document.

 

RE: Good posts (nt), posted on October 24, 2019 at 10:44:41
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Shucks...

 

RE: JA on music waveform, posted on October 24, 2019 at 13:07:05
John Atkinson
Reviewer

Posts: 4045
Location: New York
Joined: November 24, 2003
 >Tech-babble from the man himself. What "pre-conditioning" do you do?

If you read my amplifier measurements in Stereophile, you will note that
every amplifier is run at 1/3 power into 8 ohms for an hour before I do any
testing. This power level is thermally the most stressful for a solid-state
amplifier with a class-AB output stage, which almost all do. It therefore
maximizes the temperature of the output devices, minimizing any subsequent
thermal modulation by the signal.

John Atkinson
Technical Editor, Stereophile

 

RE: JA on music waveform, posted on October 24, 2019 at 13:08:44
John Atkinson
Reviewer

Posts: 4045
Location: New York
Joined: November 24, 2003
 >For those interested, I've attached a link to a proposal from the mid 70's
>for a test regime. I'm not advocating for using this approach, but there's
>a lot of good references in the document.

Thank you for the link. I'll check out the proposal.



John Atkinson
Technical Editor, Stereophile

 

Many pearls were cast before you in this whole thread, posted on October 24, 2019 at 16:56:43
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 It is obvious that you do not have good technical understanding. That you would blame Fourier for a perceived discrepancy between the subjective review and the THD measurement of one amplifier would be funny if you weren't being serious - which I strongly suspect you were. And that you now criticize JA's testing methodology when you clearly do not understand it is pretty poor.

13DoW

 

RE: Many pearls were cast before you in this whole thread, posted on October 24, 2019 at 19:48:38
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 As a declared audiophile, it should be no surprise that I'm not a techie.

And I'm not blaming Fourier for the darTZeel amp. A more general look (of Fourier) reveals that it applies to LTI's only.

I did make a separate note on the industry (drilling low-THD into our heads). A different topic I admit.

JA's test appear to be a resistive load at 8 ohms, no 'torture-testing'. The test frequencies are probably carefully-chosen, to avoid showing 'too much' of the design. All this means 30 years of deeply-flawed tests from Stereophile...

 

RE: "Throw Out The...", posted on October 24, 2019 at 20:43:22
hahax@verizon.net
Audiophile

Posts: 4306
Location: New Jersey
Joined: March 22, 2006
 If an amp has to at least be able to pass sine waves the single ended amps must be considered non hi fi since they can't accurately pass sine waves. They compress the negative half of the wave relative to the top half. Nelson Pass who does like single ended amps confirmed this for me.

 

RE: "Throw Out The...", posted on October 25, 2019 at 03:39:35
Doug Schneider
Reviewer

Posts: 881
Location: North America
Joined: April 16, 2005
 Hi,

I have to clarify that no one is staying stuff like "harmonic-character is all that counts."

Doug Schneider

 

RE: Many pearls were cast before you in this whole thread, posted on October 25, 2019 at 06:02:41
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I think you'll find that the test frequencies used by JA in his Stereophile tests are very commonly used in the audio world. Some even adhere to standards set forth some years ago.

Personally, I sense absolutely ZERO attempt at fraud or anything like it from Stereophile with regard to testing.

You can certainly argue about whether these tests are really meaningful or adequate in terms of telling you how a component sounds, but I don't believe there's any attempt to mislead the reader. (Of course, I could be proven wrong, since I'm not privy to what goes on. But, that's my view, fwiw.)

To me, Stereophile, just like, say, Road and Track, is not a technical journal. Both are magazines devoted to providing the readers information on a subject they are interested in. Some of that information is technical in nature, at least to the extent that can fit in the pages and is within reach of the technical capabilities of the audience. I even find similarities between the two magazines just mentioned in terms of the "subjective" reviews. Having several writers fast lap around a race track in a sports car and reporting their lap and segment times isn't that different from a listening observation. Each driver has different capabilities that gives different times and different opinions. Unless you happen to be one of those drivers, it's interesting to read, but may or may not apply to you. The interested in reading part is the key.

 

RE: Many pearls were cast before you in this whole thread, posted on October 25, 2019 at 06:25:23
John Atkinson
Reviewer

Posts: 4045
Location: New York
Joined: November 24, 2003
 >As a declared audiophile, it should be no surprise that I'm not a techie.

Someone can be an audiophile and a "techie." But if you are not a "techie,"
note that as an editor, my standing instruction to the people who used to
write for me was: "Don't write what you don't know."

>And I'm not blaming Fourier for the darTZeel amp.

The darTZeel amplifier typifies what I wrote in the original review's
measurements section: "While its distortion is not as low as is usually
found in modern solid-state designs, perhaps of greater importance is the
fact that that distortion comprises low-order harmonics, and that the
distortion doesn't change its harmonic character with frequency." This
amplifier's operating parameters are not correlated with the audio signal
and as I wrote in an earlier posting, I feel that this results in an amplifier
that sounds "musical."

>A more general look (of Fourier) reveals that it applies to LTI's only.

Other than amplifiers with multiple voltage rails, I have only come across
one amplifier that had a significant variation of its operating parameters
with time, affecting steady-state measurements. The original Schiit
Ragnarok was intentionally designed so that its output-stage bias current
was related to the audio signal history. I designed tests to unmask this
behavior - see figs.5 & 11 and the postscript at the link below.

>JA's test appear to be a resistive load at 8 ohms, no 'torture-testing'.

As I wrote earlier in this thread, the power level at which I precondition
amplifiers is thermally the most stressful. It is indeed a torture test
and some amplifiers turn off during this test.

>The test frequencies are probably carefully-chosen, to avoid showing 'too
>much' of the design. All this means 30 years of deeply-flawed tests from
>Stereophile...

At the risk of repeating myself, "Don't write what you don't know."


John Atkinson
Technical Editor, Stereophile

 

RE: As amplifier age do they lose their memory?, posted on October 25, 2019 at 06:40:51
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Hey! Maybe that explains "burn-in" phenomena!

 

Just to clarify, posted on October 25, 2019 at 06:54:45
Dave_K
 As I mentioned in the middle of my post, the math behind Fourier analysis requires making assumptions that aren't valid for a non-linear system. HOWEVER, that does not prevent you from using one of its tools, the FFT, to learn about the distortion behavior of an amplifier. All those FFTs you see showing noise floors and distortion spectrums are fine.

What you can't do is fully characterize the transfer function by measuring the frequency response (mag & phase), or use inverse FFT to obtain the time domain response.

 

RE: Just to clarify, posted on October 25, 2019 at 11:05:52
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 But Atkinson is making those assumptions...

His tests also don't use a multitude of frequencies. But, the torture test IS valid (as a precondition). I'm referring to what the reader complained about -no real-world, complex waveforms used DURING the test.

And a nulled output vs. input, which MAY have been an issue with available equipment, but I'm not sure of this.

JA can defend himself all he wants but he (apparently) isn't convincing anyone here. As mentioned above, I'd like to know what the 'more useful tests' are used in the pro-audio world.

 

:) (nt), posted on October 25, 2019 at 15:09:39
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 nt

 

RE: JA on music waveform, posted on October 25, 2019 at 15:40:40
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I did some more research on how spectral shaping used with multi-tone testing might give better insight into how devices might sound.

First, look at the paper in the link below.

If I'm reading this right - and I certainly hope somebody will correct any error I hold - the level of high frequency energy in a musical performance is roughly 40 dB lower than that at 100 Hz or so.

This suggests that IMD measurements performed with tones at 19 and 20 KHz might not tell much in terms of musical fidelity. Even if you have speakers that are on the lower sensitivity side - the Vandersteens we use at home certainly qualify - and, even if you are playing at far too loud volume, the amount of power required for the highest treble is measured in 10's of milliwatts.

So, perhaps IMD measurements should be made with the tones at much lower frequencies where the normal sound level power is pretty much close to its peak.

At this point, I'm going to throw in a measurement I made a while back. Without going into the details of the measurement, I found that with our own loudspeakers in our room, the most power we'd ever actually need is around 20 watts. Even that is really painful. (There's other reasons why you might want an amplifier with higher voltage supply rails and higher output current capacity, but system loudness is not one. That is, unless you have a really, really large room or you have hearing issues. Or want hearing issues.)

What all this suggests to me is that the traditional testing approaches may be great for ensuring that all units coming off the production line perform as designed, but the testing doesn't tell much about the quality of sound you'll get from they design.

Other test approaches are needed that take into account all the system factors and consider what kind of signal actually passes through. You might agree with me that most listening is done with the amplifier only pumping out a watt or three, if that, so other crap getting into system might be a much more significant detriment to sound quality than we usually think. We probably need to test preamps at the output levels they usually are run at, with the volume control set to that point. The noise levels at the output of a power amplifier should be referenced to one watt, instead of maximum output power. And so on. (OT: It can be surprising how electronics perform over a wide range of source impedances. As in, changing the volume control. That's something you really can measure well.)

A well known audio guy, who has passed from our mortal coil, once told me that working on making better tests was a waste of time. Not because you couldn't make a good set of tests - it's probably possible. His point was that time spent on making tests is time *not* spent on making better DACs, preamps, amplifiers, and what have you. I think he may have had a point.

 

RE: JA on music waveform, posted on October 25, 2019 at 15:56:49
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Mmmm....so better/fewer circuits and better parts (esp. in output stage) is where it was.

I mentioned darTZeel more than once here, but I cannot afford this amp. Other co. are innovating as well. I think Sanders was the first to use Thermal-Trak output transistors and an IC-regulated power-supply. Did it measure better than the (heavily advertised) Halcro .00001% ? Probably not, but IMO an important (sound) advance in amplification.

 

RE: JA on music waveform, posted on October 25, 2019 at 16:20:41
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I'm not sure how much darTZeel innovated. They use a diamond output stage - that's been around for half a century now. IC manufacturers used it before darTZeel did and Walt Jung publicized it as well. Maybe not for power amplifiers, but that's a matter of degree, isn't it? Other small power amplifier companies used diamond output stages before darTZeel even existed. They also use a trick at the output stage that attempts to moderate the effects of Gm doubling in a push-pull output stage.

That doesn't mean that the design is bad. Or good. I have no opinion. I've never seen, much less heard, any of their products, except I know that they stole the color scheme from Tony Stark.

What I'm trying to say is that there's a lot of design and performance details that get ignored in most testing. Single ended triode amps measure badly using traditional methods. Lots of people are willing to pay a lot for high quality examples of these. Perhaps it's worth trying to analyze what they do right technically rather than what they do "wrong."

BTW, interesting point about Thermal-Trak devices. The notion that these may give better results suggests that thermal effects are not just a function of how hot the devices get in use.

 

RE: "Throw Out The...", posted on October 25, 2019 at 16:23:23
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Or, that the distortions the traditional tests emphasize are not the same as what our auditory systems are sensitive to.

 

RE: "Throw Out The...", posted on October 25, 2019 at 18:18:10
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Would be nice to measure gate-capacitance (of the output stage). Sander's Thermal-Trak vs. conventional MOSFETs.

 

Whatever happened to, "I like the way that sounds" . . ., posted on October 26, 2019 at 16:27:12
Billy Wonka
Audiophile

Posts: 3760
Joined: April 25, 2013
Contributor
  Since: October 15, 2013
 Without referring to spectral sheet music to see if it is acceptable to like the way it sounds. Technical analyses have been oversold establishing a scientific bible ruling over a subjective matter.

Burn your oscilloscopes and be free to listen to what you want.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 26, 2019 at 20:29:03
hahax@verizon.net
Audiophile

Posts: 4306
Location: New Jersey
Joined: March 22, 2006
 Absolutely fine when choosing your own system but what has this to do with fidelity? And if it's a published review do you want 'it sounds good to me' as a review? What good is this unless you personally know the reviewer? Any objective info added to a review makes it more universal and probably more accurate. Some measurements do give useful info. I have seen(heard) it.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 27, 2019 at 10:25:51
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 To be clear, this piece was not about abolishing tests. If a test can confirm (or disprove) a claimed S/N, then fine. Testing rated-power -maybe, but how many co. lie about this ? You can test a speaker's cavity-noise in db, if not too hard to do.

The question was what John Atkinson is doing, for the last 30 years. Then Fourier, seems to be a failure in the frequency and time domains.

Even Art Dudley questions one of his tests (DAC resolution) in his Nov. column. Then says "the time has come to acknowledge a disconnect". Why it took 30 years to say this is beyond me.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 27, 2019 at 14:33:11
John Atkinson
Reviewer

Posts: 4045
Location: New York
Joined: November 24, 2003
 >The question was what John Atkinson is doing, for the last 30 years.

Perhaps you didn't see my most recent response to you.

>Then Fourier, seems to be a failure in the frequency and time domains.

Oh dear. :-(

John Atkinson
Technical Editor, Stereophile

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 27, 2019 at 15:22:56
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 I saw it, 'nuttin changes...

 

The only music waveform is from speakers. What is the audio signal anywhere else? nt, posted on October 28, 2019 at 05:31:08
geoffkait
 nt

 

RE: The only music waveform is from speakers. What is the audio signal anywhere else? nt, posted on October 28, 2019 at 10:29:39
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 My mistake -signals, not waveforms.

On the technical side, short-time Fourier (STFT) or Hilbert Transform (HT) is what should have been used. NOT Atkinson's Fourier.

For those who can digest a tech-PDF, type "an improved hilbert transform for non-linear citeseerx". I don't understand it, but someone like Dave K here probably would.

 

Seeing Fourier as a failure is like complaining to Darwin that there are still idiots (nt), posted on October 28, 2019 at 10:48:14
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 nt

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 28, 2019 at 13:48:55
John Atkinson
Reviewer

Posts: 4045
Location: New York
Joined: November 24, 2003
 >>Perhaps you didn't see my most recent response to you.

>I saw it, 'nuttin changes...

So it goes.

If you need to brush up on FFT analysis, the book linked below has been my
bible on the subject for many years.

John Atkinson
Technical Editor, Stereophile

 

Hey, you're getting close to home . . . , posted on October 28, 2019 at 14:28:25
Billy Wonka
Audiophile

Posts: 3760
Joined: April 25, 2013
Contributor
  Since: October 15, 2013
 I just walked out of the jungle and could use a Diet Coke.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 28, 2019 at 16:33:08
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 But it's the wrong book. I'm new to this but it really does seem that STFT and Hilbert are (much) more appropriate for non-linear systems.

I didn't expect this thread to open-up a deep discussion. I hope Austin and Dudley see it....

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 28, 2019 at 19:40:09
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 With all due respect, I think you are missing the point.

Fourier was hardly a one trick pony. He covered a lot of area.

As someone here already pointed out, using his transform to describe a system has some limitations and restrictions. That does not by any means make his work useless.

Besides, I've never once seen any audio magazine make even a vague attempt to define the transfer function of any audio component. Perhaps I missed it, since I haven't been a faithful and continuous reader to any of the magazines for the past 40+ years. I'm sure somebody will point out what I missed.

I'm not sure what anybody would do with that information anyway.

But, what I have seen a lot of use of is the Fast Fourier Transform, which is the basis for most modern low frequency spectrum analyzers. That transform is used by the test gear, not the device being tested - there's a difference.

My own opinion is not that these analyzers don't work properly, but that the use of averaging to find teeny weeny distortion products amidst totally random noise is a bit wrong headed. For one thing, this method presumes that the noise components are truly random and never even remotely correlated with the desired signal. It may be convenient to think like that, but it clearly is not true, especially with constantly changing source material and constantly changing device characteristics. Plus, I don't think our aural systems work that way. If they did, there would be no such thing as audible noise, since our brain would average it all out. In fact, we'd have died out by now. If you average any sound over a long enough time, it averages to some very low level, probably described by the second law of thermodynamics. So, forget about using sound to avoid an oncoming wildebeest, truck, rain storm, or swarm of bees.

Much of the testing performed today is production to a specification test and design verification. The tests have to a large degree become an end unto themselves, since they are really abstracted from what the actual purpose of the devices are - to reproduce some aspect of sound. A two tone IMD test with the tones at 19 and 20 KHz is an interesting stress test, especially at high power levels, but they really are not relevant to home audio. There just isn't that much energy at those frequencies compared to below a KHz. In addition, most listening is done at pretty low levels where cross-over distortion (in push-pull circuits) is a way bigger deal, just as one example. Thermal effects in resistors as a function of frequency are significant, too. You won't learn about these with that IMD test.

BTW, I accept that you have no reason to believe me. So, I suggest you check out some of the links I posted previously as well as Cordell's book and papers published by Richard Cabot, the founder of Audio Precision. Neither of these guys are trying to sell you anything other than books or test equipment, respectively.

So, I'm not sure what you're trying to get across here. I certainly hope that I didn't inadvertently plant any false impressions earlier.

 

Thank You, good stuff)nT, posted on October 29, 2019 at 06:12:17
Cleantimestream
Audiophile

Posts: 7550
Location: Kentucky
Joined: June 30, 2005
 ~!
The Mind has No Firewall~ U.S. Army War College.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 29, 2019 at 09:53:14
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Thanks for explaining the limits of (today's) testing. That's why I mentioned the STFT and Hilbert Transform -much more appropriate for non-linear systems. I typed-out a link, but there are more...

A post here mentioned that spectral sheet-music should be referenced when testing. I guess Atkinson isn't doing that !

Finally, it's quite obvious that test equipment was invented in the 1960s to make electronics look good. Esp. the many-stage A/B amp w/ feedback. When products are not built to those standards, out they go ! And here comes the 'high-distortion' readouts. This was a scam all the way...

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 29, 2019 at 17:26:32
hahax@verizon.net
Audiophile

Posts: 4306
Location: New Jersey
Joined: March 22, 2006
 I know it's been a few decades but probably the best audio measurements were done in the old Audio magazine by Richard Heyser. His reviews were special and a step above anything probably ever published in popular magazines. One can only wonder where his research would have led him if it wasn't for his premature death.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 29, 2019 at 17:42:33
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I don't think that's right.

The original distortion analyzers came into use in the 30's or 40's. Spectrum analyzers became available in the 60's. See the link.

The point I am obviously failing to convey is that the use of an FFT based spectrum analyzer has nothing to do with any distortion caused by the device under test or its transfer function. The FFT is used by the analyzer, not by the audio equipment. A scale does not cause one to be skinny, fat, or somewhere in between. It just reports a measure of the force gravity applies to your mass. Same thing here. The math used by the analyzer is independent of the distortion.

There probably are charlatans in the audio business, as there are in most every human enterprise. But, in general, I think the people running these tests are just sticking with tradition or whatever you want to call it. That may not be as helpful as we'd wish in analyzing whether we'd like how the product might sound, but that doesn't make it dishonest.

I think that I don't have any more to offer in this discussion, so I'll stop here. I should have stopped a while ago.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 30, 2019 at 08:40:46
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 You lost me. I'm simply reporting that other transfer-functions (for test equip.) exist besides FFT. They seem to be much more appropriate for non-linear systems.

A poster above says "there are more useful tests" from the pro-audio world. And now someone believes that Richard Heyser had more useful tests in the 70s.

Then Art Dudley in his current column, calling-out Atkinson's DAC-resolution test. Then for speakers, the extreme lack of bass-harmonic distortion in today's test-world.

This *was* a hoax -or at least became one. Your posts here revealed the shortcomings of today's tests. Thanks for your contribution.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 30, 2019 at 12:05:50
John Atkinson
Reviewer

Posts: 4045
Location: New York
Joined: November 24, 2003
 >Art Dudley in his current column, calling-out Atkinson's DAC-resolution test.

You don't seem to have understood what Art wrote. He didn't "call out" this
test, he wrote from the viewpoint of someone without a formal audio
engineering background that he did not understand it (or that I did not
explain it well enough). See the link below, where Art writes "I have asked
JA about this at least twice, which is also the number of times he has
explained it to me, and I still don't get it."

John Atkinson
Technical Editor, Stereophile

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 30, 2019 at 15:32:19
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 It can't be much of a test. I'm sure you explained it to Dudley over the years and if he can't get it by now...

Then, the average reader in Stereophile is no smarter than he, so your own readers are thrown-off.

Lastly, no signals need 20-bit devices and what good is a 20-bit DAC if the system's noise-floor cancels it out. DAC errors (there are 7 types) would be much better tests. Hard to do, but it would be nice to see (besides white papers from Analog Devices).

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 30, 2019 at 17:41:58
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 After reading this again, and again, I want to make something clear.

For those following at home, keep in mind that the test equipment should not have any non-linearities, or at least minimal non-linearities. That's kind of fundamental to the test process. So, using a standard Fourier transform is fine in the test gear.

Of course, no test gear is perfect, but Audio Precision is pretty friggin good. I am certain that if some other approach would give more accurate measurements, they'd use it. They have done that throughout their history. As better techniques become available through advances in technology, they are the first to use it. The price is high, but you get what you pay for in this case.

It's the audio gear with the non-linearities. It's the audio gear that might be imperfectly described by a standard Fourier transform.

So, you're comparing apples to hickory trees.

However, you seem to be promoting a particular argument. I'll not stand in the way of your agenda. Good luck!

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 30, 2019 at 17:57:31
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 Right about the equipment. But I said "transfer-functions for non-linear systems". Meaning AUDIO GEAR. I thought it was obvious what non-linears were, even you didn't know ?

Never said test equipment is non-linear.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 30, 2019 at 18:21:06
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 So, then, why go on about the other transforms? If the test gear is pretty linear, the other transforms aren't needed. Best I can tell, nobody is using transforms to define the performance of the audio gear.

If nothing else, you've given me a headache.

How about this, then...

You win. I give up.

Everybody please ignore my previous comments. I'd remove them from the website, but that would cause a mess to readers due to lost continuity - my apologies for that.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 30, 2019 at 18:57:32
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 I gave a link -and mentioned more than once, other transforms like STFT & Hilbert. Which seem to be more appropriate to visually-represent audio signals (vs. standard FFT).

Please Google these before you come back with strange, unaware responses.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 31, 2019 at 18:57:23
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Despite your accusation, I spent the time to read the paper you pointed to. Twice, just to be sure.

I completely get why the modified Hilbert transform gives more accurate data over a long period of phenomena like damped and non-linear mechanical vibration. Regular Fourier transforms are known to be "leaky" and therefore are not entirely accurate in presenting the frequency spectrum.

OK...

Since I'm well known to give strange, unaware responses - as the following will demonstrate to you, yet again - please explain to me how and why this will provide better insight into harmonic spectral analysis of an electronic amplifier. In simple terms, please.

What does knowing that the seventh harmonic is really at -78 dBc instead of the previously measured -80 dBc tell us? If you average multiple sweeps to reduce the noise floor, which also averages the signal values in each bin, how does the modified transform provide additional, not just potentially more accurate, information? My impression is that the issue with audio measurements is not the absolute accuracy of the measurements, but whether they provide adequate information to really describe the system.

Enquiring minds want to know!

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 31, 2019 at 19:37:32
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 A deep question. For an overview of the limits of FFT and (possible) other methods, try Dave K's post on Oct 23 11:52.

I also see new research (published in the past decade) that vie for a way out of FFT. A Google search leads to PDFs that young-techs might understand.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on October 31, 2019 at 19:48:02
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 I read his well thought out post.

And, I'm asking you since you were so vocal about my strange, unaware responses.

Please explain what additional information can be gathered by a "wave analyzer" using any other transform for a simple harmonic analysis, especially when the traditional averaging technique is used to reduce the noise floor. Not more accurate, but more informative.

Even a link would give me a good start.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on November 1, 2019 at 10:42:12
J. Phelan
Audiophile

Posts: 810
Joined: May 12, 2009
 The "unawareness" was your repeated assertions that I was talking about the linearity of test equipment. I never said that and then, this thread went into various ways to create a spectrogram.

You can ask someone else here (or go to another forum) for questions about the math/visual rep. models. I've said it before -I'm still new to all this...

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on November 1, 2019 at 10:46:11
13th Duke of Wymbourne
Audiophile

Posts: 2793
Location: Orange Co., Ca
Joined: September 19, 2001
 Hi CG,
I fully agree that a more informative test is needed but the elephant in the lab, IMHO, is the transfer function from sound-in-the ears to enjoyment. I think that will be hard to pin down. Even if perception is understood there are layers of taste and preconception that can never be universal.

Regards
13DoW

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on November 1, 2019 at 10:51:47
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Well, you certainly are assertive - and accusatory - about areas you admit to being unfamiliar with.

And, once again, I'd like you to point me to anywhere where the transfer functions for audio gear have been described or measured. Specifically how Fourier math enters into that.

 

RE: Whatever happened to, "I like the way that sounds" . . ., posted on November 1, 2019 at 10:53:45
CG
Audiophile

Posts: 432
Joined: October 11, 2000
 Well, I'm not sure "never" applies. That's a long time.

But, I certainly agree that it's a knotty problem that probably doesn't have much emphasis placed on it.

 

RE: Whatever happened to, "I like the way that sounds" . . .+1 re Both Your Postings Here (nt), posted on November 1, 2019 at 12:22:53
goldenthal
Audiophile

Posts: 1001
Location: Ontario
Joined: March 28, 2003
 n

 

. . .Richard C. Heyser, required reading for all ... a humble man whom dwarfs every other audio man..., posted on November 3, 2019 at 20:25:06
Cleantimestream
Audiophile

Posts: 7550
Location: Kentucky
Joined: June 30, 2005
 To whit:
The Mind has No Firewall~ U.S. Army War College.

 

RE: The only music waveform is from speakers. What is the audio signal anywhere else? nt, posted on November 18, 2019 at 08:14:02
ishmael39@earthlink.net
Reviewer

Posts: 10
Location: Brooklyn
Joined: March 11, 2017
In my view, the only valid test signal is music playing through the amp or speakers. To see it and evaluate it is EASY -- all one must do is compare what the signal looks like (on an ociliscope) at the input and output. Any differences (besides gain) are either subtractive or additive distortion and VERY obvious. If you can see it you can hear it.

The reason "testers" don't want to do this is it does not produce numbers and "testers" need numbers.

just sayin'
herb

 

RE: The only music waveform is from speakers. What is the audio signal anywhere else? nt, posted on November 18, 2019 at 10:51:36
Inmate51
 I think some people here, including myself, would be interested in learning more about how you, or folks you know, are currently accomplishing this.

 

RE: The only music waveform is from speakers. What is the audio signal anywhere else? nt, posted on November 19, 2019 at 21:13:36
Krav Maga
Audiophile

Posts: 2350
Location: Texas
Joined: October 19, 2017

Isn't that what Technics did back in the Seventies?
"All thoughts are prey to some beast" - Bill Callahan

"I'll be your mirror
Reflect what you are" - Lou Reed

"Blind belief in authority is the greatest enemy of truth." - Albert Einstein

 

Never been to a live acoustic music concert, then? , posted on November 29, 2019 at 02:40:16
Timbo in Oz
Audiophile

Posts: 23221
Location: Canberra - in the ACT - SE Australia
Joined: January 30, 2002
 Sheesh!

Do try to tell me how you can exist, walk around, and eat with your head back-down-and-in-there.

Our speakers should not get in the way of the original signal.




Warmest

Tim Bailey

Skeptical Measurer & Audio Scrounger


 

Sooo, - Acoustic music concerts have NEVER happened, never Will happen? , posted on November 29, 2019 at 02:55:07
Timbo in Oz
Audiophile

Posts: 23221
Location: Canberra - in the ACT - SE Australia
Joined: January 30, 2002
 ROTFLMAOff WTOLRDMCheeks.

You have managed in one post, to ensure I do not bother with any future posts by yourself.

Well done.

SIASDoes.


Warmest

Tim Bailey

Skeptical Measurer & Audio Scrounger


 

I hear musical waveforms from acoustic instruments and human voices at least once a week. , posted on November 30, 2019 at 21:55:24
Timbo in Oz
Audiophile

Posts: 23221
Location: Canberra - in the ACT - SE Australia
Joined: January 30, 2002
 What you have posted is silly.

What speakers do to that, comes after what the engineers and producers did to it.

Most, of what most of you listen to, has so little fidelity to music-making as I prefer to experience it, that I wonder why most of you bother!




Warmest

Tim Bailey

Skeptical Measurer & Audio Scrounger


 

It seems to me..., posted on February 27, 2020 at 13:38:39
EdAInWestOC
Audiophile

Posts: 6828
Location: Glen Burnie, MD USA
Joined: December 18, 2003
 The thing that seems obvious to me is that we have lots of very intelligent people who have different opinions on what should be measured in audio equipment reviews. What we obviously do not have is some way to perform meaningful measurements. A way to measure something that corresponds to the way a component actually sounds.

I am not a fan of measuring components with steady state signals. It really doesn't tell us very much about how a component sounds. It is basically important because a piece of gear should be able to reproduce steady state signals reliably.

The problem arises when people assume that these measurements actually tell us how any piece of gear sounds. Of course frequency response measurements can reveal obvious issues but any decent manufacturer should be able to make a piece of gear that reproduces the audible spectrum without any frequency response anomalies and with little to no appreciable distortion. That is a basic.

I do not blame the reviewers for these measurements. It is the reader who jumps off and assumes that these measurements tell us anything about how the component sounds. This is where good high end audio magazines come into play.

A good magazine must include basic measurements to ensure that the piece of gear is properly designed and manufactured. Where the magazine adds value is with the reviewer adding the listening impressions of the gear. This is basically the quantitative and qualitative assessments of the gear.

Neither stands alone and you cannot depend on any measurements or any reviewer alone. The reviewers assessment of the gear has all of the pitfalls of system synergy and room interaction issues. That plus the reviewers point of view vs your own leads to a qualitative assessment that you have to take with that grain of salt again.

The measurements and the reviewers impressions are what I use to add or remove a piece of gear from my audition list. I utilize reviews in Stereophile and TAS to help me setup an initial list of components I may be interested in. Then I seek out auditions.

I do not think any measurement is some indication of how a piece of gear sounds. IMO we are still years away from being able to describe that. We must understand how human hearing works better and then how gear reproduces music better. Steady state signals are a simple sanity check...again IMO.

Ed
Life is analog...digital is just samples thereof

 

Page processed in 0.059 seconds.

[ Top Top of Page ] [ Contact Us ] [ Support/Wish List ] [ Copyright © 2024, Audio Asylum, all rights reserved ]
[ General ] [ Speakers ] [ Tubes ] [ Vinyl ] [ Digital ] [ Hi-Rez ] [ Video Asylum ] [ Cables ] [ Tweaks/DIY ] [ Music ] [ Films ]