I can let you hear of it - in my experience with digital audio (which goes back more than 20 years, but is not contiguous over that time, I've taken breaks for several years) controlling common mode noise in a system is more of a challenge than controlling jitter. And common mode noise does tend to make the sound fatiguing to listen to (sibilant generally) whereas jitter does not. So when engineering a design I'll always focus on CM noise first and only after that move on to jitter issues. YMMV 😀
Hello guys, long time no see 😀
Speaking about CM noise, supposing the following :
- DAC has a clean local clock
- either source is slaved to DAC
- or DAC's clock is tracking source clock (a la Thorsten's PLL which isn't really a PLL) with good jitter rejection and FIFO
Then, why not use TOSLINK or other optical links ? Those are pretty good against common mode noise I guess. They have the reputation of crummy jitter performance, but that should only be a worry if the clock recovered from that is actually used in the DAC (which is against the suppositions above).
Could you give more details about the measurement setup and gear please ?
Speaking about CM noise, supposing the following :
- DAC has a clean local clock
- either source is slaved to DAC
- or DAC's clock is tracking source clock (a la Thorsten's PLL which isn't really a PLL) with good jitter rejection and FIFO
Then, why not use TOSLINK or other optical links ? Those are pretty good against common mode noise I guess. They have the reputation of crummy jitter performance, but that should only be a worry if the clock recovered from that is actually used in the DAC (which is against the suppositions above).
Could you give more details about the measurement setup and gear please ?
Indeed great to see you back again peufeu after all this time 🙂
Yep, I had been thinking along similar lines myself 😀 Can't fault your reasoning there. How current hungry is a TOSLINK receiver?
Hi,
Yup. My biggest issue is that Toslink is hit and miss even with supposedly 192KHz components to actually work at that. And that is before we start accounting for all the fakes in the market that somehow make it intoproducts made by "competitors" (we found out the hard way when products we purchased as 192KHz SPDIF sources including optical simply did not work at quad speed until we changed the Toslink transmitters).
Using transformer isolation can work well, it is just necessary to consign the datasheet straight to the dustbin and to characterise the transformer yourself.
You need to find the LC resonance formed between leakage inductance and parasitic capacitance, this determines you maximum bandwidth (with any transformer really, knowing audio transformers well helps tremendously when dealing with them elsewhere).
You need to then load the secondary winding with the critical impedance that correctly damps this resonance (which I have found is NEVER EVER anything close to 75 Ohm, incidentally, in transformers the manufacturer rated at 75 Ohm).
You now have a very interesting input impedance that is not flat with frequency, which you need to apply SMD (the smaller the better and only do this in the final PCB layout) components to make it look to source and cable like a nice 75R load. At the minimum it means a resistor and a Zobel, more may be needed, depending on bandwidth and the quality (or lack thereof).
But here is the one to boil your noodles. Set things up as described above, clip on your 1:10 scope probe on and the SPDIF waveform (if source, connectors and cable are correctly 75 Ohm) even at 192KHz looks so textbook, so perfect, it is pretty much just like a pair of 0805 150 Ohm resistors soldered directly to a to a BNC socket as termination.
Now connect your receiver chip (even if a mere fraction of an inch from the whole input shebang) and watch your nice looking SPDIF textbook signal go STRAIGHT TO HELL. It is really illuminating.
If you are blessed with a strong stomach, repeat the same test using the above transformer terminated at it's secondary with 75 Ohm and the Receiver attached. The use the transformer driven from 75 Ohm in the source if you do not believe it can get worse.
"Lo and behold, a great wailing and gnashing of teeth was heard in the land when they beheld the waveform."
Using most commercial/commodity grade gear with SPDIF is a really bad proposition. If you don't believe it, get a nice fast scope and have a look.
Normally, for DIY'ers who lack the means to deal with all that RF stuff well (and do not wish to aquire the test gear and knowledge needed) and just need a quick and easy "patent recipe" I recommend:
1) skip transformers just 75R and the required caps for the input
2) use either very short or very long SPDIF cables (so the unavoidable impedance mismatches are somewhat disarmed in their effect on the actual trigger point)
3) Use a simple purely Resistor/Capacitor output from the source
4) If using Receivers that are AES/EBU by design and abused as SPDIF receivers (Cirrus Logic, AKM) allow as much level as possible, normally 2.5V PP when the source runs of 5V lines and 1.65V PP when it runs of 3.3V.
If the receiver uses "trick circuitry" such (Jocko Homo's) that actually works (most don't and are best ripped out) then use a divider that gets as close as possible to the 0.5V PP Level (into a 75 Ohm Load) that is the official standard.
5) Find another way to deal with the ground loops, and related issues (this is DIY after all), which is actually not that hard (certainly less hard than making a decent dual transformer coupled SPDIF link).
Ciao T
Yup. My biggest issue is that Toslink is hit and miss even with supposedly 192KHz components to actually work at that. And that is before we start accounting for all the fakes in the market that somehow make it intoproducts made by "competitors" (we found out the hard way when products we purchased as 192KHz SPDIF sources including optical simply did not work at quad speed until we changed the Toslink transmitters).
Using transformer isolation can work well, it is just necessary to consign the datasheet straight to the dustbin and to characterise the transformer yourself.
You need to find the LC resonance formed between leakage inductance and parasitic capacitance, this determines you maximum bandwidth (with any transformer really, knowing audio transformers well helps tremendously when dealing with them elsewhere).
You need to then load the secondary winding with the critical impedance that correctly damps this resonance (which I have found is NEVER EVER anything close to 75 Ohm, incidentally, in transformers the manufacturer rated at 75 Ohm).
You now have a very interesting input impedance that is not flat with frequency, which you need to apply SMD (the smaller the better and only do this in the final PCB layout) components to make it look to source and cable like a nice 75R load. At the minimum it means a resistor and a Zobel, more may be needed, depending on bandwidth and the quality (or lack thereof).
But here is the one to boil your noodles. Set things up as described above, clip on your 1:10 scope probe on and the SPDIF waveform (if source, connectors and cable are correctly 75 Ohm) even at 192KHz looks so textbook, so perfect, it is pretty much just like a pair of 0805 150 Ohm resistors soldered directly to a to a BNC socket as termination.
Now connect your receiver chip (even if a mere fraction of an inch from the whole input shebang) and watch your nice looking SPDIF textbook signal go STRAIGHT TO HELL. It is really illuminating.
If you are blessed with a strong stomach, repeat the same test using the above transformer terminated at it's secondary with 75 Ohm and the Receiver attached. The use the transformer driven from 75 Ohm in the source if you do not believe it can get worse.
"Lo and behold, a great wailing and gnashing of teeth was heard in the land when they beheld the waveform."
Using most commercial/commodity grade gear with SPDIF is a really bad proposition. If you don't believe it, get a nice fast scope and have a look.
Normally, for DIY'ers who lack the means to deal with all that RF stuff well (and do not wish to aquire the test gear and knowledge needed) and just need a quick and easy "patent recipe" I recommend:
1) skip transformers just 75R and the required caps for the input
2) use either very short or very long SPDIF cables (so the unavoidable impedance mismatches are somewhat disarmed in their effect on the actual trigger point)
3) Use a simple purely Resistor/Capacitor output from the source
4) If using Receivers that are AES/EBU by design and abused as SPDIF receivers (Cirrus Logic, AKM) allow as much level as possible, normally 2.5V PP when the source runs of 5V lines and 1.65V PP when it runs of 3.3V.
If the receiver uses "trick circuitry" such (Jocko Homo's) that actually works (most don't and are best ripped out) then use a divider that gets as close as possible to the 0.5V PP Level (into a 75 Ohm Load) that is the official standard.
5) Find another way to deal with the ground loops, and related issues (this is DIY after all), which is actually not that hard (certainly less hard than making a decent dual transformer coupled SPDIF link).
Ciao T
(Here's me wading into areas I don't know again.)
With the fibre optics used for ethernet links etc. we're talking about gigabits per second aren't we? Why mess about with anything else?
A couple of reasons come to mind - price, and power consumption. Not to mention compatibility with existing kit.
Well people have generally been talking about SPDIF and USB here. Both of those are interesting because they're standards. Can't quite work out where you're coming from with that question?😱 Care to elucidate?
The Hiface Evo usues ST optical out - which is said to be superior to standard implementation - but only a few receivers use ST optical.
How much of a real problem is common mode noise in practise? - I don't believe I've experienced it myself - what would it equate to sonically?
How much of a real problem is common mode noise in practise? - I don't believe I've experienced it myself - what would it equate to sonically?
Last edited:
Yeah, that's an issue of standards is it not? Pointless using some whizzy new fast optical link if there's nothing to connect it to. 😛
I've found it to be the biggest challenge as a designer - see my post a little further up. Common mode noise, not 'common noise' btw 😀
All you're trying to do is to get a stream of pulses from A to B with electrical isolation, and from Thorsten's piece above, seemingly making very hard work of it. Couldn't you just stick a proper fibre optic transmitter and receiver in the way and be done with it? Some run at hundreds of gigabits per second, so I doubt jitter would be much of a problem, either.
I can't imagine the power consumption is worth worrying about in the scheme of things, and as for compatibility with existing kit, people round here seem only too happy to rip things apart and mod them...
Just a thought, anyway.
Is it standards of connector types or standards of signal protocol?
I know you have but I'm looking for other practitioners opinions as mine is not in agreement with your yet 🙂 Yep corrected my post already
Ah yeah, I forgot your original preface of 'wading into areas you don't know'. All becomes clearer in that context. Now do you want to learn or are you just content to pontificate? 🙂
Hi,
Well, Toslink is one of the two standards for interconnecting consumer digital audio devices that are not PC's or Mac's (if you have these add IEE1394 also known by the fancy marketing speak "Firewire" which seems far from the flaming fire Apple would wish and USB).
Of course as we do DIY here we have no direct requirement for compatibility with existing equipment. In that case of course using I2S from the source to the DAC and sending a clock back to the DAC, perhaps even using gigabit optical stuff (quite expensive though) is an option, there are many others.
However, it seems few want to give any compatibility with commercial equipment up so radically (a commercial manufacturer obviously CANNOT do so, even if they want to).
Hence SPDIF electrical (RCA or BNC, 75Ohm Coax, 0.5V PP), AES/EBU electrical (XLR with 110 Ohm twisted pair, 5V PP) and Toslink (optical transmission at 1980's consumer standard levels) plus USB is what is available, from the sources that exist commercially.
The challenge to DIY'er and commercial manufacturer alike is to get low jitter from such sources, by whatever means...
Ciao T
Well, Toslink is one of the two standards for interconnecting consumer digital audio devices that are not PC's or Mac's (if you have these add IEE1394 also known by the fancy marketing speak "Firewire" which seems far from the flaming fire Apple would wish and USB).
Of course as we do DIY here we have no direct requirement for compatibility with existing equipment. In that case of course using I2S from the source to the DAC and sending a clock back to the DAC, perhaps even using gigabit optical stuff (quite expensive though) is an option, there are many others.
However, it seems few want to give any compatibility with commercial equipment up so radically (a commercial manufacturer obviously CANNOT do so, even if they want to).
Hence SPDIF electrical (RCA or BNC, 75Ohm Coax, 0.5V PP), AES/EBU electrical (XLR with 110 Ohm twisted pair, 5V PP) and Toslink (optical transmission at 1980's consumer standard levels) plus USB is what is available, from the sources that exist commercially.
The challenge to DIY'er and commercial manufacturer alike is to get low jitter from such sources, by whatever means...
Ciao T
Last edited:
In the case of ST, dunno. Is it the same as the AT&T glass connection that was around in the very old days? If so then its still SPDIF but just a different medium. I think that one had a bandwidth of `50MHz so was much better jitter-wise than TOSLINK. Much more expensive too.
I know you have but I'm looking for other practitioners opinions as mine is not in agreement with your yet 🙂 Yep corrected my post already
Cool then I'll take a step back for others to take the limelight 😎
Wow! You are quite a piece of work, Abraxalito.
First I 'self-deprecate' by mentioning that I am dipping a toe into an area I'm not sure about. Then I lace my remarks with questions, not pontifications, and you come back with that!
It must be a massive chip you're carrying on your shoulder, for whatever reason.
Yes, T, but what is the compromise here -as was said electrical engineering is about compromises? Jitter seems to be ubiquitous & worth reducing for better sonics - this is a given for me. Is common mode noise as "common" or ubiquitous & do we know how detrimental it is & at what levels, etc.?
Indeed I am, well spotted 😀
You see something wrong with what I wrote? Are you going to share what exactly that is you're hinting at?
No, its 14nm feature size, not massive in the scheme of things.😛
Hi,
Well, let me put it that way, I have seen common mode noise cause so much "jitter" that receiver chip's with wide open PLL's lost lock. This is arguably an extreme case, but one that illustrates how pernicious it CAN be.
Ciao T
Well, let me put it that way, I have seen common mode noise cause so much "jitter" that receiver chip's with wide open PLL's lost lock. This is arguably an extreme case, but one that illustrates how pernicious it CAN be.
Ciao T
- Status
- Not open for further replies.