S/PDIF termination; practical guide?

[new2hifi]

Hi All:

I have a kit DAC in a metal box and I want to provide it with working S/PDIF input.

You all might say that this topic has been beaten to death. But still I could not find a guide/textbook/method for beginners like me .

I have read the posts here and then on diyhifi and the theoretical talk of Gurus’ just goes over my head. Please guide me in steps what to do.

This is what I currently plan to do. Take 6 inches of 75ohm coaxial cable, solder or crimp one end to 75ohm female chassis mount BNC. BNC is chassis isolated, i.e. BNC outer frame does not electrically connects to chassis. Solder the other end of this cable directly to PCB where it says S/PDIF or use existing terminal block. I read on posts about need to use 75R resistor, where should I place it if this is required step?

I just have a DMM and no oscilloscope or other equipment. I know some of you will write that they do not know what is on the PCB, “characteristic impedance” of PCB trace,…neither I know the answers to these nor I think DMM can help. Thus please give me your best practical guess of installing a BNC for S/PDIF in.

Regards

[SY]

You want the termination of 75R to be at the board end of the coax. You also want the traces to the receiver to be relatively short. Make sure the board doesn't already have an input resistor on it!

If you want to be anal, you can also use a properly terminated input transformer- there could be some benefit in certain situations. Don't worry too much about the gurus- it's not hard to make the termination Good Enough if what you care about is the audio output rather than some sort of theoretical perfection of the digital signal.

[marce]

Most PCB's are not controlled impedance builds, and very few designs on this site warrant a controlled impedance build. I regularly have controlled impedance PCB's made, the controlled impedance traces though are the really high speed parts of the layout, DDR2/3 ram, gigabit Ethernet etc. For SPDIF, I2S, I2C and the rest I wouldn't worry to much about.
The lengths of the PCB traces are minimal compared to the length of the cable (and depending on the PCB build, will be in the 110-65 ohm range) and slapping a 75ohm resistor to ground is common practice and works. This is how the CATV (cable TV network ) is terminated, and those cables are higher bandwidth and longer than any SPDIF cable and they work fine.
More important is keeping the SPDIF signal and return path together, ie don't terminate the ground/shield of the cable at some star point, as SPDIF is digital they need to be closely coupled. If you are worried about ground noise then use a transformer and isolate the SPDIF signal as it enters the receiver.
Actually controlling the impedance of PCB traces is a pain, and the cost of the boards goes up quite dramatically. When you do have a controlled board for digital designs you are looking at usually 100ohm for diff pairs and 50 ohm for single ended, and as said we only look at them if we have to, and then only when we are getting clock speeds in the high MHz (ie 100MHz plus) so I wouldn't worry about the SPDIF interface. I have found that to corrupt digital signals to the point where they cannot be received correctly takes a lot of effort, at the end of the day the whole point of digital transmission is to be tolerant of a whole host of nasties such as noise, etc.

[new2hifi]

SY and marce, Thanks!

The board has just two small capacitors before + and – input soldering pads, there is no resister or transformer of any kind.

SY, this is what I understand from your suggestion. I should solder a 75ohm resister across the + and – soldering pads and also solder the coaxial cable on the same pads.

marce, I appreciate your help and understand your point about digital. However, I was talking about the posts that refer to real phenomena of increased clock jitter even when data remains intact if there are many reflections due to Impedance mismatches or termination problems.

[Speedskater]

I think that marce has a much better understanding of how the system works, than the people that go to extremes with cables and terminations.

[infinia]

The board has just two small capacitors before + and – input soldering pads, there is no resister or transformer of any kind.

SY, this is what I understand from your suggestion. I should solder a 75ohm resister across the + and – soldering pads and also solder the coaxial cable on the same pads.

If it's a proper kit, then 99% odds there is already a termination resistor on the PCB.
post a schematic and PCB image for better help.

marce, I appreciate your help and understand your point about digital. However, I was talking about the posts that refer to real phenomena of increased clock jitter even when data remains intact if there are many reflections due to Impedance mismatches or termination problems.

sigh, really not an issue unless you try esp hard to mess it up.

[marce]

marce, I appreciate your help and understand your point about digital. However, I was talking about the posts that refer to real phenomena of increased clock jitter even when data remains intact if there are many reflections due to Impedance mismatches or termination problems.

what like taking a probably perfectly good crystal off a board with a short trace, and replacing it with an add on board, with a super dooper crystal, then add a twisted pair wire link a yard long and a different impedance, that will reduce jitter.
Only joking (slightly), I use SIV (signal Integrity Verification) and High Speed routing software (allows us to put constraints on signals such as allowable skew etc) software as part of my CAD system (full blown Cadstar), all clocks on a board are checked using the SIV software and a termination scheme chosen that works. This software takes into account the topography of the PCB as well as the drive and receiver characteristics (IBIS format). I can also model connectors and cables so we can simulate a whole system set up.
I actually worry more about a signal being monotonic than most of the other artifacts you see on digital waveforms, such as ringing etc, but even then we are talking about digital signals in what most would call the RF range. Keep things simple, make all clock lines as SHORT as possible, have a contiguous closely coupled ground plane (or co-ax/twisted pair cables), avoid excessive rise times in relation to base frequency and for most digital design you will see in the DIY audio world it will work; In fact quite often it takes real audiophile beliefs and determination on digital signalling and layout

Where it says SPDIF input on your PCB, put the DMM across the terminals on a resistance setting (no power to the board), you will probably find that it reads 75ohms.
Have Fun

[new2hifi]

Originally Posted by infinia

If it's a proper kit, then 99% odds there is already a termination resistor on the PCB.

Thanks infinia. It is no-name kit that I bought from eBay and trust me there is no resistor before receiver.

[new2hifi]

Originally Posted by marce

Where it says SPDIF input on your PCB, put the DMM across the terminals on a resistance setting (no power to the board), you will probably find that it reads 75ohms.
Have Fun

Thanks marce!

The things your wrote just went over my head , I am a beginner. However last lines did make sense to me.

[marce]

This app note from Ti should explain the basics of digital signal propagation, without getting to deep!
Even though it states high speed design, the rules are applicable for ALL digital, if you follow best practice you will get the best results, and with a lot of today's devices having fast rise times it is becoming a necessity.
If you want any more info or a simpler explanation of any data let us know, it was one of the simpler texts I could find, and gives a good overview.

http://www.ti.com/lit/an/scaa082/scaa082.pdf

Monotonicity, is a mathematical function, but when used in conjunction with digital signals it refers to a change in direction on the rising or falling edge of a signal. (Figure 2 in the link below shows a picture of this). Having a monotonic rising or falling edge is critical to digital signals especially on clocks; as it is on these edges that the logic switches and the non-monotonic edge can cause false switching. A big part of signal integrity and digital PCB design is ensuring that the rising and falling edges are clean, this is more critical than other artefacts' such as ringing.

Design for Signal Integrity: Unit 02 Integrity of Digital Signals