Is digital audio intrinsically "broken"?

[NorthStar]

Where would we be without Frank?

 

[fas42]

Amir, so you suggest that people shouldn't have a $9 part in their DAC to improve jitter performance, but it makes perfect sense to invest in $2,000 CD transport to do the same thing? Hmmm ... or maybe we should reduce the thickness of the metal plate on the front of the DAC box from the 1/2" which makes it look good, to save money.

Also, electrically, a clock is a clock: it has nothing to do with S/PDIF, it has to do with fixing the signal pulse just prior to being fed into the D/A circuitry. A device has no knowledge of what purpose the designer had for it, if it does the job correctly then that is all that matters. We'd better throw all those tube amps using glassware designed to be used in Russian military equipment in the bin then, they can't possibly work as audio amplifiers, I guess ...

I've already pointed out why relying on the target device for the clock was never an option as an industry standard: once two devices are linked to a source then it becomes a problem; recording studies would have had a meltdown for a start.

Frank

 

[amirm]

Amir, so you suggest that people shouldn't have a $9 part in their DAC to improve jitter performance, but it makes perfect sense to invest in $2,000 CD transport to do the same thing?

No, once more this thread is about *architecture* of digital audio. It is not about $2,000 CD players, tape or LP.

Also, electrically, a clock is a clock: it has nothing to do with S/PDIF, it has to do with fixing the signal pulse just prior to being fed into the D/A circuitry. A device has no knowledge of what purpose the designer had for it, if it does the job correctly then that is all that matters. We'd better throw all those tube amps using glassware designed to be used in Russian military equipment in the bin then, they can't possibly work as audio amplifiers, I guess ...

Even more off-topic.

I've already pointed out why relying on the target device for the clock was never an option as an industry standard: once two devices are linked to a source then it becomes a problem; recording studies would have had a meltdown for a start.

Frank

I have already answered that. And recording industry can and has used external master/house clocks for decades.

 

[fas42]

I think I'll close out on on this now, Amir, we don't seem to be getting anywhere; leave it for another day ...

Frank

 

[fas42]

What is multiple targets? In an AVR? Then there can be one master clock driving all the DACs. If it is in separate devices, then a house clock is used as in pro equipment. Or else, you don't care like playing music in different rooms.


I explained above. As another example, I can have 1000 DAC playing the same identical file on a NAS shared over Ethernet. Nothing breaks. No complexity either.

Sorry, last call. Somehow missed that response, but what you said is not a solution. The master or house clock is nothing more than being equivalent to the first clock, the one in the CD transport or player. Just because it's inside a particular device doesn't make it any less a master clock, and just because its pulses are turned into data of the S/PDIF stream shouldn't make its transitions any less precise, assuming decent engineering. Plus, once outside the case of an external master clock we're back to the same old ball game of worrying about cables, connectors, interference pickup introducing jitter into our beautiful, expensive master clock ...

Regarding that visual playing with the waveform, I guarantee that I'll be able to play with it and the AP analyser won't pick the difference in spectrum and THD ...

Frank

 

[amirm]

Sorry, last call. Somehow missed that response, but what you said is not a solution.

Why? Are they not able to play audio? If they are, then you can't say the system breaks. That is of course putting aside the fact what you describe is not a useful scenario for home playback of music.

Regarding that visual playing with the waveform, I guarantee that I'll be able to play with it and the AP analyser won't pick the difference in spectrum and THD ...

Frank

What are you going to do again?

 

[Orb]

Frank, if you are going to feed us Google snippets without understanding and reading them I won't be wasting time with you. I have no interest in answering tidbits produced by Google.

The Si5317 is a $9 part *without* the external parts it needs. It is in no way designed to be used for audio applications. This is what Si says about the part: "Highly Integrated Si5317 Jitter Attenuating Clock Filters Unwanted Noise from High-Speed Networking and Telecommunications Systems". The app notes are about SONET, not S/PDIF. So clearly you don't appreciate what these parts do and the nature of the problem.

No one has said that you can't clean up jitter. You can. But it costs you complexity and part cost. The entire point of the thread has been that we would not be worrying about jitter nearly as much if it didn't rely on source input so much. To the extent you keep throwing these expensive parts out there as solutions, then you have helped prove the case.

I highly suggest you read Don's articles instead of Googling jitter.

Good point on Don's articles, which I did mention the primer is excellent to read and combine with posts from both you and Don.

Problem is I feel some are just skimming the Tektronix primer, which I strongly suggested should be read carefully so as not to be taken out of context.
That said the primer is easier to follow than the tech articles of Don (due to being a primer), so definitely goes well with Don's.
Just to clarify for others, it does not matter the document does not cover audio-jitter levels raised by Frank as its purpose is to show the behaviour-cause-pattern of jitter, the eye pattern is crucial to jitter transmission and very close to Don's posts (which I mentioned earlier this document should be used with if interested but not as specific as now).
And importantly as I mentioned, these match very closely to real world audio products, if one follows reviews and measurement a lot by John Atkinson and Paul Miller.
That said the primer is excellent for following architecture of digital signals and jitter.
Thanks
Orb

 

[Orb]

Sorry, last call. Somehow missed that response, but what you said is not a solution. The master or house clock is nothing more than being equivalent to the first clock, the one in the CD transport or player. Just because it's inside a particular device doesn't make it any less a master clock, and just because its pulses are turned into data of the S/PDIF stream shouldn't make its transitions any less precise, assuming decent engineering. Plus, once outside the case of an external master clock we're back to the same old ball game of worrying about cables, connectors, interference pickup introducing jitter into our beautiful, expensive master clock ...

Regarding that visual playing with the waveform, I guarantee that I'll be able to play with it and the AP analyser won't pick the difference in spectrum and THD ...

Frank

Frank, as I mentioned the differences are quite noticable for a measurement tool in terms of distortion when looking at -60dbfs (and this is still a relevent level), I know this as some reviews analyse digital gear at -20 -60 -80, while another magazine shows the whole distortion-noise against dbfs (0 to -120dbfs).

Cheers
Orb

 

[NorthStar]

Digital Audio, is it broken from the get go? Yes, timing is everything, and was never taken care of.
Picoseconds of delay are polluting our hearing, even if we don't hear it!

And it has nothing to do with cables or/& all that external Jazz, but everything to do with the digital internal timing clock.
Digital in itself is intrinsically complex with all the paths & parts, and their implementation requires extreme attention...
Totally different of analog.

Euphony, that's where we're at with digital (upsampling, quantization, jittering, etc.), to the point we are now in auditory disillusionment.
We became addicts; our own timing in life is influenced by it!

Time to return to our true source, our real origin; Analog.

* It's like the difference between an electric guitar with all those digital delay pedals and distortion and sustain and flanging, as compared to an acoustic or classical one in the nude (natural).

P.S. Amir & Frank, I read everything you posted, every single word with the utmost attention.

 

[Vincent Kars]

The groove of an LP

Don't think I will go back to analog

 

[NorthStar]

The groove of an LP

Don't think I will go back to analog

Looks fantastic to me; nature at its very best, like the world's nicest crevasses.

* And all of this read by a sophisticated multi-cut diamond; how more cool can it be?

 

[microstrip]

Amir,
Do you want to educate us on how jitter is usually measured in DACs? I was looking at the jitter measurements of the Simaudio Moon Evolution-650d cd player but I must say I can not understand them .


http://www.stereophile.com/content/simaudio-moon-evolution-650d-cd-player-measurementsthe

 

[amirm]

Amir,
Do you want to educate us on how jitter is usually measured in DACs? I was looking at the jitter measurements of the Simaudio Moon Evolution-650d cd player but I must say I can not understand them .


http://www.stereophile.com/content/simaudio-moon-evolution-650d-cd-player-measurementsthe

Oh gosh. For months I have been thinking about writing a tutorial on what the stereophile measurements of digital audio means and I keep forgetting . I actually wanted to recruit Don to help me write it. Don, are you game?

For now, let me say they are the toughest things for people to interpret. I know, I used to be among them . My eyes would just gloss over them. I mean what the heck do those graphs mean? As to not steal the thunder from that article we have yet to write , let me show the bit that we are discussing regarding jitter:

Let's get past the obvious: the DAC was fed a digital stream that represented an 11 KHz tone. That is the tall peak reproduced in the middle in blue. An ideal system would just have that and nothing else. Instead, we see a number of pulses surrounding it in red. Those are all distortion products.

That's not the full story though. We have to understand the amplitude of said errors. The noise floor of the system is around 145 db so we can assume he is testing 24-bit audio samples. This is confirmed in the captions of the figure above:

"Simaudio Moon Evolution 650D, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 16-bit data via 15' TosLink S/PDIF from AP SYS2722 (left channel cyan, right magenta), 24-bit data (left blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz."

The system then has its maximum response at 0dbfs and lowest at -145db (or thereabouts). The distortion products of this product "rise" up from the noise floor to -125db. In other words, they are 20db worse than a theoretical DAC. But let's put them in context. CD's 16 bit system has a noise floor of -96db. So this DAC handily beats that specificaton by a whopping 30 db! Clearly it is hard to argue that such distortion is audible if it is so much better than the CD. We get confirmation of this from John Atkinson running the test:

"all that can be seen in the spectrum are the residual harmonics of the low-frequency squarewave. This is state-of-the-art jitter rejection."

Stated of the art indeed.

But there is more to test. We see that in the graph caption and the start of the above paragraph. The input test tone is NOT just a pure sine wave. But rather, the sinewave has been manually manipulated by toggling its right hand bit on regular intervals. This is done at a precise moment that causes all the bits to flip. In decimal it would be like going from 9999 to 10000. There is only 1 number differentiating those two values but all the digits change at once.

The above is the so called J-Test. An ideal DAC would not be sensitive to its input bit pattern. What should it? A DAC outputs its input values and should not care if all the bits toggle at once or don't. Real life DACs don't work that way. A drastic change in digital input has enough of an impact on the DAC to cause distortion.

Critiques of J-test claim this is a "worst case" signal and the results not representative. I don't buy their argument. We want to know how our systems differ from the ideal "wire" test. We want to know if a DAC is data dependent in its distortion as such distortion can be very insidious and hard to pinpoint. Music signals can do what they want. There is no way to assure that most if not all the bits in an audio sample won't change. If a DAC all of a sudden puts out a lot of distortion then, we want to know what. If the DAC is sensitive to this, then it may produce such distortion even if all the bits are not toggled.

Did this make sense?

 

[muralman1]

Late to the party, the digital crowd may have been thankful. My sentiments are with the vinyl lovers. However, truth tells it, the CD is vastly superior! It's failing lies squarely on the shoulders of digital engineers. You see, I have proved the super microscopic and complex music signal's properties fall under the purview of Quantum Physics. One cannot ascertain the location of the signal without measurement. That, in turn, freezes the signal in an altered form.

Vinyl reading is direct. There are no following filters. My system has no chip other than the DAC and clock. Yes, that freezes the flowing signal, but only in it's basic analog form. Now, treated with kid gloves to the ultimate truth teller, the Apogee Scintilla, passed to the experienced ear. music attains a state of realism never imagined possible.

 

[microstrip]

Thanks Amir!

I am picking another example from Stereophile. The main difference I see is are two extra peaks at the central zone. Are they due to jitter?
Do you think that the 5 dB mean difference in the background is of any importance?

 

[Orb]

Just expanding upon the J-test,
the squarewave signal is used as it creates spurs relating to the signal-data, these should be the harmonics and so at set frequencies.
How well a system deals with these affects the level of the spur.
On top of this we have other jitter affects and noise.
The problem is Micro you need a system that allows one to differentiate between the various spurs for the noise and jitter types, this is why Paul Miller's tool is good as it can break these down into greater detail.
The following graph of the DAC64 is perfect as it shows the behaviour of jitter for data, random,noise like-uncorrelated jitter:

The data related jitter-harmonics sidebands are the red markers, the random noise is the hill with buffer disabled, and with buffer enabled shown in black and with blue-green markers (looking closely the J-test data marked red sidebands can be seen to have two heights with the lower one being with buffer enabled)
So those that are not red markers, are noise like or uncorrelated jitter.
This matches very closely to the Tektronix primer with the section I mentioned earlier, plus a much later one which is entitled Putting it all Together.

One thing that I find interesting is that a reviewers perception suggests that quick look at the low figures still means they can be heard in terms of influence.
As a very good example is the Yamaha NP-S2000 that was reviewed in HifiNews.
This has a reasonably clean jitter outline with the floor at around -140 or -135 (if 96khz) but with a narrow and sharp skirt hill (like in the image I posted but starting within 500hz or 300hz depending upon sampling rate) that reaches a very narrow point at 0hz centre signal like in the figure but at -100db.
The result for several reviewers was the effect thad "taken the edge off the sharp stereo focus of vocal and other images".
But as mentioned different patterns due to the jitter type or behaviour can or may have different affects to what is perceived.

Edit:
For reference here is the measurement for Yamaha NP-S2000, will need to login or register if 1st time, scroll down for the 48k and then 96k jitter test to see how subtle but anecdotally perceived by reviewers.
http://www.milleraudioresearch.com/download2011/reports/feb11/yamaha_nps2000.html
Cheers
Orb

 

[fas42]

The point with all these graphs are that jitter and other distortion products at the -100dB level or better. This means they are INAUDIBLE!! Now reviewers and other listeners can hear that the edge can been taken of the stereo focus and similar impressions, and claim this is due to jitter, but as far as I'm concerned this has now become the easy peg to pin every "problem" with digital sound on. Just like feedback in a power amp: hmmm... doesn't sound too good, must use too much feedback ...

I can take the edge of the stereo focus extremely easily in my system, by reversing any 1 of 100 things, weaknesses, I've fixed over a period of time: it would be pretty dumb of me to say that every one of those things only benefit was to improve the jitter performance.

Frank

 

[fas42]

Frank, as I mentioned the differences are quite noticable for a measurement tool in terms of distortion when looking at -60dbfs (and this is still a relevent level), I know this as some reviews analyse digital gear at -20 -60 -80, while another magazine shows the whole distortion-noise against dbfs (0 to -120dbfs).

Cheers
Orb

Orb, the only thing that matters is the effective, absolute distortion level: if a -60dB signal has 1% distortion that sounds terrible to say, but 1% is -40db, add that to to -60dB signal, gives you -100dB actual sound level: inaudible.

Try testing a vinyl signal at -60dB: if you're lucky you'll get 100 - 1000% distortion and noise, good thing LP distortion is inaudible on a good system ...

Frank

 

[fas42]

Why? Are they not able to play audio? If they are, then you can't say the system breaks. That is of course putting aside the fact what you describe is not a useful scenario for home playback of music.


What are you going to do again?

Move data points on the waveform in various, random ways to mimic exactly what quite severe jiter would do to alter the signal, say equivalent to 100 nsecs of peak to peak, random jitter.

Frank

 

[amirm]

Move data points on the waveform in various, random ways to mimic exactly what quite severe jiter would do to alter the signal, say equivalent to 100 nsecs of peak to peak, random jitter.

Frank

You can't do that. Jitter impacts the entire waveform. Not one or two data points. And we are not worried about random jitter because it simply manifests itself as noise. Have you read the jitter tutorials yet? It doesn't seem like you have Frank.