[B] NEED HELP - Diganosing audio differences between source components [/B] - Page 2

No, not CD errors. But yes, it definitely is possible to have audible differences even though the digital streams are error free. Here is the explanation: http://www.madronadigital.com/Librar...dioJitter.html

It is actually three things:

1. Amount of jitter in the source.

2. Amount of jitter induced by the interconnect.

3. Amount of jitter rejection and jitter addition in the receiver.

They all add up to eventually cause variations in the DAC output. Characterizing where it is happening is very hard without the right measurement gear. But you can narrow it down some by the testing I mentioned. For example, use the same player with HDMI and Toslink and see which sounds better than the other. Do that blind by closing your eye and randomizing which input is which. And then see if you can repeatedly identify which one you think sounds better.

This came up in another thread:
http://www.avsforum.com/avs-vb/showthread.php?t=1409443 (which I just realized you also started and have probably read, ).

Price doesn't dictate anything here, and as was explained there, the reason for those DACs and ADCs is specific to the needs of production; it is not even in the same sport, let alone ballpark. Arny isn't out of his head in making that claim.

Amir is correct in talking about measurable differences between the two (I will say right here though: you're making a quantum leap with "assuming the proper ABX showed differences to exist" statement though; it basically takes us out of reality and into a hypothetical discussion), but the question then comes back to: are these differences relevant to listeners?

I'm not denying that yes, you can measure more or less jitter (or THD, or whatever else) and that his charts and graphs are usually great at showing that. But where I'm at least skeptical is whether or not this actually would manifest in a listening test. There's an AES workshop that Ethan Winer gave a few years ago, and he asked a question along the lines of "how do you know jitter influences the sound - do you have some box that you can dial jitter up and down on, and compare the results?" And that's basically where I'm hung on the issue too - lots of data supports systems overall (so putting EVERYTHING in a black box) being more or less transparent to one another. That makes me want to say the jitter issue is a non-issue unless it gets to the point of signal drop-out.

However, I'm aware of no test that basically sets up a scenario with "very little jitter" and "lots of jitter" and then does a proper ABX or ABC/HR to determine what's what. So I feel that it's sort of like THD - sure we can measure it, and we can show that A has .00000001% and B has .0000000000000000000000000001%, but does this make B better in a way that us humans can appreciate?

Gave Amir's link a quick look, and there's a line near the end:


While this is a fairly logical premise (in other words, if you have the choice between unit A and unit B above, and they cost the same, buy B, who cares if it's inaudible on both), I'm still somewhat skeptical if it involves spending substantially more money, mostly because these kinds of measurements aren't available for lots and lots of equipment.

Checking for matched levels is going to be like 1% of the effort that would need to be expended to make the OP's evaluation into a proper test. It's not a chase, its a step in an organized approach to doing good subjective tests.

I know I'm dealing here with people who have no clue about actually doing Science hands-on. A lot of proper scientific experimentation is about jumping through necessary hoops. If you don't have the patience to approach your audio experiments in a logical and systematic way, I'd say stay in retailing!



Right, by Walbert. So why didn't you chew him out for mentioning level matching? Why aren't you chewing me out for repeating the other three?

Isn't it true Amir that you have a weak dog in this fight - one composed of your past, repeated, unretracted statements about you repeatedly doing listening tests without level matching? Isn't level matching a sore point with you?


There was a good reason. On the face of it, there appears to be no obvious technical explanation for the perceived differences. When faced with a mystery, the best procedure is to start from scratch and make no presumptions.


I didn't know that I had to give a reason for any of the 4 points. So are you now saying that I need to give my 487th, (more or less) recitation of how to do proper listening tests? GMAB!

I posted what I did to back up Walbert and give a hoepfully clearer (numbered list versus paragraph of prose) recitation of the very same points.


LOL!


Level matching is easy for many of us to do (but apparently not you, Amir), and there may be a hidden gotcha.


I did provide enough of an explanation Amir, but you want to draw this out and go on and on about level matching, apparently because of your past denials of its necessity. Let's face it Amir, you are one of AVS's strongest advocates of badly done listening tests.


No, Amir this is all about you trying to lecture me about how to do listening tests, which is really pretty funny not to mention highly ironic.


No, Amir this about you going on and on about a simple point. This is about you trying to find a technical explanation for what is probably the consequences of a total lack of relevant experimental controls.

Heck, the guy doesn't even have two CDs to play! He plays the disc in one player, he takes it out, flicks at least one switch, loads the other player, waits for it to become ready, indexes the part of the track he's using to do the comparison and listening. Several minutes may have passed. Then he listens. Does he even listen to the identical same musical passage?

Of course he hears a difference!!


Amir, I don't think you even know where the objectivity camp pitch their tents, and I've never seen any evidence that you actually have a tent that you can sleep in whenever you want to.


That's just it Amir. When you are doing listening tests you leave as little to chance as you can. Just another example of how you don't get it.

It is not possible. Unfortunately, there will be industry leaders who will freeze their understandings at or about the time they received their degrees, or within a decade of it (I've assumed a degree here, but have no knowledge of that fact). After that, everything becomes a nail to which their only hammer is applied to. Those types will stall many technological advances as long as they are involved, and it takes their retirement to get the ball rolling again.

I see this in my chosen field. Giants who have committed their understandings to paper....and those writings followed without question..even though it may be old technology, or even incorrect understandings.

AK has no understanding of DSP. So ignores it as well as advances in that field. Given this, you are correct..



A declaration of utter shock. It is unfathomable that you have no knowledge of DSP, finite impulse filters, infinite impulse filters, recursion filters, reconstruction math..

I've read many posts by you, and have always had a rather high opinion of your knowledge. Your recent clarification has tempered one aspect of my opinion of you.

On certain topics, perhaps it's best you get out of the way??

As in, lead..follow, or get out of the way.

j

Possible, yes. But highly unlikely.

Jitter as mentioned is more likely to be a source of measurable differences. Still unlikely to be a source of audible differences, but just how unlikely is an unsettled question (to some) and the debate continues ad nauseum.

As for professional dacs, don't discount the possibility that professional musicians and recording engineers may be just as susceptible to the marketing and perceptual biases that also cause mere consumers to spend more money for the "name" or paper specs or pretty case or whatever. There are certainly cases of overpriced overhyped studio jewelry just as is the case for home audio.

That being said, many pro dacs costing thousands of dollars are multichannel devices offering a variety of input types, conversion options, master clock sync, etc and are technically well built devices that on closer inspection are a true value compared to an equally priced home audio two channel dac. Try to build a competent 8 or 16 channel dac yourself from parts and see what the damage is.

JN, the above is false. You seem to be in some weird state of self-induced semi-consciousness that I may never fathom. TMI.

What you don't seem to get is that in audio, if it doesn't make an audible difference or can't be traced to an audible difference, it doesn't matter.

In an consumer-oriented online forum context (e.g. AVS), detailed knowledge of DSP, finite impulse filters, infinite impulse filters, recursion filters, reconstruction math is TMI for just about everybody. People here want to know about what makes a difference, in terms of sound that they can hear. If you disagree, show where I am wrong.

In a way, saying that my posts show no evidence of what I know about those things or even that I know anything at all about those things, is music to my ears.

So thanks for the kind thoughts! ;-)

Sorry guy, its a rabbit hole.


True and surprisingly old news. Component interfacing used to be a big discussion all over people's audio systems. Today, its mostly only audible in the interface between amps and speakers.

Back in the day, the very idea that digital component interfacing and cables could introduce potentially audible difficulties was a novel idea, particularly for people who were living in analog land up until just lately.

Their lives up until that point had been 100% analog, and digital was falsely claimed to be "Perfect Sound forever". Digital never was perfect, it never will be perfect, but it can be a great improvement.

When all the dust settled, people upped their game a certain amount and component interfacing problems related to digital interfaces became far less of a real world problem.

I still remember the first time I saw a partially trashed display that used a HDMI interface. HDMI is about as modern and digital as it gets. In principle, HDMI should either work or not work. A noisy screen should not be possible. But it was....

Now getting back to your music player comparison. God only knows all of the signal processing that happens inside your new Blu Ray player until us mere mortals get ahold of the relevant documentation and understand it. I think we know that it at the least resamples lots of media for DRM reasons. It might have its own digital volume control.

Now I could either buy the service manual and study it for a day or a week, try to steam some info out of the vendor; or I could instruct you to do some simple checks with a cheap meter to see if the levels are matched... ;-)

Well, it could be errors on either side. Could be jitter—much more likely on the digital connection than within the CD player. Could be something wrong with the analog section of the CD player. Could be a lot of things.

But it almost never is.

No one's got a shred of evidence that the magnitude of differences we commonly find between consumer (or professional!) digital products is audible—no listening tests, no measurements, nothing. It's all baseless opinion, mostly from people with either money or ego on the line. Often both.

In fact, the most logical conclusion to your hypothetical test above is that if you were able to hear a difference, you probably screwed up the test!

I may indeed be... However, that doesn't change the fact that in response to a discussion of the advances that have been made in D to A conversion over the decades, specifically the output algorithms and the increasing number of coefficients in an IIR reconstruction scheme, you blast out with "Audio isn't about math, its about sound quality". I'm sure that I am not the only one who looked at that and laughed..


While you may wish others believe such balderdash, it is another of your strawmen.

Learn about the advances DSP has made over the last two decades. Specifically, the horsepower advances which allow realtime increases in the DSP coefficient number (quantity, or depth). Either you understand what I'm talking about, or it will be hopeless to get you to learn.


If you are referring to my statement:

Then you are welcome. I stand by that statement. It is an opinion I have formed on my own from casual reading of your posts. And it exists despite any statements of others I have read to the contrary. I call it as I see it.

My opinion is independent of whether we agree on a topic or not. Always has been.

Cheers, j

Often it's the altered position of the listener's ears. In typical home setup (not so well treated), the audible difference can be caused by just an inch change in ear position even if no component was changed. Try it yourself, while playing music or tone, tilt, twist or sway your head little bit. You will be able to tell.

Thanks JN for reiterating that you think that audio is about math and not about sound quality.

Also, please explain how output algorithms and increasing number of coefficients in IIR reconstruction schemes that were discovered after 1986 (the date of the article I cited) enabled previously impossible levels of sonic accuracy, IOW whose uniqueness, necessity and effectiveness can be readily demonstrated with a DBT.

I provided measurements in the article. Did you not read it? Here they are again:http://www.madronadigital.com/Librar...dioJitter.html

"Here is an example measurement for the Onkyo TX-NR5007 AV Receiver:

S/PDIF: 0.79 ns
HDMI: 4.87 ns

As you see, the jitter over HDMI is not only more than six times higher than S/PDIF, it is also way above the maximum threshold for 16 bits of fidelity. No wonder then that Paul gives the product a failing grade on that interface. Here are the measurements on another AVR, the Yamaha RX-V3900, so that you don't think the above is the exception:

SPDIF: 0.183 ns
HDMI: 7.7 ns

Here we have excellent response on S/PDIF but HDMI is a whopping 41 times worse! There is not one measurement on Paul Miller's site that has better measurements for HDMI vs. S/PDIF. The common ratio is 10:1 in favor of S/PDIF."

We can translate these numbers to equiv. audibility/perceptual levels and show how they compare to threshold of hearing and thereby be audible. No it is not a listening test per-se but the comparison is based on such.


I don't consider Audio Engineering Society standards which cover this very topic to be about either. I suggest reading AES11 and AES9 which are recommendations for digital audio interfacing in professional space. Both provide strict limits for jitter. Great analysis of the same is in the joint paper between Professor Hawksford and Julian Dunn: http://www.scalatech.co.uk/papers/aes93.pdf.

"ABSTRACT

It is a requirement of high quality digital audio systems that all digital interfaces in the signal path exhibit signal transparency. The widelly adopted AESEBU/SPDIF interface has received criticism from some quarters for a lack of signal transparency; this paper addresses possible problems with such interfaces and presents methods for improving the interface standard."


When there are measurable differences of as much as 40 dB as I showed in the article, what you say is premature conclusion based on one's bias and not the facts. We need to investigate to see if we can rule out the difference that we can measure. I provided some specific recommendations to do that on making the test easier and more reliable to do.

All things are in some sense possible.

For example very short unrecovered disc tracking errors are often not heard as interruptions to the music, but rather as a slight loss of loudness or impact.

Of course, these losses are measurable, even with a cheap averaging-type voltmeter.

This kind of problem is more probable with older optical players.

Which should be meaningless to most audiophiles because:

(a) They are just numbers


(b) 16 bits of fidelity is far more than is necessary for most people to experience listening pleasure.

Proof: Look at all the years that people enjoyed LPs and consumer analog tape and never had more than the equivalent of 12 bit fidelity.

Another strawman...please read with more accuracy.


I am not sure how rudimentary I have to get for you..

First, the textbook I mentioned where I first learned DSP was 1977 vintage. At that time, the IBM 360 I worked with was incapable of real time processing of any digital audio stream.

Then a PDP and a Data General Eclipse in the early 80's...still insufficient horsepower.

In 1986, there were no DSP chips capable of real time emulation of an 8 coefficient 24 bit or even 16 bit digital (edit: aha! I forgot to say reconstruction here, no wonder arny posted a brick later..my bad) filter for consumer product..

Which leaves one with an output sample and hold followed by the filter from "heck".

It's changed drastically, dude...far fewer tradeoffs are necessary at the brick wall. And, zero crossings across the frequency spectrum are closer to the origonal source... Guess I hafta explain what zero crossings have to do with how we hear, right??

cheers, j

ptsawyer, amirm is just repeating his marketing in disguise. The key phrase is "marketing in disguise". Why do I say that?
Read this first: http://www.avsforum.com/avs-vb/showp...&postcount=267
Then read this thread (what he is trying to do here has been exposed already): http://www.avsforum.com/avs-vb/showt...php?p=20255022

The math can determine audibility. If I have a 5 bit system, the math says we hve ~30 dB of signal to noise ratio. This math indeed tells us the system has audible distortion.


You can start by reading this AES paper:

"Sound Quality Evaluation of 96kHz Sampling Digital Audio
Authors: Yoshikawa, Shokichiro; Noge, Satoru; Ohsu, Masatoshi; Toyama, Soichi; Yanagawa, Hirofumi; Yamamoto, Takeo
Affiliations: Kanagawa Institute of Technology, Atsugi, Kanagawa, Japan ; Pioneer Electronic Corporation, Tokorozawa, Saitama, Japan(See document for exact affiliation information.)
AES Convention:99 (October 1995)

This paper refers to the experiments the authors conducted to verify whether
it is possible to discriminate the difference of sound quality of 96kHz sampled signals
from that of 48kHz sampled counterpart utilizing a DAT(digital audio tape recorder)
as a playback tool. The subjects were 11 adult male with normal hearing (age 21 to
24). As the test materials, a "fusion" type music and a popular music were used.
Experimental results employing the RXY method suggest that the subjects can be
classified into 3 groups; significant group (3 subjects, 27.3%), likely to be significant
group (2 subjects, 18.2%) and no significant group (6 subjects, 54.5%) in the case
of the fusion type music. They also suggest that the subjects were divided into 2
groups; significant group (2 subjects, 18.2%) and no significant group (9 subjects,
81.8%) in the case of the popular music. The results turned out to be that there exists
significant group (2 subjects, 18.2%) which is able to discriminate the difference of
sound quality of both test materials, although the different classification of the
subjects were obtained depending on the test materials. The results confirm the fact
that 96kHz sampling will be effective for the high quality digital audio in the future."

If I were you, I would hang my hat on filtering differences resulting in audibility of 48 Khz vs 96 Khz sampling and not them hearing ultrasonics.

Most? I take what is left.


So is one's salary .


I enjoyed music on cassette tapes too. That doesn't mean I want on purpose to use a digital system whose distortion sinks that low when I can use proper knowledge of digital audio to avoid it. Why put one's head in the sand based on what people did a few decades back?

The above is irrelevant because it is about 1977 at the latest, and not 1986 or later.


The restriction of consumer products is gratuitous, and therefore excluded even though it makes no difference.

There were such things as digital filters that were implemented without DSPs - they were impemented with ordinary combinatorial logic.

For example the Philips SAA7220 has been around for decades (first production use 1985) and implements a 120 coefficient digital filter according to http://html.alldatasheet.com/html-pd...7/SAA7220.html

Page 7.

This was preceeded by the SAA 7030 (96 coefficients) in the original Philips CD63 (1982). The SSA 703 came before that, but I don't have a lot of details about it. They all supported data words that were at least 16 bits long.

Bottom line JN is that your claim that proper digital filtering can't be implemented in mass-market consumer gear without higher-performance DSPs appears to be a product of your lack of knowledge of how digital audio actually developed.

Amir you should read your own posts about the relationship between bits and dynamic range and put 2 and 2 together!

Cassette tape is more like 10 bits or less.


Right and as I pointed out in my little history lesson for JN, we had highly competent digital filter chips in consumer digital music players as early as 1982. This is prior to my 1986 cut-ff date for sonically signfiicant improvements to digital audio. BTW there is nothing wrong with analog filters in digital audio, except for their cost. If you can write a proper transfer function for a filter, you can implement it in the analog domain. Just costs a lot!


Indeed, Amir. Why do you do that? ;-)

I don't understand why they should sound identical. I would question the "depth and detail" but I wouldn't be surprised that they sound different. I also find the level-matched ABX recommendations not realistic. Come on now. Just listen whichever sounds better to you. Yes they can sound different as others mentioned mostly likely due to internal sound level differences or quality of recordings in different CDs.

You didn't even test identical CDs. There is no reason why they should sound identical.

It is amazing you keep bringing jitter back in these discussions while ignoring the elephant in the room. Did you miss the part where he said CDs he tested were not identical?

Measurements alone are not evidence of audibility. Measurements correlating to previous, high-quality listening tests could be. But you didn't do that. You explicitly ducked the question entirely.

No because he didn't say that. My assumption and I think that of everyone else posting but you is that he was shuffling the same CD between the two transports. This is what he said that led us to that: "I havent made a duplicate CD yet so I can quickly switch between the two.." Otherwise he would note that he would need to duplicate them to hear the same music.

He is here and he can clarify.

Then why did you bring it up by saying: "No one's got a shred of evidence that the magnitude of differences we commonly find between consumer (or professional!) digital products is audibleno listening tests, no measurements, nothing."


The purpose of the article was the address OP's original question and common myth: that differences cannot exist in digital transports. Once we agree that science and measurements do show differences in digital transports, then we can get to the next chapter. Seeing how you said all of this remains an opinion with no science or measurements, we are still not there