USB vs. Toslink

[noogle1]

Some questions from a puzzled newbie:

 

Was looking at dCS DACs online and saw they use USB inputs rather than Toslink optical. Doesn't this inject a lot of electrical noise from the PC into the DAC? Wouldn't the optical isolation of a Toslink eliminate this problem? Or am I missing something?

 

[audioengr]

The problem with most Toslink feeds from PCI cards or imbedded motherboard audio circuity is the jitter. These are generally low-cost solutions using relatively high-jitter oscillators. Toslink itself is also an extra layer of interface that adds jitter. Computers are generally electrically noisy environments, with lots of high-frequency clocks and thousands of traces all carrying high-frequency signals. The power supplies are usually switchers and generate their own noise. Space on motherboards usually precludes the best power delivery topologies, so it is generally a lot of compromises that ultimately impacts sound quality by making a low jitter signal very difficult to achieve.

 

USB done well, such as asynchronous mode with separate power supply, eliminates virtually all effects of the computer itself. The computer is only a means to transfer the dataset to the USB interface. Then, the USB interface establishes a master clock with low jitter and generates a low-noise S/PDIF or I2S output. I2S goes directly to the D/A. S/PDIF is the "coax" cable that goes to the DAC.

 

If you are not familiar with jitter, read this:

 

http://www.positive-feedback.com/Issue43/jitter.htm

 

Toslink does offer galvanic isolation between the computer and the audio system, but most USB interfaces also offer this.

 

Steve N.

Empirical Audio

 

[Codifus]

......given that DACs are getting better and better at eliminating jitter, would TOSlink "redeem" itself as a more viable connection given its complete electrical isolation? After all, apparently no matter how jittery the transmission may be (of course we're not talking about losing lock scenarios), the data can still remain bit perfect.

 

CD

 

[noogle1]

@Steve N -

 

Thanks. I didn't realise that USB could achieve galvanic isolation - I assumed the screen of the USB cable electrically connected the PC chassis to the DAC chassis.

 

On Toslink jitter - I believe my DAC (Naim) generates its own low-jitter master clock on board and buffers the data input from the PC, rather than recover the clock from the Toslink data feed.

 

[audioengr]

"I believe my DAC (Naim) generates its own low-jitter master clock on board and buffers the data input from the PC, rather than recover the clock from the Toslink data feed."

 

Not possible. If there is an internal clock that is active, it MUST synchronize to the incoming datastream. This means recovering the clock.

 

I have heard that some DACs use a VCO or two clocks to "bracket" the frequency of the stream by toggling and keeping a buffer half-full, etc.. The problem with these schemes is the large range of the input stream frequency that is required. It's not just a few PPM. It prevents one from using a single free-running clock, which will always have the lowest jitter. Also, switching clocks that are radically different frequencies can have adverse effects on the D/A conversion. I have tried it. If you are going to bracket like this, the oscillators should be tuned to a very narrow range to prevent adverse D/A effects.

 

If the DAC has a word-clock output, then it may be possible to put the DAC in an internal clock mode where the source must synchronize to the outgoing word-clock of the DAC. This is a good technique for minimizing jitter, but depends on the quality of the internal clock in the DAC.

 

Steve N.

Empirical Audio

 

[audioengr]

"given that DACs are getting better and better at eliminating jitter, would TOSlink "redeem" itself as a more viable connection given its complete electrical isolation?"

 

This is a commonly misunderstood notion. It is true that there is some level of jitter rejection in most modern DACs (not NOS DACs), however the rejection even in the best of these is not sufficient to allow a high-jitter input to be tolerated without sonic degradation. It is still necessary to send a DAC a very low jitter signal, whether with S/PDIF, AES/EBU or I2S.

 

Some DAC's use asychronous upsampling hardware to reduce jitter. Contrary to popular belief, this technique is not immune to incoming jitter and these upsamplers tend to add their own jitter as well as distortion of the upsampling algorithm. I have yet to hear any hardware upsampler that is as good as the unsampled data.

 

Steve N.

Empirical Audio

 

[skyline]

"This is a commonly misunderstood notion. It is true that there is some level of jitter rejection in most modern DACs (not NOS DACs), however the rejection even in the best of these is not sufficient to allow a high-jitter input to be tolerated without sonic degradation."

 

Thank you for clarifying this. I have to say in my experience this is completely true. I have a Chord QBD76 DAC that advertises to completely remove data related jitter. But I can easily hear a big drop off in sound quality going from Halide Bridge to toslink on my Mac.

 

[Codifus]

That warped glass analogy from your jitter article is perfect:)

 

CD

 

[Miska]

Toslink from high quality PCI/PCIe/ExpressCard card is in many cases better than USB, exactly for the reason of galvanic isolation from EMI/RFI noise from the computer. Most of the current USB audio interfaces don't seem to properly isolate USB from rest of the electronics, especially from the clocking circuitry. With Toslink the master clock is regenerated within naturally optically isolated device.

 

And also for the reason that PCI/PCIe/EC are technically much better suited for audio, since the card can directly read the samples from RAM, based on it's own clock. No need for all the extra pile of **** from USB packetization and transfer procotol.

 

Even inexpensive cards like ASUS Xonar STX manage to achieve fairly impressive results for the price.

 

Still, neither one is good, I'm still yet to see large scale adaptation of properly made double-fiber (two-way comms) connections. My approach is PCIe -> double fiber -> DAC with DAC running the clock.

 

 

[wappinghigh]

This Jitter and interface talk just goes around and around in circles IMHO.

 

Now whe have a debate going on within other threads as to how you actually measure jitter...and which type you measure and when.

 

How about CA pass the hat around and sponsor a PhD student to sort this once and for all.

 

What we need is proper scientific research with double blind trials, and standardized measuring equipment. Done by people with non commercial interest. And reviewed in peer respected scientific journals. With all due respect to the authors, the "Journal of Stereophile", or "6moons" are not proper peer reviewed scientific journals.

 

So lets get some science onto all this shall we. Otherwise it might as well all be alchemy, as far as I can tell.

 

[Miska]

How about CA pass the hat around and sponsor a PhD student to sort this once and for all.

What we need is proper scientific research with double blind trials, and standardized measuring equipment.

 

AES is usually publishing this sort of material. Pure non-commercial PhD at university would most likely go heavily feets-off-the-ground hands-waving.

 

Usually the useful parts are still company-sponsored publications, made by PhD's or others.

 

There are already bunch of scientific papers on jitter, published by AES. (and getting debated by AES members)

 

 

[wappinghigh]

Would you mind providing the links to the AES. Who they are, who sponsors them, What their journal is. And if they have an annual world scientific meeting. Then forum members might be able to judge for themselves if this body represents a truly impartial scientific organization. And is the world authority on these matters.

 

BTW I strongly disagree about your opinion on non commercial PhD's. This is precisely why we need them

 

Thx.

 

[wappinghigh]

http://www.aes.org/

 

And is their most current statement on Jitter?

 

http://www.aes.org/publications/standards/search.cfm?docID=57 (its dated 2007.)

 

Thx

 

 

[Miska]

this who you mean?

http://www.aes.org/

 

Yes... The mission statement at http://www.aes.org/about/ pretty well describes it. And to not discredit academia, there are lot of people and academic organizations involved in the work.

 

Probably the most well-known Finn there has been Dr. Matti Otala and his work on TIM distortion. (This is how I personally became familiar with the organization.)

 

And is their most current statement on Jitter?

 

I don't think there's such thing, since there are various aspects to the jitter. Some of the papers discuss ways of measuring jitter, others on it's effects on various different converter architectures, and some other different ways of reducing it. And all of these are equally important.

 

 

[noogle1]

@ Steve N:

"Not possible. If there is an internal clock that is active, it MUST synchronize to the incoming datastream. This means recovering the clock."

 

See http://www.naimaudio.com/userfiles/modules/attachment/naim_dac_august_2009.pdf

 

The Naim DAC measures the incoming data rate and selects one of ten fixed-frequency master clocks that most closely matches this data rate. There is a RAM buffer to allow for difference between the two clocks.

 

 

 

[Miska]

The Naim DAC measures the incoming data rate and selects one of ten fixed-frequency master clocks that most closely matches this data rate. There is a RAM buffer to allow for difference between the two clocks.

 

This works quite well if any of the available frequencies cause buffer to under/overflow by less than one sample once per session, which is usually one album. If the buffer is large enough compared to amount of drift, this can be achieved.

 

Not very suitable approach for home theater amp where sync with video is needed, but very much suitable for pure playback.

 

 

[audioengr]

"Toslink from high quality PCI/PCIe/ExpressCard card is in many cases better than USB, exactly for the reason of galvanic isolation from EMI/RFI noise from the computer."

 

Of course. There are many poor USB interfaces out there, just like there are a lot of substandard phono cartridges.

 

Then there are the good ones. I have NO problems whatsoever with async USB interface. Works perfectly. Until you have heard one of these, I believe it is difficult to make comparisons.

 

My biggest problems have been with the playback software that is buggy.

 

Steve N.

 

[audioengr]

Wapping wrote:

"Now whe have a debate going on within other threads as to how you actually measure jitter...and which type you measure and when."

 

I would be happy to share some direct jitter measurements (not through some DAC), but these cannot be compared to indirect measurements through a DAC.

 

This is the problem with these measurements and publishing them. I'm the only manufacturer doing direct measurements, and I measure P-P worst-case across all frequencies. The others are selecting the DAC with the best jitter rejection and then measuring RMS jitter around a single frequency tone. This is apples and oranges.

 

Steve N.

Empirical Audio

 

[audioengr]

"The Naim DAC measures the incoming data rate and selects one of ten fixed-frequency master clocks that most closely matches this data rate. There is a RAM buffer to allow for difference between the two clocks."

 

Okay, so what happens when I play a long playlist and the buffer overruns or underruns?

 

The buffer cannot be infinitely large.

 

Steve N.

Empirical Audio

 

[Miska]

I have NO problems whatsoever with async USB interface. Works perfectly. Until you have heard one of these, I believe it is difficult to make comparisons.

 

A fully opto-isolated async USB perhaps? I still find USB being non-elegant architecture even when it doesn't have impact on sound quality.

 

My biggest problems have been with the playback software that is buggy.

 

It's not any of the parts alone, it's how all the bits and pieces work as whole... I haven't got so much problems with bugs in software, or in hardware. But more like systems not extracting full potential of all the components, ie. system being less than sum of the components.

 

Oh well, I'm a difficult-to-live-with perfectionist anyway.

 

[noogle1]

@Steve N:

 

As long as I don't get buffer overrun within a listening session then I'm a happy bunny. And if the buffer does overrun, the DAC reverts to an asynchronous sample rate conversion.

 

[audioengr]

"As long as I don't get buffer overrun within a listening session then I'm a happy bunny. And if the buffer does overrun, the DAC reverts to an asynchronous sample rate conversion."

 

I see. Interesting approach, although not one that I would pursue. So many clocks would add a LOT to the cost, particularly low-jitter clocks.

 

Thanks,

Steve N.

Empirical Audio

 

[Miska]

I see. Interesting approach, although not one that I would pursue. So many clocks would add a LOT to the cost, particularly low-jitter clocks.

 

Depends if the clocks are programmable or not. Using a single programmable clock synth doesn't increase the cost except for the clock synth itself. All the frequencies then are free...

 

 

 

[noogle1]

AFAIK the number of clocks is not about having a "scatter-gun" spread around the target frequency, but to support multiple different encoding data rates (44.1kHz, 48kHz, 96kHz, 192kHz...).

 

[audioengr]

"Using a single programmable clock synth doesn't increase the cost except for the clock synth itself. All the frequencies then are free..."

 

Sounds easy doesn't it? The BIG problem is that anything other than free-running oscillators have high jitter in comparison.

 

This is why Adaptive Mode USB fell into disfavor.

 

Steve N.

Empirical Audio