Why do digital cables sound different?

Not only do cables sound different from each other, but they are also directional.

OK.....I know some of you are thinking that I have spent too much time out in the hot Texas sun. Yes, I have, but that is besides the point. For many years, I was the "wire and cable guru" in the Advanced Technology Lab for a major telecom outfit. To support this claim, I have provided links to some data on different cables that I have here at the lab.

For those of you who aren't interested in looking, here is the short version:

All cables have impedance perturbations along their length. Where they are (in relation to the load and source end) and their magnitude, along with anomalies were the shield and centre conductor are separated to go into the connector will have *some* impact on impedance. And therefore, how they will measure. Hey......if they measure differently, it is not a hard stretch to conclude that they will sound differently. On SPDIF, this is one place where "If it measures different............."

Granted, the differences, both measured and audible are minuscule, they do exist.

Before I go any further, I should explain something about coax cables in general. A bog standard 75 ohm coax will have an impedance variation of +/- 3 ohms. IOW, the cheap stuff that you may buy at "rat shack". A decent 75 ohm cable will have a variation of +/-1.5 ohms. A precision one will have a variation of +/-0.75 ohms. Most of you will never see one that good. The only place they turn up in my work is for calibration kits for network analysers.

So.....for those of you who want to see what I am talking about, here you go.....

First is the cheap cable that we used to include with our D/A convertor box. The U-Byte 1 cable was not intended to be the greatest thing ever invented (we sold them for $20 for anyone who wanted one), but it outperformed most cables in the<$200 price range. (This was back in '92 or so. I think.......) Anyway, the reason it worked so well was not its absolute impedance (which was is the +/-1.5 ohm range), but due to its 6 m length. It was that ol' "If the reflections arrive after the decision point" philosophy at work.

If you look closely at the graph, you will see how there are minor differences between the green and black traces at the higher frequencies. One is one way.........the other the other way. (Which direction is right?.........both or neither, depending on your outlook in life.)

A further bit on the graphs..........

As the length of the cable (6 m, in this case) becomes significant in relation to the frequency, you get lobing on the graphs. For this reason, using a Vector Network Analyser is not the preferred way to accurately measure cable impedance. That is best done on a Time Domain Reflectometer. If I get the time, I will make those measurements available.

So......it is best to read the impedance at the lower frequencies. In this case, the reflection is -40 dB, and that translates to an error of +/-1.5 ohms.

OK, next is the U-Byte 2 cable. This is a better cable. It has an reflection of -50 dB at the low frequencies, which gives us an impedance error in the +/-0.5 ohm range. Not bad for something that I bought from Belden for a buck a foot. On this one, while the errors are less, they are more apparent with direction. This is easily explained, as a readily available (and affordable) 75 ohm BNC connector was not to be found. I used a good BNC jack that I had to modify (by hand) each and every one, in order to make it fit the Belden cable. Yes, done that way, there is an obvious lack of precision. Still.........it was and is a damn good cable.

*Which is no longer made or available. Please do not ask to buy one, as there are none to be had, except the few that are still floating around in the lab. And no, I am not parting with any. Their inclusion is for reference purposes only.* In fact, we are not in the cable business at all. I will measure any cables supplied for testing, but will not make any, for anyone, at any price.

OK, the obvious question now becomes "What do other cables look like?"

Well, here are two. Not designed for SPDIF, but 2 cables that I have sitting on my bench. So I measured them.

First is a generic RG-59 cable. It is only 1 m long, so there is much less lobing at the higher frequencies (due to its shorter length). The accuracy is around +/-0.85 ohms, which is pretty darn good for something that just happens to be sitting around. If if were a bit longer, it may sound pretty good. Maybe. (The TDR would tell more.)

Then we have a Suhner cable that is clearly marked on its jacket that it is +/-1.5 ohms. It also has a return loss of -45 dB. Better than advertised. However, you will notice that it has less lobing, and is more accurate at the higher frequencies. Hmmm......maybe I should connect a few of them in series and listen to them.................

Well, maybe not. I have other things to do. None of which involve audio.

Hope this gives some of you something to ponder next time you plunk down $$$$$$$ on digital cables.

Enjoy, and Happy Listening!

Hey, U Byte! Oh yeah, U Byte 2! Well I'm thinking about going one box. It's the only solution for this insanity.

A guy who worked for our dealer in Chicago came up with the names. Yeah.....that is pretty much what he was thinking when he did.

SPDIF is a less than optimal solution. To get the jitter on the clock to the level of a one-box solution, it takes a lot of extra parts, in the form of secondary phase-lock loops. And/or feeding the clock back to the transport.

We made one outboard D/A box. Went back to making one-box players after that.

Ar_t...One interesting experience in my career as a systems engineer on Submarine-Launched Ballistic Missiles was a visit aboard a 20 year old boat to evaluate the ability of the existing wiring from tube to fire control (as much as 300 feet) to transmit digital data at the higher bandwidth to be used by a new system. We injected pulses at one end and looked at them with a scope at the other end. My God! Were they ever distorted. All kinds of spikes and overshoots. But, and here is the point, the information transfer over the wires using those sorry-looking pulses was flawlwss.

You have described how the digital waveforms are distorted by improper impedance, stub terminations, etc. but it is still unclear how an analog wareform reconstructed from digital information could be affected by the shape of the digital pulses.

It has to do with the way the clock is extracted from the SPDIF signal. There is a high degree of correlation in it. This leads to a great deal of data-related artifacts in the recovered clock.

(If someone was to hook up some sort of listening doo-hickey to the point in the circuit where the PLL filter is, you will hear a very distorted version of the programme material.)

Any reflections in the data stream manifest themselves into a change in the data-dependent jitter. Not so much in the actual amount, but the frequency distribution. Absolute jitter numbers by themselves are of little good without the corresponding spectral distribution. Close-in jitter, say <10 Hz, is more detrimental than jitter at 1 kHz. So, as the reflections alter the decision point, they alter the spectral distribution.

I know.....a lot of technical mumbo-jumbo, but that is it in a nutshell.

If the clock and data were sent via separate cables, this sort of problem would not occur. Which is why one-box solutions will always be better.

What Eldartford describes is basically a Time Domain Reflectometer. I hope to have some pictures of different cables soon. (I need to construct a hood for my camera, so that I can photograph the screen. The TDR I use was made in '63. Back before they had a data port on the back to pull out the data in a form you can make a JPEG with.)