[sdiy] S/PDIF cable.

[Magnus Danielson]

From: "Batz Goodfortune" <batzman at all-electric.com>
Subject: Re: [sdiy] S/PDIF cable.
Date: Sun, 28 Apr 2002 03:06:02 +0930

> Y-ellow All.
>          And thanks to everyone for that. Yeah. Sounds like 75ohm coax is 
> the go here. I was kinda hoping the 50 would have been suitable simply 
> because it's a narrower diameter but it really isn't overly traumatic. I'm 
> not exactly spoilt for choice at the moment.

S/P-DIF according to IEC 60958-3 (Consumer variant) shall have a cable
of 75 Ohm +/- 35 % in the frequency range of 0.1 to 6.0 MHz. This is
not far of the mark for consumer video, which is probably the
intention. The output impedance shall be 75 Ohm +/- 20% for the same
frequency range while the input impedance shall be 75 Ohm +/- 5% for
again the same frequency range. The rise and fall times of the
transmitter is only given as maximum values, and they may be at
maximum 0.4 UI. A warning is that too quick rise-times cause more
trouble than they solve, at least most of the times - you just beg for
more reflections and more ringings - neither is desireable.

> At 09:01 AM 4/27/02 -0700, John L Marshall wrote:
> >Use the 75 ohm coaxial. Do you have RG-59 foam? That would be my first
> >choice.
> 
> I always thought RG59 was thinner than this stuff but I don't have much 
> choice. The sample I have here doesn't have anything printed on it so it's 
> unclear what it is. But it IS 75ohm. It's actually designed as aerial cable 
> but it works OK for video. I have surveillance cameras running on 20 metres 
> of the stuff with out any problems.

20 meters, your maximum reflection time round-trip would be about
2*20/0.2 ns = 200 ns. This isn't more than 200 * 0.0135 = 2.7 pixels
in standard PAL or NTSC television systems. You have to have a sharp
edge to see any mirroring effect, and I doubt that you will. Just a
reality check...

The risetime of a S/P-DIF may be something of 0.1 UI, which turns out
to be 0.1/(2*32*2*48kHz) = 0.1/(6.144 MHz) = 16.27 ns. At it's
slowest, it can be 0.4 UI, which is 65.1 ns. A 20 meter cable-run is
certainly longer than the rise-time, even at it's slowest. A 5-meter
run is about the same length as the slowest rise-time, but to be able
to ignore it, you need to be much shorter. A slow-risetime source can
have a pretty bad cable, but only if you keep it below half a meter,
since then it is too short to cause much trouble. A faster source will
however not like this cable as much. I.e. in effect you want to do
"the right thing" all the times.

> >Transformers exist to convert from AES/EBU 110 ohm balanced to 75 ohm
> >unbalanced. The unbalanced lines can run long distances.
> 
> The problem at that end is that the only option on a Yamaha O1V is a 
> 4in/4out AES/EBU podule. I'm not even 100% sure it won't spit the dummy but 
> since S/PDIF is a superset of AES/EBU (logically) and none of the 
> extraneous bits will be sent anyway, it should work.

S/P-DIF isn't that different from AES/EBU. The AES/EBU just use the
bits differently. I'd rather look at it as S/P-DIF being a partly
incompatible variant of the AES/EBU.

> And Biphase mark doesn't require polarity concerns unless of course
> there's some non-isolated electrical imperative. Which of course,
> the transformers take care of.

If possible, I would run balanced mode for longer runs. Propper
twisted pair isn't bad at the frequencies we are talking about. Some
of the old telco cabling standards are a problem thought, and I've
even seen NEW installations of it (sigh!).

> Some of the sources have transformers but the RCA sockets are still 
> grounded. Makes no sense to me but there you go.

Signal ground and chassi ground is two isolated things. I've found
that out the hard way many times. You are 5 decades up from the line
fundamental with S/P-DIF and AES/EBU, and a lot can happend impedance
wise over such a range.

> But one in particular has 
> no isolation at all so I'll be relying entirely on the transformer in the 
> back of the podule for that. The podule apparently takes care of 
> synchronization so I don't have to worry about a central word clock 
> generator.

Good for you! I guess nobody here ever heard of AES-11 synchronisation?

> I have assurances from the supplier but Yamaha's documentation 
> is less than clear. They'll hear me shouting abuse at them all the way over 
> in Japan if this is not the case.
> 
> Believe it or not it's taken me 2 years to get to this stage. I remember 
> discussing how to make a suitable transformer should I need to with Peter 
> Ullrich just prior to having to leave the list. Hello Peter if you're out 
> there. But it actually took them this long to get back to me with assurance 
> that it will handle asynchronous inputs.

Ah.

> The only other problem I've just realized. The I/O on the back of this 
> podule is actually a DB25. That's going to be a fiddley cow of a thing to 
> hook coax to. Still, we have to do what we have to do. And I've had to do 
> worse. And if it really is unworkable, I guess it wouldn't hurt to run a 
> short length of some liter shielded cable out to an RCA socket. Or 
> something along those lines.

You could make a small board to distribute the signals out to coax
tappings. However, a friend of mine did his own "VGA-extender" with a
pair of 15-pin Mini-DSUBs and 5 RG-59s. Just soldered the damn cables
down the DSUBS and thermo-glued it fixed inside the inner-roll of
toilet-paper. Looks like shit, but he has no reflections, and then he
is tossing over 100 MHz of graphical bandwidth at it... which is much
worse than S/P-DIF.

Cheers,
Magnus