The title is:"There's No Such Thing As Digital..."


Subtitled: "A Conversation With Charles Hansen, Gordon Rankin and Steve Silberman". It's an interesting read if you're not yet familiar with this particular topic...or have only considered it briefly. I wouldn't call myself a digital expert, but I can see no reason to quibble with it one bit:

www.audiostream.com/content/draft

Enjoy.
128x128ivan_nosnibor
Can our ears really detect say 400 pico seconds of jitter? If not then even a basic CDP is acceptable.

We can detect perhaps as much as 50ps of jitter while basic CDP can go as high as few nanoseconds.

The reason why jitter is so audible, in spite of small levels, is because it creates sidebands that have no harmonic relation with root frequency. If jitter comes from 60Hz noise then these sidebands will be +/- 60Hz apart from the root frequency and not that audible, but if jitter is caused by higher frequency resulting sidebands will be further away hence more audible. In reality there is some uncorrelated jitter coming from random noise and correlated jitter caused by particular interference frequencies. Also, instead of one root frequency we have whole bunch of them (music) and jitter turns into hash that is proportional to amplitude of the signal (undetectable without signal).

Computer data has no jitter because it has no timing. Data is stored on hard disk without timing. It goes thru all sorts of buffers before it is send out. It can be send out as data when we have wireless or network based DAC but it can also be converted to asynchronous S/Pdif stream. The very moment of this conversion creates jitter.

I also have quibbles with this article. It seems to concentrate on DAC clock, that is usually not that bad, placing less attention to delivery of the signal. It should state that both are equally important.

Assuming perfect buffering of the CD stream signal has to be delivered to DAC with very short transitions to reduce threshold uncertainty - requiring perfect source/cable/dac characteristic impedance match (to avoid reflections on impedance boundaries) or perfectly quiet system with perfect shield on the cable to avoid noise induced jitter when transitions are slow. Since both are, being system dependent, very difficult to do possible solution is to reclock the signal just before the DAC. I have DAC with reclocking built in and it is very clean sounding but Steve found that external reclocking works better.
"I didn't think there was any such thing as 100% jitter free? I thought all timing sources had jitter?"

There is no such thing. Marketing BS and an outright lie.

"At what point does that become audible? Can our ears really detect say 400 pico seconds of jitter?"

Depends on the system. Resolving systems can easily demonstrate difference between 20psec and 100psec of jitter, and it's not subtle. Resolving means ultra-low noise floor and distortion. Usually no active preamp to achieve this.

"I also have quibbles with this article. It seems to concentrate on DAC clock, that is usually not that bad, placing less attention to delivery of the signal."

That is interesting, given that many DACs don't have an internal clock, except maybe for the Async USB interface master clocks. These are usually the important ones. This is where the jitter starts in a USB system. It will even have an effect on additional reclocking.

"I have DAC with reclocking built in and it is very clean sounding but Steve found that external reclocking works better."

That is primarily due to the separation of power systems, putting the master clock on its own power system, separate from the DAC circuits. If you can do this effectively inside the DAC, that is fine too. Pretty awkward to have two power cords though...

Steve N.
Empirical Audio
That is interesting, given that many DACs don't have an internal clock, except maybe for the Async USB interface master clocks

Steve, you're missing all network DACs including Ethernet, Firewire, Wireless etc. You also forgot about asynchronous reclocking DACs.

Typical DACs contain Phase Locked Loop (PLL) that contains adjustable oscillator and phase detector. Phase detector compares average phase of incoming signal and adjusts internal oscillator (a clock) to match it. DAC is clocked from this adjustable clock and not from the input signal. Quality of this internal clock (jitter) is very important.
"Steve, you're missing all network DACs including Ethernet, Firewire, Wireless etc."

These are usually not called DACs, but music processors or servers etc., but you are correct.

"You also forgot about asynchronous reclocking DACs.

Typical DACs contain Phase Locked Loop (PLL) that contains adjustable oscillator and phase detector."

These may be typical, but I didn't forget about these, I totally ignored them. I don't even consider resampling DACs because the effect of resampling is so damaging to audio quality, even using a high-quality clock, which most don't have. I would never design such a DAC. Once you have this, no matter how good the input source is, the internal clock and resampling wrecks it IME. I have modded more than a dozen such DACs in the past 12 years, so I know how this sounds.

Steve N.
Empirical Audio
You probably think about oversampling and that is different. Pretty much every CDP contains PLL. It can be used straight just to suppress jitter providing stable clock (by means of averaging) or can be used to create oversampling. Such oversampling is done by comparing in phase detector incoming stream to divided down frequency of higher frequency internal clock. Because this division is integer these oversampling DACs always operate on multiples of incoming frequency while upsampling DACs can work on pretty much any non-integer ratio. Usually PLL is inserted somewhere. Even asynchronous rate converter based DACs (upsampling DACs) like my Benchmark have some form of PLL to make signal stable enough for upsampling. Such PLL is fast responding and single stage while most of CDPs have dual stage PLL that operates at different time constants.