Digital XLR vs. Analog XLR - Balanced Cables

Kijanki, I hear you. But one of my concerns is with many of these cable companies is when does the marketing end and the real engineering begin. Even with digital cables, there are certainly some cable manufacturers who clearly point out that the cable is a "true 75 ohm" or "true 110 ohm" cable while many others just call the cables digital cables.

Ckoffend - I would say that most, if not all, digital cables have particular targeted characteristic impedance - in case of audio 75 ohm for unbalanced and 110 ohm for balanced. But just look at typical 75 ohm video coax - it is so much different than analog cable. First of all analog cable uses separate wire for the ground. Carrying ground thru shield is really bad idea and grounding shield at both ends is even worse (XLR is but it was mistake). Metal is important only on the surface (plated) because of the skin depth at these frequencies. Lower dielectric constant is important but not as much as in analog cables. Dielectric constant of polyethylene, most likely used in digital cable is 3.3 while foamed teflon is getting close to 1.5 (and oversized air tubes can bring it even bit closer to 1)).

I understand the need for exploration and experimentation but when I buy cooking oil I am not tempted to try motor oil instead just because there is zero cholesterol and no saturated fat in it(and no other parameters to differentiate them). It was designed for the car and I have no reason to question it. It is a matter of taste, so to speak, but we can get easily lost here since changes between cables are very small. Going by the book shields us from many mistakes.
Jitter will transfer from time domain as a noise. It won't change the sound other than making background less black. That was what I noticed with jitter rejecting Benchmark. Its jitter bandwith is in order of few Hz and at frequncies of interest (kHz) gives -100dB rejection of the noise that was at -80dB to start with - practically complete rejection. Cables here don't make any difference - similar with your Purcell but if you have instead of upsampling DAC oversampling DAC or even NOS DAC than digital cable will make huge difference.

There is an excellent article in Stereophile (available on line) on the jitter explaining how sidebands are created, why they are audible and showing everything in numbers with typical transport/CD. Just educational - you can use any cable with Purcell (if I understand it right)

As to characteristic Z and BW: First, the reason to set a characteristic impedance of a cable is to reduce transmission line effects. T-line effects amount to standing waves. These only become important when the wavelength of the signal approaches the length of the cable – how far the signal has to travel. So, whether or not a given and specified characteristic impedance of a cable will matter depends on the cable length. So does the bandwidth of the cable, for that matter, because the total capacitance is determined by the length of the cable. As between the two, as I will explain below, the bandwidth is going to be more important, for the lengths we are talking about.

As to the reference to the Stereophile article – not exactly a reference that is going to add validity to a technical position when posing such position to an engineer. Next time try something a little more accepted in the scientific / engineering community – such as an IEEE journal, or even something published by the AES or the ARRL.

As to the transmission line effect, we are talking about interconnects here. I made the assumption that the lengths are somewhere in the neighborhood of less than 10 feet. T-line effects only kick in when the wavelength of the highest signal component approaches the length of the cable. Standing waves, if they are present, will tend to round off the edges of the square pulse, this is what causes the jitter due to T-line effects. The purpose of selecting the characteristic impedance to match the source and the load impedances is to get rid of T- line effects.

The highest signal component in the case of digital audio will be about 10 times the fundamental frequency of the signal because at that frequency you have a nicely shaped square wave.

The wavelength of a 100 MHz signal is just under 10 feet, so you really aren’t getting T line effects until you approach that cable length, if we are talking about a signal with 100 MHz components. A safe rule of thumb is a 1 to 10 ratio, so there one could argue that to completely eliminate the possibility of T-line effects the cable should be less than 1 ft long. However, the transmission rate of digital audio at a 96 kHz sample rate isn’t 100 MHz. If you go out two decades, you are still at only 10MHz, which is a wavelength of just under 100 ft. Hence, a 6 foot interconnect will not be a source of jitter due to T-line effects.

More likely (but still not very likely) is that the rounding of the pulse will be due to bandwidth limitations. If the cable has too high of a capacitance value, it is possible to create a low pass filter that will start rounding the square wave and create jitter. The chances of that happening are also slight at the lengths we are talking about, but more likely and does not depend on the creation of a standing wave. For that reason the bandwidth of the cable is more important. A subtle difference, but there is a difference.

But, on a practical side – it just doesn’t matter – a cable made for analog transmission will work find up to about 50 feet and most interconnects for home audio are not that long.

Remember also that jitter only becomes a problem at the conversion. Circuitry at the convertor should reconstitute the clock and reject jitter that is not extreme.

Is it a big deal, no – not if you are purchasing the IC’s new, I haven’t priced digital vs analog IC’s but there is no reason that one should be significantly more expensive than the other. Furthermore, since the 110 ohm low capacitance cable is not going to cost significantly more for 6 foot lengths and it will work just as well for analog, my guess is that reputable sellers simply make up all their cables out of the same cable and connectors and just charge a small amount more to sell you one rather than two cables; i.e. $ 60 /pair vs $35 each. Not unfair.

Whenever an electromagnetic wave encounters a change in impedance some of the signal is transmitted and some is reflected (impedance boundary). Reflected signal creates all sorts of shape distortions making overshoots, oscillations and staircase (Bergeron diagrams). Rule of thumb says that you can consider that line (cable) is in the low frequency domain when trise>6t where t is line delay. Signal travels thru conductor at about 70% of the speed of light making 1m in 4.8ns. Multiplying this by 6 gives us 29ns. for 2m interconnect it will be 58ns and for 3m it's 87ns (50ft would be disaster - 438ns) . Most of the output drivers switch below 29ns (much less 438ns)therefore we have transmission line effects. Selecting slower driver by designer wouldn't do any good because it creates noise induced jitter on the receiving end. Receiving end has either asynchronous reclocking in upsampling DACs or dual PLL in the rest of them. PLL, even dual, works poorly for fast jitter.

I still recommend Stereophile article - it might be not up to your standards (as an engineer and/or scientist) but at least it is not as boring as IEEE stuff and one can even understand it for a change. And it is audio related - have I mentioned that?

Funny, I have found IEEE journals to be understandable. As to the attempt at an explanation in your last post- it makes no sense.