Find Source Re Stabilizing Data Stream

[canoosa]

Good Morning Everybody.
It was at least 6 yr. ago that I first came across Amir's mention of this aspect (paraphrased), regarding transmission of computer-based audio data:

there is a minimum length of cable (5' or 1.52 m) required to stabilize the stream due to inherent jitter-generating reflections within shorter lengths.

I've proven the perceived truth of this (to myself, via listening) a number of times, but now need to be able to quote a specific, scientific source for this fact.
Can anyone point/link me to such a definitive source (Amir is probably/hopefully still sleeping at this time)?

Best Regards.

 

[DonH56]

That does not make sense to me but I can think of a couple of things that might be going on:

1. If source and load are not matched to the cables, i.e. properly terminated, then reflections can corrupt the data stream. The optimum length to avoid the impact of reflections depends upon the data rate.
2. Longer cables and poor (or just mismatched a little) terminations result in cable parasitics (primarily capacitance, resistance, inductance) rolling off the signal bandwidth. This can reduce the impact of fast edges and the resulting reflections. However, longer cables also increase ISI (intersymbol interference) which degrades the data stream.

Either of those could cause "jitter-generating reflections" but it's hard for me to believe it would be bad enough, or the receiver poor enough, to result in bit errors. But some audio data schemes do not include error recovery so I suppose anything is possible.

I think there is an article in the technical area about transmission lines.

Assuming the system can properly recover the data there should be no audible change from a short vs. long cable assuming the cable is not long enough to cause data loss (e.g. signal too small for the receive to recover, or introduces significant jitter and duty-cycle-distortion, which may cause bit errors or clock errors if the recovered clock drives the DAC).

I think you might want to arrange a blind test yourself before declaring it as fact, though ideally such a test would use several different devices for Tx and Rx to see if it is a general rule or a function of particular components. Components with poor terminations and/or data recovery schemes would be more sensitive to reflections and other signal perturbations of the data stream.

 

[canoosa]

Good Morning Don,

Thanks for the reply!
As I recall, improper terminations were not specifically mentioned by Amir at the the time on WBF - but memory CAN play tricks .
Termination impedance was something I learned about later, when a "Company insider", also on WBF, mentioned that there were a large number of brands whose connectors he tested, only to discover that none of them even came close to their nominal value; I think it was also at that time that you mentioned how very expensive the connectors are that you use at work, specifically because they had the proper ohm value.

Here is my case: inserted a Cybershaft word clock into the system, driving an Antelope Zodiac Gold. Did not have a 50 ohm cable to begin with, and accepted the statement on the Cybershaft site that using a 50 ohm cable to connect to a 75 ohm component made no appreciable difference in audio quality. To be fair, however, Kenji Hasegawa (Cybershaft owner) is a true scientist and open to discovery/learning. Although puzzled by the fact that Kenji only offered 0.5 m and 1 m 50 ohm cables (puzzled because I remembered Amir's statement), I purchased the .5 m cable (Teflon coated, near as I can determine), thinking that because the signal per se is only a 10 MHz pulse, no audio data is involved, and length of cable might not be an issue under those circumstances.
After letting the clock and its .5 m cable run for quite a large number of hours, I sat down to listen and noticed significant, obvious benefits in the reproduction quality. A few months later, I was given the opportunity to exchange the short cable for a 6' custom-made Kubala-Sosna 50 ohm cable. The increase in quality was obvious. And yes, there is the additional variable introduced which I have no way of evaluating, i.e. the K-S "quality" vs. the short cable "quality" - within the framework of solely a pulse transmission, not to mention a possible/probable difference in the quality of connector used on the K-S..........

However, it bears repeating at this point, that the cable still connected to Antelope's BNC connector, a connector whose quality I was forced to assume was of the normal, generic, "poor" type, given that another type of connector on that DAC I KNEW to be exceedingly cheap.
Decided to investigate the Cybershaft statement about the 50 to 75 ohm negligibl- impact transfer.
Purchased a Pasternak PE7007 [75 Ohm BNC Male to 50 Ohm BNC Female Matching Pad Operating From DC to 2 GHz RoHS Compliant].

The Results On Always-Used Test Tracks

The attack of notes was OBVIOUSLY much better regimented: the initial thought was that I had inadvertently turned up the normally-used test volume by about 1 dB, possibly 2. Turns out this is the perceived result of increased clarity/definition.
Notes on album tracks (Red Book cd's) that were always poorly presented now stand out clearly.
The benefit is universal, i.e. the sound-reproduction quality of all types of music is improved, albeit more evident for acoustic instruments, presumably because the signature of electronic instruments' notes is ? "broader" (for lack of a better word), more "distorted".

Yes, I agree most whole-heartedly that proper impedance matching is priceless! However, as already mentioned, there was also a perceived benefit going from a short 50 ohm cable to a 6' 50 ohm cable, and in terms of what Amir had pointed out years back, I would also like to refer to that parameter (and its presumed importance) when discussing the Pasternak PE7007 within the above context. Hence my question about a source or link to a scientific test of that length vs coherence parameter, if such actually exists.

Cheers,
Ingo

P.S. The system here: APC 1000 > Balanced Transformer (Toroid B2000) > home-built server using a great deal of interior shielding including a Lynx AES 16e > xlr breakout cable > Antelope Zodiac Gold with Antelope Voltikus p/s > Focal SM9 (via xlr)

 

[DonH56]

I have not posted on WBF for a while. Not banned AFAIK, but my last few PMs went unanswered, and the last couple of posts I made were deleted, so I took the hint. I believe Amir is banned.

I remain skeptical of sighted listening tests, mainly because I have fooled myself so many times in the past. Maybe that makes me unscientific, but OTOH open-minded can go too far as well. Opinions vary. If the cables make a difference it is almost certainly measurable, both in terms of clock performance and the accuracy of the output bit stream. Triple-transit times and transmission line effects are well-known. In free space the wavelength of a 10 MHz clock is about 100' (30 m) so there is very little a connector can do in a fraction of an inch. If there are problems I would suspect the impedance of the source and load more than the connector itself. Someplace I have simulations showing what discontinuities can do to a clock signal; it can be a problem, especially if there are stubs involved (caused big problems for a 100 MHz reference clock on an eval board), but that reflects a bad design.

I had some bad experiences with Pasternak awhile (years) ago so do not use them. Fairview Microwave is one source for inexpensive RF passives. Huber-Suhner is a source of high-quality connectors, but really even a bad BNC at 10 MHz should not be an issue -- the frequency is low enough that a discontinuity would be too short to matter. One caveat is I do not know the edge speed of your clock source, and that impacts how large a signal glitch can be generated.

At this point I'll let Amir chime in if he remembers the article; not sure I have much to contribute.

 

[Vincent Kars]

Reminds me of this: https://positive-feedback.com/Issue14/spdif.htm

 

[DonH56]

Reminds me of this: https://positive-feedback.com/Issue14/spdif.htm

Looks like it, thanks Vincent!

Note that the transit time and relation to signal frequency means different lengths will cause the signal to behave differently in the presence of a discontinuity. A 1 m cable might work well, a 2 m not, then a 3 m may work well again, etc.

 

[Vincent Kars]

Note that the transit time and relation to signal frequency means different lengths will cause the signal to behave differently in the presence of a discontinuity.

You won't have this problem if you buy Steve Nugent's Bitmeister SPDIF cable at only $630

 

[RayDunzl]

Note that the transit time and relation to signal frequency means different lengths will cause the signal to behave differently in the presence of a discontinuity. A 1 m cable might work well, a 2 m not, then a 3 m may work well again, etc.

So, we need stretchy cables now?

 

[canoosa]

Thank you both!
And Vincent: certainly the time of that article would easily fit into the scheme of things, something that Amir might have been referring, to many years after that was posted, while trawling through his memory banks!

Don: thanks for the recommendations as to other sources of supply, much appreciated.
And I certainly see that there are numerous other variables involved here, esp. the integrity of the 75 ohm path within the dac, beginning with its connector!
As for the "edge speed" of the clock, I'm unaware (polite way of saying ignorant) of what that refers to, although that might be the rise time referred to in Vincent's link.

The Certificate of Calibration that comes with the Ultra Precision OCXO Clock reveals due diligence as well as sophisticated/proper instruments, which include:
Agilent 53132A; symmetricom TSC5115A; Agilent E4402B.

Parameters Tested/Listed:
Frequency Value - 9,999,999,999.63 MHz; Allan Deviation - 1s=3.17E x 10 (to the 13th), 10s=5.77E x 10 (to the 13th); Output Level under conditions of Res. BW 100Khz VBW 3 KHz Span 0Hz; Harmonic Distortion 20 MHZ=40.10dBc, 30 MHZ=46.91 dBc, 40 MHZ=75.21dBc; Phase Noise 1 Hz=111.7dBc/Hz,
10 Hz=133.3dBc/Hz, 100Hz= 142.5dBc/Hz

Another heightened quality revealed after the insertion of the PE7007, but implied by the greatly improved attack times, was that ALL notes are more clearly revealed on an individual basis; I thought the system was pretty good before, but using bluegrass played at "lightning speed" as a useful indicator, those notes are even clearer now, and noticeably so.

I, too, am hopeful that Amir might be here this weekend, but agree that the positive-feedback reference, or something similar, is the most likely underpinning of Amir's original statement.

What is a very welcome aspect of the PE7007's contribution is the relatively huge return on investment on the $51 cost of the unit

Cheers!

 

[canoosa]

One Last Thing: it just occurred to me, after reading the PFO article that what Nugent is referring to, and most likely what Amir was also referring to, is an SPDIF cable (with its inherent compromise), whereas I am dealing with balanced cables (whether 50 or 110 ohm, or a supposedly balanced 75 ohm input on the dac. Somewhere along the line, I seem to have done a bit of mental fusion and lumped ALL digital cable-types into one category that makes them all vulnerable to data speed and cable length interacting variables....no wonder this "didn't make sense" to you, Don !!!

Best Regards,
Ingo

 

[DonH56]

S/PDIF uses single-ended (unbalanced) 75 ohm links and RCAs; AES3 (the "pro" version) uses 75 ohms unbalanced and BNCs, or balanced 110-ohm cables and XLRs. S/PDIF also has an optical (TOSLINK) option. Some proprietary interfaces may use different impedances... RF clocks are generally 50-ohm single-ended or 100-ohm balanced though I have not looked at clock modules for audio stuff lately.

Any signal will be sensitive to cable length and termination; it is usually a non-issue at low frequencies, however.

 

[Speedskater]

Decades ago, the industrial audio consult Dick Pierce noticed that some high-end digital audio components with poorly designed output and input stages were very sensitive to digital cables. One of those components was what Stereophile used to rate cables.