Timing error/jitter reduction

[Boy Howdy]

Folks:

I'm new to computer audio and I'm trying to understand how to control (reduce) jitter between the computer and the DAC.

 

I'm using a Mac (iMac, PowerBook, Air) and so I have USB out or mini-TOSLINK (for the iMac and PowerBook). My understanding is that USB output is better than the TOSLINK (certainly willing to hear discussion on either side of this) so I am assuming USB out from the computer for now.

 

Given that timing issues (i.e., jitter) are the most serious compromise to quality audio, I've heard that some kind of "USB converter" (implementations vary apparently) reclocks (retimes?) the data coming out of computer before it gets to the DAC. Apparently "asynchronous" is not a panacea sufficient to address the jitter issue but it is not clear to me whether this is a damnation of ALL asynchronous implementations or only some... Some DACs also claim to deal with the jitter problem but apparently some (even high end DACs) have implemented a canned USB solution that doesn't deal with jitter through the USB input.

 

As a beginner, I do not have the experience to be able to parse the various perspectives on jitter, so I'm asking for some insight from those of you with experience. Also as a beginner (as opposed to a cutting edge adopter) I am trying to understand what is going on with digital audio rather than seeking an immediate state-of-the-art solution. Cost is a consideration but I am much more interested at this point in understanding the logic and implications of one solution over another for jitter reduction.

 

My questions include:

1. What are "USB converter"s that re-clock the data? I don't know if this is another box or an implementation strategy or...

 

2. Are there "asynchronous" solutions (e.g., MF V-LINK) that effectively address jitter?

 

3. Are there DACs which effectively address jitter?

 

An example or two would be helpful accompanied by an explanation for HOW the jitter is reduced.

 

Thanks.

 

[darascal]

1. USB Converters don't re-clock the data. There are re-clockers - check Empirical Audio's website - but these are different beasts.

 

2. The main motivation behind asynchronous converters is to reduce jitter. (Obviously also helpful if your dac does not have USB input.) Same is true of asynchronous implementations on a USB dac. Check Wavelength's website for some technical information on asynchronous USB theory.

 

3. Yes different designers have addressed the jitter issue in different ways. Check out Benchmark & Naim's websites for example. I believe both have white paper descriptions of their dac design and how they are the best way of reducing jitter.

 

Rascal

 

[barrows]

and you will need to do reading and studying to come to a complete understanding of this topic. Here with are some things to consider, and to start you on your way.

 

 

"My questions include:

1. What are "USB converter"s that re-clock the data? I don't know if this is another box or an implementation strategy or...

 

 

Re-clocking is generally a band aid approach, to reduce jitter that has already been created. Many DACs, re-clock internally, usually using an asynchronous sample rate converter (ASRC), not to be confused with Asynchronous USB. While re-clocking this way can reduce measured jitter caused noise artifacts in the DAC's analog output, the re-sampling (often referred to as "upsampling") of the data done in the ASRC can cause sonic problems of its own. Personally, my feeling is that most, if not all, ASRCs cause sonic problems, and are best avoided if one is looking for the best sound.

 

2. Are there "asynchronous" solutions (e.g., MF V-LINK) that effectively address jitter?

 

Aynchronous DACs are usually the best solution to keeping jitter at a minimum level. The most common of these use asynchronous USB receivers to avoid jitter problems. It is important to understand that this approach avoids jitter by not creating it in the first place, rather than by trying to reduce jitter after it has been created (as re-clocking does). In a properly designed asynchronous USB DAC, the clock(s) in the DAC control the rate at which data is sent over USB, and they also time the data out of the USB receiver and into the DAC chip itself. This approach allows the use of single, fixed frequency, high accuracy clock to keep jitter as low as possible. By only having a single, low jitter clock, which controls the timing at every step, the jitter is kept to a minimum level. It is advantageous to have a DAC with a built in asynchronous USB receiver (as opposed to an external USB-SPDIF converter) as this avoids the use of the jitter prone SPDIF path. But, if one has a DAC with only an SPDIF input, or with a poorly implemented USB input, a well designed Async USB-SPDIF converter can offer improved performance (Wavelink, Alpha USB, Offramp 4, M2 tech evo).

 

3. Are there DACs which effectively address jitter?

 

Examples: dCS Debussy, Ayre QB-9, Aesthetix Pandora, The Calyx DAC, etc.

 

An example or two would be helpful accompanied by an explanation for HOW the jitter is reduced.

 

Once again, it is important to disntguish between "jitter reduction" which is usually accomplished by some kind of re-clocking, and using an inherently low jitter design, which does not produce jitter in the first place (async USB).

Additionally, none of these approaches are a panacea-how well they are implemented will have huge implications in their performance: just because something says it uses "asynchronous USB" is no guarantee of great sound. Very careful design, attential to the details of implementation, power supplies, oscillator selection, board layouts, etc... all will affect the final performance.

 

I would recommend going to ayre.com to read their technical explanations regarding jitter and asynchronous USB.

 

 

 

[Boy Howdy]

Rascal:

Thanks for the homework. I'll be back.

 

[Boy Howdy]

barrows:

Very clear distinction between jitter reduction and low jitter design.

and more reading. I sincerely thank both of you for your responses. I'll go read and return. Ciao.

 

[darascal]

This one's a little hard to find, so here's a link to Naim's white paper on their Dac design. It is a non-asynchronous design, and relies on buffering and re-clocking data. It's a good argument that there's more than one way to skin a cat. Also, it sounds great. Everything is going asynch now, but it's no guarantee of quality. One of the highest regarded (and most expensive) dacs, the dCS, was, until very recently, non-asynch.

 

Rascal

 

[Fireblade]

Reducing transmission jitter.

1st school of thought: Traditional, tries to justify exhorbitant audiophile HW prices, or knows better.

2nd School: A simple DAC with asynchronous streaming from PC files beats audiophile players (USB cable from PC to DAC beating Toslink, even). Digital means numbers array (0's and 1's), and a clock for precise music timing of that array. If a DAC is capable of asynchronous streaming (i.e., dictating the receiving distribution between PC data and itself, buffering, and enough multiple sampling), it should do the trick. I have the same problem, and need to decide which way to go (it would be big business for non-rich potential audiophiles like me if the 2nd school of thought is right.) Comments anyone?

 

[darascal]

Dunno about that, I think there's different designs that can accomplish the same thing.

 

However, asynch USB is in vogue now because of the ubuiqity of USB devices & because of the low jitter that is possible, given a good implentation. So there is a plethora of these dacs on the market now, and everyone wants asynch, so this has the effect of driving prices down.

 

One thing that seems true, it is possible to acheive very good sound at a low cost with asynch USB, for example HRT Music Streamer II, $150.

 

Rascal

 

[Mark Powell]

'Given that timing issues (i.e., jitter) are the most serious compromise to quality audio'

 

Is that a given? I am not so sure. But minimising it cannot be a 'bad thing' even though it may not be as important as many believe. But minimising 'obvious' causes of problems is true of every other aspect of audio, and in fact all branches of engineering.

 

There have been 'White Papers' (become an instant expert, write a white paper) saying that even high levels of jitter have no effect on the perceived sound quality. In 'high end' audio there will be an equal number of people saying the complete opposite, simply because 'quality' in this field is entirely subjective. In fact, except for jitter, many people don't like the thought of any kind of objective measurement whatsover.

 

All that said, let us assume that low jitter is important.

 

Asynch USB is one way of reducing it. Currently fashionable, so much so that it will probably become the norm. The direct effect on sound quality is again not measureable, nor even responsive to subjective opinion, as changing from synch to asynch requires a different DAC or the addition of a converter.

 

But you do ask some specific questions, though you appear to have 'locked on' to one aspect, jitter.

 

Avoid the mini Toslink output of the Apple machines, it has very high jitter.

 

The M F V-Link can do four things. Give you asynch rather than synch, give you galvanic isolation via its Toslink output, Give you 96K where your existing DAC is limited to 48K on its USB input, and give you conversion from USB to S/Pdif. Many existing DACs already have these things. Regarding jitter, their advertising blurb says 'No Jitter'. Presumably they think 'No added jitter' is not forceful enough for an advertisement. I bought one for all four reasons. It sounds different. I am unable to tell whether it sounds better or worse, simply because I was not present in the studio when the recording was made. That of course applies to all of us, though many seem to overlook it. I do agree with a previous appender in that it is better to do it in the DAC. It eliminates the weaknesses of S/Pdif, and the (rumoured) high jitter of ANY toslink connection. Pricewise, you do not have to pay much for these things, the V-Link is probably as good as any.

 

Be wary of DAC adverts that say 'Immune to jitter' or similar. None are, and if they use such blurbs their sound quality is probably not as good as they would would have you believe either.

 

As others have said, there are many ways to skin a cat, and many synchronous DACs sound every bit as good as many asynchronous ones. It is all subjective, choose the DAC that gives you the sound that you most like and forget about the technicalities. After all, we are talking about entertainment here, not anything technical or 'scientific'.

 

A final comment. The Naim dac has been mentioned. It avoids the issue, simply by not supporting USB from a computer at all. dCS did the same until recently.

 

[Elprior]

Speaking of (older) dCS products, I'm using them with an external clock that is also feeding the usb (asynchronous) interface.

 

On theory that should avoid jitter in the first place (although jitter is never suppressed). Real field tests are conclusive on my system.

 

The asynchronous mob is always prone to forget about this approach too, which is quite surprising to me...

 

Elp

 

[Boy Howdy]

Mark:

I am still reading recommended literature so I don't consider myself "armed" to re-enter this discussion yet. However, I was talking with one of the folks at Schiit about the Bifrost DAC and I asked whether the USB or the TOSLINK out of my iMac would have the lowest jitter. His response was: "The jitter performance of USB is better, but it tends to have more artifacts in the audio band. TOSLINK sounds better to us." He did tell me to use glass cable rather than plastic for best results.

 

What strikes me is that while they don't disagree with your assessment of the TOSLINK connection, they still prefer it because of "artifacts in the audio band" through USB. This obviously supports your view that jitter isn't everything.

 

Now I need to deal with understanding "artifacts in the audio band." Sigh... Back to my reading...

 

Thanks for your comments,

Mike

 

[drystream]

Give you 96K where your existing DAC is limited to 48K on its USB input

 

Mark,

Could you elaborate? I'm unsure how a DAC with a 48K receiver can somehow handle 96K data fed from a V-Link. I would think you could feed the DAC any sample rate you'd care to and it would still be limited to 48K.

 

[Wavelength]

Howdy,

 

Welcome to the confusing world of Computer Audio. The problem is that the marketing arm of companies wants to throw their 2 cents out to make themselves look good.

 

First why async? As barrows points out the idea is that we don't fix or have to reduce jitter instead we start with a platform for the least amount of jitter. Now that everybody has jumped on board a couple of warnings! First the idea behind what I came up with is that you have two fixed oscillators as close to the dac as possible. These oscillators become the master clock which then is feed to the dac and then feed back to the USB controller to create the I2S Serial Audio feed which is the area of concern about jitter. Two oscillators are required because of the sample rates which group 44.1, 88.2, 176.4 on one oscillator and 48, 96 and 192 on the other.

 

The problem with some asynchronous implementations is they use instead of high quality low jitter oscillators that they use Frequency Synthesizers instead. These have a ton of jitter already and therefore the output becomes poor.

 

The other problem is that some companies understand software and maybe not so much hardware. Too keep an oscillator to have low jitter there is a ton of stuff that has to be done correctly. Companies that make canned oscillators use really low noise power regulators to test their products. Crystek for example claims 0.5ps of jitter at 12KHz, but they use a 7nVrms power supply at 1Hz to test with. Most common three terminal power regulators are between 1uV and 50uV of noise, that's a 77dB differential in noise which will mean that this oscillator will probably have about 50ps of jitter using a standard three terminal regulator. But the big thing is that Audio related jitter happens below 10Hz not 12KHz. So really finding or making a really good oscillator is key, but so is the power supply.

 

~~~~

 

Reducing jitter has never worked. There are basically a number of common ways... upsampling, fifo reclocking, standard reclocking.

 

Almost all high end stuff, even asynchronous USB stuff done correctly will reclock with gates before the dac. FIFO reclocking means you push the data in with the input clock and then have a cleaner clock on the output to pull the data out. Kind of the same thing with an upsampler though you are changing the sound and these have all kinds of other issues.

 

The problem with USB, Firewire and SPDIF that require jitter reduction is that it acts like a filter and therefore the more that comes in the more that goes through.

 

Starting with as little as possible and then assuring that jitter does not raise up is really the best answer for solving the jitter problem.

 

Thanks

Gordon

 

[Boy Howdy]

Gordon:

Thank you for your explanation and elaboration on the jitter issue. I think I am clear on the idea of avoiding jitter rather than generating jitter and then reducing it. I'm still trying to get my head around some of the different strategies of avoiding jitter. I appreciate your input.

Mike

 

[Mark Powell]

In quite a few of the lower cost DACs the USB receiver chip is restricted to 48K. The C-Media chip which mine uses, for example. There are others, but I do not offhand recall their numbers/manufacturer. Texas Instruments lower-cost chips are some of them.

 

But you can bypass this limitation by using a V-Link or similar which uses a higher speed USB receiver chip (max 96K on the V-Link and V-Link II, 192K on its 'not yet released' alternative). These boxes then feed a 96K Toslink and/or coax output which you connect to the Toslink or coax S/Pdif input of the DAC. So have bypassed the 48K limitation.

 

In my case I did it because I had not thought it all through fully before I purchased the DAC. The V-Link is also asynchronous, which many say is an advantage. Also you get electrical isolation from the noisy computer if you use Toslink. Not too sure about 192K capability on Toslink though.

 

The V-Link also reclocks, in a manner that Gordon said below.

 

It is a workaround at best.

 

[drystream]

I sent a 96K data stream via V-Link to an older 48K-capable DAC and all I got was static. Is it possible that your DAC is downsampling to 48K?

 

[Mark Powell]

No, my DAC has indicator lights for all speeds. A VERY useful feature. Setting my system up precisely as Chris says in his Windows/JRiver guide, including making the default speed in Windows 48K, which is often not used, gives the result he says. Play a ripped CD and it says 44.1, play a 96K file and it changes to 96K, switch it all off and back on, play nothing,and it shows 48K.

 

[drystream]

Have you tried using JRiver to upsample a ripped CD to 96K and send it to your DAC without using the V-Link?

 

Quality issues aside, I would think you'd get the same results as with the V-Link.

 

I guess I'm missing something here, but I don't understand how a 48K-limited DAC can handle hi-rez files just by upsampling the data stream, with either hardware (V-Link) or software.

 

[barrows]

To using the V-Link. One, using any external USB-SPDIF converter is only a "second best" solution in comparison to using a DAC with an internal asynchronous USB receiver because the SPDIF interface will always cause additional jitter, and does not need to be used internally in an async USB DAC.

Secondly, the V-Link specifically, does not use two fixed frequency oscillators to time the data. It uses a single oscillator reference, and therefore must use some kind of digital synthesis to derive the needed two base frequencies in order to support both the 44.1 kHz and 48 kHz bases. This approach will have more jitter than using two fixed frequency oscillators.

 

The design choices used in the V-Link I suspect were done in order to keep the price relatively low. This is a good example that just because something works asynchronously, does not guarantee the best possible performance-all asychronous USB interfaces are not equal, and how they are implemented will matter to the final performance.

 

 

 

[Mark Powell]

It's limited to 48K on its USB input ONLY. The S/Pdif inputs are limited to 96K. And you plug the V-Link output to one of those. Not forgetting to change the input selector switch to match, of course.

 

Just out of curiosity, just now I removed the V-Link, connected my laptop directly to the DAC 48K limited USB input, moved the selector switch back to its USB position, and played a 44.1 ripped CD file. It worked fine, as expected.

 

Then I tried to play a 96K file. Nothing came out, also as I would expect. I also got a nice message from the ever-helpful JRiver saying it could not play as my DAC did not support the speed of the file.

 

All as I would expect.

 

[drystream]

I didn't know your DAC was only limited to 48K on its USB input. My apologies for the misunderstanding.

 

[Mark Powell]

I did say that any of these converters are at best just a work-around. But the V-Link has come out well in a review on this site and in two magazines I have read. Personally I don't have anything else to compare it with.

 

Mine is a V-Link II, which may be different inside than the original. It was a touch more expensive.

 

Tomorrow I may take the case off to see how many crystals it has got. I will report back!

 

[Elprior]

@Barrows :

One, using any external USB-SPDIF converter is only a "second best" solution in comparison to using a DAC with an internal asynchronous USB receiver because the SPDIF interface will always cause additional jitter.

 

Clocking the dac and the interface should (in theory) remove this limitation. I understand that practice is another story (for several reasons), but I'm having very good results here.

 

Elp

 

[Mark Powell]

I think we all agreed that. I am only using a converter because I forgot to 'Engage brain before buying DAC'.

 

[4est]

I think in your situation, much of the issues with spdif are not applicable. If I understand it, your clock being fed to both the receiver and the DAC indicates that the DAC does not have to recover the clock signal from the spdif feed. I would not assume that to be inferior. dCS did it that way on purpose. Contrary to what Gordon proclaims (yes, how dare I), a clock amidst the noisy environment of the DAC may not always be the best/lowest jitter. The DACs environment (power supply, ground, RF, EMI and ...) is very harsh, and as Gordon mentioned, will effect the clock jitter. If dCS or Esoteric or whomever wanted to have the clocking internal, they could have done so. For whatever reason, they chose to have it be outboard.

 

The proof is in the sonics after all.