How immune to jitter are modern dacs?

[jostp]

If modern dacs, like the bel canto dac3, are immune to jitter, what is then an explanation for the difference I hear between 3 connections:

CEC transport -> AES/EBU -> DAC3

CEC transport -> Wireworld glass toslink -> DAC3

Mac Mini -> Wireworld glass toslink -> DAC3?

 

Especially the difference in the case of the same Wireworld with the Mac Mini and CEC puzzles me.

Is there another element in play then 'bit-perfectness' and 'jitter', or is it jitter anyway?

Johan

 

[Thomas J]

Hi Johan, I would caution against using words like "immune" which for its part projects a state of absolute-freedom-from. I personally prefer to begin from the assumption that, in any circumstance in which jitter is questioned, and unless strictly proven otherwise, there is and will be no zero-state of jitter. Any DAC can therefore be improved as no DAC is immune to jitter.

 

[Encore]

Johan,

 

As you can see in some of the other threads, electrical isolation of the DAC from the bit source also seems to play a role. I don't which of the three options in your list gives you the best sound? Which of the toslink setups sound best? How are you connecting the power cords to you setup? The computer may pollute back through the power cables to the DAC. Something that may not happen to the same degree with the CEC drive, which has less processes going. Different factors are at play here ...

 

[jostp]

Thomas,

Actually, I took the word 'immune' from the Benchmark DAC1 HDR advertisment at the top of this page at the moment of my writing.

Although I'm aware that perfection doesn't exist, I'm still surprised by the differences I hear.

 

Encore,

The Bel Canto has galvanic and transformer isolation at it's digital inputs. I too first thougt about the computer power isolation, but the difference was still there while listening to the cec with the mac mini on (and putting it's noise on the net). However, I recently have discovered that the bel canto is very sensible to power conditioning, power cords and so on in general.

I can't tell which connection is the best, but I'm sure that the mac mini toslink is worst (against my expectation). In so far that I suspended the ripping of my cd collection untill I find a better solution for playback.

Difference between CEC toslink and CEC ASE/EBU is clear from the first second you switch every time. But so far I cannot decide which is best. The toslink sounds fuller, gives more bass and maybe somewhat more reverb. But the ase/ebu seems more detailed, f.i. when listening to large orchestral works with choirs (complex textures). Which one is right? I don't know, maybe neither is....

johan

 

[ar-t]

No, just different levels of tolerance, and schemes for accomplishing that level of tolerance.

 

Pat

 

[barrows]

you have noted, no DAC is truly "immune" to jitter. Some DACs are better at rejecting incoming jitter than others, I have never encountered a DAC that did not respond with better sound to a low jitter source at its input.

The bel canto uses an ASRC to filter incoming jitter, while this approach does usually result in very good measured jitter performance, I have heard some digital engineers speculate that the incoming jitter is just turned into a different type of artifact on output. In any case, I have a highly modified PS Audio DL-III, which also uses an ASRC to reject incoming jitter; the measured jitter at the output of this DAC is very good, but it still sounds horrible via toslink connection to a MacBook in comparison to being connected to a decent CD transport. BTW, I think it was Gordon Rankin of Wavelength Audio who noted in a post here that the jitter level of the toslink output from Macs was very, very high (I think I recall over 2000 pS).

In my opinion, if you want to get sound out of a computer audio set up that is going to equal (or better) the sound you get from your transport, you are going to need to look into low jitter options like: Async USB DACs or USB to SPDIF converters, ASYNC Firewire DACs or Firewire to SPDIF converters, or a true re-clocker that will reduce the jitter between the computer and DAC (Empirical Audio Pace Car, PS Audio Digital Lens).

 

[Thomas J]

For me, the word tolerance is little more descriptive than immune. I prefer what Barrows says:

 

I have never encountered a DAC that did not respond with better sound to a low jitter source at its input.

 

Referring back to our discussion regarding the Diverter, everything in audio is a (cough) filter. Some percent of jitter-in will be passed to the output, either as jitter-out or as a jittered analogue signal. Johan, I'm unsurprised you got the word immune from the Benchmark ad: more manufacturer misrepresentation. The filter analysis is based on Ohm's law, and from what I can tell, Ohm's law ain't too hard to understand. I infer that any engineer has a good grounding in Ohm's law.

 

[bordin]

I had a discussion with a friend of mine about low-jitter transport, particularly using computers. Some audiophiles always argue that their $$$$$$ CD transports sound the *best* and never believe in the computeraudiophile approach.

 

If we could make jitter down to zero (or almost zero) then would the music sound pleasing to human ears ?

 

It is a fact that human perception is very subjective in nature. We love distortion, don't we ? An anechoic chamber (no sound reflection) makes audio gear worthless.

 

My friend said pro audio industry has acheived the level of zero-jitter for a long time. This's doubtful to me.

 

Perhaps, such expensive CD transports add some jitter into the system to make it sound pleasing.

 

[Thomas J]

Hi Bordin, your friend who claims zero-jitter is simply wrong. Zero jitter, for one, cannot be measured because the measurement is beyond the resolving capabilities of the best measurement devices. And even using current best-available such devices, I know of no gear that measures no positive level of jitter.

 

Fwiw, I own a Weiss AFI1 firewire to AES converter. This would be considered a top-line piece of pro equipment. The unit's stock power delivery setup looks like this:

 

cheap $4.99 inboard switcher ---> cheap 3v3 regulator ---> cheap 1v8 regulator

 

I replaced the inboard switcher with a Hynes shunt supply, and the sound improved nicely. That would be a jitter reduction, no? I'll soon replace the cheap off-the-shelf regs with Hynes shunts, and probably will add a dedicated transformer with an extra step-down regulation layer for the ever-so-sensitive 1v8 regs, which supply the ever-so-jitter-sensitive clocks. I'm expecting a larger improvement than the previous upgrade.

 

Ain't no zero-jitter for the stock Weiss, that's for sure.

 

[Encore]

"If we could make jitter down to zero (or almost zero) then would the music sound pleasing to human ears ?

 

It is a fact that human perception is very subjective in nature. We love distortion, don't we ? An anechoic chamber (no sound reflection) makes audio gear worthless."

 

Bordin, while this may be true with analogue signals, in the digital world it is highly unlikely. For the simple reason that the distortion resulting from jitter has a very sharp sonic character, compared to the distortion from tube gear, especially. While in the latter the distorstion is harmonically related to the signal--in the case of tubes, even-order harmonics--jitter distortion has no harmonic relation to the signal, but is instead related to the jitter. Thus, distortion products may for instance occur at 1 kHz sidebands, completely unrelated to the music. This may be what is causing poorly implemented digital equipment to sound so harsh.

 

[ar-t]

All, repeat, all clocks will exhibit a certain amount of Gaussian uncertainty. Noise sidebands, of said Gaussian nature, exist on either side of the clock frequency. In a good clock, it will be very low, but not zero.

 

So how low is low? Let's say that you have a clock at 256 x Fs (11.2896 MHz). A good clock will have jitter less than, say, 10 pSec. But that also depends on how it is spec'ed. So, let's say that it is spec'ed for the frequency range (of sidebands) of 10 Hz

[bdiament]

I'm skeptical, to say the least (and to put it nicely) of any claims of jitter "immunity" or "zero jitter".

 

Funny too, how those claims are never made by the folks making the very best DACs.

 

Others attempt to attack jitter by resampling the audio on-the-fly.

They make no mention (and perhaps are not aware) of the sonic price exacted by real-time sample rate conversion (at least for my ears, in every instance I've experienced it, regardless of the price or design).

 

Then again, some (including a number of reviewers) hear the brightening and hardening of the sound from the spurious harmonics added by the conversion as added "detail". As I see (hear) it, if it isn't contained in the original, it isn't added detail, it is added distortion.

 

Just my perspective.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[Thomas J]

I agree, Barry.

 

Actually, the Benchmark ad says "completely immune to jitter." That borders on a deliberate lie.

 

[idiot_savant]

As has been pointed out, all oscillators will exhibit some kind of jitter - sometimes referred to as "intrinsic" jitter.

You may then have jitter artefacts due to the power supply e.g. injecting some mains-related stuff into the local oscillator.

On top of this, you then have to deal with interface jitter - the jitter at the local clock that is added by attempting to synchronise the local clock with the clock in the source - the the DAC has to consume data at the same rate the source is feeding it, or buffers will overflow.

 

These are ordered from least to most ( in a competently designed system )....

 

There are a few main techniques for dealing with interface jitter:

Ignoring it altogether.

The interface itself.

Using a PLL to filter the interface jitter, providing a cleaner version of the clock reconstructed from the interface to the DAC.

Using an Asynchronous Rate Converter ( ASRC ) to manipulate the data such that it "fits" in the two clock domains

Using the DAC clock to drive the clock in the the source ( "asycnchronous" modes - async USB, async firewire, most ethernet solutions, some kind of link from DAC to source )

 

The last is the best - assuming the same level of competency is applied in all cases, but all bets are off otherwise...

 

your friendly neighbourhood idiot

 

 

[pj]

I'm glad to see Pat making a contribution on this forum. Having read Pat's interventions across many forums over the years I think it's safe to say that he is an (if not "the") authority on spdif interface design and related issues of jitter, as well as an outstanding audio designer. Pay heed to what he has to say

 

[ar-t]

I think it is safe to say that I have spent way too much time working on these issues, for a number of years. Probably more than most anyone else out there. So, the safest thing to say is that I have a lot of experience. Whether that makes me an expert, that is you to decide.

 

Thanks again.

 

Pat

 

[cfmsp]

Thomas

 

"Actually, the Benchmark ad says "completely immune to jitter." That borders on a deliberate lie."

 

Sounds like it's completely on the other side of the "border" to me.

 

As Barry said - and I agree - there seems to be an inverse correlation between claims (of jitter-free implementation) and highest quality implementations. There's a clue for you!

 

Reminds me of the old saw:

 

"Good golfers talk about their bad shots, bad golfers talk about their good shots."

 

Replace golf with investments, etc, .... you get the idea.

 

 

caveat emptor,

clay

 

 

 

 

 

 

[The Computer Audiophile]

"Good golfers talk about their bad shots, bad golfers talk about their good shots."

 

So apropos!

 

One company that I've questioned about its claims of no jitter is Playback Designs. Very expensive and very high performance, yet still claims of no jitter. PD seems to be an exception to the above "rule."

 

 

http://www.playbackdesigns.com/

 

[Thomas J]

Hi Chris, I own a PD unit. It is very good, but it's either not no-jitter, or the sound it produces is compromised somehow else. For instance, the USB input, as Jonathan Tinn said himself, is the "worst input," which he recommends not using. How could any input be worse than another in a no-jitter machine?

 

On the upside of their technology, and this is a bit of a guess, I suspect they've taken jitter reduction into the software realm. I think their design somehow 'reads' incoming jitter, creates an algorithm corresponding to the negative of that jitter, then applies that algorithm to the signal digitally. Voila, or so the story is ideally supposed to go. It is very effective, I must say, but my iPod + modified Pace Car beats it in the digital realm.

 

I've not listened to the DAC portion through speakers. I suspect the DAC is very good (everything's upsampled to DSD).

 

[icebreaker]

"Others attempt to attack jitter by resampling the audio on-the-fly.

They make no mention (and perhaps are not aware) of the sonic price exacted by real-time sample rate conversion (at least for my ears, in every instance I've experienced it, regardless of the price or design)."

 

A very interesting topic: to NOS or to Oversample. I have a very basic understanding and a number of questions.

 

Are you referring to both asynchronous SRC and synchronous SRC?

 

Do you maintain that the deleterious effects such as "the brightening and hardening of the sound from the spurious harmonics added by the conversion" outweigh the problems caused by not using SRC to deal with imaging and aliasing?

 

I have read comments that SRC eliminates high frequency energy that can affect the performance of amplifiers and the spatial presentation of the music.

 

Thanks.

Regards,

 

[bdiament]

Hi James,

 

I'm referring to every real-time SRC process I've heard - my personal response to them of course.

 

"Do you maintain that the deleterious effects such as "the brightening and hardening of the sound from the spurious harmonics added by the conversion" outweigh the problems caused by not using SRC to deal with imaging and aliasing?

I have read comments that SRC eliminates high frequency energy that can affect the performance of amplifiers and the spatial presentation of the music."

 

I'm not sure what you mean by "the problems cause by not using SRC to deal with imaging and aliasing". I have never thought of SRC as mandatory (unless I'm going to master from a 44.1k source or of course, when I'm creating a 44.1k master for CD replication). Outside of that, I can't think of a good reason to subject the signal to the additional processing.

 

Not sure where the "SRC eliminates high frequency energy..." stuff comes from but it sounds to me like that would be the low-pass filter's job and have nothing to do with SRC. Again, I question what sort of amp the source of this quote is using, that its "performance or spatial presentation" is so compromised.

 

Just my perspective, of course.

 

Best regards,

Barry

www.soundkeeperrecordings.com

www.barrydiamentaudio.com

 

 

[Wavelength]

Gang,

 

I have to agree with Pat with a few additions....

 

With SPDIF and an Asynchronous Upsampler ARSC (not to be confused with Asynchronous USB) you not only have the jitter related problems of the SPDIF. But you also have to take into account the intrinsic jitter between and including the Sample Rate Converter, Master Clock and the DAC.

 

But I think what Johan is experiencing is what I have been talking about for the last year. When you try and fix jitter be it using an ASRC or Reclocking, FIFO Reclocking or PLL/VCXO the problem is what happens to the effect of the Digital Audio Stream out of the SPDIF receiver and into these jitter fixing devices?

 

Well I think this is were Johan is experiencing a difference in sound. I think there is more going on here than engineers are afraid to ask about. I did some testing here by changing code in a dac that uses a similar technique to the DAC3 and it was pretty easy to hear the difference when I changed the jitter into the ARSC from 2800ps to 675ps.

 

If an ARSC truly makes a dac immune to jitter then I would have to beg to differ.

 

ANY jitter reduction circuit acts like a filter. The more in the more out...

 

Also ANY company that claims zero jitter is governed by marketing as zero is not a reachable goal.

 

Thanks

Gordon

 

[Thomas J]

I suppose it doesn't help things when reviewers hope onto the zero-jitter fantasy-wagon. Here's John Atkinson from his review of the dCS Puccini:

 

Almost all USB DACs operate in what's called adaptive isochronous USB mode, which means they have to adjust the frequency of their master-clock oscillator every millisecond to match the rate of the data being streamed from the host PC. This inevitably leads to increased jitter that potentially degrades ultimate sound quality. By contrast, the asynchronous mode slaves the PC to a high-precision, fixed-frequency master clock in the DAC (Ayre, Wavelength) or the U-Clock. In theory, this eliminates jitter entirely.

 

It doesn't "in theory," and it doesn't "in practice."

 

[BobH]

I'm now a long way past modern and, at least the last time I looked, I wasn't actually a DAC at all - but I am completely immune to jitter.

 

All things are possible!

 

[jostp]

OK, I understand this from all what I read:

there is a source (transport, computer,...) which has inherent jitter, there is the interface (spdif coax, spdif toslink, aes/ebu, ... which adds jitter to the source signal, then there is the receiving dac, reconstructing the source signal - albeit with all cumulative jitter.

But what's the problem if this reconstructed signal is stored in a small buffer?? Even the apparently very bad mac mini has only 2000 ps of jitter, a large buffer is not needed to middle this out. Once the bits/bytes are waiting in the buffer, all previous (jittered) clock information doesn't matter anymore. It's up to the dac now to send the bits/bytes to the AC converter on the pace of it's own ('gaussian' jittered) clock.

My question: why would one have at the output of the dac other jitter then the jitter of the dac's own clock?

But then again, I hear three clearly different sounds from my bel canto in the case of cec-toslink, cec-aes/ebu, mac mini-toslink. So, I must be wrong. Or is there 'something else' in the game? Every one talks about 'jitter', but no one can exactly explain what is going on based on this concept.

By the way, see:

http://stereophile.com/digitalprocessors/1107bc/index4.html

(Fig. 9, jitter performance of the bel canto dac3, toslink input)