Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Jitter, AIR and USB
#1
(26-Apr-2017, 20:02)yabaVR Wrote:
(26-Apr-2017, 17:37)thumb5 Wrote:
(26-Apr-2017, 17:08)yabaVR Wrote: My understanding of AIR is that everything that reaches the AIR-Driver (the front door of AIR) is afterwards sorted and send jitterfree/timecorrect through every signalpath of the computer and the network until it reaches the front door of the Dxxx. Then it is converted back into a stream without a time correcting protocol and therefor is prone to jitter again.
The problem as I see it is the signal path from the HDD/SSD up to the front door of AIR. I don't think there is a jitterfree time dependent protocol that holds jitter off this signalpath. But once this jitter contaminated signal passes the front door of AIR the AIR-software takes it as an original signal. The induced jitter can not be seen by AIR.

I don't see why or how you draw a distinction between what happens "in front of" or "within" AIR: it's all computer hardware and software, it all works in the digital domain, and it is all either bit perfect or not.  As Confused said rather concisely, the audio samples can arrive at the Devialet more or less any time relative to each other so long as they are in the correct order and not missing, both of which errors can be detected, and the Devialet will re-clock them synchronously to it's own internal clock.  So no inter-sample time variation before the re-clocking should make a difference to the output from the amplifier.

So I misinterpreted one line of your post...sorry  Angel .

If I sum up your comments is it right that you say there is no such thing as jitter in the audio chain from a computer to a Devialet or DAC espacially if we use AIR (over ETH)? Right?

No need to apologise!

Being pedantic, in order to answer your question I need you to define what you mean by jitter.  But it is true to say that I don't believe that the timing of what the computer sends via AIR or USB can affect the Devialet's audio output.  I'll explain why I believe that a little later in this post.  (I am assuming that all the data actually arrives, of course, otherwise you would hear drop-outs.)

(26-Apr-2017, 20:02)yabaVR Wrote: I get the feeling you want to get me wrong and did not read my post depending on AIR. I didn't say it is AIR who is the culprit here when we're talking about jitter. It's (mostly) the rest surrounding AIR in the digital chain that is causing jitter.

Not at all, I did read your post and think I understood it, but disagreed with it and wanted to challenge you a bit on why you said what you did!

(26-Apr-2017, 20:02)yabaVR Wrote: One question, if you open any program does it always need the same time to start? If you open a jpeg from your HDD does it always take the same amount of time? Depends on what your OS is doing at the moment doesn't it?

Those are rhetorical questions of course, but yes, of course I agree that the time will vary.

(26-Apr-2017, 20:02)yabaVR Wrote: There might be a delay. This is crucial for a stream.

Here we get to the meat of it.  The data delivered by the computer to the Devialet via AIR or USB is /not/ a stream in the sense of being a series of individual samples each delivered independently at specific times.  Rather the samples are delivered in packets asynchronously to the sample rate, which is only implicit (agreed in advance).  The Devialet re-clocks the samples out of a buffer at the sample rate driven by its internal clock.  This process is not affected by the kind of delays you're talking about above.

(26-Apr-2017, 20:02)yabaVR Wrote: And if it was not why is it that tuning software as 'Audiophile Optimizer' 'jPlay' 'Process Lasso' 'Fidelizer' and the like lead to fantastic better sound when they are installed?

My view is that in the case of streaming via AIR or USB they can make no objective difference to the audio output.  In other words, any changes you hear are due to expectation bias or some other psycho-acoustic effect (which is absolutely fine, so long as you don't then try to retrofit a technical justification).  If you have some concrete evidence of objective changes then I'm all ears Smile

(26-Apr-2017, 20:02)yabaVR Wrote: I'm no software engineer nor a chip designer I'm only thinking out loud because there's to be a reason why we are hearing differences within firmwares (ah, got the topic here  Big Grin ), computers and tuning softwares. Many here on the forum are speaking about different AIR versions are sounding very different. Why is that when AIR does everything right and is bitperfect by design?

Good question.  Apart from people naturally "wanting" to hear differences, there may be changes to the signal processing done by different firmware versions.  That is independent of the fact that AIR delivers bit-perfect samples into the signal processing system.

(26-Apr-2017, 20:02)yabaVR Wrote: If you say that jitter doesn't exist in a digital computer audio chain why is it that all the different computers using the same AIR have a different sound? Or are you saying this is not the case? Then we can stop the discussion here  Sleepy .

You seem to be assuming that jitter is the only possible source of sound differences.  I don't agree: noise can be transmitted from the computer through wired Ethernet and USB connections to the Devialet, for example.  I certainly don't know that all computers using the same version of AIR have a different sound, as perceived by the same listener - and I am not sure whether anyone has done an experiment to answer that question (I might have forgotten if it has been posted here).


OK, I feel the onus is on me now to explain why I've come to these conclusions.

Let's assume we have a computer sending a digital audio stream to a DAC.   For the sake of argument, let's assume that it's CD-quality stereo: in that case the audio stream consists of two, 16-bit samples at a rate of 44.1 kHz (roughly 22.7 microseconds per sample).  The same logic applies for different bit depths and sample rates, of course, just with different numbers.

One way to transmit this stream would be for the computer to run software which very accurately measured time and sent each stereo sample at a very regular interval of about 23 microseconds (22.6757... microseconds to be more precise).  In this case the DAC might be directly connected to an output port on the computer such that each time the software sent a sample, the DAC would change its analog output exactly at that moment.

In reality this would only be workable for a dedicated computer running only the audio streaming software (and probably not running a traditional desktop operating system).  As soon as the computer had to do any other work, the audio streaming software could not guarantee to send the samples at intervals of exactly 22.6757... microseconds, and the output of the DAC would not change at the expected times.  The variation in time between transmissions of successive samples (and the corresponding variation in the time at which the DAC output changed) is what I understand as "jitter".

Now let's talk about AIR (wired or wireless) and USB audio streaming.  Both of those are packet-based protocols, meaning that they will deliver not one sample at a time but bunches of several samples at the same time.  So at time t a packet containing N samples arrives, then at time t + T another packet arrives, this time containing M samples.  The significant things here are:

* the number of samples per packet (N, M) is not always the same for each packet

* the time T between packets is not always the same for each pair of consecutive packets

* the time T between packets is not directly related to the audio sample rate (44.1 kHz in our example)

For USB, the time between packets is fixed by the USB host clock and is nominally 125 microseconds.  This is independent of the audio sample rate.  For CD-quality audio, each packet normally contains five or six samples.  On the other hand, when I decoded AIR Ethernet packets some time ago, I found that for CD-quality audio each packet contained 200 samples and the time between packets was normally about 4 ms.

The point in each case is that the computer is not delivering a steady stream of samples, and the DAC can't rely on any timing based on arrival of samples from the computer.  The original sample rate (44.1 kHz) is simply not reflected in the arrival times: it is only implicit -- more precisely, it's agreed in advance between the computer and the DAC that they're both working at the same, known sample rate independently of when and how the samples actually appear "on the wire".

Generally this means the DAC has to have some kind of buffer memory into which the incoming samples are dumped whenever they arrive.  It must then pull samples out of this buffer at the right interval (22.6757... microseconds, defined by the agreed sample rate) which it must determine on its own - that is, by means of an internal clock.

With this method there should be no possibility for timing variations in arrival of the samples (generated by the computer) to have any effect on the output of the DAC.  Unless the computer just doesn't keep up, in which case the DAC's memory buffer empties and we'll get an audio drop-out.  Or unless the DAC is badly designed, I suppose.

The way to think of this is that the DAC's buffer memory is like a bucket of water.  The bucket has a hole in the bottom where the water leaks out at a fixed rate - this corresponds to the DAC pulling samples out and converting them to analog.  When the computer sends a packet of samples to the DAC, it's like pouring a glass of water into the bucket.  The point is that the rate at which water leaks out of the bucket is not affected by when water is poured into the bucket, or indeed how much water is already in the bucket.  All that matters is that over a period of time, the rate at which glasses of water are poured into the bucket more or less matches the rate at which water leaks out at the bottom of the bucket, to make sure the bucket doesn't brim over or run dry.

As Confused pointed out, this buffer is quite obvious when you're streaming by AIR over Ethernet: if you pull the Ethernet cable out, music continues to play for a noticeable time (determined by the target buffer size setting).  The buffer for USB is not likely to be so large because the the rate at which packets arrive is more predictable, as it's defined by the USB audio class specification.

Put back into the context of the original discussion: my view is that it is meaningless to use the term "jitter" to refer to anything that happens in the path from the digital source (e.g. CD rip, Qobuz) via AIR or USB before the point where the Devialet clocks samples out of its buffer, because before that point there is no particular, fixed "expected" time between consecutive samples.  For samples in the same packet, the inter-sample time is effectively zero, whereas for samples in different packets it could be many times the nominal period (based on the audio sample rate).  The variations in actual sample arrival time are all smoothed out by the Devialet's buffer, and any jitter depends entirely on the Devialet's internal clock which is not affected by the computer.
Roon (Mac Mini), Wilson Benesch Full Circle, Expert 1000 Pro CI, Kaiser Chiara
Warwickshire, UK
Reply
#2
A word to the wise; never try to explain the phenomenons that cause SQ differences in computer audio , that aren't yet well understood/widely accepted, using basic computer science/electrical engineering knowledge. You'll dig yourself into a very deep hole! Wink

It's such a complex topic with very rapid developments, I haven't come across anyone who can properly (as in completely) explain what's going on in all these cases. Proof or measurements are often lacking, at best in these cases there's theories as to some of the why's.

Explaining this all as placebo/expectation bias etc. is way too easy and something at least I won't accept as an explanation. Often differences are "night and day" as in don't need eleborate A/B testing.

Much credible theories/explanations are written on the Internet, I should keep a list where I read what for these types of discussions. Smile Last year I linked to some posts of people in the industry who give some pointers, they're here: https://devialetchat.com/showthread.php?...9#pid41689
PS Audio P3, Shunyata ΞTRON Alpha Digital and HC/Furutech power cables, Paul Hynes SR7EHD-MR4, DIY Roon Server & Roon Endpoint running AudioLinux Headless, Phasure Lush^2 USB cable, Audioquest Diamond RJ/E ethernet, Uptone Audio etherREGEN, Mutec MC-3+ USB, Shunyata ΞTRON Anaconda Digital XLR AES/EBU, Devialet Expert 250 Pro CI, Nordost Tyr Reference LS cables, Von Schweikert VR-5 SE Anniversary Edition, Anti-Mode Dual Core 2.0, JL Audio Fathom F112. More detail here.

The Netherlands
Reply
#3
(06-May-2017, 16:55)Antoine Wrote: A word to the wise; never try to explain the phenomenons that cause SQ differences in computer audio , that aren't yet well understood/widely accepted, using basic computer science/electrical engineering knowledge. You'll dig yourself into a very deep hole! Wink

I agree with you; that is rather the point I'm making!   In particular, I'm not trying to explain differences in sound quality, I'm trying to explain why I believe there should not be any differences.  It's great to hear people report subjective differences in sound quality - and I sometimes do so myself - but if that's accompanied by a technical-sounding explanation then I expect it to hold up to some examination rather than simply taking it at face value.

So, for example, if someone says "firmware v10.1.0 sounds better than 10.0.5" I think: fine, I might or might not agree but I have no problem with that.  If someone says "firmware v10.1.0 sounds better...because...jitter [or whatever]" I think: what's the evidence for that?
Roon (Mac Mini), Wilson Benesch Full Circle, Expert 1000 Pro CI, Kaiser Chiara
Warwickshire, UK
Reply
#4
This is indeed a fascinating topic, and overall I tend to agree with Antoine's statement that there are phenomena that influence digital sound quality that are not well understood.  It is a curious business, the entire western world (and much of the rest of the world) is using digital technology in a range of areas to vast to list, and it pretty much all works exactly as science would predict.  Then we have the computer audio world, which appears to have many theories, very little consensus regarding these theories, and hard science and testing of theories tends to lead to more debate rather than providing firm evidence leading to conclusions. Consider that many of us are happy to download music and store it on a hard drive.  I am sure two people in the world could do this using different computers, networks, and so on.  Compare the two files and they would be identical.  In the domestic environment, this could be sent over a network and buffered and exist in temporary memory local to the DAC, how could anything upstream of this make any difference whatsoever?  We have sent probes to Pluto, discovered the Higgs Boson particle, built functioning quantum computers, but computer audio still confounds us on quite a fundamental level.  Puzzling, but fascinating nevertheless. Consider the Higgs Boson particle was discovered in part using computer equipment just a prone to noise and jitter as anything in the hifi world.  Maybe the Devialet hardware designers armed with their oscilloscopes are smarter than we think and they will crack the whole thing with the new streamer board! Shy  (or maybe not, and the software will probably ruin it anyway, but you never know) Sad

I have to be blunt here, but an awful lot of what I read on this subject just does not stand up to logical scrutiny, in addition, a lot of stuff I read does show clear signs of people falling victim to expectation bias, confirmation bias, placebo effect and so on, to me it is plain to see in very many cases.  This stuff exists and is very real. However, I am not saying that this explains everything, in some cases their is a consistent body of evidence to suggest that something is going on, which would indeed appear to be a phenomenon that is not yet widely understood or accepted.  Unfortunately the phycology of this game and endless pseudo science from some of the people that play this game can create a very obstructive fog to finding and understanding these key phenomena.  

My view on this is that we need vastly better methods of testing this stuff.  To put this simplistically, we need some kind of test equipment that analyse what matters to a higher resolution than the human ear can detect. There is the old mantra, 'not everything that can be measured matters, not everything that matter can be measured.'  (a twist on a Einstein quote, I believe)  My view is that in audio and with 21st Century technology it should be possible to measure everything that matters, and more.  The Hubble telescope does a better job of seeing things than the human eye, by some margin in fact.  OK, the Hubble telescope is an expensive bit of kit, but the same is true of a pair of binoculars.  So why not come up with something that can hear better than the human ear?  The guys with oscilloscopes do not seam to be able to resolve this one, I think we need something new.  Wouldn't it be great to have some kind of high resolution listening device that measured what matters?  You could set it up on a system with three different USB widgets, and declare that widget A and B are doing nothing, but widget B is.  That would be splendid I think.  All we have to do now is invent such a device.  Any ideas?  Believe it or not, I have a few!
1000 Pro - KEF Blade - iFi Zen Stream - Mutec REF10 - MC3+USB - Pro-Ject Signature 12
Reply
#5
I sold everything and bought a Devialet to turn in a much simpler direction, and now... this.
Reply
#6
(08-May-2017, 18:55)BoyScout Wrote: I sold everything and bought a Devialet to turn in a much simpler direction, and now... this.

LOL! The easiest way to be happy is to stay away from any forum or social media altogether. Don't read on and stay blissfully unaware. Smile
                                                    Lifetime Roon, Mac mini, int. SSD, ext. HDD, tv as monitor, key board and track pad on bean bag as remote,Devialet 200, Od'A #097, Blue jeans speaker cable,                                     
                                                                                                                                                                            Dynaudio C1 MkII.
                                                                                                                                                                              Jim Smith's GBS.
                                                                                                                                                                        Northern NSW Australia.
Reply
#7
(09-May-2017, 02:41)Pim van Vliet Wrote:
(08-May-2017, 18:55)BoyScout Wrote: I sold everything and bought a Devialet to turn in a much simpler direction, and now... this.

LOL! The easiest way to be happy is to stay away from any forum or social media altogether. Don't read on and stay blissfully unaware. Smile
A Devialet, a digital source and a pair of speakers. Who has this working properly, please?
Reply
#8
(09-May-2017, 12:52)BoyScout Wrote:
(09-May-2017, 02:41)Pim van Vliet Wrote:
(08-May-2017, 18:55)BoyScout Wrote: I sold everything and bought a Devialet to turn in a much simpler direction, and now... this.

LOL! The easiest way to be happy is to stay away from any forum or social media altogether. Don't read on and stay blissfully unaware. Smile
A Devialet, a digital source and a pair of speakers. Who has this working properly, please?

Worry not!  I ran my D240/250 and then D800 for a long time using AIR.  Since then I have been trying many things, streamers, endpoints and so on.  I now run a microRendu, Mutec front end.  Is it better than AIR?  Yes it is, but not by a massive margin.  Last weekend I tried AIR with the 1000 Pro, it was absolutely fine, I could sit listening to it all day.  (all year even, if I had the time)

So if you have your Devialet and speakers, why not just enjoy the music and forget about all the stuff like optimising for jitter reduction?  For me, I do find this stuff very interesting, but sometimes I just want to forget about all the 'audiophile' tweaking and just enjoy some tunes.  So if you are happy with how your set-up sounds, then just enjoy the music, that would make you the smart one!
1000 Pro - KEF Blade - iFi Zen Stream - Mutec REF10 - MC3+USB - Pro-Ject Signature 12
Reply
#9
(09-May-2017, 13:27)Confused Wrote: A Devialet, a digital source and a pair of speakers. Who has this working properly, please?

Worry not!  I ran my D240/250 and then D800 for a long time using AIR.  Since then I have been trying many things, streamers, endpoints and so on.  I now run a microRendu, Mutec front end.  Is it better than AIR?  Yes it is, but not by a massive margin.  Last weekend I tried AIR with the 1000 Pro, it was absolutely fine, I could sit listening to it all day.  (all year even, if I had the time)

So if you have your Devialet and speakers, why not just enjoy the music and forget about all the stuff like optimising for jitter reduction?  For me, I do find this stuff very interesting, but sometimes I just want to forget about all the 'audiophile' tweaking and just enjoy some tunes.  So if you are happy with how your set-up sounds, then just enjoy the music, that would make you the smart one!

Now we´re talking the right language!
The music is the most important element.
I´m almost there, i guess. Tunning the gear properly without tons of stuff (money and time consuming...) and then hitting the foot against the carpet, if you know what i mean.
Reply
#10
(09-May-2017, 12:52)BoyScout Wrote: A Devialet, a digital source and a pair of speakers. Who has this working properly, please?

...insert emoticon of me holding my hand up and waving it... though when I add up what I've spent on hi-fi I'm a bit disconcerted. And yes, I realize that I'm in at the shallow end of the pool on this point.

Jitter? Never heard of it, and don't even know what street it lives on.
Firmware? I leave it firmly in place and don't think about it or fuss with it.
Damon
Powernode, NAD M32, Cambridge CD transport, Analysis Plus, Nordost, iFi Nova, CSS Criton 1TDX, KEF C62
Vancouver, Canada
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)