PDA

View Full Version : Calling all digital cable experts - SPDIF termination question



HighFidelityGuy
22-09-2009, 12:14
Hi, sorry if this has been discussed before.
I've been thinking about digital coax cables today and in particular the argument of proper 75 Ohm termination.

To set the scene, the basic argument is that for the best SPDIF signal transfer, all the plugs, sockets and cables in the signal path should be true 75 Ohm to make sure that reflections in the cable are minimal. I hear that most phono plugs and sockets are not true 75 Ohm and even the ones that claim to be are not likely to actually be true 75 Ohms. To get true 75 Ohms I hear you need to use BNC connectors.

So, if BNC connectors are a fairly certain way of making the best possible connection, then the most productive upgrade to SPDIF equipment would not be buying expensive SPDIF cables with expensive phono connectors, it would be by swapping the phono sockets for BNC and getting a BNC to BNC coax cable. Surely this would be cheaper and better in the long run as it would provide better signal transfer?

I then started to think outside the box a bit. When thinking about this from an electrical conductivity point of view, putting any socket and plug connection in the signal path will degrade the signal slightly. Some are better than others. The best ones are usually expensive. So removing the phono or BNC connection out of the signal path all together would surely be better and cheaper.

So my main question is this; Would de-soldering the sockets from the source and receiver and directly soldering a good quality true 75 Ohm coax cable to the PCB at either end be the best solution or would this not provide a true 75 Ohm connection due to some reason I'm not aware of?
Obviously the results wouldn't be very practical due to the permanent connection but we'll ignore that for now.

I'd also tike to understand how much difference in sound quality there is likely to be between using a Phono connection, BNC, or a directly soldered connection.

I'm hoping to get some solid scientific facts about this to help me understand what's actually best. Hopefully this won't turn into a big argument like I know it would on other forums. :lol:

Thanks. :)

NRG
22-09-2009, 12:51
OK I'll chip in my 2pence worth...in answer to your first question, yes, for sure.

The answer to your second question may not be so firm, I'm sure STan or Dave can shed some insight here but without being able to see what the effect would be of a direct cable to PCB track solder connection via a TDR (Time Domain Reflectometer) I think the result would be very hit or miss.

For me the inconvenience of a direct end to end connection would outweigh any benefit...

Tony Moore
22-09-2009, 13:26
Well I suppose it depends on how the current phono connectors are wired to the transmitter and receiver at the two ends. If they are pcb mounted and have lengths of pcb track snaking around the place, how can that have similar characteristics to the cable? Is it too short to matter? Are the termination components critically chosen to encompass the effects of the pcb tracks to the socket? (and the socket itself maybe?)

Or do most implementations simply ignore the effects of the link to the socket?

:scratch:

HighFidelityGuy
22-09-2009, 14:18
All of the SPDIF gear I own have the phono sockets soldered directly onto the PCB. I'm not sure on track length after that. I'm also not sure whether the length and composition of the conductor inside the socket is factored into the design. I guess this may vary from product to product. From a simple conductor length point of view, I would have thought directly soldering the cable to the PCB would result in a similar conductor length as the cable would need to be slightly longer to get inside the case and this would balance out the loss of conductor from removing the socket. However, the loss of the socket would no doubt reduce the resistance of the circuit despite the extra cable length. I don't know whether that could have a negative effect on the signal.

electric beach
22-09-2009, 14:22
Hi Dave,

Try Mike Homar's digital cable, then say if you can even imagine needing anything better! I've used a Trichord Research Pulsewire with BNC connectors into my MF DAC21 before but it's never sounded anything like this!

(Sorry Mike, just one more...) :respect:

HighFidelityGuy
22-09-2009, 16:29
Thanks, I've read a bit about Mike's cables and from what various people have said they seem like an excellent product. I'd be interested to find out if they sound better with the phono plugs he uses or BNC plugs. It would be rather difficult to do a comparison though. From a scientific point of view any phono plug should be inferior to a BNC in this application so it's difficult to imagine that the cable would sound better with phono plugs. But audio is an odd beast and things that should improve the sound quality don't always and can sometimes make it worse. I'm just trying to find out what approach would be best from a scientific point of view to improve my understanding. I'll then have to decide if it's worthwhile actually trying any different methods.

I'd be interested to get Mikes input on this subject.

The Grand Wazoo
22-09-2009, 16:36
From a scientific point of view any phono plug should be inferior to a BNC in this application so it's difficult to imagine that the cable would sound better with phono plugs.

I'd suggest that although we all use them, any phono plug is inferior to almost any connection you can make short of a quick twist of two wires & a bit of insulating tape!

You can put all the locking devices, exotic metallurgy, solder & crimping you like into them, but you're just polishing the turd of a fundamentally bad design.

alb
22-09-2009, 16:58
I would use BNC connectors. I could hear a difference when changing from phonos.
Also, be aware that not all BNC connectors are 75 ohm. No point in using 50 ohm connectors.
Keep the cable length shorter than 0.75m or longer than 1.25m.

NRG
22-09-2009, 17:25
Keep the cable length shorter than 0.75m or longer than 1.25m.

That only really applies / seems to work if using phono's. With 75R BNC and a good RG6 cable you can choose what length you want...within reason.

Mike
22-09-2009, 17:26
Or do most implementations simply ignore the effects of the link to the socket?

:scratch:

It wouldn't surprise me at all, TBH. :)


Thanks, I've read a bit about Mike's cables and from what various people have said they seem like an excellent product. I'd be interested to find out if they sound better with the phono plugs he uses or BNC plugs. It would be rather difficult to do a comparison though. From a scientific point of view any phono plug should be inferior to a BNC in this application so it's difficult to imagine that the cable would sound better with phono plugs. But audio is an odd beast and things that should improve the sound quality don't always and can sometimes make it worse. I'm just trying to find out what approach would be best from a scientific point of view to improve my understanding. I'll then have to decide if it's worthwhile actually trying any different methods.

I'd be interested to get Mikes input on this subject.

I can make those cables with BNC's, no problem. I agree wholeheartedly that they'd be a superior connector for this sort of application. The RCA phono should have quietly disappeared decades ago!

As for soldering the cables directly... no idea. I'd imagine the results could/would vary depending on the multitude of different DAC and transports kicking around. Almost impossible to predict the outcome if you ask me! :confused:

Cheers...

Mike
22-09-2009, 17:28
I'd suggest that although we all use them, any phono plug is inferior to almost any connection you can make short of a quick twist of two wires & a bit of insulating tape!

You can put all the locking devices, exotic metallurgy, solder & crimping you like into them, but you're just polishing the turd of a fundamentally bad design.

Excellently put! :exactly:

Dave Cawley
22-09-2009, 18:08
What is wrong with them?

Dave

Marco
22-09-2009, 19:12
I'd suggest that although we all use them, any phono plug is inferior to almost any connection you can make short of a quick twist of two wires & a bit of insulating tape!


Indeed :)

If only we could hard-wire our whole hi-fi system from start to finish!

Marco.

P.S Banana plugs on speaker cables are just as bad, but of course practicality and convenience naturally win the day.

Mike
22-09-2009, 19:20
What is wrong with them?

Dave

Erm... they're shite! :lolsign:

Mike
22-09-2009, 20:13
Sorry Dave,

That was a bit, err, blunt!

Now, not for a minute am I calling your technical expertise into question, but... I'm a little bit surprised by the question, I thought you were a bit cleverer than that! ;).

Let's start with the simple things, first of, physically it 'makes' signal before screen, and breaks screen before signal.

Not great!

Next? :)

alb
22-09-2009, 20:18
That only really applies / seems to work if using phono's. With 75R BNC and a good RG6 cable you can choose what length you want...within reason.

I didn't know that!
But then again, there are many things i don't know.
I have to admit that i can't hear any difference between one metre BNC leads and any other short length.
I was just playing safe.:)

Mike
22-09-2009, 21:24
The answer to your second question may not be so firm, I'm sure STan or Dave can shed some insight here but without being able to see what the effect would be of a direct cable to PCB track solder connection via a TDR (Time Domain Reflectometer) I think the result would be very hit or miss.

I've never used a TDR with sufficient resolution that could differentiate a few mm of pcb track!... got any pointers?

I use these things very regularly. Admittedly, mostly OTDR's... but I'd still be interested. :)

Cheers...

Dave Cawley
23-09-2009, 08:24
I thought you were a bit cleverer than that! What, because I asked a question!!!!!!

physically it 'makes' signal before screen, and breaks screen before signal. No, I'm not being funny here, I don't see that as a technical issue? But I am interested in why people think they are a problem?

On all the digital stuff I have here, the manufacturers fitted BNC's as standard on the digital bits and XLR's for balanced audio.

Regards

Dave

Marco
23-09-2009, 08:34
Surely the most obvious explanation why any form of connector is poorer than hard-wiring is because every time you introduce another interface into the chain the signal is broken? In connections where electricity is being passed, this raises resistance - always a no-no; hence why hard-wiring is often the preferred route.

Or are we discussing something else? :)

There's a very good reason (as mentioned) why people hard-wire circuits in valve amps, as opposed to using PCBs, and also for example why I've hard-wired mains leads directly into the MCBs of the dedicated consumer unit powering my hi-fi system ;)

IECs, phonos, BNCs, DINs, XLRs, bananas, 13A plugs, fuses, - they're all 'shit', some more so than others. But it's unfortunately "shit" most of us have to put up with.

Marco.

Mike
23-09-2009, 16:22
I thought you were a bit cleverer than that! What, because I asked a question!!!!!!

physically it 'makes' signal before screen, and breaks screen before signal. No, I'm not being funny here, I don't see that as a technical issue? But I am interested in why people think they are a problem?

On all the digital stuff I have here, the manufacturers fitted BNC's as standard on the digital bits and XLR's for balanced audio.

Regards

Dave

Perhaps the 'manufacturer' also thinks phonos are shite then? :eyebrows: :)

They were designed back in the 40's, mainly to be cheap I reckon. Though, as usual, the audio industry has found ways and means to make them expensive! :lol:

They're just poor, and well past their 'best before date'.

Dave Cawley
24-09-2009, 06:34
I'm not in a position to judge what the manufactures think. When BNC is the broadcast standard why would they fit phono? But for audio nothing on this thread has suggested why phono connectors are not OK, not even any anecdotal evidence. Simply saying they are poor without qualifying it, isn't really good enough. My original question "What is wrong with them?" still stands, but I don't expect to be personally criticised, I expect a rational discussion?

Can we be a bit more focused guys?

Regards

Dave

NRG
24-09-2009, 08:06
Because they are not impedance matched?

The Grand Wazoo
24-09-2009, 08:10
May I refer the honourable gentleman to the thread I made earlier about Camac connectors:
http://theartofsound.net/forum/showthread.php?t=2740

In it there's a list of some of the features of Camacs. Read it & count the differences with RCA phonos as they were originally conceived:


They allow the cable to rotate through 360 degrees in relation to the plug body.

Contact areas are heavily gold plated.

They allow for earth connection to be made before live connection.

They have a self-locking mechanism.

The strain relief system doesn't pinch or distort the signal conductor/s.

The sockets are flush mounted

As long as you can use a soldering iron you can make up cables with Camacs.

I make it 6 out of 7.

NRG
24-09-2009, 08:16
...and not forgeting Barry's slating of RCA's in his post. ;)

Dave Cawley
24-09-2009, 08:25
Well, that shows, is that there are better connectors if you pay more, but not what is actually wrong with phono connectors for audio? I think we take it as read that for digital data and clocks, that equipment already uses different connectors, all my stuff does.

Dave

NRG
24-09-2009, 08:30
Your stuff may Dave but I'd lay money that the majority of manufacturers producing digital consumer products use RCA...

Dave Cawley
24-09-2009, 08:36
Your stuff may Dave but I'd lay money that the majority of manufacturers producing digital consumer products use RCA...

OK, that is bad! Phono's were never designed for that, and it is inappropriate to use them outside of their specification.

Dave

The Grand Wazoo
24-09-2009, 08:41
My original statement was that they're a fundamentally poor design. I apologise for not qualifying my statement - I thought it was generally accepted that phonos are pretty shonky.

You asked why I thought they were inferior & I've pointed out 6 reasons for this. Doesn't that answer your question?

(Please let it be noted, Dave that I have not personally criticised anyone)

Dave Cawley
24-09-2009, 08:51
Phono connectors were never designed for digital, so using them is silly, it is the manufactures that should be criticised, not the connector!

Yes there are better specified connectors, but that doesn't suddenly make phono bad, poor or inappropriate. Most phono's are gold, it is good practice never to unplug anything live, there is no strain so self locking is not required, many have non invasive strain relief, who cares about flush, and you can solder phono's easily.

I'm struggling to find and reason why phono's are not good for audio. A better argument is why don't more manufacturers adopt the recording studio approach and use three wire fully balanced systems/connectors?

Dave

Tony Moore
24-09-2009, 09:01
Some useful stuff?

http://www.hardwarebook.info/S/PDIF

It's in the IEC 60958-3 spec to use RCA phonos so I guess the manufacturers are only doing what they all agreed to do when they created the standard, which was for a cheap consumer alternative to AES/EBU. Still, they _could've_ done a better job of something that's had to last so long! :steam:

Cheers,
Tony

The Grand Wazoo
24-09-2009, 09:05
Well, we'll have to agree to differ.
My position was that they are a flawed design, not that they don't do the job - we all know they do the job. Somewhat parallel to your last sentence.....


A better argument is why don't more manufacturers adopt the recording studio approach and use three wire fully balanced systems/connectors?

........all I am trying to do is point out that we use a flawed design while there are better alternatives. Never mind the other advantages, I have heard differences between phonos & Camacs - I preferred the Camacs. Now that may be because of something so stupid as that I made a better job of soldering the Camacs than the RCA's, but I heard a difference that I preferred.

Dave Cawley
24-09-2009, 09:06
Phono's were never ever designed for digital, so trying to retrospectively impose standards on them is ludicrous! only Dr Who could achieve that, but then it would violate the prime directive. BNC's though, were always capable of digital and "should be" the preferred and only standard today.

Dave

Dave Cawley
24-09-2009, 09:08
I preferred the Camacs And I suspect if you tried a properly designed balanced system you would prefer that, but it wouldn't make the Camacs a poor design?

Dave

The Grand Wazoo
24-09-2009, 09:16
I preferred the Camacs And I suspect if you tried a properly designed balanced system you would prefer that, but it wouldn't make the Camacs a poor design?

Dave

No, they wouldn't suddenly become a poor design. That's because they do everything I expect a well designed connector to do.

Phonos, however, do not do everything I expect a well designed connector to do, therefore, that indicates to me that they are poorly designed.

Dave Cawley
24-09-2009, 09:20
The phono is a well designed connector, it's just not designed for what you think it should be. The Camacs are merely a half way house to a properly designed balanced system. And as I have explained not any better than a premium phono connector.

Dave

The Grand Wazoo
24-09-2009, 09:36
As I said, we'll have to agree to differ. Obviously, the place where we differ is in our expectations of what is 'better'. Either way, we're pretty much stuck with the RCA phono connector.

Dave Cawley
24-09-2009, 10:00
stuck with the RCA phono connector Yes we are! But there are XLR and BNC and Camac!

Dave

Spectral Morn
24-09-2009, 15:11
As most of my gear is balanced I use them that way thus using XLR. However a pet hate of mine is the way in which there is no industry standard for the mounting of the sockets, by this I mean the way they face. I don't know about you guys, but having to bend XLR cables to fit, can and is a pain in the ass; and sometimes leads to breaks or plugs moving (Audience cables are notorious for this. I have had a number break on me, no matter how careful I have been). With the RCA this is not an issue.

WBT Next Gen plugs which were hailed as a major break through in RCA design are IMHO no such thing. Yes they may cut down on metal and allow contacts to be made and broke at the right time (as well as being available as a genuine75 Ohm plug), but they are flimsy and easy to break. Their innards can also move out of alignment thus making them hard to fit. Sound better yes, perhaps, but the flaws IMHO make them no saviour for audio connections.

I would like to see an XLR industry wide mounting standard so all XLR sockets face the right/same way. All digital connections should be BNC...the only true standard.

Just my few pence worth.


Regards D S D L

Dave Cawley
24-09-2009, 15:17
Neil

I 100% agree with 100% of your post. I think I ought to go and lie down now.....

Dave

markf
24-09-2009, 16:09
The RCA phono is good to 10Mhz ,if you need higher go for BNC at 2Ghz,higher still go for SMA that's good to 15Ghz.

Spectral Morn
24-09-2009, 16:28
Neil

I 100% agree with 100% of your post. I think I ought to go and lie down now.....

Dave

Me too :)


Regards D S D L

Mike
24-09-2009, 16:34
I'm not in a position to judge what the manufactures think. When BNC is the broadcast standard why would they fit phono? But for audio nothing on this thread has suggested why phono connectors are not OK, not even any anecdotal evidence. Simply saying they are poor without qualifying it, isn't really good enough. My original question "What is wrong with them?" still stands, but I don't expect to be personally criticised, I expect a rational discussion?

Can we be a bit more focused guys?

Regards

Dave

Where has this happened?

Primalsea
24-09-2009, 17:51
Put it this way,

SPDIF is probably one of the worst ways to go about transmitting a digital signal.

Using a voltage to transmit an audio signal is probably one of the worst ways to go about transmitting an audio signal.

IEC plugs and sockets are probably one of the worst ways to go about connected the mains.

However they have become the standard.

Why?

Simple!

They never asked us first ;)

My opinion is that Phono sockets are not ideal from a longevity standpoint. They were never designed with this in mind. They are simple to connect and (can be) cheap.

Barry
24-09-2009, 19:35
Gentlemen,

I have been advised of this thread by Chris (The Grand Wazoo) and since a post of mine has been cited, I perhaps ought to add my two pence worth.

This thread seems to consist of two parts: the desirability of a 75Ω digital system impedance and suitable connectors, and the quality or otherwise of RCA phono style connectors.

I shall try to make my comments roughly in the order of posting.


An impedance of 75Ω seems to have been chosen as the de facto system impedance for digital transfers, and as such, it makes sense that any interconnects used have a characteristic impedance of 75Ω, subject to caveats that I will mention later.

In theory, hardwiring all items of equipment ought to be preferable, however I believe the audible improvement would be miniscule, if not inaudible, and would hardly justify the practical inconvenience created in removing the connecting interface. Reference to the practice of ‘hardwiring’ circuitry is misleading; point-to-point wiring is sometimes chosen rather than the use of PCB techniques, to avoid problems of circulating return currents. These problems will also occur if the chassis is used as a return and are not unique to the employment of PCBs. To circumvent these problems either a ground bus bar is used or star earthing; techniques which can be also be employed in PCB designs.

The connecting tracks between a 75Ω impedance connector and the circuit connected to it can easily be made to have a 75Ω impedance if required, provided the dielectric properties of the board is known (for example FR4 has a dielectric constant of 4.7 at low frequencies), but again rigorous adherence to a characteristic impedance is not essential; small deviations over a short distance will have little effect. Incidentally the use of time domain reflectometry to measure mismatches separated by short distances depends on the pulse rise time of the TDR. To resolve two mismatches separated by 2mm on a circuit made on FR4, would require the TDR to have a pulse rise time of 45ps!

Having accepted a system interface impedance of 75Ω, then not only the cable but the connectors ought to have this characteristic impedance. The cable is not a problem; either RG6 or RG59 coaxial is suitable, or Huber and Suhner G 02330-HT triaxial cable (wherein the impedance between the inner conductor and the first screen is 75Ω) can be used.

Ideally BNC connectors having a characteristic impedance of 75Ω should be used, as I don’t believe that RCA phono plugs can ever have a high enough characteristic impedance. My argument for this was made in another thread:

http://theartofsound.net/forum/showthread.php?t=3519&highlight=75ohms&page=5

(post 46), and points out that even for a sampling rate of 192kHz, the concept of characteristic impedance is questionable for reasonably short cable lengths.

I am suspicious of the claim that Tiffany makes for their 75Ω phono plugs. If the deviation from a coaxial geometry is used to raise the impedance, then I would like to see measurements and I believe that this introduces problems of its own; similar to that of the ‘bullet plug’ variant of the RCA phono plug. Which brings me to the second part of this thread.


In a thread started by Chris (TGW):

http://theartofsound.net/forum/showthread.php?t=2740

I supported his opinion that the RCA phono connector is a poor design that unfortunately has been adopted in the (domestic) audio industry. In my response, posts #2 and #4 (and I ought to apologise here for the typos and spelling mistakes – I wrote it in a hurry), I stated that my complaint of the phono plug was not that it doesn’t work; of course it works, but then so would the screw terminals or wander plug connections that used to be used in the ‘40s and ‘50s. In fact they were designed to be a cheap and simple connector to allow a phonogram (record player) to be connected to the audio amplifier of a domestic radio. Rather, having adopted a mediocre design, the industry has ‘tarted it up’ by offering exotic plating and improved build quality, and often charges an excessive amount for their efforts. (Have a look to see how much Furukawa charge for their versions). In Chris’s words: ‘You can’t polish a turd’!

My championship of the CAMAC connector is based on the fact that it is a well designed precision connector, used in nuclear, medical, avionics, and instrumentation equipment where a reliable (i.e. life dependant for medical equipment) high quality connection needs to be made. No, they are not cheap, but cheaper than Furukawa, Tiffany or WTB, and are no more expensive than say a Vishay bulk foil resistor or any exotic capacitor that enthusiasts insist upon.

It is indeed a shame that they were not adopted by the audio industry; however since we are stuck with the phono connector, I use the best I can find: Neutrix ‘ProFi’, designed to make and break the outer conductor first, and when mated grip like there’s ‘no tomorrow’. I also dislike ‘bullet plugs’, which to my mind take a poor design and make it worse.

Balanced line operation is a good technique (especially for long cable lengths), however it is not obvious that a priori it is a superior system. Fortunately the XLR connectors are a good design; unfortunately there is no uniformly adopted convention on how to use them in unbalanced systems.

Mention has been made of the SMA connector. I have seen these used on some audio equipment but they are not to be recommended, for the simple reason that they are not designed for multiple matings.

If interested, there is an item on audio connectors in the AoS Library, where more details can be found.

Regards

NRG
24-09-2009, 20:31
Barry, in your post on the other thread (#46) you seem to be saying that impedance matching is not important and then in the next paragraph you seem to say it is....please clarify....as most devices capable of SPDIF output (that I know of) will have made some sort of attempt at matching a 75ohm o/p impedance...

I'd also be interested in your thoughts on this thread from some time ago...

http://www.audiocircle.com/index.php?PHPSESSID=2mt2vp5ccp76n3oqrofal5dtivi7o4 r8&topic=45330.0

...as to my ears after implementing the ideas posted in that thread impedance matching does make a difference...a very noticeable difference for the better....

Mike
24-09-2009, 21:18
Put it this way,

SPDIF is probably one of the worst ways to go about transmitting a digital signal.

Using a voltage to transmit an audio signal is probably one of the worst ways to go about transmitting an audio signal.

IEC plugs and sockets are probably one of the worst ways to go about connected the mains.

However they have become the standard.

Why?

Simple!

They never asked us first ;)

My opinion is that Phono sockets are not ideal from a longevity standpoint. They were never designed with this in mind. They are simple to connect and (can be) cheap.

Indeed!.. whats wrong with AES/EBU anyway?... It would have been much better to stick with the one standard rather than muddying the waters and introducing yet another 'standard'! :confused:


Gentlemen,

I have been advised of this thread by Chris (The Grand Wazoo) and since a post of mine has been cited, I perhaps ought to add my two pence worth.

This thread seems to consist of two parts: the desirability of a 75Ω digital system impedance and suitable connectors, and the quality or otherwise of RCA phono style connectors.

I shall try to make my comments roughly in the order of posting.


An impedance of 75Ω seems to have been chosen as the de facto system impedance for digital transfers, and as such, it makes sense that any interconnects used have a characteristic impedance of 75Ω, subject to caveats that I will mention later.

In theory, hardwiring all items of equipment ought to be preferable, however I believe the audible improvement would be miniscule, if not inaudible, and would hardly justify the practical inconvenience created in removing the connecting interface. Reference to the practice of ‘hardwiring’ circuitry is misleading; point-to-point wiring is sometimes chosen rather than the use of PCB techniques, to avoid problems of circulating return currents. These problems will also occur if the chassis is used as a return and are not unique to the employment of PCBs. To circumvent these problems either a ground bus bar is used or star earthing; techniques which can be also be employed in PCB designs.

The connecting tracks between a 75Ω impedance connector and the circuit connected to it can easily be made to have a 75Ω impedance if required, provided the dielectric properties of the board is known (for example FR4 has a dielectric constant of 4.7 at low frequencies), but again rigorous adherence to a characteristic impedance is not essential; small deviations over a short distance will have little effect. Incidentally the use of time domain reflectometry to measure mismatches separated by short distances depends on the pulse rise time of the TDR. To resolve two mismatches separated by 2mm on a circuit made on FR4, would require the TDR to have a pulse rise time of 45ps!

Having accepted a system interface impedance of 75Ω, then not only the cable but the connectors ought to have this characteristic impedance. The cable is not a problem; either RG6 or RG59 coaxial is suitable, or Huber and Suhner G 02330-HT triaxial cable (wherein the impedance between the inner conductor and the first screen is 75Ω) can be used.

Ideally BNC connectors having a characteristic impedance of 75Ω should be used, as I don’t believe that RCA phono plugs can ever have a high enough characteristic impedance. My argument for this was made in another thread:

http://theartofsound.net/forum/showthread.php?t=3519&highlight=75ohms&page=5

(post 46), and points out that even for a sampling rate of 192kHz, the concept of characteristic impedance is questionable for reasonably short cable lengths.

I am suspicious of the claim that Tiffany makes for their 75Ω phono plugs. If the deviation from a coaxial geometry is used to raise the impedance, then I would like to see measurements and I believe that this introduces problems of its own; similar to that of the ‘bullet plug’ variant of the RCA phono plug. Which brings me to the second part of this thread.


In a thread started by Chris (TGW):

http://theartofsound.net/forum/showthread.php?t=2740

I supported his opinion that the RCA phono connector is a poor design that unfortunately has been adopted in the (domestic) audio industry. In my response, posts #2 and #4 (and I ought to apologise here for the typos and spelling mistakes – I wrote it in a hurry), I stated that my complaint of the phono plug was not that it doesn’t work; of course it works, but then so would the screw terminals or wander plug connections that used to be used in the ‘40s and ‘50s. In fact they were designed to be a cheap and simple connector to allow a phonogram (record player) to be connected to the audio amplifier of a domestic radio. Rather, having adopted a mediocre design, the industry has ‘tarted it up’ by offering exotic plating and improved build quality, and often charges an excessive amount for their efforts. (Have a look to see how much Furukawa charge for their versions). In Chris’s words: ‘You can’t polish a turd’!

My championship of the CAMAC connector is based on the fact that it is a well designed precision connector, used in nuclear, medical, avionics, and instrumentation equipment where a reliable (i.e. life dependant for medical equipment) high quality connection needs to be made. No, they are not cheap, but cheaper than Furukawa, Tiffany or WTB, and are no more expensive than say a Vishay bulk foil resistor or any exotic capacitor that enthusiasts insist upon.

It is indeed a shame that they were not adopted by the audio industry; however since we are stuck with the phono connector, I use the best I can find: Neutrix ‘ProFi’, designed to make and break the outer conductor first, and when mated grip like there’s ‘no tomorrow’. I also dislike ‘bullet plugs’, which to my mind take a poor design and make it worse.

Balanced line operation is a good technique (especially for long cable lengths), however it is not obvious that a priori it is a superior system. Fortunately the XLR connectors are a good design; unfortunately there is no uniformly adopted convention on how to use them in unbalanced systems.

Mention has been made of the SMA connector. I have seen these used on some audio equipment but they are not to be recommended, for the simple reason that they are not designed for multiple matings.

If interested, there is an item on audio connectors in the AoS Library, where more details can be found.

Regards

Barry, thank you for a superbly written post. I wish I had the patience, however, I shall endeavour to be more eloquent in future.

Barry
25-09-2009, 00:02
Barry, in your post on the other thread (#46) you seem to be saying that impedance matching is not important and then in the next paragraph you seem to say it is....please clarify....as most devices capable of SPDIF output (that I know of) will have made some sort of attempt at matching a 75ohm o/p impedance...

I'd also be interested in your thoughts on this thread from some time ago...

http://www.audiocircle.com/index.php?PHPSESSID=2mt2vp5ccp76n3oqrofal5dtivi7o4 r8&topic=45330.0

...as to my ears after implementing the ideas posted in that thread impedance matching does make a difference...a very noticeable difference for the better....

Hi Neal,

First let me apologise for the apparent and unintended contradiction of my post (# 46). What I was trying to say was that regardless of the system impedance or characteristic impedance of a transmission line (Zo), the effect of small deviations from this impedance (ΔZ) occurring over a short distance (Δx) would have only a small effect. The reflection coefficient (|ρ|) caused by this deviation can be shown to be given by:

|ρ| = (2πΔx/λ).(ΔZ/Zo),

where λ is the wavelength in the perturbed region. For a 192kHz signal, λ = 1561.42m in air. In general Δx is only a few mm, small compared with 1.5km! I claimed that a phono plug has a calculated characteristic impedance of ~ 41Ω, and this deviation from the 75Ω characteristic impedance, occurs over about 25mm at most. Thus ΔZ = 34Ω, Zo = 75Ω, Δx = 25mm and λ = 1561.42m, so

|ρ| = 4.56 10^-5,

or a mismatch of -86.8dB. Now you might argue that ΔZ = 34Ω is not small compared with Zo = 75Ω. However if your do the calculation properly the result is little different, simply because in this case Δx, really is small compared with λ. Physics lecture over!

What I was also saying that for cable lengths that are very short compared to the wavelength, the concept of characteristic impedance looses its validity. I went on to say that even at 192kHz where the wavelength in a Teflon insulated cable is 1077.5m, then for cable lengths of a couple of metres, the concept is somewhat moot. I did go on to say that having settled on a system impedance of 75Ω, it would make sense to maintain that impedance through the interconnect and, if possible, the connectors.

From your experience and others, the adherence to 75Ω would seem to be important and I would not disagree with you. Choose the cables to be of the correct impedance, but if the connectors are not, it shouldn't matter too much.

I ought to say that I have virtually no knowledge of digital circuitry so cannot really comment on the thread given by your link. All I can say is that the author presents a TDR trace showing the effect of separating the inner and outer conductor of a 75Ω coaxial cable (RG179 and RG187 are mentioned) in order to connect it to a (presumably) 75Ω BNC socket. Assuming that this separation occurs over an estimated distance 6.5mm and the wavelength (in air) is 0.3m (he mentions a bandwidth > 1GHz ), then Δx/λ = 0.022. Again I have assumed where the conductors are attached to the BNC, they can be treated as a two wire pair having diameter of 2mm and separation of 4mm; the characteristic impedance of such an arrangement is 211Ω. Performing the above calculation gives a reflection coefficient of 0.25 (at 1GHz), compared with the displayed value of 0.06. To fully resolve a distance of 6.5mm in air, his TDR would need to have a bandwidth of > 4GHz.

Regards

HighFidelityGuy
28-09-2009, 16:44
Wow, thanks for taking the time to go into this in so much detail Barry!! :)

As maths isn't one of my strong points, I don't fully understand everything you said but I think the conclusion is that we don't need to worry too much about slight impedance mis-matches caused my connectors etc as they would only cause problems with very very long cables. However, seen as the transmission method is designed to work at 75 Ohms, we should try to make as much of the signal path as possible 75 Ohms to avoid any problems. Please correct me if I've mis-summarized. :)

So from that information and what others have said, it seems like the best compromise is to use BNC connectors wherever possible, as using directly soldered cables would be too difficult to live with to make the small benefits worth the hassle.

Primalsea
28-09-2009, 18:04
Hi Barry, Nice post I bow before your oracle-like knowledge. I also thank you as I was alos looking for something to read in bed to help me sleep. lol

Dare I ask but what effect does output impedance have on the signal transmission. Are we to assume that all items of hifi equipment actually transmit spdif signal at 75ohm OP impedance?

Barry
28-09-2009, 18:12
............

I think the conclusion is that we don't need to worry too much about slight impedance mis-matches caused my connectors etc as they would only cause problems with very very long cables. However, seen as the transmission method is designed to work at 75 Ohms, we should try to make as much of the signal path as possible 75 Ohms to avoid any problems. Please correct me if I've mis-summarized. :)

So from that information and what others have said, it seems like the best compromise is to use BNC connectors wherever possible, as using directly soldered cables would be too difficult to live with to make the small benefits worth the hassle.

Hi Dave,

I should have summarised my points, but you have more or less got it:

Use 75 ohm cable if possible, but don't lose any sleep if you terminate them with RCA phonos rather than BNCs.

The mismatch in using phonos is small because the distance over which the mismatch occurs is small relative to the wavelength (in the perturbed region).

Regards

Tony Moore
28-09-2009, 18:14
Are we to assume that all items of hifi equipment actually transmit spdif signal at 75ohm OP impedance?

That's the ideal but in reality probably not!

:doh:

ps. Sorry to butt in!

NRG
28-09-2009, 18:43
Yes, thank you Barry for the detailed reply. However, I think there is more to this than you have posted...and certainly from my experience and others here (using Mikes Cables) attention to the impedance pays rewards. Some searching on the internet about cable reflections threw up a few interesting links, this one in particular is an easy enough read and explanation without going into heavy technical detail

http://www.positive-feedback.com/Issue14/spdif.htm

The same conclusions are reflected (!) in a number of DIY Audio threads from a few years ago where it was said signal reflections, due to very fast leading edge rise times, can cause problems even with the, in theory, very short impedance mismatch that an RCA connector represents to the signal.

Barry
28-09-2009, 19:36
Hi Neal,

As I said, I have virtually no knowledge of digital circuitry, my experience is with microwave circuitry operating >1GHz.

Your link suggests CD transport readers have a rise time of 25ns (10% to 90%). This would suggest that the frequency invoved is 5.74MHz, then the wavelength in Teflon insulated cable is 36.1m. The wavelength/(2*pi) = 5.73m, so sensible cable lengths (1 - 1.5m) are going to be a sizable fraction of this, therefore reflections from impedance mismatches are going to be important.

My mistake seems to have been assuming the frequency involved was the sampling frequency of 192kHz (at most). I did not consider the fast rising pulse edge.

I think I had better step down from this discussion as my radar knowledge would seem to be inappropiate.

So, use 75 Ohm cable and 75 Ohm BNCs !

Regards

Tony Moore
28-09-2009, 20:00
44.1Khz * 2 (channels) * 24 bits = 2.1Mhz

44.1Khz * 2 * 32 bits = 2.8Mhz

Most are encoded with 32 bits.

Both * 2 for biphase mark encoding = 5.6Mhz

48 Khz would be 6.2Mhz

192Khz would be 25Mhz! :eyebrows:

Barry
28-09-2009, 23:05
Don't understand, Tony what that means. Is that the bit rate you've calculated?

44.1kHz * 32 bits * 2 channels = 2.8Mbps ?

Said I didn't know much about digital circuitry.

Regards

NRG
29-09-2009, 07:02
Barry, that is the raw signal rate including overhead. SPDIF data is transmitted in a 32bit data word (sub frame)...Each is made up of 20 bits for the audio data and the rest for sync/preamble, parity, status etc. These sub frames are interleaved and used to make up a single data frame...192 for left channel and 192 for right channel.

As each sub frame is 32bits transmitted at 44.1Khz and bi-phase encoded the actual signal bit rate is 2.8MHz....the data rate is therefore 20bits * 44.1Khz or 882Kbs

Tony Moore
29-09-2009, 07:30
Hi Barry,

Sorry for my rushed post last night! :o

Actually, I believe I stated the raw bit clock signal frequency of the I2S data rather than the s/pdif frequency. As Neil says though the s/pdif actually only transmits 20bits of data and the rest of the frame are control bits. Sorry for any confusion, I'll hold off posting after drinking wine in future! :cool:

;)

Cheers,
Tony

Primalsea
29-09-2009, 18:26
Have a look on the TNT website at the DAC project on there. It explains how you get the frequency from the sample rate and no of bits.

Barry
29-09-2009, 20:12
Hi Guys, (Neal, Tony and Paul)

Thanks for the explanation. One of the nice things about AoS is the educational aspect - I'm learning all the time.

My background (nearly 30 years) has been in electronics research (I took early retirement three years ago), my speciality being RF and microwave filter design for use in radar and communication systems. It was purely analogue and my knowledge of digital techniques virtually non existant.

I was unaware of the fast rise time of the sampling waveforms, and so had not appreciated the very high effective frequencies involved. My argument up until now has been that even at frequencies of 192kHz (now known to erroneous), the treatment of connecting cables as transmission lines, the concept of characteristic impedance and the consideration of reflection from impedance mismatches was not relevant. With effective frequencies in the MHz range, and for cable lengths of a metre or so, then the afore mentioned considerations do indeed have to be taken into account. I should also add that expressing bit rates in Hz also confused me.

So now I understand that adherence to the system impedance is very important and that this ought to include the connectors involved.

I'm now going to bow out of this thread as I don't think I'll be able to contribute anything of use. I shall of course continue to read any future posts as I'm still learning!

Thanks

Regards

Mike
30-09-2009, 15:48
Hi Guys, (Neal, Tony and Paul)

Thanks for the explanation. One of the nice things about AoS is the educational aspect - I'm learning all the time.

My background (nearly 30 years) has been in electronics research (I took early retirement three years ago), my speciality being RF and microwave filter design for use in radar and communication systems. It was purely analogue and my knowledge of digital techniques virtually non existant.

I was unaware of the fast rise time of the sampling waveforms, and so had not appreciated the very high effective frequencies involved. My argument up until now has been that even at frequencies of 192kHz (now known to erroneous), the treatment of connecting cables as transmission lines, the concept of characteristic impedance and the consideration of reflection from impedance mismatches was not relevant. With effective frequencies in the MHz range, and for cable lengths of a metre or so, then the afore mentioned considerations do indeed have to be taken into account. I should also add that expressing bit rates in Hz also confused me.

So now I understand that adherence to the system impedance is very important and that this ought to include the connectors involved.

I'm now going to bow out of this thread as I don't think I'll be able to contribute anything of use. I shall of course continue to read any future posts as I'm still learning!

Thanks

Regards

Hence why I test my cables at 645Mbps... It's nice to have a bit of 'headroom'! :)

Dave Cawley
30-09-2009, 17:12
645Mbps Wow! could you post a waveform for us analogue folks please?

Thanks

Dave

Mike
30-09-2009, 17:52
I'm afraid not! :)

I test the cables with a 'BERT' (Bit Error Ratio Tester). It pours data in and measures what comes out and gives a result in the form of a number, I either see a zero or it goes in the bin! ;)

Simples! (squeek)

Mike
30-09-2009, 17:58
Oops!

Just noticed I typed 645 when I meant 622... or more accurately 622.080Mbps. :(

Dave Cawley
30-09-2009, 18:44
Nice one, which model tester do you use, I'm a sucker for test gear!

Regards

Dave

Primalsea
30-09-2009, 18:49
Mike,

Do you think there is any mileage in SATA cables, they're traditionally balanced but go upto 3GBPS!!

Marco
30-09-2009, 19:43
Nice one, which model tester do you use, I'm a sucker for test gear!


Oh you don't say - tell 'em about the vintage 'collection' in your workshop... :eyebrows:

Marco.

Dave Cawley
30-09-2009, 19:46
Oh you don't say - tell 'em about the vintage 'collection' in your workshop... :eyebrows:

Marco.


Exactly! Hence my question to Mike. Come on Mike, spill the beans!

Dave

Mike
30-09-2009, 19:58
Nice one, which model tester do you use, I'm a sucker for test gear!

Regards

Dave

I can't remember the model numbers off the top of my head, there are a couple I have access to. I'll note them down when I'm at work tomorrow and let you know. Theres a couple of HP jobs and 'a big blue and black thing' ;)... might be W&G. :scratch:

They're 'Telecoms' testers we use to error test SDH links. 622Mbps for STM4 and 155Mbps for STM1

Dave Cawley
30-09-2009, 20:22
I'll note them down when I'm at work tomorrow and let you know.

Great Mike, I love measurements, they are soooo traceable as Avril Lavigne (http://www.avrillavigne.com/)whispered in my ear recently.

Thanks

Dave