Axiom Home Page Forums General Discussion Home Theater RCA or Coaxial

Bren, I assume that Tom is talking about a connection from his CD player just a few feet long. Using the 44.1KHz sampling rate and applying the common 128 times sampling rate suggestion for flawless digital transmission, we get about 5.6MHz. For wavelength take speed of sound (300,000,000 m/sec) divided by 5.6 MHz to get a little over 50 meters. Using the common standard of characteristic impedance of an interconnect being significant when it's longer than about 1/4 wavelength, we get maybe 12 meters or 40 feet before characteristic impedance of a CD interconnect needs to be taken into account.

John - I've never heard of that calculation before and it's either accurate or some incredible double-talk. Does it take into account shielding from EMI sources and is it network-centric (assumes twisted pairs)?

Your calculation would have to take being a 44.1 two channel signal into account, effectively halving the length at which one starts getting concerned about impedence... and by the time you get into multichannel, that maximum cable length is pretty short... also, since it's a serial and not a parallel connection, wouldn't you immediately also have to multiply by 16 (for bits per sample)? If both of these are correct - that would make the longest run without worrying about impedance 40'/2/16 = 15 inches.

Again, we're talking a difference of a few bucks. Even if just for piece of mind... splurge!

Bren R.

No, Bren, that's actually the way that they figure it. For example, here's the rather informal explanation of the point by Belden Cable engineer Steve Lampen that I've linked before(I believe in conjunction with the the link to the guy who used a coathanger and the other one who used rusty fence wire for a digital connect).

Then it's he that made what seems like the mistake... he admits he was writing that on a plane.

Before I call him a liar, and I'm not... just that he missed quite a bit of bandwidth in his mental calculations. It is my understanding that S/PDIF works in a 1 bit serial - on/off. His calculations are correct for a 44.1K/1 bit/mono signal. CD audio is 44.1/16/2, so he should be off by a factor of 32 (plus data overhead). For each CD sample, a total of 32 bits (plus overhead) are transmitted one at a time.

If someone with technical understanding of the S/PDIF standard would like to weigh in here... it IS my belief that it is a 1 bit pipe.

Bren R.

This is not something I ever thought I would see--BrenR and JohnK arguing about cable. Enjoy, folks!

Wow, too many numbers. Here's my take on it :

The original post say "44.1 KHz sampling, rule of thumb is 128 times sampling or 5.6 MHz". No idea where the rule of thumb came from, haven't read the links yet.

Bren properly objected to what seemed like "funny science". It's that 128 multiplier that makes it possibly valid though...

The bit rate for a stereo CD signal (would be 44.1 KHz x 16 bits/sample x 2 channels, or 1.41 Mb/s. That is a raw digital bitstream (I think there is some encoding but not sure) so let's say we need frequencies up to the 5th harmonic to get a recognizeable signal at the other end, so minumum bandwidth required is about 7 MHz (a bit higher than the rule of thumb indicated).

5th harmonic is a bit low for a good square wave -- my guess is that the rule of thumb takes that into account, so I think the intended application was 44.1 KHz x 128 x 2 channels, or about 11 MHz bandwidth required.

This all assumes that the 1.411 Mb/s bitstream is sent at the lowest possible bitrate, ie that a higher bitrate is used for HDCD etc -- I think this is true but not 100%.

Anyways, it's that "128 times sampling rate rule of thumb" which makes the numbers reasonable.

In reply to:

---

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).

---

This is the line, the 128 multiplier makes no sense... it's like he pulled it out of thin air. A 16 bit stereo signal would be 32x, a 16 bit 5.1 signal would be 96... 24 bit x 5.1 is 144. I'd like to meet up with this guy for coffee one day.

Too late to keep thinking about this stuff...

Bren R.

Too early for ME to be reading this stuff.

Something that came to me while I was brushing my teeth.

His equation assumes that the signal moves along the wire at the speed of sound... well, sure, it's audio in those bits, ain't it? Nope. It's DC current. That means it moves at the speed of... well, DC current.

Bren R.

Typo, I think. He said "speed of sound" but used "speed of light" (which is actually a bit high for current in a wire, electrical current is normally closer to .7 light speed or thereabouts).