HDMI vs. Coaxial Digital Interconnects
- Thread starter RBFC
- Start date
I got a few hints from watching the founder of Wikileaks on 60 minutes last week!
What frequency would you then measure cable impedance at? The clock frequency? No cable is a perfect conductor, and the impedance is a combination of resistance (probably insignificant), inductance and capacitance (which might probably be more important at the higher frequencies). If so, then we end up having to design almost completely different cables at different lengths and for different clock frequencies!! What a swamp!
I did a search and ran into this fun set of pictures of S/PDIF outputs of a few CD players. Any notion that the signal is "ones and zeros" and that the problem is "easy to solve" at the receiver, should be dispelled
Amir, could those pictures also be showing reflectance due to incorrect termination of the output because the probe cable and oscilloscope load impedance is not equal to the source impedance? Depending on the CD players that they are measuring, the source impedance might not even by correctly 75ohms.
I'm sorry that I'm asking a lot of questions because this will impact the analog/digital comparison that we are going to do next week. I have the Burmester Ph100 which has digital on a RCA output, and the Weiss Minerva which has digital input on an RCA input. I think that the swamp is going to be very deep, smelly, and crocodiles will abound..... and may be even a few anacondas.
Correct. It is measuring sum total. BTW, I have a head a lot of horror stories of bad output. Building a good transmitter also takes good design skills which some of these companies lack.
Please ask Gary. That is how these threads become useful. Best way to get Don and I to explain things is to ask questions. Hard to get motivated to write something long without it
.
OK. So here's the question - it looks like we have an imperfect digital interface going into my analog/digital comparison. Both source and load have RCA connections - which are not as good as if both had BNCs.
What would be important if I were to build a cable to go between these two pieces of equipment? Build it exactly for 75ohm (in the hope that both ends are also about 75ohm) and make it 2m long?
Would the length change if the sample rate were 48kHz, 96kHz and 192kHz?
What would be important if I were to build a cable to go between these two pieces of equipment? Build it exactly for 75ohm (in the hope that both ends are also about 75ohm) and make it 2m long?
Would the length change if the sample rate were 48kHz, 96kHz and 192kHz?
You left out the shunt conductance, surface roughness, and all the nonlinear effects, not too mention hysteresis from DA (dielectric absorption) and such...
Just to present it for everybody else (I know you know it already), the simple impedance formula for a cable is Z = sqrt(L/C) ohms.You have indeed followed and identified the problem -- the "proper" length depends upon a myriad of factors, including clock rate, edge speed, even which terminations (source or load) are "worst".
When I test a cable, I use a network analyzer, usually capture S-parameters, and try to measure to a decade beyond the highest frequency (which is normally set by the edge, i.e. fmax = 0.35 / tr where tr is the rise time). What else I do depends upon what I see and the application. It is almost always the case that a cable that measures good to well (perhaps an octave, factor of two) beyond the highest frequency of interest will do fine for lower frequencies. As I said, this relates to the "magic" length only peripherally. That is, how good the cable is, does not relate directly to how long it should be to avoid those nasty ill-placed (ill-timed) reflections.
As for length, ideally you want any reflections (which are periodic in time) to land in the "quiet", unchanging part of the data bits. Thus, how long the "perfect" cable is does indeed depend upon the bit rate as you said. In practice, and I have done this in test systems, you may trim the cable in tiny fractions to place the reflections where you want them. That means trimming over a range of about 1/2 the bit period. If we have a 10 Mb/s stream (for example, I don't know what it actually is in an HDMI or coax cable), then the bit period is 1/10M = 100 ns. In a cable, the velocity is about 1 ns in 9", so at low bit rates it's not too hard to dial in a cable. Way up in my world, trimming is nearly impractical because it takes just a few thousands of an inch (we use other techniques).
Finally, I want to remind everyone that the impact of all this may be inaudible in most systems and to most people. Of course, to those that want What's Best...
I was writing whilst you posted, sorry Gary.
Unless you have some way of determining the exact input and output impedances and are prepared to design and build suitable matching networks, I would use the best 75 ohm cable I could get (foil shield, low-loss, etc.) The main issue with reflections comes from the source and load (including the conenctors), not the cable. Even the connectors are often indetectable in the mess of mismatch at source and load, simply because the connector mismatch, while perhaps ugly if viewed alone, is very short compared to the bit period and actually ends up being a minor contributor. Just make sure the attachment is good from cable to connector and connector plug to jack, including the shield.
Yes, theoretically the ideal length would vary, but since the sample rates are all multiples of two one cable should work. You need to know the actual bit rate, which is not the sampling rate (Amir -- are there extra bits for scrambling and error detection/correction in the S/PDIF format?) and from there can estimate the cable length. I would design for 192 kHz on the theory that shorter is better but it probably won't matter.
If you have a decent 'scope you can look at the waveforms at each end. We have a nice 50 GHz model that includes a TDR unit, but that's a bit much for audio rates... Of course, you'll have to get a matching pad to go from 75 to 50 ohms into the 'scope.
Unless you have some way of determining the exact input and output impedances and are prepared to design and build suitable matching networks, I would use the best 75 ohm cable I could get (foil shield, low-loss, etc.) The main issue with reflections comes from the source and load (including the conenctors), not the cable. Even the connectors are often indetectable in the mess of mismatch at source and load, simply because the connector mismatch, while perhaps ugly if viewed alone, is very short compared to the bit period and actually ends up being a minor contributor. Just make sure the attachment is good from cable to connector and connector plug to jack, including the shield.
Yes, theoretically the ideal length would vary, but since the sample rates are all multiples of two one cable should work. You need to know the actual bit rate, which is not the sampling rate (Amir -- are there extra bits for scrambling and error detection/correction in the S/PDIF format?) and from there can estimate the cable length. I would design for 192 kHz on the theory that shorter is better but it probably won't matter.
If you have a decent 'scope you can look at the waveforms at each end. We have a nice 50 GHz model that includes a TDR unit, but that's a bit much for audio rates... Of course, you'll have to get a matching pad to go from 75 to 50 ohms into the 'scope.
amirm, WBT claims 75 ohms characteristic impedance for at least some of its nextgen RCA plugs and sockets. If correct, problem solved if manufacturers use these or similar products?
Frank
Frank
Don and others, I haven't been keeping up just lately with what's out there in terms of cleaning up the S/PDIF signal. There was the Digital Lens, which uses a different technique, but what is out there now that works in real time, that is plug and play, to take a dirty S/PDIF in, and spit out a spiffy, laundered version, through minimal cable length into the DAC?
Frank
Frank
That I cannot answer -- not something I have done any research about. The only S/PDIF stuff I have and use semi-regularly is pro recording gear and a pair of Roland monitors (they use digital modeling and coax to hook to the mixer). My home components in the main media room hook up through HDMI. However, I suspect a number of others may help out with various and sundry opinions...
Still curious, so did a bit of digging and found Audient Technologies Audit and Tactic -- from another down to earth discussion on jitter: http://audio.peufeu.com/node/7. Unfortunately, tons of spam at the end of the page ...
Frank
Frank
Well, many companies claim that too although WBT does build good product. Problem is, we need an *in-circuit* impedance of 75 ohms end to end. Meaning the combination of wire termination+male connector+female PCB connector and PCB traces to all add to a perfect 75 ohms. BNC connectors have been designed for this application for decades so you can achieve that target much better than just a male connector offered this way with an unknown coupling by the user (BNC connectors are locking, RCA are not).
As an aside, and not to belittle then, I absolutely hate BNCs as RF connectors... Poorly controlled impedance (though better than RCA!) and positive contact is still a little difficult to ensure (again, better than RCAs, and as Amir said at least they have do a true positive capture mechanism). SMA's are much better, but of course people would overtighten and ruin them.
Whilst on my rant, whatever person designed the RCA pin and jack such that hot is made before ground should be rotting in... well, you know! It would not be too hard to adjust the mechanical assembly so the pin is recessed and ground is made first, though it would require a longer plug and jack. Maybe I'll "invent" and patent it...
Back on-topic, I'll echo Amir in saying good assembly technique is most of the battle. It's easy to attach cables to connectors; it's harder to do it well. And, the PCB launch (incoming and outgoing) and traces themselves are a significant issue...
Whilst on my rant, whatever person designed the RCA pin and jack such that hot is made before ground should be rotting in... well, you know! It would not be too hard to adjust the mechanical assembly so the pin is recessed and ground is made first, though it would require a longer plug and jack. Maybe I'll "invent" and patent it...
Back on-topic, I'll echo Amir in saying good assembly technique is most of the battle. It's easy to attach cables to connectors; it's harder to do it well. And, the PCB launch (incoming and outgoing) and traces themselves are a significant issue...
quick coaxial comparison results
I borrowed a 1 meter and a 2 meter UltraLink coaxial cable (Challenger II), both low/mid priced from the same line. I also borrowed a 2 meter Purist Audio 75 Ohm cable (retail ~ $330) to compare.
Results: The 1 meter exhibited the midbass tubbiness and lack of high-frequency extension I mentioned with my own coaxial cable. The 2 meter reduced that tubbiness a bit, but still lacked the high end extension. The Purist Audio cleaned up the midbass a bit more, but still left some tubbiness. It had a bit more extension, but still not the transparency of the Purist HDMI in my system.
It is quite possible that the coaxial output of my Denon 3930CI has an intrinsically high level of jitter, which no coaxial cable can rectify. In the end, I still preferred the HDMI from my Denon 3800 Blu-ray player for CD playback.
Lesson learned: Each device probably has unique levels of jitter and also differing levels of attention paid to different circuits in that device. A little experimentation is probably necessary in any case to achieve the best results.
Lee
I borrowed a 1 meter and a 2 meter UltraLink coaxial cable (Challenger II), both low/mid priced from the same line. I also borrowed a 2 meter Purist Audio 75 Ohm cable (retail ~ $330) to compare.
Results: The 1 meter exhibited the midbass tubbiness and lack of high-frequency extension I mentioned with my own coaxial cable. The 2 meter reduced that tubbiness a bit, but still lacked the high end extension. The Purist Audio cleaned up the midbass a bit more, but still left some tubbiness. It had a bit more extension, but still not the transparency of the Purist HDMI in my system.
It is quite possible that the coaxial output of my Denon 3930CI has an intrinsically high level of jitter, which no coaxial cable can rectify. In the end, I still preferred the HDMI from my Denon 3800 Blu-ray player for CD playback.
Lesson learned: Each device probably has unique levels of jitter and also differing levels of attention paid to different circuits in that device. A little experimentation is probably necessary in any case to achieve the best results.
Lee
Very interesting, Lee. By all rights, the HDMI connection should have more jitter than the coax link, often much more. There may be more than jitter going on... I suspect SI (signal integrity) is higher on the HDMI link for various reasons.
The Genesis Digital Lens has been discontinued for over 10 years, and we still have people emailing and calling us about it - it must be doing something right! However, with modern DACs with input buffers, jitter rejection, etc. etc. I don't know if it is still relevant. The only other device of this type I know is the Monarchy Audio DIP. Even brand new, it's still cheaper than a second-hand 10 year old Digital Lens. I have several DIPs running Squeezeboxes at home and in the office - they certainly improve the sound in that application, but the main reason I use them is that I take optical out from the Squeezebox, and translate this to AES/EBU to the DAC.
Don - Neutrik fixes this with their Pro series RCA's - and they are far cheaper and sound better than most of the "high-end" ones.
http://www.parts-express.com/pe/showdetl.cfm?Partnumber=092-114
Thanks for that Gary! Geez, I have Neutrik XLR, TRS, and Speakon connectors all over various installs, never considered their RCA line... Duh!
Similar threads
 | Steve Williams Site Founder | Site Owner | Administrator |  | Ron Resnick Site Co-Owner | Administrator |  | Julian (The Fixer) Website Build | Marketing Managersing |