DIYHiFi.org

I have seen here that longer SPDIF cable length (2 meters or longer) is recommended (on properly terminated 75ohm cable) in order to minimise signal reflection problems. I believe the reason was along the lines of avoiding reflected signal showing up during the logic transistion of the SPDIF waveform.
I have also seen this statement

Can all you RF engineers let me know how reflections actually operate on a SPDIF cable which has an impedance mismatch at the receiving end? Does the signal get through to the comparator & is interpreted as either 0 or 1 and a reflection of fraction of the signal begins it's journey back down the cable towards the transmitter? Does this get then get partially reflected back from the transmitter end towards the receiving end? In real world cables, how many round trips before this reflection dies to insignificance & why aren't these interpreted by the receiver as valid signals?

If this reflection is bouncing back & forth can it not interfere with the timing of subsequent transition detection causing jitter? If, on the other hand it makes enough round trips that it dies out before the next signal is transmitted - it cannot cause jitter? This would seem to favour shorter cables as more trips happen faster.

Which of the two positions above is correct depends on the details of how transmitter, receiver & reflected signals behave in digital cables & I don't know squat about it - can anybody enlighten me?

_________________
There is a crack in everything, that's how the light gets in!

There's the long answer, which you probably don't want, and there's the shorter more practical answer.

Here's a link to the latter that I can post:

http://www.analogresearch-technology.net/ubyte.html

No, it is the long answer that I'm after

I read that link before & just when it got interesting here's what it said

So it is the detailed explanation that I would love if you have the time! Or if you have a link, I would be delighted but I've searched & can't find one that details this!

_________________
There is a crack in everything, that's how the light gets in!

There's two reasons you'll want to lengthen a cable... (1) cable attenuation and (2) time delay.

Reason #1 makes sense at first; the main signal gets attenuated by the cable reflection, the 1st reflection gets attenuated 3x the cable reflection, the 2nd reflection by 5x, etc...

But run the numbers and it's pretty much pointless. RG59 is 0.5dB/100ft at SPDIF frequencies, so a 20 foot SPDIF cable only gives you 0.1dB of attenuation with each pass. And lets say for this example that you've got "reasonable" source and end impedances with a 25dB return loss on each end:

- At first pass through your cable, and neglecting reflection loss, your end receives your source power minus -0.1dB.
- The first reflection bounces off your end (-25dB), passes through your cable (-0.1dB), bounces off the source (-25dB) and passes through the cable again (-0.1dB). Your first reflection is -50.2dB relative to end power.
- Second reflection is -100.4dB relative to end power by the same method.

You'll see quickly that the magnitude of the reflections is dominated by return loss, not by cable attenuation. So if you borrow a piece of test equipment (TDR, RF impedance analyzer, whatever) and tweak up your source and end impedance so that your return loss is -30dB on each end, you've just made a 10dB improvement on your first reflection - equivalent to 1000 feet of cable! So, there isn't much point really.

...

On to reason #2, delaying reflections. This is actually common stuff on high speed serdes designs (PCIe, SATA, etc) where frequencies are high and parasitics keep you from making really good impedance matches, so you deal with the reflections by delaying them.

Think about the SPDIF eye pattern going into your receiver. During the transition phase, you don't want anything to interfere with the SPDIF signal since added crap here will introduce jitter. But between transitions, added garbage isn't a problem as long as the magnitude isn't high enough to trip the comparator.

With a short cable, reflections will happen almost instantly and will affect the transition window. But lengthen the cable and you'll introducing extra delay on the reflections, and they'll hit the receiver after the transition. The longer the cable, the more you'll delay the reflection - Just don't make it too long, or your reflection will hit the next transition period.

To start with, unless you know the reflection coefficient of each end, you won't know how much energy is reflected. So, let's assume in a crappy set-up (and most are), that you will have between 10-20% of the energy reflected. So, you do the math: how many round trips will a reflection need to make until it is insignificant?

If you build something that has a reflection coefficient of -40 dB, only 1% is reflected. After 1 round trip, it is down to 0.01%.

If you can't make stuff with a decent reflection coefficient, you make a delay line that keeps them from arriving at the decision point.

Jocko

_________________
"Because, I hate you guys. I hate you guys so very, very much.

Yours,
General Cartman Lee"

This is hardly a textbook explanation - you'd need to already have the math and the E/M fields course work - but it's pretty good.

http://ece.wpi.edu/analog/resources/tra ... online.pdf

The important thing really is what is described on that linked page that describes the UByte cable. You could spend time doing a detailed analysis of exactly what the forward and reflected waveforms are like, but you'd need to know the scattering parameters of the cable and terminations as well as the characteristics of the launched waveform. Then it just tells you how bad your results are. Better to solve the problem, I'd think.

200$ such a deal? Whoaaaaaaaaaaaah

"I can get if for you wholesale. For you, I will do you this..................."

Mantra of the people that CG works with!

Feh.

Jocko

_________________
"Because, I hate you guys. I hate you guys so very, very much.

Yours,
General Cartman Lee"

Thanks guys, some good info. The quote in my first post comes from Dan Lavry whom I thought had some experience? . He started into this when I posted this: And he laid into this with:

He seemed so bullish that I had to get to the bottom of this. I find what he says confusing (as I'm on unsure ground) & contradictory So according to this the cable length is immaterial as the original signal is instantly overlaid by any reflection? But he is stating that short cables are the ones that show the least reflections ??? He also claims that the signal can bounce back & forth: Pure bunkum, by the sound of it - how can he be so off base?

_________________
There is a crack in everything, that's how the light gets in!

Because he is an idiot. And wrong.

I have run 100', with no problems. I have also done pulse dispersion measurements at that length.

I guess that I am some kind of self-deluding douche bag, as opposed to "serious" EE. RF or otherwise.

Yes, if the length is just the right length where the reflection then falls at the point that you are trying to make sure that it doesn't hit, yeah, then you are stupid.

You notice that he says nothing about reflection coefficient. I assume that he has no knowledge of that concept. Without that, he is deluding himself.

Jocko

_________________
"Because, I hate you guys. I hate you guys so very, very much.

Yours,
General Cartman Lee"

I guess that I should have pointed out that I don't have a cable under 30' at my house. Three systems, in 3 different rooms, cables back and forth between them.

But according to some guy that I don't know, it can't work.

Jocko

_________________
"Because, I hate you guys. I hate you guys so very, very much.

Yours,
General Cartman Lee"

"I guess that I am some kind of self-deluding douche bag, as opposed to 'serious' EE. RF or otherwise."

We've heard that about you. People talk...

very interesting , find out why the length of digital cable, be "better"
I have found this article:
http://www.positive-feedback.com/Issue14/spdif.htm

I believe that we already had that discussion with him.

Jocko

_________________
"Because, I hate you guys. I hate you guys so very, very much.

Yours,
General Cartman Lee"

Are we saying the cable length has to be calculated based on the rise time of the signal so as to fall in the gap between signals? Hmmmm.... bit of a problem, methinks! So the recommendation for long cables is misleading & should be qualified by the above. Jocko, how are you using all sorts of different length cables - recclocking, maybe?

BTW, here's a short answer from Eric Bogatin, that contradicts the longer cable idea & suggests :
http://www.thefreelibrary.com/Riddle+me ... 0123202835
Hmmm.........

_________________
There is a crack in everything, that's how the light gets in!