Objectivist or Subjectivist? Give Me a Break

[jkeny]

I did look at that video, but I'm not willing to sit through more than an hour just to learn that you likely quoted him out of context, or find out how he's mistaken. If you tell me specifically what part to watch I'll gladly do so. YouTube makes it easy to skip around. I seem to recall asking you for the time many pages back and you ducked that.

I never remember any ducking on my part? Everything from about 20 mins on should be of interest but here are some relevant excerpts:
At 23:45 "It is very hard for us engineers who are used to looking at FFTs and seeing 150Db of SNR, it is actually very hard to believe, to accept that people can hear something that none of instrumentation an measure............ the one that's not so good is the sigma delta modulator despite having so much better technical specs"
At 26:00 "Dustin Forman, who is one of my best engineers, has trained himself to listen for what the audiophiles listen to."
At 27:00 "Even though you can show, Frank Sinatra's New York, New York, has a fairly horrible SNR, it still exposes errors in a sigma delta modulator that is a hundred times better than the original recording & it's because the human ear can detect signals well below the noise level present"

Ok, so I hear you say that this is all anecdotal & furthermore it's plain marketing. So skip to the measurements section from about 34:00 on. Just to help the slides are to be found here http://www.esstech.com/pdf/noise-shaping-sigma-delta.pdf

Said the guy who sells expensive converters.

Yep & a similar accusation could be levelled at you as you sell room treatment products! Your line of all DACs, amps sound the same could be considered to neatly dovetail into your focus on speakers & room treatment. So should this same cynicism be levelled at your statements?

Your link is broken but I figured it out anyway.

Sorry, here is the correct one, I hope http://arxiv.org/pdf/1208.4611.pdf

Regardless, that paper says no such thing. It compares "FFT uncertainty bounds," not what is possible to measure. You should run for president of the US.

--Ethan

Yes, maybe you don't understand what you are saying? The paper says that hearing can simultaneously evaluate frequency & phase to a more precise & deeper level than the mathematics allows & also beats any existing current measurement instrument, in this regard! Is this not what you understood from reading it? Take either one in isolation, frequency or phase, as you will no doubt want to do & measurements can beat hearing but only in isolation. We listen to sound envelopes & not to frequency or phase in isolation! I don't know what you don't understand about your quote "FFT uncertainty bounds" but if you say what it is that you are uncertain of maybe I or others can help?

 

[jkeny]

Myles, I assumed that Ethan was talking about ESS selling expensive DAC chips but I might be wrong. Even if this is the case, he is still trotting out un-researched & unfounded comments as their product range starts at about $2 up to $60 (for the hyperstream technology range of DACs).

If he is talking about my products, yes let's hear his comments.

 

[jkeny]

I did watch the video, and enjoyed it. What it told me was that the algorithm you use for a Sigma Delta DAC can be very poor, or very good - and early Sigma Delta DACs were simply not as good as they could be; it didn't need any more expensive hardware, merely for someone to sit down and think through how to optimise the algorithm for the best noise performance and stability. As I said in a previous post, it's all in the 'Intellectual Property' and once refined and perfected it can be churned out by the million at no extra cost. Like the man said, ESS are fabless and simply pick the "cheapest" manufacturer to produce the silicon. It is also apparent that ESS's algorithms are licensed to many other DAC manufacturers - or have been copied by them!

I don't think your last statement is correct. Who has copied or licensed the ESS algorithms? One other question - are you saying that only early sigma-delta DACs had a problem but this is sorted in the current examples?

Very interesting point about how typical frequency domain measurements didn't show the problem up clearly, and it made our discussion on time domain null testing seem quite relevant. However, in this case you wouldn't need to make actual physical measurements, merely analyse what the Sigma Delta algorithm was doing.

Oh? That's an interesting statement. Care to tell us how this "mere analysis" could be achieved?

A good video I thought.

 

[Soundminded]

What I don't understand about objectivists is that since their tests don't correlate with what people hear,why don't they develop or at least demand that others develop tests that do?

What I don't understand about subjectivists is that since they claim to be able to hear the difference between one amplifier and another, one wire and another, one vacuum tube and another,one power cord an another all of which must be very subtle, why can't they hear the difference between all recorded music played on any stereo system and all live music which is a gulf as wide as the Grand Canyon? Why do they keep supporting an industry whose products fall so far short of expectactions at prices so out of line you'd think nobody in their right mind would buy them? But not only do they keep buying them, they keep swapping, trading up, down sideways and always at a substantial loss. They always jump for the newest silver bullet, the next carrot on a string and it's always jerked out right in front of them just as they think they're about to grab onto it. You're a real loser Charlie Brown

Audiophiles of either stripe are a strange lot. I'm just glad I'm not one of them.

 

[treitz3]

Finally, a "on topic post". How refreshing.

Tom

 

[Johnny Vinyl]

Audiophiles of either stripe are a strange lot. I'm just glad I'm not one of them.

me neither. I'm just addicted to the music!

Finally, a "on topic post". How refreshing.

When was the prior one?

 

[Raffles]

I don't think your last statement is correct. Who has copied or licensed the ESS algorithms?

From memory he says that 30% of all DACs (or is it CD and DVD players?) contain the 'Hyperstream' technology licensed from ESS, but not publicised. And apart from that I'm just repeating the bit where he says that anything they ever put on the web (like that video) is copied.

Oh? That's an interesting statement. Care to tell us how this "mere analysis" could be achieved?

Simply analyse what the algorithm does. For example pass some real sampled music data through the algorithm and run tests on the output purely in software. No hardware needed. I would have expected anyone working on ADC or DAC algorithms to have done this, and been curious enough to have developed the algorithms to the Nth degree. Even if no one could have been expected to hear a difference, I would have expected them to have done this anyway. If it turns out that many modern DACs are not improved beyond the early days I withdraw my assertion that all DACs are all the same!

 

[NorthStar]

---Are we sometimes turning in circle? :b

 

[jkeny]

.......
Simply analyse what the algorithm does. For example pass some real sampled music data through the algorithm and run tests on the output purely in software. No hardware needed. I would have expected anyone working on ADC or DAC algorithms to have done this, and been curious enough to have developed the algorithms to the Nth degree. Even if no one could have been expected to hear a difference, I would have expected them to have done this anyway. If it turns out that many modern DACs are not improved beyond the early days I withdraw my assertion that all DACs are all the same!

And are you saying that this mathematical analysis would show the noise modulation & phase modulation revealed in this video? If so why wasn't it addressed before now? Or are you saying that using steady state signals as input to the analysis is why nobody had picked up on theses issues, to date?

 

[Raffles]

And are you saying that this mathematical analysis would show the noise modulation & phase modulation revealed in this video?

Yes.

If so why wasn't it addressed before now? Or are you saying that using steady state signals as input to the analysis is why nobody had picked up on theses issues, to date?

I have a sense that in electronic engineering there are sometimes gaps in between people's specialities. Chip designers have probably been educated in pure digital electronics; they can understand and analyse a sigma delta building block and know how to implement it in a chip, but have little knowledge of how 'heuristics' and adaptive algorithms can make it even better. I have experienced this sort of thing myself where, in developing a device inextricably combining hardware and software, merely explaining how it works to the specialists who are then going to productionise it is very difficult. The chap at ESS understands it, though, and has had the satisfaction of finding that he could improve an existing system using a bit of ('empirical'? 'informal'?) DSP, and that it could then be sold to a lot of people for real money. Respect!

 

[jkeny]

Yes.

I have a sense that in electronic engineering there are sometimes gaps in between people's specialities. Chip designers have probably been educated in pure digital electronics; they can understand and analyse a sigma delta building block and know how to implement it in a chip, but have little knowledge of how 'heuristics' and adaptive algorithms can make it even better. I have experienced this sort of thing myself where, in developing a device inextricably combining hardware and software, merely explaining how it works to the specialists who are then going to productionise it is very difficult. The chap at ESS understands it, though, and has had the satisfaction of finding that he could improve an existing system using a bit of ('empirical'? 'informal'?) DSP, and that it could then be sold to a lot of people for real money. Respect!

I'm interested in whether you are speaking from experience or as a commentator on how you "think" this should work?
In the thirty or so years that sigma-delta DACs have been in existence, nobody until now has run this analysis & picked up the noise & phase modulation issues that you say are easily identifiable? Even today, how many people agree that S-D DACs have these modulation issues? Is this not gross incompetence, if it so easily identified according to your statement by "merely analysing what the S-D algorithm is doing"?

 

[microstrip]

What I don't understand about objectivists is that since their tests don't correlate with what people hear,why don't they develop or at least demand that others develop tests that do?

You will notice that unhappily objectivists will focus on your the next sentences, and forget this one. Much easier to do ...

What I don't understand about subjectivists is that since they claim to be able to hear the difference between one amplifier and another, one wire and another, one vacuum tube and another,one power cord an another all of which must be very subtle, why can't they hear the difference between all recorded music played on any stereo system and all live music which is a gulf as wide as the Grand Canyon?

You are distorting what most subjectivists say just to write soundbyes. No subjectivists worth this designation will tell you that they can not hear or there is no difference between all recorded music played on any stereo system and all live music. They will tell you that under appropriates circumstances, optimized systems can reproduce enough of the reality to create an illusion that it will approach the experience felt in real conditions systematically, not just for an occasional moment. They know not that 100% is not possible, but that the subjective approach is also needed to know what can be sacrificed in the process of sound reproduction, and make it more rewarding.

Why do they keep supporting an industry whose products fall so far short of expectactions at prices so out of line you'd think nobody in their right mind would buy them? But not only do they keep buying them, they keep swapping, trading up, down sideways and always at a substantial loss. They always jump for the newest silver bullet, the next carrot on a string and it's always jerked out right in front of them just as they think they're about to grab onto it. You're a real loser Charlie Brown

In every category there are winners and loosers. Some people focus on the loosers to justify their choices. I prefer to see the enthusiasm and dedication of true audiophiles, some at WBF, and do my best to join them. Because of this hobby I made a lot fo friends, some of them are in the high-end industry. I enjoy the music and the systems themselves. BTW, you can forget about the carrots, silver bullets and Charlie Browns - they will bring you nowhere in this search about objectivism and subjectivism.

Audiophiles of either stripe are a strange lot. I'm just glad I'm not one of them.

Perhaps you are and you do not know. Just think that an objectivist is a guy who knows enough of the subjective matters do analyze them and make a synthesis of them in objective statements, aware of the limitations of his studies. I can not pretend to have knowledge enough about audio to become an objectivist - but enjoy reading those I consider the real good ones and talking about it.

 

[Raffles]

I'm interested in whether you are speaking from experience or as a commentator on how you "think" this should work?
In the thirty or so years that sigma-delta DACs have been in existence, nobody until now has run this analysis & picked up the noise & phase modulation issues that you say are easily identifiable? Even today, how many people agree that S-D DACs have these modulation issues? Is this not gross incompetence, if it so easily identified according to your statement by "merely analysing what the S-D algorithm is doing"?

I have no direct experience of the interior design of sigma delta DACs, but I do often work with DSP and not un-related stuff; I frequently do a PC-based simulation of an algorithm before doing the real time version. In the case of a sigma delta-style DAC algorithm, I think it would be possible to take some real music, feed it into the algorithm and do a null-type subtraction of its output from the input (in software) and have a very close look at it. I guess you could also boost it up and listen to it. I'm guessing that something that sounded like pure white noise would be preferable to something less random, modulated by the music. At the same time, if the peak noise was 100dB down from the peak output, it might reasonably be supposed that no one could possibly hear it. But the man from ESS says that some people can hear it. I'm not entirely convinced, but if it was easy to do, I'd do it anyway.

Maybe this problem (if that's what it is) has been known for a long time, and ESS are not the only people with a solution to it..? Is this patent from Microchip aiming to do something similar, for example?

http://worldwide.espacenet.com/publ...P&FT=D&date=20110216&CC=TW&NR=201106636A&KC=A

 

[cnpope]

What I don't understand about objectivists is that since their tests don't correlate with what people hear,why don't they develop or at least demand that others develop tests that do?

I realise I'm a bit late to the party in this particular thread, but I chanced upon this remark while browsing today, and couldn't resist responding. I think I would be considered an objectivist, and my answer would be the following:

First, one needs to establish if people are actually hearing the differences that they say they hear. The human ear/brain combination is notorious for hearing what it "expects" to hear, and so one has to disentangle this from any "real" differences before there is anything substantial to discuss. There is really only one way to do this, and it is accepted in almost all other areas of scientific enquiry (I use "scientific" in a rather broad sense, as a quest, through logical investigation, for knowledge and understanding), namely by means of double-blind testing.

Having established whether an audible difference can actually be substantiated by such listening tests, one can then discuss the question of whether these correlate with measurements. I am not aware of any cases where they do not. In fact there are numerous documented tests where it has been pretty convincingly demonstrated that listeners are unable to demonstrate any ability to distinguish between any two audio amplifiers playing from the same music source provided that the sound levels are closely matched, the frequency response is matched by means of an equaliser, and the distortion is below a couple of percent. In other words, the precision with which measurements can be made vastly exceeds the discrimination of the human ear.

Chris

 

[Phelonious Ponk]

I realise I'm a bit late to the party in this particular thread, but I chanced upon this remark while browsing today, and couldn't resist responding. I think I would be considered an objectivist, and my answer would be the following:

First, one needs to establish if people are actually hearing the differences that they say they hear. The human ear/brain combination is notorious for hearing what it "expects" to hear, and so one has to disentangle this from any "real" differences before there is anything substantial to discuss. There is really only one way to do this, and it is accepted in almost all other areas of scientific enquiry (I use "scientific" in a rather broad sense, as a quest, through logical investigation, for knowledge and understanding), namely by means of double-blind testing.

Having established whether an audible difference can actually be substantiated by such listening tests, one can then discuss the question of whether these correlate with measurements. I am not aware of any cases where they do not. In fact there are numerous documented tests where it has been pretty convincingly demonstrated that listeners are unable to demonstrate any ability to distinguish between any two audio amplifiers playing from the same music source provided that the sound levels are closely matched, the frequency response is matched by means of an equaliser, and the distortion is below a couple of percent. In other words, the precision with which measurements can be made vastly exceeds the discrimination of the human ear.

Chris

Welcome to the forum. Really. Welcome. Thank god you're here.

Tim

 

[microstrip]

I realise I'm a bit late to the party in this particular thread, but I chanced upon this remark while browsing today, and couldn't resist responding. I think I would be considered an objectivist, and my answer would be the following:

First, one needs to establish if people are actually hearing the differences that they say they hear. The human ear/brain combination is notorious for hearing what it "expects" to hear, and so one has to disentangle this from any "real" differences before there is anything substantial to discuss. There is really only one way to do this, and it is accepted in almost all other areas of scientific enquiry (I use "scientific" in a rather broad sense, as a quest, through logical investigation, for knowledge and understanding), namely by means of double-blind testing.

Having established whether an audible difference can actually be substantiated by such listening tests, one can then discuss the question of whether these correlate with measurements. I am not aware of any cases where they do not. In fact there are numerous documented tests where it has been pretty convincingly demonstrated that listeners are unable to demonstrate any ability to distinguish between any two audio amplifiers playing from the same music source provided that the sound levels are closely matched, the frequency response is matched by means of an equaliser, and the distortion is below a couple of percent. In other words, the precision with which measurements can be made vastly exceeds the discrimination of the human ear.

Chris

Chris,

Just to know what tests you are addressing and what were their conditions, could you give direct references to them?

 

[andromedaaudio]

Welcome to the forum. Really. Welcome. Thank god you're here.

Tim

 

[j_j]

"Subjectivist vs. objectivist" - a completely false dilemma.

Some so-called "objectivists" (like me) base their results on SUBJECTIVE results.

Given the very, very fragile nature of auditory (and visual, etc) perception, having a falsifiable test is required, of course. I am well aware of the nonsense that others have spread about DBT's, mostly motivated by the fact that they didn't like this or that DBT, and sometimes motivated by a complete denial of proven human perceptual behavior, but, frankly, a DBT or an SBT with carefully designed computer administration are a first requirement. That's not the only one, since obviously, like anything else in the universe, somebody can design a bad DBT as well, and I'm rather certain some people have.

 

[Robh3606]

Ding, Ding, Ding

And her we have in this corner for 12 rounds

Some so-called "objectivists" (like me) base their results on SUBJECTIVE results.

Hello JJ

So you use measurements as a tool to help get you there?

Rob

 

[j_j]

Ding, Ding, Ding

And her we have in this corner for 12 rounds



Hello JJ

So you use measurements as a tool to help get you there?

Rob

Well, yes. What was your point? A subjective test is a measurement, albeit a complex and somewhat tricky one.

But I also use standard kinds of measurement, just not anything as useless as THD or SNR.