Jitter and 75ohm cable length

"The Stereophile test showed it to be awful."

Not exactly. High jitter appears on analog outputs only. Digital is about 10x better (258ps vs 2400ps) and that's what Stereophile stated:

"The noise floor has dropped by 4–5dB, the word-clock jitter to a respectably low 258ps, which is actually better than the case with the standalone D/A processor driven directly by my PC's S/PDIF output (provided by an RME PCI card).

Considering that the AirPort Express's analog output is basically a freebie function added to a computer Wi-Fi hub, jitter aside, its measured performance is quite good. The beauty of this unassuming component, however, is its S/PDIF data output, which allows the AirPort Express to assume a respectable role in a true high-end audio system."

Rpg,

Jitter is basically a noise in time domain. Applied to one frequency it creates sidebands at very low levels. In spite of low levels (less than -65dB) sidebands are audible being not harmonically related to root frequency. Now, take whole bunch of frequencies (music) and you'll get whole bunch of other frequencies at very low level - basically a noise. Amplitude of this noise is straight proportional to amplitude of music and without music (gap) is zero - therefore undetectable.

My first impression of Benchmark DAC1 that suppresses jitter was that sound was too clean (some people call it sterile or analytical). I had impression that some instruments had to be missing from the recording. I also understand that noisier (or distorted) signal sounds more lively the way that distorted guitar sounds more dynamic than clean jazz guitar at the same volume. Other than that sound is on neutral side - I would not attribute any warmth or lack of it to jitter. Imaging is more focused but perhaps a little narrower.

I don't see warmth as desirable quality. Benchmark technical director John Siau said that overly warm gear can negatively affect sound of instruments with complex harmonic structure like piano making it sound almost like out of tune. On the other hand cold sounding (expanded odd harmonics) gear is much worse. I had problem of brightness until I replaced speakers with aluminum dome tweeters. New soft dome Hyperion HPS-938 are wonderful - neutral and never bright on any CD. Sibilants are still very audible but always clean and natural.

I think that main difference between oversampling or upsampling DACs and NOS DACs is not the sampling itself but filtering. Traditional linear filtering adds pre-echo to impulse. Our hearing is very sensitive to it and getting rid of filter altogether (NOS) or using apodizing filter (extending post echo) might be a good thing. Here is some info on the subject:

http://mrapodizer.wordpress.com/2011/08/16/technical-analysis-of-the-meridian-apodizing-filter/

Thank you Al, and thanks for the links especially the second one. I suspect that balanced cable might be superior in noise rejection simply because of common mode rejection on the receiver side but also because wires are twisted what might be superior to any shielding. Twisting wires exposes them evenly to interference (capacitive or electromagnetic) leading to cancellation while shielding have serious limitations. I mentioned it before, but non magnetic shielding does not protect against EMI (magnetic in nature) but fortunately induced noise travels (to ground) on the outside of the cable - shield, because of skin effect. This beneficial skin effect is good at high frequencies but less than perfect at the lower frequencies where cable is still long enough to become effective antenna (antenna is practically ineffective below 1/10 of wavelength). Understanding of this should lead to understanding that shortest cable is the best cable. Statement that digital cable should be at least 1.5m is not complete - it should state instead "as short as possible but not shorter than 1.5m". In ICs (or speaker cables) twice shorter means twice better. Sales people often recommend 1m IC vs 0.5m IC because that's what they have in stock - absolutely no other reason.

Let me add to Al's great post. As he stated, typical transport has transition times in order of 25ns. Threshold resides most likely at the half of that - 12.5ns while impedance boundary, that causes reflection will reside on the other end of the cable. Sharpest slew rate change, causing reflection, is usually at the very beginning (knee). From that point signal travels forth and back (reflection) over distance of 2x1.5m=3m with speed of about 60-70% of light speed - let say 0.2m/ns. Reflection will return in 15ns missing time-wise threshold point. I would use 1.5m-2m length or less than a foot where transmission line effect is non-existent yet. Rule of thumb says that we're dealing with transmission line when transition time is less than 8 times propagation delay (one way). It would imply that typical 25ns transport digital cable becomes transmission line when propagation is longer than about 3ns being equal to about 0.6m. It sounds strange but good cable should be very short or 1.5m-2m. When transitions are slow we don't have much of reflection induced jitter problem but rather noise induced jitter (noise affecting threshold point). When transport has fast transitions noise induced jitter is reduced but reflection induced jitter is dominant requiring very good cable. Long cables in addition add to noise pickup so whole thing becomes system dependent. Same cable might sound great with one system but no so great with the other. Many people report better results with Toslink, in spite of slow transitions, perhaps because of noisy environment or ground loops that coax might create.

Magfan, I commented only on Stereophile findings. AE sounds very clean in my system but I use it with Benchmark DAC1 that is jitter suppressing. It is possible that your AE is bad but it is also possible that one sent for evaluation to Stereophile was extremely good (selected?). It is even possible that Stereophile measured wrong - who knows.

Magfan, CD data stream is asynchronous. It is also jittery because of less than perfect CD printing and reading plus quality of the transport and system noise. What is needed to reduce jitter is either to create stable clock for D/A converter based on average datastream rate locking both with PLL (Phase Lock Loop) - solution used in most CDP or ignore completely datastream rate and reclock it with fixed stable clock in Asynchronous Rate Converter (Benchmark DAC1).

We can add to this jitter introduced in A/D process, that cannot be removed no matter what you do. At the very beginning a lot of analog recordings got digitized with less than perfect (jittery) clock and the only way out is to digitize it again if analog master tapes still exist.

The cable length of 0.6m without transmission line effect, that I calculated, applies to 25ns transition time assuming that driver delivers constant slew rate. There are drivers that do that but very often leading "knee" has higher slew rate. Because of that I would perhaps limit such cable to half of that (0.3m). Above that careful matching of characteristic impedance is recommended. This characteristic impedance has very strange definition. It is impedance of infinite cable or finite cable terminated with its own characteristic impedance - which sounds a little like Catch22. For all practical purpose it is simply SQRT(L/C) implying particular geometry.

I'm not sure what happens with balanced cables. Impedance is 110 ohm and voltage levels are much higher but at the same time slew rate is likely higher and reflection induced jitter taking over noise induced jitter. Maybe Al can help here?