Close in phase noise

[jkeny]

Oh great. Just what we need, another pissing contest.

Gentlemen, with all due respect, few of us here know what the hell you are talking about. This is a very esoteric conversation. At the very least, can you please dumb it down for the rest of us?

Marty & Bruce
I'm not sure how simple to make it?
It's simply that clock jitter of a certain type is claimed to be audible at very low levels - I have heard this myself & hence am speaking from experience. A number of people whose ears & opinion I respect report the same.
Yet others such as Amir deny it's possible & I was laying out how I thought it was possible from a psychoacoustic persepective & from an analysis of how current measurements are missing it.

 

[RogerD]

Back to the normal programs as it seems Elvis has left the building

It has been said that the Jtest is not a suitable test for examining jitter in an environment which doesn't involve SPDIF or AES. The Jtest was developed by Julian Dunn to examine a particular weakness in this protocol which involves Inter Symbol interference in SPDIF (& AES) receiver chips. This weakness has long ago been attended to - mostly because of this test focusing attention on the weakness (& the weakness being identified in the first place by Dunn). Now why has this Jtest remained as the most used test for jitter? Mostly due to laziness. Is it the sensitive enough for jitter testing on modern digital audio devices, particularly USB connected digital audio devices? No, it's not sensitive or specific enough for these use cases & here's an example of why & how it's not suitable.

Let me introduce you to another way of looking at mistiming in digital samples & how the jitter test & FFT analysis is unsuitable. It may also have a bearing on close-in phase noise

I introduced this community to this test here in 2013 - it's a test from Jim Lesurf called IQ-test. Oh & btw, the measurements are taken on the analogue outs of the DAC - the most difficult place to examine such very small timing deviations in the signal.

I'll try to present it briefly here but the full description can be read in my link
Here's a measurement, using this test method, of a Cambridge Audio DACMagic (in red) receiving audio digital samples via USB & the Halide bridge (in blue) USB to SPDIF converter feeding the same DACMagic but this time via SPDIF

What we see in this graph is that DACmagic receiving its digital audio signals via the USB signal is not stable - it does not provide a stable replay on the analogue outs - jumping in replay rate every 25 seconds or so - seen in the red plot. The DACMagic replay in blue when receiving its digital audio signals via SPDIF, is stable in replay rate as measured at the analogue outs.

Stereophile (& users) has this to say about using the USB channel for audio replay "Although its USB input is really of only utility quality and shouldn't be used for serious listening, the Cambridge Azur DacMagic otherwise offers superb measured performance"

Now in that link the Jtest FFT for the DacMagic is shown & it says this "The DacMagic obviously features superb jitter rejection via its conventional data inputs. The USB input, however, performed significantly worse on this test, with both a raised noise floor and significant sidebands apparent (fig.12)."
View attachment 32402

"Fig.12 Cambridge DacMagic, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit USB data. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red)."

So what we see in this graph is an increase in 'noise floor', a slight widening of the base (skirt) of the main jitter signal of 11.025KHz & some sideband spurs, none of which are above -05dB
Now in terms of the usual arguments used about audibility - would this FFT be judged as showing "USB input is really of only utility quality and shouldn't be used for serious listening"?

Or an alternative viewpoint may exist that if we did have a zoomed in FFT close to the main signal spur would we see a vast difference in the width & height of the skirt at the base of the signal & would this be something worth investigating further?

The other aspect to this is that FFTs which are regularly cited & claimed as showing that there is nothing of audible consequence to be seen, need to be queried/examined further.

John, have you ever tested or heard the DacMagic? Btw thanks for the thread.

 

[jkeny]

John, have you ever tested or heard the DacMagic? Btw thanks for the thread.

No, I haven't - have you?
You're welcome, Roger

I know this stuff may be too technical for some people but I hope there are enough who are following it - I try to write it in as succinct a manner as I can but what a number of us (including you) are involved in is relating what we hear to what we have done to our systems & attempting to derive some explanation/logic from this.

The problem with all of this is that it involves an understanding of two very technical areas - technology involved in audio playback (usually digital audio) & auditory perception - in other words relating what the technology does (or doesn't do) to how we perceive (or not) such behaviour of the technology.

You have focused on addressing ground pathways between connected device in the audio playback chain & by your reports have achieved great success. I'm pretty sure that you are reporting real audible benefits as they concur with my perceptions of reducing various types of noise intrusion into digital audio devices.

What strikes me with all these approaches is that there is a set of reported common audible attributes that all the approaches are coalescing towards - it's the same with close-in phase noise - more clarity & a more solid sound stage - I call it a more realistic illusion. All of which suggests to me that there are some common underlying psychoacoustic mechanisms at play & I'm interested in trying to tease out what they might be.

Edit:Looking at your quote of my post I see the second graph is not displayed - it's displaying for me, is it not displaying for you?

 

[marty]

Marty & Bruce
I'm not sure how simple to make it?
It's simply that clock jitter of a certain type is claimed to be audible at very low levels - I have heard this myself & hence am speaking from experience. A number of people whose ears & opinion I respect report the same.
Yet others such as Amir deny it's possible & I was laying out how I thought it was possible from a psychoacoustic persepective & from an analysis of how current measurements are missing it.

Thanks for your patience and explanation. It's helpful.

 

[jkeny]

It appears Amir's not faring any better in my absence. Looks like there may be 1 smart member left on the forum. I wonder how long till he's ousted?

http://audiosciencereview.com/forum/index.php?threads/close-in-jitter.1621/page-9

Yes, Jakob1863 is a knowledgeable guy (he correctly categorised Amir's testing results, the early work of a technician not suitable for release).

Ken Newton is another voice of reason and Amir will always come off worse against people who know what they are talking about.

It's just a shame that Amir shows no capacity for learning - he appears to be stuck in his MS training & takes every opportunity to show this in his ABX results.
Unfortunately this & his ownership of an old Audio Precision analyser seems to have reinforced & further ossified his understanding.

When probed his understanding is found to be superficial which doesn't cut it when discussions are technical.
His claims of knowledge regarding psychoacoustics are often found to be wanting as seen here.
His understanding of technical matters are often found to be erroneous as in:
- his claim that USB packets are retransmitted in the isochronous USB transmission protocol used in audio (worse that when challenged he claimed John Swenson told him this)
- his claim that FFT was oversampled
- his inability to answer the questions I asked him above in the thread

His usual mo is to disappear when probed as has again happened on this thread.

 

[jkeny]

Yes it's very sad when people are just programmed to be soldiers for corporations, and haven't developed the facility's of independent thought in their minds. The ability to reason becomes non-existent. But what can we do. Often over-education can be much more harmful than plain ignorance.

Yes my sig says much the same "The greatest obstacle to discovery is not ignorance – it is the illusion of knowledge"
The world is beginning to react to IYIs (intellectual Yet Idiots)

Two telling examples from that article:
"their main skill is capacity to pass exams written by people like them"
"They can’t tell science from scientism?—?in fact in their image-oriented minds scientism looks more scientific than real science."
"The IYI pathologizes others for doing things he doesn’t understand without ever realizing it is his understanding that may be limited."

And finally this telling characteristic - it applies to Amir & to all objectivists on ASR & other forums:
"Typically, the IYI get the first order logic right, but not second-order (or higher) effects making him totally incompetent in complex domains."


We see the result of this backlash such as Brexit (although this makes much more sense) & Trump - two obvious examples of overeacting & a swinging of the pendulum to the other DIRE extreme (Dumb & Idiotic Reaction Example) - DIRE in so many ways & so many consequences!

 

[jkeny]

A typical IYI comment from ASR:

"I do think a modicum of effort is all that is needed. The ways that jitter manifests itself are very clear cut. Either a random noise or sidebands. Once you know that then picturing what the effects of sub-picosecond jitter with periods below 10 hz is this leaves one really scratching their head as to how this could result in an audible difference."
​

So here we see the IYI characteristic summed up succinctly. They 'know' what is audible based on their "limited" understanding (which they don't recognise is limited) & so they don't try to investigate the phenomena empirically (they have "no skin in the game").

I always find a characteristic of these types is lack of inquisitiveness as exampled by KlasuR on this forum - he heard a difference between two cables but wasn't bothered investigating why & then proceeds to argue that there is no difference between cables & what he heard could have been easily explained if he had bothered

There is no self-realisation about the irony of this.

 

[RogerD]

No, I haven't - have you?
You're welcome, Roger

John, no I haven't...was just wondering how it would compare with others.

I know this stuff may be too technical for some people but I hope there are enough who are following it - I try to write it in as succinct a manner as I can but what a number of us (including you) are involved in is relating what we hear to what we have done to our systems & attempting to derive some explanation/logic from this.

The problem with all of this is that it involves an understanding of two very technical areas - technology involved in audio playback (usually digital audio) & auditory perception - in other words relating what the technology does (or doesn't do) to how we perceive (or not) such behaviour of the technology.

You have focused on addressing ground pathways between connected device in the audio playback chain & by your reports have achieved great success. I'm pretty sure that you are reporting real audible benefits as they concur with my perceptions of reducing various types of noise intrusion into digital audio devices.

There are many types of noise and one type that can be measured is shield induced cable noise and that does impact the sound performance. There are a few cable manufacturers that have technology that addresses these issues,but the majority don't seem to understand or think it is irrelevant,which is fine.
I have never increased the chassis ground size and had it not make a audible difference. I listened to Judy Garland @ Carnegie Hall the other night and on disc 1 there's a part where Judy sings jazz. Well the group of musicians unpacking their gear,zipper cases and all was like they were in the room,up close and personal. So what all this does is that the music just comes closer to the listener or is it in fact closer to the microphone? It is both and there must be a profound effect on the electric current that is flowing through every device. So simply if you can find a scientific paper on the what effects current flow,maybe that can get us closer to understanding phase,emi,and other non signal artifacts.

What strikes me with all these approaches is that there is a set of reported common audible attributes that all the approaches are coalescing towards - it's the same with close-in phase noise - more clarity & a more solid sound stage - I call it a more realistic illusion. All of which suggests to me that there are some common underlying psychoacoustic mechanisms at play & I'm interested in trying to tease out what they might be.

Edit:Looking at your quote of my post I see the second graph is not displayed - it's displaying for me, is it not displaying for you?

John answered and some thoughts in italics. Thanks

 

[jkeny]

What I have heard Roger is the? sound of a low jitter (low close in phase noise) clock on one of my devices & the audible improvement is exactly as described by all those who have heard such. I know this was solely due to the clock as it was a measured clock selected for low close in phase noise from a batch. So the only thing different between one DAC & the other was this clock.

 

[amirm]

Gentlemen, with all due respect, few of us here know what the hell you are talking about. This is a very esoteric conversation. At the very least, can you please dumb it down for the rest of us?

Here is an explanation. To convert digital audio to analog sound, you need three ingredients: 1) PCM audio samples, 2) Reference Voltage, 3) Clock signal. #1 comes from the recording and you have no control over it. #2 can actually create jitter like noise and distortion but is not the subject of this discussion. The topic here is a narrow aspect of #3, namely, what happens if you have random variations of the clock signal at very low frequencies. And when I say very low frequencies I mean below tens of hertz.

Mivera/Mike started this argument by saying there are high-precision clock sources designed for radar applications and such that have lower amounts of random jitter at low frequencies. As you can imagine, in a radar application where you are measuring timing, and you are operating at very high frequencies, these clock sources become important. The manufacturer is not touting the use of such clocks for audio but as is the case in DIY audio world, they chase parts that have better specs with no regard as to advantages/disadvantages for audio.

So measurements of said oscillators are put forward and you are expected to take a leap that what goes into the DAC, is what comes out of it and therefore you better one the best oscillator there is. At some level, i.e. engineering excellence, this is fine as long as measurements are show that the output of the DAC meaningfully changes. No such measurements have been put forward. Worse yet, claim has been made that they don't show up in the best audio measurement tools we have. This is complete hogwash and no evidence has been put forward of any such inability.

Another such claim has been that such random, low frequency jitter noise is readily audible. Both John and Mike say they have heard it yet neither has put forward details of any such experiments. They then put forward word of design engineer Bruno saying such noise is easily audible (both in AES paper and elsewhere). Alas, he too fails to provide any evidence whatsoever of such audibility tests were conducted. If they are conducted at all. In his case, he says the problem manifest itself in such things as smearing of the stereo image. To a lay person that kind of makes sense: timing error must be mean errors in stereo imaging.

Unfortunately lots of crimes are committed there. Specifically they are ignoring the fact that such listening tests have been performed and threshold of hearing random jitter is way, way higher than any DAC regardless of price! More importantly, years of research into how we hear explains the listening test results both on basis of threshold of hearing and masking. Please see the second link I provided in my first post.

Now, it is easy to dismiss all of this as technical gobbledygook and continue to cling to notion of timing error causing audible problem. For that, I actually have thousands of audiophiles to testify otherwise, with vast number from this forum! Yes, if you are into analog sound, you already have ample evidence that all of this is nonsense. Analog formats have horrible timebase accuracy as compare to digital. The reason we don't consider the outcome anything remotely horrible is because of the psychoacoustics. Here is a post I just wrote on this topic: http://www.audiosciencereview.com/forum/index.php?threads/close-in-jitter.1621/page-9#post-40863

Let me see if I can copy and paste it here:

====

One other way to intuitively evaluate the audibility of such random, low frequency jitter if you are into analog sound. Let's look at the measurement of a Linn turntable: http://www.stereophile.com/content/...power-supply-measurements#bA0gIcPIt6uEPkxt.97

Here an LP with a 1 Khz tone was used as the source signal for the measurement. Again, in an idealized situation we would have a single, super sharp spike at 1 Khz and nothing else.

We clearly do not have that here. There are "shoulders" or skirts around our main 1 Khz tone and that indicates random speed fluctuations or in digital lingo, "jitter." We know they are random because if they were not, they would show up as spikes as I have indicated (deterministic/periodic jitter).

Focusing back on the random, close-in jitter that is the topic of this thread, we see massive amounts of it here. The main 1 Khz tone is broadened even at levels of just -10 db! We have large amounts of it by the time we get to -50 db.

The measurements of jitter we have been showing use source signal of 10,000 Hz, not 1,000 Hz like is used above. The higher the frequency, the more pronounced the level of jitter. It is a simple matter to compensate though. To have -50 db of jitter, we need to have a timing error of 2 microseconds at 1 Khz! This is 2,000,000 picoseconds!!! Far, far higher cry than 0.5 picoseconds advocated by Mike.

As we all know, there are tons of advocates of LP playback. None complain of their stereo image being smeared even though they suffer from random jitter that is two million times higher than being advocated.

Why is that case? Reasons we have mentioned before: threshold of hearing and masking. Without these two psychoacoustic effects, no one would be able to enjoy analog formats. They have horrendous timing errors yet with good content, they are delightful to listen to. And even more so for their advocates who consider it better than digital.

The levels in analog jitter can be so high that we can just analyze them in time domain where the actual level of sound can noticeably change. More on this in another post.

=====

So please don't listen to these campaigns. These characters abuse objective audio science to promote myths by creating fear and doubt. They give the whole audio science a bad name. It is classic "measurebating" in my book to show oscillator clock noise spectrum instead of showing us what comes out of our DAC.

 

[jkeny]

So please don't listen to these campaigns. These characters abuse objective audio science to promote myths by creating fear and doubt. They give the whole audio science a bad name. It is classic "measurebating" in my book to show oscillator clock noise spectrum instead of showing us what comes out of our DAC.

Oh, god, please stop - I'm splitting my sides laughing at the irony of these statements. Please no more, you're killing me.

Let's see what accusations you have thrown:
- abuse objective audio science
This from the arch abuser himself
- a guy who is so stuck in in his artifact training days that he produces an ABX test result at every opportunity as a boast & challenges everyone else to do the same. A guy who is so proud that he owns an Audio Precision analyser (about 20years old) that he drops it into every conversation.
A guy that accuses Bruno, Grimm Audio & others (who own far more modern & capable equipment) & have reported the audible differences heard with a low phase noise clock - accuses them of not bothering to use a "modicum" of effort to run a measurement which shows the effect. A guy whose kack handed measurements have been shown in the past to be unworthy of a neophyte technician

- They give the whole audio science a bad name
- From a guy who doesn't know how isochronous USB audio works & makes claims that packets are retransmitted if an error in occurs. When his mistake is pointed out to him he states that he heard this from a competent electrical engineer John Swenson & audio designer. A guy who doesn't know enough about FFT to know that oversampling is not part of it & when told he is wrong tries to claim that DSP engineers use this term all the time. A guy who disappears when asked for evidence be produced to back up these claims. A guy who tries to deny something & then runs away instead of exposing his lack of knowledge. A guy who hides behind soundbite science - whose only able to comprehend first order effects & can't handle complexity.
A guy who accuses me, Bruno Putneys, Grimm Audio & others of giving audio science a bad name!

- And this final one just floored me with laughter
-"It is classic "measurebating" in my book"
Please no more, it's classic self flagellation. the guy is accusing others of what he has been so often accused of himself
You know the more he posts the more he reminds me of your great conman in charge, the chief tweeter - the lack of self awareness, the hilarity of the outrageousness of the statements, the complete lack of any admission of being mistaken, the doubling down when in a tight corner

As to the specifics of the statement "to show oscillator clock noise spectrum instead of showing us what comes out of our DAC" again only shows what first order thinking leads to - an inability to deal with complexity (audio playback systems & their interaction with auditory perception) & hence a reversion to immature & childish approaches to problems, simply because it's more comforting to have 'an answer' even if it is the wrong answer, than to be left in doubt & have to think through how best to approach the problem & what measurements are needed to tease out the underlying mechanisms at play. He uses this simplistic logic - the analogue out is what we listen to, to suggest that if something doesn't show up in simplistic measurements then it proves there's nothing there - as I see all the time the phrase used from the Whitehouse "there's no there, there"

But if you want to prove that you are the only shining beacon in audio science then stop disappearing & stay here to debate your points until conclusion.
Oh, btw, I could feign umbridge at your abusive language towards me & skulk off in feigned disgust, like you always do but I don't do that. Let's see if your a man who can take it as well as give it or are you going to feign disgust & ............

Whadya say?

 

[Robh3606]

As to the specifics of the statement "to show oscillator clock noise spectrum instead of showing us what comes out of our DAC" again only shows what first order thinking leads to

Hello Jkeny


Well I must be dumb. I was wondering pretty much the same thing. Why would the clock phase noise show up in the output of the DAC and why would it be audible? It's not like you are using the clock as an oscillator for a carrier that you are modulating with audio. It's just a gate pulse to keep things synchronized. That is basically what it's doing no? I can understand that the frequency moves around a bit so that could change the period a little but you are running the clock at a much higher frequency than the sampling rate so wouldn't any small time variations basically almost average out except for the crystals aging rate which is long term. With oversampling you are doing all kinds of manipulation and getting clocked thousands of times depending on the latency. So is it cumulative through the process or at the end as you said in your original response with the example of a 20Khz signal??

Rob

 

[jkeny]

Hello Jkeny


Well I must be dumb. I was wondering pretty much the same thing. Why would the clock phase noise show up in the output of the DAC and why would it be audible?

You're not dumb, you're just being misled by Amir's lack of knowledge & his misunderstanding of technical matters - here's what Amir said ""to show oscillator clock noise spectrum instead of showing us what comes out of our DAC" Note that he doesn't say "clock phase noise" but "clock noise spectrum". His lack of knowledge & misunderstanding is confusing for sure but you seem to have corrected this & stated "Why would the clock phase noise show..." but I don't think you understand what "phase noise" signifies & I'm sure Amir's confusion hasn't helped your understanding?

The use of the term phase noise could be misleading - a better way of looking at clock timing errors is with timing measurements - Allan Variance or Time Interval Error (TIE) are two recognised ways of measuring timing of clocks. Allan Variance is a measure of frequency stability in clocks, oscillators and amplifiers. But the easiest to understand is TIE measures - essentially an algorithm calculates the clock rate from many samples to derive a "perfect clock" & then compares the timing of each actual clock tick to the perfect clock tick timing.

In this diagram we see represented the ticks of a perfect clock overlaid on the ticks of an actual clock & how this all maps to the samples of a signal.
First, as you can see there is no averaging over many samples - the timing error @ sample1 can be in the + or - direction & similarly the timing error @ sample2 + or - so there is additive & subtractive possibilities between samples - there is no averaging happening.

Secondly, what has happened to the sinewave when the samples are being converted to an analogue value at the wrong time - the peak to peak period of the wave is shifted - it's no longer a 20KHz period - it's shifted by the timing error between the samples. Now go to the next period & a similar process is happening but the timing error is different so we get a different period - again it's not a 20KHz period. And so on throughout all the tones in the dynamic signal of music - each tone is off by some amount due to the timing errors. Note that this isn't some random noise down at -120dB this is reshaping the sinewaves that make up the musical content of music.

Remember at the recording stage of the music a clock was being used to take an amplitude sample at 44100 times per second (or 96,000/192,000 times per second). the perfect recreation of this signal would be to use exactly the same timing to reconstruct these amplitude samples into the analogue waves of music. Timing errors give us a different analogue output.

The number of occurrences of mistimings is shown by the phase noise plot of clocks - a given clock will seldom be off in timing by a large amount but will be much more likely to be off timing by a small amount - this is what the phase noise plots of clocks are showing - the occurrence of mistimings grow as the mistimings get smaller (closer to the 0Hz)

It's not like you are using the clock as an oscillator for a carrier that you are modulating with audio. It's just a gate pulse to keep things synchronized. That is basically what it's doing no? I can understand that the frequency moves around a bit so that could change the period a little but you are running the clock at a much higher frequency than the sampling rate so wouldn't any small time variations basically almost average out except for the crystals aging rate which is long term. With oversampling you are doing all kinds of manipulation and getting clocked thousands of times depending on the latency. So is it cumulative through the process or at the end as you said in your original response with the example of a 20Khz signal??

Rob

I hope the above answers your questions?

 

[jkeny]

"As to the specifics of the statement "to show oscillator clock noise spectrum instead of showing us what comes out of our DAC" again only shows what first order thinking leads to"

So let me deal with this in a bit more detail.
Why is it so difficult to measure this shifting of the periodicity of the signal on the analogue outputs of the DAC?
Well maybe it's not, maybe it's already there but is just being ignored & categorised incorrectly on the Jtest FFTs?
Everyone sees the increase in the skirt at the base of the signal spike & many just ignore it as immaterial?
Why? Because the FFT isn't of high enough resolution -as I showed in the zoomed in portion of the FFT that the 'skirt' reaches much further up the signal spike as you begin to see frequencies closer to the spike.
Now what if we had a very fine FFT showing levels down to 0.1Hz & you could see that the top of the spike 0dB was spread across .1Hz & @ 3dB down it was spread across .5Hz & 6dB down it was 1Hz - in other words the signal is spreading out in frequency - as it spreads out it's amplitude reduces.

Now we are dealing with two channel stereo - the signal in the other channel is different which gives us the illusion of soundstage & realism in our playback. So if we slightly blur (by this frequency modulation) the characteristics in each channel that gives us this illusion of sound stage & solidity to the illusion - is it any surprise that these are the characteristics that are reported with low jitter clocks?

BTW, these are also the perceptual attributes that many report of the recent improvements in audio playback. These are the very same characteristics that are being derided & sneered at over on ASR. As I said it's typical IYI (intellectual yet idiots) behaviour - they are not equipped in how to handle complexity & think their way through it - instead they are threatened by it & turn to deriding anybody who reports this - they need their comfort blankets.

 

[jkeny]

It would appear that Amir will not stay & defend his side of any technical argument because his lack of technical knowledge will be exposed as wanting - he m.o. is to fire out accusations, post some technical charts gleaned off the internet, make statements which bear no relationship to what he posted & then leave.

It is indeed ironic & shameful that he accuses others of "These characters abuse objective audio science to promote myths by creating fear and doubt. They give the whole audio science a bad name." and wont defend this accusation!!

 

[amirm]

View attachment 32407

You have lost your own plot.

That graph clearly says it is about periodic jitter. The kind of jitter we are talking about here is random, not periodic. Random jitter is not called "wonder."

And I got a kick out of getting a graph from "fiber optics for sale.com!" Perils of googling for graphs....

And none of this answers what Rob asked you. He is asking you why you are assuming that the clock jitter translates directly into output of the DAC. A pedantic explanation of what jitter is, does not work here.

Yes, jitter will modulate the PCM samples. No, there is no 1:1 relationship to clock jitter.

Clock source in a DAC runs at higher frequency than our audio sample rate. A divider is used to lower that value to the sample rate we need for audio. While jitter spectrum remains the same, the energy in clock jitter proportionally gets reduced by that divider. Here is a simple plot that shows it from my post on ASR Forum

Notice that when we divide the clock by 2, we have fewer of those grayed jitter area. If you keep dividing as Rob said, the energy keeps getting reduced further and further. Generalizing, your jitter gets reduced by the tune of 20*log(N) where N is the division factor.

Due to above, depending on internal architecture of the DAC (i.e. its internal sampling frequency), what goes in as far as clock, is certainly not what comes out.

Furthermore, the DAC itself has intrinsic noise. Depending on how much that is, what the clock contributes may or may not be significant. In other words, the intrinsic DAC jitter at low frequencies may dominate the output, not the clock jitter.

 

[amirm]

Forgot the punchline. You must, must show us measurements of what comes out of the DAC. No amount of stolen powerpoint graphs online is going to substitute for that. We don't listen to digital clock source. We listen to the analog waveform out of the DAC. You must show that design changes that are measureable at component level, also are measureable in the output of the DAC.

 

[amirm]

Is it the sensitive enough for jitter testing on modern digital audio devices, particularly USB connected digital audio devices? No, it's not sensitive or specific enough for these use cases & here's an example of why & how it's not suitable.

Nonsense. Your own example below uses it. More on this later.

I introduced this community to this test here in 2013 - it's a test from Jim Lesurf called IQ-test. Oh & btw, the measurements are taken on the analogue outs of the DAC - the most difficult place to examine such very small timing deviations in the signal.

I'll try to present it briefly here but the full description can be read in my link
Here's a measurement, using this test method, of a Cambridge Audio DACMagic (in red) receiving audio digital samples via USB & the Halide bridge (in blue) USB to SPDIF converter feeding the same DACMagic but this time via SPDIF

What we see in this graph is that DACmagic receiving its digital audio signals via the USB signal is not stable - it does not provide a stable replay on the analogue outs - jumping in replay rate every 25 seconds or so - seen in the red plot. The DACMagic replay in blue when receiving its digital audio signals via SPDIF, is stable in replay rate as measured at the analogue outs.

So? USB interfaces on high-end devices worth a penny perform as good or better than Halide bridge. In addition, this is a periodic issue related to locking to incoming USB bus, not low frequency random clock issue that is the topic of this thread.

But let's take this at face value. So what that something happens every 25 seconds? When listening to LP, pops, glitches, track noise, etc. happen with far higher frequencies and amplitudes. Don't see those folks complaining about the sound.

Stereophile (& users) has this to say about using the USB channel for audio replay "Although its USB input is really of only utility quality and shouldn't be used for serious listening, the Cambridge Azur DacMagic otherwise offers superb measured performance"

Now in that link the Jtest FFT for the DacMagic is shown & it says this "The DacMagic obviously features superb jitter rejection via its conventional data inputs. The USB input, however, performed significantly worse on this test, with both a raised noise floor and significant sidebands apparent (fig.12)."
View attachment 32402

"Fig.12 Cambridge DacMagic, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit USB data. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red)."

So what we see in this graph is an increase in 'noise floor', a slight widening of the base (skirt) of the main jitter signal of 11.025KHz & some sideband spurs, none of which are above -104dB or thereabouts
Now in terms of the usual arguments used about audibility - would this FFT (showing spurs n greater than -104dB down) be judged as showing "USB input is really of only utility quality and shouldn't be used for serious listening"? What exactly is audible based on this FFT graph?

Let's post the other graph that was using Toslink (?) input versus above that was with USB:

Right away we see that our instrumentation is exceptionally revealing. We see that USB interface has jumped up 20 db to cover up the spikes that were in the signal at -125 dB. From engineering hygiene point of view, that is nasty. The DAC silicon used was capable of higher performance than this primitive USB implementation allows.

But again, our instrumentation demonstrated that.

As to JA's comment, that is a subjective remark. I doubt that he could could hear the raised noise floor of USB. The close-in deterministic jitter is all at -105 db or better. Our hearing system has a limit of 106 db signal to noise ratio. So even if you listen at full reference level, these distortion products are not audible.

And remember, that tone is at full amplitude at 11 KHz. We don't have any music that we listen to with such high amplitude. Real music has far, far less energy in high frequencies which would push those measured sidebands well into noise floor.

I am happy to give anyone a test who thinks they could hear the difference between USB and Toslink. I am confident they won't be able to tell the difference if they don't know which interface is being used.

So your thesis is quite wrong. We can and do measure performance of DACs well beyond our hearing thresholds. And psychoacoustics helps us figure out audibility. What a person says otherwise is neither here, nor there unless they submit to controlled listening tests.

 

[jkeny]

You have lost your own plot.

That graph clearly says it is about periodic jitter. The kind of jitter we are talking about here is random, not periodic. Random jitter is not called "wonder."

Well, let's see - there is nothing called "wonder" on the graph - I guess this is just your outward expression of your inner state?
Can you possibly think logically, instead of plucking stock labels which you know little about?
"Wander" is just a category label put on jitter which is slipping in timing at a frequency below about 10Hz

And I got a kick out of getting a graph from "fiber optics for sale.com!" Perils of googling for graphs....[/quote]I was looking for the least technical presentation I could find to make it understandable - it obviously isn't understandable to you.
Does this help you understand?

You see how the samples are not represented by a point on the plot of the sinewave, they are represented by a rectangle, representing that the slippage in time between the ideal clock & the jittered clock.

And none of this answers what Rob asked you. He is asking you why you are assuming that the clock jitter translates directly into output of the DAC. A pedantic explanation of what jitter is, does not work here.

Firstly, you confused him by this ""to show oscillator clock noise spectrum instead of showing us what comes out of our DAC".
Secondly, what can I help you with - ask me a question which shows you understand something about this & I will try to answer.
Rob is the best one to state whether what I posted answered his question or not - your lack of understanding is not his, I hope?

Yes, jitter will modulate the PCM samples. No, there is no 1:1 relationship to clock jitter.

Clock source in a DAC runs at higher frequency than our audio sample rate. A divider is used to lower that value to the sample rate we need for audio. While jitter spectrum remains the same, the energy in clock jitter proportionally gets reduced by that divider. Here is a simple plot that shows it from my post on ASR Forum

Notice that when we divide the clock by 2, we have fewer of those grayed jitter area. If you keep dividing as Rob said, the energy keeps getting reduced further and further. Generalizing, your jitter gets reduced by the tune of 20*log(N) where N is the division factor.

Again, you have no ability to think clearly - it's the grey area (jitter) on the clock tick which is timing the sample (the bit clock BCLK) that is important - the intervening clock tick mistimings are irrelevant to this timing. I already explained this to Rob & you obviously didn't understand - why don't you read it again & ask me any questions about points you find difficult to understand?

Due to above, depending on internal architecture of the DAC (i.e. its internal sampling frequency), what goes in as far as clock, is certainly not what comes out.

Again you make no sense! Who claimed this was the case - typical strawman argument.

Furthermore, the DAC itself has intrinsic noise. Depending on how much that is, what the clock contributes may or may not be significant. In other words, the intrinsic DAC jitter at low frequencies may dominate the output, not the clock jitter.

Ah, so you want to invoke a crap DAC as an excuse for close-in phase noise being inaudible! Perhaps this is your situation but I can guarantee you that there are many who don't suffer your ills.

 

[amirm]

Perceptually noise is perceived differently to signal as it is a broadband signal which spans across the frequency bands into which the hearing mechanism divides the signal.

Plot is lost again. We are only talking about differences in clock oscillators that differ in performance at very low frequencies. Here is a graph that was put forward by Mike in the Rutger's write-up:

Notice that by the time we get to 100 Hz, the "old" no good clock source has noise level that is a whopping -140 db down. We are talking about part of the last bit in a 24-bit audio converter which is always noise anyway.

Above 200 Hz, they both have noise levels below -150 db which goes beyond what an ideal 24-bit DAC could produce (best case performance of an audio DAC is about 20 bits). So this discussion is not at all about "broadband signals." It is about narrowband but random noise. This is why it broadens the skirts of our test tones. Otherwise it would have elevated the entire noise floor.

My reel to reel deck has signal to noise ratio of just 80 db. Even if all of that clock noise was pumped into the output of the DAC, its contributions would have been 60 db less than my Reel to Reel deck. While I do hear the background noise in my R2R deck at times, no way, no how can you talk about audibility of such immensely low level noise profiles in digital systems.

You simply have no case. None based on engineering. None based on measurements (which you lack). None based on controlled listening tests.

It is all about inventing a problem, then building a supposed solution, which itself has not been verified to solve the original "problem." The hope is that by using fancy technical terminology, you lose the reader and they resort to lay assumptions and believe you. Not going to work here.