Question on transformer coupling S-PDIF

[Porksoda]

Hey all,

So I am designing a DAC and I keep hearing that transformer coupled SPDIF is the greatest thing since sliced bread, so I was wondering:

Adding a transformer creates a 90-degree phase lead, right? Turning a square wave into a sawtooth. So the logical solution would be to have a transformer at both ends, so you get 90-90 = 0 degree phase shift.

That's great in theory and all, but doesn't that mean your digital source and DAC are now a matched set and probably not compatible with any other components unless they are configured similarly? Or, is there some sort of standard for this sort of setup?

Or, is there a simpler solution I am missing here?


Thanks!

[Jocko Homo]

I don't even know where to start on this one, so I won't.

Jocko

[Porksoda]

Well, antagonism is always helpful.

[B Cullingford]

I'll play nice guy.
The transformer is the 'simple' solution and the best, but maybe not easy to get right.

May I humbly suggest a search on s-pdif interfaces, there is plenty to read, and perhaps a reread of those theory books, or a new book (sorry I do not have a suggestion for a book, though I have run across a couple reasonable internet resources).

Or maybe even better: if you have a 'scope build a simple cmos oscillator (inverters, R & C) get a pulse transformer or two, load it right and look, nothing like the real thing to learn from.

Oh and if you're going to try the latter course you may as well do a search and some reading and order transformers that will do the job well, I'll give you a start: they should be 1:1 - any more and I'd have to do some research and I don't want to steal your fun!

[danb1974]

A transformer gives you (1) galvanic izolation (good if a common ground is not desired) and (2) the posibility to scale the voltage (for example between 0.5V and 5V spdif).

If neither are necessary, don't use it.

[Jocko Homo]

Not really.

SPDIF transformers work the same way any other transformer does. Name one that turns a square wave into a sawtooth, when operated withing its range.

Besides galvanic isolation, it gives you some common-mode rejection. You could use it to scale levels, but then I will get antagonistic.

There are more threads on SPDIF transformers here than you probably have time to read, but it would be a better starting point.

Stick my name in the search function to help weed out the useless ones.

Jocko

[audioengr]

Originally posted by Porksoda
Hey all,

So I am designing a DAC and I keep hearing that transformer coupled SPDIF is the greatest thing since sliced bread, so I was wondering:

Adding a transformer creates a 90-degree phase lead, right? Turning a square wave into a sawtooth. So the logical solution would be to have a transformer at both ends, so you get 90-90 = 0 degree phase shift.

That's great in theory and all, but doesn't that mean your digital source and DAC are now a matched set and probably not compatible with any other components unless they are configured similarly? Or, is there some sort of standard for this sort of setup?

Or, is there a simpler solution I am missing here?


Thanks!

No sawtooth. The things that you need to worry about are:

1) no DC to the transformer - this will cause it to saturate - always AC-couple with at least .1 uFd low ESR cap
2) Drive the transformer with a fast driver, not a Flip-Flop
3) The impedance must be matched to 75 ohms - this does not mean putting a 75 ohm resistor in series - the sum of the output impedance of the driver PLUS the resistor must equaly 75 ohms
4) the P-P level from the output must be in the 500-600mV range, so you must reduce the voltage somehow, usually with a voltage divider
5) The output winding of the transformer need not be AC-coupled

Read this:
http://www.epanorama.net/documents/audio/spdif.html

and this:
http://www.rane.com/note149.html

Steve N.
Empirical Audio
Manufacturer

[Porksoda]

Thanks for the comments!

When you say that I have to "drive the transformer", I take it you mean on the transmitting end?

As far as impedence matching, last time I got into this discussion at head-fi things got pretty unclear as to the best way to match impedences across a transmitter, cable, and reciever.

So for transmitters, you simply put a resistor in series with the output buffer so that their total impedence is 75 ohms? How does this work for the reciever, since it is differential?

One topic that came up a lot on head-fi is the issue of parasitic capacitance and inductance in the jacks and wiring effecting the impedence of the signal path. The only consensus I recall is that BNC connectors are the only "true" 75-ohm connector one can find. How big a deal is this issue for other connectors, and is there any good way to go about dealing with it. Moreover, I intend to add a toslink input. I take it I should adjust the toslink's output to 75 ohms aswell?

Thanks!

[Bgt]

Originally posted by Porksoda
Moreover, I intend to add a toslink input. I take it I should adjust the toslink's output to 75 ohms aswell?

you can use 100 ohms in series to the 75 ohm terminated coax input.

[Jocko Homo]

You set the load on the RX to look like 75 ohms back to the TX end. What you do to the RX chip is your business, but I advise keeping it differential.

As for connectors, yes, RCAs, XLRs and the like, look like crap at SPDIF frequencies, and are totally useless. You could make some network to compensate for all the stray capacitance and stuff, but why? Few here have the test equipment to measure this stuff. It would be far easier, and in the long run better, just to use a true 75 ohm transmisson line.

Don't adjust the TOSLINK output at all. It is a cheap logic gate. Do not leave it connected to the RX chip when you are not using it. Why you would consider using one is beyond me, but it is your system.

Anticipating your next question:

Build a some sort of linear stage to go between the line and those lousy CS841x series RX chips. You can then use a switch/relay/mux chip/whatever to switch between coax and optical inputs.

Jocko