Clock transmission from DAC to CDP

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
There have been a few posts on this subject here and other places, but few specifics.

I'm pondering a DAC for my SACD player, and I was planning on using a clock (the kwak clock) in the DAC itself, and feeding the clock back to the player. The DAC (BB/TI DSD1702), doesn't require any particular phase relationship between the system clock and the bitclock, thankfully.

Basically what I'd have now in the DAC is the output of the kwak clock, and I'd need to do galvonic isolation (transformer? BB ISO150? Whatever will work), and then send it out to the transport. My understanding was to use 50 or 75 ohm coax (w/ BNC connectors). Put a 50/75 ohm resistor in series on the transmitting side, and to ground at the receiving side.

Any recommendations for an unbalanced line driver? Or perhaps a balanced line driver & receiver pair? I'd like to keep the cost for the driver & receiver below $20, if possible. Also, they must run on a single-ended supply. (5v preferred, but 3.3v would be workable.)

Also, if you have any thoughts on sending the data from the transport to the DAC... I need about 12 signals, and max speed is probably the DSD bitclock at 2.8Mhz. I think I can probably get the signals OK just with unbalanced signalling in a shielded cable, something along the lines of a printer cable. Jitter here shouldn't really be a problem.


Found 'em!

Allied Electronics has them. Unfortunately they're around $15 each. It's not too bad, though.

The only thing is, as long as it is relatively clean when it reaches the transport, jitter and such isn't a big issue with the clock that goes to the transport, since it doesn't effect the DAC.

Although if this is the best way to transmit the 16mhz clock reliably, that's the way to go.

btw, cable length will be less than a foot.


Actually, I'm looking at this idea:

Take a SPDIF transmitter & receiver, and use them, feeding the other way. Just have it transmit digital silence, and throw away the data recovered at the output. The PLL in the receiver recovers the system clock, and feeds it to the transport.

The quality of the clock isn't great, but it should be stable, and consistent.

Might be a cheap way to solve my problem.

Thanks Elso. That's a pretty cool trick :)

Do you think decent signalling can be done with a non-impedance-controlled wire (like a wire in a shielded printer cable for a computer) up to around 2.8 mhz? Would this schematic work for that?

I need to feed the clock (16mhz) back to the transport, and that will be done with a 75 ohm cable (probably Belden 1506A), and good BNC connections and such.

But then I need to send the I2S signals, the serial control signals, and also the DSD data from the transport to the DAC. The fastest thing here is the DSD bitclock & data lines at about 2.8 Mhz.

I'd like to avoid having to build 10 high quality 75 ohm digital cables, as that will get pretty expensive with all the connectors.

Clock Transmission

Hi Charlie,
My simple scheme is to be used with 75 Ohm characteristic impedance coaxial cable. The trick is that a properly terminated cable presents only a ohmic resistance to the driving transistor and no reflections occur at the ends of the cable. My scheme would not work with printercable.
I would send all signals mentioned by you through coax.
I am using cheap but double shielded cable. $ 0.50 per meter.;)

The problem is not the cable. It's not too expensive and I need like 20 feet at most. The problem is all the connectors!

I think I am just gonna use nickel-plated RCA's (with a metal body, for shielding) Buying BNC's for all those would be very expensive. I may still use a BNC for the clock.

I know BNC connectors are often listed as 50 ohm or 75 ohm, would the same apply to RCA connectors? I have never seen industrial ones rated at anything, though I have HEARD that Canare makes a specific '75 ohm' RCA plug for digital cables.

On the other hand,
Another thought I had was I could design the DAC so that it stacks on top of the transport, and put all the connectors (RCA's, obviously BNC's would be impossible) so that when you stacked the DAC on top of the transport, the RCA's would line up and make the connection. This would probably require no special circuitry, too. There are probably other connectors that would work here, and would have much lower capacitance & inductance and whatnot than RCA's, too. If I could get something that regular CMOS could drive across w/o problem.. (In this layout, track length would be less than 2 inches, probably through a simple pin header & socket connection.)

Am I gonna have to start another "new law" thread?

There is no such thing as a 50, or 75 ohm RCA connector. Anyone who tells you otherwise is either stupid or a pathological liar. That goes double for Canare.

Why would BNCs be "impossible"?

You can't have lower inductance and capacitance at the same time; no matter what the NBS cable guy says.

BNC's wouldn't neccesarily be impossible. It would be difficult to twist them to lock them while lowering the metal case down on to the CDP. You'd have to turn & attach all 12 of them at the same time.

And I'm not talking about cable. What I mean is, if I make the two connect directly, I could use something like D-Sub connectors, which have a nice metal shield all the way around them. Or there are a variety of high-speed (lower pin inductance, I suppose) connectors available with a 0.5mm or 1.27mm pitch. But no shield on those, usually.

I think 3.3v CMOS wouldn't have much trouble driving across maybe 2" of trace and a D-Sub connector. I could also use chips that let me lower the slew rate, if neccesary. Jitter on these lines shouldn't matter too much. Even the clock going back to the transport isn't critical, as long as it is clean enough to be read properly.

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.