CAS 3: Clocking & Connections

After last 2 articles, I have gather more questions from audiophiles and hopefully this document can discuss more on these areas. 


(1) I saw there is USB sound card, what is the different between USB and a normal sound card with digital output? Which one is better?

It’s hard to determine which one is better because the details lies on the design. USB is user friendly connection, but in general USB sound devices produces higher jitter than normal digital output. USB relies on ‘slave’ mode. It does not has its own crystal and buffering. Sound card is like a master. It has its own pool to store data from anything ahead and generate its own timing signal for SPDIF, TOSLINK, AES etc with its on board oscillators.


(2) There are great reviews of USB DACs, if your theory is correct, how can that be?

Everything should based on the design. In general, USB connection is more jittery than normal digital audio connection, but jitter can be cleaned/recovered. As long as the jitter is recovered before it gets to the DAC section, the jitter will not be a problem. The most widely used de-jittering in today market is ASRC (asynchronous sample rate conversion). The incoming signal re-clocks with the DAC internal crystal to form a new sample rate. This method isolates the incoming jitter. Anagram Technologies licenses its ASRC to many consumer DACs from Nagra, Manley to Cambridge Audio. Benchmark Media uses ASRC method with very good results also. Another reason that ASRC is widely used in hifi marketing term – “upsampler” is an important selling point.


(3) What is the benefit of upsampling?

A huge audiophiles misunderstanding is “higher upsampling” rate is “better”. This is entirely wrong. Almost all DAC chipsets are worked in Mhz and provides 8x or above oversampling. The reason why upsampling improves quality is based on the custom filter design. It has been proven that this filter design (around Nyquist freq.) affects our hearing the most.


(4) We love high sampling is not because of the extra bandwidth.

Human ears cannot hear frequency above 20khz. High sampling rate recording sounds better because the more relax filter design in the Nyquist frequency. The first famous filter design in high end audio history is Wadia Digimaster. Weiss custom algorithm on Sharc DSP with 40bit floating point produces industry leading performance, which is hard to compare with usual chipset solution.


(5) Weiss Minerva/DAC2 uses firewire IEEE1394 as computer connection, does it works better than USB?

Yes. Weiss High End website has a very clear document for its reasons:

“Why Firewire? Firewire is a peer-to-peer protocol, meaning that every device on a Firewire network is equally capable of talking to every other device. Two video cameras on a Firewire network can share data with each other. A Firewire audio interface could save sound data directly to a Firewire hard drive. Your computer is just another peer on this network, and has no inherent special status.

Firewire is always implemented in hardware, with a special controller chip on every device. So the load it puts on your CPU is much lighter than USB communications load, and you’re much less likely to lose any sound data just because you’re running fifteen things at once. Specialized hardware usually makes things faster and more reliable, and this is one of those times.

But the real reason Firewire is more reliable than USB is more fundamental than that. It’s because Firewire allows two operating modes. One is asynchronous, similar to what USB uses. The other is isochronous mode, and it lets a device carve out a certain dedicated amount of bandwidth that other devices can’t touch. It gets a certain number of time slices each second all its own. The advantages for audio should be obvious: that stream of data can just keep on flowing, and as long as there isn’t more bandwidth demand than the wire can handle (not very likely) nothing will interfere with it. No collisions, no glitches.

From a practical perspective, this also makes it safer to send a lot more audio via Firewire. That’s why most of the multichannel interfaces (16 channels, 24 channels, etc.) are Firewire devices, and USB devices usually just send a two-channel stereo signal.

For hooking up your mouse, keyboard or thumb drive, USB is plenty fast and plenty cheap. For hard drives, either one will do (although Firewire is somewhat more reliable). For audio devices, USB will do fine if no other devices are competing with it and if you have processor room to spare. But Firewire will always be able to handle more load with lower latency and no glitches, because it has resources it can set aside to make sure your audio gets where it needs to go.”


(6) What about WordClock connection?

Some soundcards have wordclock input. It can improve the performance if you slave the soundcard with a high quality master clock. Remember that every digital audio setup should contain 1 reference clock only.

For example, there is a CD transport, an upsampler and a DAC digital playback combo. The CD transport reads the CD data and outputs with 44.1kHz timing information. The upsampler will be slave to this incoming clock signal, and generate a new sampling rate with a new clock signal. The DAC will again locks to this timing information. In this system, the CD transport is the master.

DAC master mode: Some devices allow DAC to work as master clock. The DAC outputs reference clock signal via Wordclock output. The CD transports/upsampler sync with this timing information. This mode produces better quality because of the DAC crystal is used for reference. It is closest to the DAC section (shorter signal path), hence a better result compares with multiply locking stages.

Master Clock mode: If you have a master clock, you can hook it up with all 3 devices. Every device works under this sync should produce more accurate timing.

These are general comments for various connections. It really depends on different design and approach. For example the original Weiss DAC1 has wordclock I/O. But after in depth researches, the wordclock input will never works better than its own DSP reclocking PLL performances. So the workclock input will always produce inferior result. Therefore Weiss decided to take this feature off.


(7) What about atomic rubidium clock? The precision is so many times more accurate than the normal oscillators. 

Atomic clock is highly accurate. It takes 1000 years to shifts a second. There are indeed high end manufacturers use this method. If you look at these atomic clock internal structure, you will find there are quite some heavy PLL sections around. Atomic clock works in 10Mhz output. The master clock that needs to pull this reference frequency back to usable range, is certainly needs some extra devices. In another meaning, the PLL that may clean or pollute the clock source is in the signal path. We can hardly say atomic accuracy provides better result. Actually the result may even worse after those multiply PLLs. 

As you may read from these articles, audiophiles are easily agree on certain audio theory with their traditional audio knowledge. Many of these wrong thoughts have restricted their own potential for higher quality playback. 

The computer technologies has gone mature with digital audio. As of this writing, Apple just launched iTunes 8, which gives much better album display and GUI. As an audiophile, I do wish technologies can bring us higher quality result. For example high resolution 24bit and high sample rate recordings and playbacks. 

It does not has to be more affordable because people who seek for “better” sound should always willing to give “more”. But I wish audiophiles can spend their money on something that really improves their sounding, otherwise our new generation audiophiles will never exist because of the contradictive theory.