In article , Lou Anschuetz
writes:

OTOH, I can understand how he comes to the conclusions he
comes to. As someone with 25 years in IT, I have thrown away dozens
of supposedly competently designed SCSI/serial
cables that don't work. SCSI cables in particular are darn
hard to to get right for the newer higher speeds - and it
all seems to come down to how the cable(s) is insulated.

I also agree that this doesn't necessarily have anything to do
with audio transmission. However, IMHO, this business about
insulation continues to bedevil me. Since it does play havoc
with digital signal transmission, it seems to me hard to leap
to an absolute conclusion that it has no effect in the analog
audio world. No hard science on this yet, but it is a very active
field of research...

Well, I'm on a plane flying from Beijing to Hong Kong (and on to Melbourne,
Australia) listening to Chinese Pop music in my headphones, and nothing to do,
so I thought I would just jump in here!

Why do some cables work on signal X but not with signal Y? The answer is
simple: wavelength. (The answer is simple, the explanation is a little more
complicated.) Every signal, data, audio, video, etc, etc. occupies one or more
frequencies. Each frequency has a wavelength, that is, there is an
electromagnetic wave moving down the cable that has a specific length. This
can be calculated by dividing 300,000,000 by that frequency. (The answer comes
out in meters, so us backwards USA types have to multiply by 3.28 to get to
feet.)

If you look at 20 kHz (or you can pick any frequency you want), that gives you
a wavelength of 15,000 meters (about 9 miles). What that means is that the
impedance of the cable (how it reacts to frequencies) is of no consequence.
Most engineers agree that the ACTUAL critical distance is 1/4 wavelength which
is, in this case 2.25 miles, still too far to have any effect. This distance is
also affect by the plastic around the conductors. This affects the speed
"velocity" of the signal. (Signals only move at the speed of light in a vaccum,
i.e. outer space. If you ever have a chance to check out your speakers there,
be sure and give me a complete test report!)

Let's say your interconnect cables were the worst ever made (a "velocity of
propagation" of 50%). You would multiply the wavelength by 50% so we're down
to 1.12 miles. Anybody with one mile speaker cables? data cable? mic cables?
video cables? audio cable? Well, yes, there are people with AUDIO cables of
that length or more. They call themselves the TELEPHONE COMPANY. And the
maximum distance between your home and your central office is 13,000 ft. (over
2 miles).

But the twisted pairs they use (even fancy new ones, Category 3) are not the
correct impedance (too expensive). Luckily, the audio on your phone ends at
3500 Hz. You can calculate the bandwidth and see you can go VERY far before
you need to have impedance-specific twisted pairs. I'll be honest and tell you
that I don't know the actual occupied bandwidth of a 56k data signal. (I think
it's NRZI Manchester Coded, that would make it 28kHz bandwidth.) I'm sure you
note that this puts it on the edge for performance if you are far from the
central office, and you can't always get dialup that fast. And you'll
understand why in the USA, the FCC now mandates that all telephone wiring must
be Category 3 or better. Category 3 is impedance-specific (100 ohms) data
cable. But there are other reasons the FCC made this a standard besides
distance. The signals either go back to the source or simply stop in the cable
(called "standing waves") and then radiate all that reflected energegy into
cables around them. That's why the phone company is trying to use data cable,
to avoid "alien crosstalk" (between cables).

That's why you can use crappy interconnect cables that come free with your
receiver and it all sounds pretty good. They aren't long enough to make a
difference. As long as you have continuity (electrical signal flow), you'll be
just fine. But what if you use that cable for DIGITAL (i.e. S/PDIF). Suddenly
we're sampling that audio at 44.1kHz (like a CD), and the bandwidth is defined
as 128 times the sampling (5.6448 MHz, if memory serves). Let's just use 6 MHz
to make it easy. (If you don't want easy, do it yourself.)

300,000,000 divided by 6,000,000 = 50 meters. 1/4-wavelength = 12.5 meters or
about 44 feet. How long is your cable? 3 ft? 6 ft? No problem. It's still
not LONG enough to make a difference. Now, if you want to send S/PDIF to the
other end of the house, that might be 40 ft. or even more. What would you do?
Be sure and use cable which is the correct impedance. S/PDIF requires 75 ohm
cable (75 ohm cable has the lowest signal loss "attenuation" so it is used for
many applications where low loss is required).

What if you get a really long piece of crappy cable? What impedance is it?
Who knows!! And what will happen is that the signal, which will require a 75
ohm "transmission line" and not see it. The farther it is away from 75 ohm the
more the signal will be REFLECTED back to the source and not get to the other
end. This is called "return loss" in the cable world. You send me the actual
impedance of your cable and I can calculate the mismatch and the return loss.

But, if you get the right 75 ohm cable, how far can you go? It depends on the
size of the wire (gage) in the middle. Let's assume it is a 20 AWG center
(such as Belden 1505A, the world's most popular coax cable and a nice generic
size).. Then the cable could go 716 ft. (This is based on a source voltage of
0.5v and a minimum received voltage of 0.2v, at a bandwidth of 5.6448 MHz.)
Yup, 700 feet!!! (Snippy aside: try going 700 ft. on a piece of toslink
plastic fiber!)

And in the digital format, you can be real close to the maximum and the signal
(i.e. data/audio) sounds just perfect. But add in one patch cord, or one less
than perfect connector, or just a few more feet of cable, and you could be down
the slippery slope of the "digital cliff". So, how many of you have to run 700
ft.??? Not many, I assume.

This is not to say you shouldn't have cable with low capacitance. Digital
signals need low capacitance to keep their sharp edges (i.e. the clock). But
(amazing fact that no cable manufacturer ever told you) once you choose the
construction (i.e. plastic/dielectric/velocity) and the impedance (i.e. 75
ohms) the capacitance is AUTOMATIC. High velocity foam coax 75 ohm = 15 pF/ft.
Old solid polyethylene 75 ohm coax = 20 pF/ft. And this is why you should
avoid CATV coax, because it's not all copper in that center conductor (it's
copper-clad steel for high-frequency-only applications where only the skin of
the conductor is working.) This means the resistance is 5 to 7 times higher
than an all-copper center and the distance you can go is 5 to 7 times shorter.
(OK, we're down to 100 ft. and you're going 6 ft., but I wouldn't use it.)

And all those SCSI cables (at the beginning of this thread) suffer from the
same problem. Most of them use generic multiconductors (just pull one apart if
you don't believe me). The impedance of a pair of wires is determined by their
size (gage), the distance between them and the quality (dielectric contact) of
the material inbetween. So, while the plastic is important, these cables fail
at high frequencies because of the DISTANCE between conductors, which varies
all over the place. Open up a SCSI 2 or 3 cable, what do you see? Ahhhh,
twisted pairs!!! And the better the pairs (tighter impedance specs) the less
the mismatch, the less the return loss, the greater the received signal
strength, the less the bit errors. Flat cable can also help maintain impedance
specs, at least better than just a bunch of loose wires.

As frequencies go higher and higher, every dimension becomes more critical.
Just take your car on a racetrack if you don't believe me. You need a race car?
Buy a race car! You need a high-frequency cable? Buy one made, tested,
verified, and guaranteed for whatever frequency band you wish.

OK, class, how did I do?

Steve Lampen
Belden Electronics Division