[ezkcdude]

Hey, folks. I was very interested in Art's recent SPDIF output thread. I figured since I have a Squeezebox and a scope, why not try doing some of these measurements on my ezDAC. If you haven't seen my DAC, here's a pic (and yeah, I've been too lazy to case it up):



Here's a closeup pic of the two points I took the scope pics (A = spdif input; B=across the 75R on the input to CS8416):



So, without further adue, here's a scope trace at point A (with the power supply turned off and no music playing):



Now, here's the trace at point B, with the psu off:



And now the psu plugged in:



If you look closely, you can notice some difference with the psu turned on (more "ripply"), wihch makes perfect sense. Well, that's about all I got for now. I wanted to start this thread...I'm not sure how it will finish. Since other people have built ezDAC's, I may ask around to get comparison pics of their boards (which presumably have sources other than the SB). I don't have a TDR - so I won't be doing those measurements like Art did. Anyway, if anyone wants to chime in with suggestions for other measurements one can do with a scope (maybe I should try playing a signal throught the SB?), or submit there own measurements, that would be peachy. Take care. Oh, one last thing (as Steve Jobs likes to say)...My scope is a Tek 465B, in case anyone wants to know.

[art]

You have the input to your RX coming through X" of mystery cable.

You have capacitors in both primary wires.

You have the load resistor on the secondary side.

It goes through lots of PCB trace.

(Hint: look at the SB output thread for ideas on how to do it, and why what you have is a kludge.)

And you stick a 'scope probe, with a ground lead who knows how long, onto a circuit that is supposed to be 75 ohms.

Any wonder why it rings so bad?

Any attempt to measure LF tilt would have to be done on the output side, and through a buffered stage. Unless you turn that 465 into a TDR-like unit, there is not much else that you can do with it. I'm sure that Google can show you lots of places on how to do that. Use the SPDIF signal for the pulse, but it would help to make it as fast as possible.

The BW on mine is >1 GHz, which is ludicrous for your purposes. Suits mine, as I need to be able to see all the kinks and bends in PCB traces, and what they do.

Pat

[art]

Aren't you novices glad that I am bored by tonight's baseball game, and I have time to waste doing this stuff?

Rhetorical question.

OK, here is a progression of the right way and wrong way to make a SPDIF input. Only deals with the cable and termination. (I already did enough on transformers.)

First off, I stuck a 75 ohm SM resistor onto a 75 ohm BNC. As you can see on this TDR trace, there is a slight inductive (upwards) spike, as explained in some other thread.

Translation: Hard to make it perfectly resistive, with no reactance. Yes, I could make a zobel with about 0.5 pF in series with 75 ohms. Someday when I get bored, I will show you how to make a capacitor that small.



Next, I took 6" or so of RG-187 coax. Terminated in same SMD resistor. Indeed, you can see small spikes at the input BNC, and the end of the coax. And the coax is not exactly 75 ohms.............



Now, here is the way most of you think that you can make a SPDIF input. (There are days that I actually enjoy this job............)



Lovely, isn't it?

OK.....how to read the trace, from left to right.

First, we have a capacitive spike downwards, where there is a BNC-RCA adaptor, and the RCA jack from Dan's SB. Immediately after that, the trace shoots up rapidly, to a level of around 110 ohms.

Which is about what we would expect from a twisted pair. The trace stays at that level for about 1 horizontal division, where it has another inductive spike from the 75 ohm resistor, just stuck on the end any ol' way. (Similar to what most of you guys do.) The trace then decays down to the 75 ohm level, and that is that. You can use the decay time to calculate what the amount of the inductance of that terminating resistor is, but some other year on how that works.

Below are a series of pictures of the wiring schemes used to make these traces. First, all of them together. Followed by enlarged views of the 2 SMD resistors stuck on the end of their respective hook-ups.




OK.....just what is the point? The point is that our research shows that you need to have an input return loss >30 db. Or a rho of around 0.03.

The simple hook-ups with just a SMD 75 ohm resistor on the end are around 0.06 rho. (0.1 rho/div, as usual on these traces.) So, you can see how hard it is, even with good attention to detail. There are tricks of the trade to get it lower. But without gear like this, no one will get there.

However, that is not the point. A typical kludge thrown together can easily have a rho that is off by a factor of 10. In each direction! At the same time! Until you guys get rid of the RCA jacks, mystery wire, and inductive traces all over the place on your PCBs, there is not much hope.

But, I wish you all well. Just remember that it is RF, and not audio. You can get there.

Pat

[ezkcdude]

Thanks, Pat. I appreciate your constructive criticism. I can certainly start by making the pcb traces as short as possible.

[art]

You have tons more issues to remedy first. Wiring being the first. You have to think in RF terms, not audio. You are never going to get rid of all of the trace inductance, but it is minuscule in terms of its contribution to the rest of what you have.

Until you start using 75 ohm coax and decent connectors, you won't make in dent in the problem.

I believe that I have clearly shown how critical this stuff is, even when you use the right parts. Start with that first. Without that, you aren't going to get far.

Pat