SPDIF
In theory......? I'm sorry I actually measure and listen to things. Digital is as tweaky as analog to do right. Bandwidth of interface is usually a funtion of the output drivers used, usally HC CMOS, rise time about 3nS. Schmitt-Trigger receivers cause reflections. A Schmitt-Trigger dumps some of it's output back into the input to make it switch faster. The positive feedback signal gets in the interface as reflections. Das ist Verbotten! Reflections in the digital interface cause R.F. noise and jitter in the data stream. Partial grounding is a small impedance between CD transport and DAC grounds.
In theory......? I'm sorry I actually measure and listen to things. Digital is as tweaky as analog to do right. Bandwidth of interface is usually a funtion of the output drivers used, usally HC CMOS, rise time about 3nS. Schmitt-Trigger receivers cause reflections. A Schmitt-Trigger dumps some of it's output back into the input to make it switch faster. The positive feedback signal gets in the interface as reflections. Das ist Verbotten! Reflections in the digital interface cause R.F. noise and jitter in the data stream. Partial grounding is a small impedance between CD transport and DAC grounds.
Give 'em hell, Harry
Couldn't have said it better myself, but what the hell........
In theory, I'm a nice guy. So much for theory.
SPDIF is band-limited probably only to reduce EMI. I really don't think they ever thought someone would be stupid enough to use it. So then they decided to "square-up" those rounded-off noise glitches that were supposed to be square waves.
A Schmitt trigger is the worst thing you can hang on a transmission line where you have to recover the clock from serially encoded data.
Jocko
Couldn't have said it better myself, but what the hell........
In theory, I'm a nice guy. So much for theory.
SPDIF is band-limited probably only to reduce EMI. I really don't think they ever thought someone would be stupid enough to use it. So then they decided to "square-up" those rounded-off noise glitches that were supposed to be square waves.
A Schmitt trigger is the worst thing you can hang on a transmission line where you have to recover the clock from serially encoded data.
Jocko
It's Evil...let me rant the ways
Since we are in DIY land, we can do our own master clock schemes. We don't need no stinkin' standard...
Regarding the S/PDIF standard, er, clusterf***, I had two words to say about it years ago:
It's evil.
Having tried all sorts of tricks to "massage the last picosecond out of it" since saying those words: droop compensation, preamble edge extraction, sub hertz corner freq dual loop PLL/VCXOs...I have somewhat broadened my respect for it:
It's still evil.
I have yet to hear a derived/synthesized clock setup sound as magically good as a master clock arrangement. It's that last few tens of picoseconds that do make the difference. Kinda like that last inch...
A big contributing factor is the high pass function, or low freq roll off...typically in the 50kHz range which is a setup with the inherent high freq rolloff, typically 10 to 50 MHz for correlated jitter. The sag, or droop, shifts the launch point at each transition in a manner highly correlated with the data. Also known as deterministic jitter as opposed to random jitter.
I was thinking the Lundahl 1570 with it's 20Hz low end might be a good choice if I *had* to S/PDIF.
Also, the data sheets of optical ST tx and rx modules give it away: even with 256 MBd ecl/pecl modules like the HP 1119T/2119T the deterministic jitter is listed as 0.3nS/0.9nS. Ouch. The cheezier ones used by most, even "statement" top of the line products are worse. Adding insult to injury, the 5 volt supplies are usually the noisy digital supplies, seemingly almost as an afterthought. Whatever the sonic gains from galvanic isolation are more than worsened by that kind of correlated jitter.
What the ST modules are excellent for is master/slave feeds, where the master clock in the D/A feeds the transport which feeds S/PDIF to the D/A, with registering/reclocking along the way.
WMS
Since we are in DIY land, we can do our own master clock schemes. We don't need no stinkin' standard...
Regarding the S/PDIF standard, er, clusterf***, I had two words to say about it years ago:
It's evil.
Having tried all sorts of tricks to "massage the last picosecond out of it" since saying those words: droop compensation, preamble edge extraction, sub hertz corner freq dual loop PLL/VCXOs...I have somewhat broadened my respect for it:
It's still evil.
I have yet to hear a derived/synthesized clock setup sound as magically good as a master clock arrangement. It's that last few tens of picoseconds that do make the difference. Kinda like that last inch...
A big contributing factor is the high pass function, or low freq roll off...typically in the 50kHz range which is a setup with the inherent high freq rolloff, typically 10 to 50 MHz for correlated jitter. The sag, or droop, shifts the launch point at each transition in a manner highly correlated with the data. Also known as deterministic jitter as opposed to random jitter.
I was thinking the Lundahl 1570 with it's 20Hz low end might be a good choice if I *had* to S/PDIF.
Also, the data sheets of optical ST tx and rx modules give it away: even with 256 MBd ecl/pecl modules like the HP 1119T/2119T the deterministic jitter is listed as 0.3nS/0.9nS. Ouch. The cheezier ones used by most, even "statement" top of the line products are worse. Adding insult to injury, the 5 volt supplies are usually the noisy digital supplies, seemingly almost as an afterthought. Whatever the sonic gains from galvanic isolation are more than worsened by that kind of correlated jitter.
What the ST modules are excellent for is master/slave feeds, where the master clock in the D/A feeds the transport which feeds S/PDIF to the D/A, with registering/reclocking along the way.
WMS
Harry, if I didn't know better, I would think you are trying to provoke an international incident.
Oh hell, I do know you better. What has it been, almost 2 weeks since I got someone down under mad at me? Must be losing my touch.
So here goes:
No wonder the crocs are so mad. I would be too if someone made me drink that stuff.
Next time try a real beer:
Oh hell, I do know you better. What has it been, almost 2 weeks since I got someone down under mad at me? Must be losing my touch.
So here goes:
No wonder the crocs are so mad. I would be too if someone made me drink that stuff.
Next time try a real beer:
Attachments
Simple Direct Transformerless Interface
Hi Wildmonkeysects and all other esteemed inmates of this lighthearted Forum,
Over a year we discussed my direct coupled interface in conjunction with the Wildmonkeysects loopfilter on the Audio Asylum forum.
http://www.audioasylum.com/audio/tweaks/messages/30462.html
A forum now passing away slowly because of obesity.....I can't even find back my own posts. Here we see passing along genetically manipulated cats, a dog addicted to booze & smoking, forummembers wearing very weird hats etc.....
Quite recently I improved the circuit, now using the J310 N-Channel JFET from Fairchild in place of the 2N2369 bipolar transistor. I am using double shielded coaxial cable used for video and TV transmission.
The Interface works fine for me and should have adequate bandwith. It has of course no ground isolation as the transformer.
Hi Wildmonkeysects and all other esteemed inmates of this lighthearted Forum,
Over a year we discussed my direct coupled interface in conjunction with the Wildmonkeysects loopfilter on the Audio Asylum forum.
http://www.audioasylum.com/audio/tweaks/messages/30462.html
A forum now passing away slowly because of obesity.....I can't even find back my own posts. Here we see passing along genetically manipulated cats, a dog addicted to booze & smoking, forummembers wearing very weird hats etc.....
Quite recently I improved the circuit, now using the J310 N-Channel JFET from Fairchild in place of the 2N2369 bipolar transistor. I am using double shielded coaxial cable used for video and TV transmission.
The Interface works fine for me and should have adequate bandwith. It has of course no ground isolation as the transformer.
I am a little puzzled - are we talking about digital transformer which functions as an alternative to 10nF capacitors at the spdif input to the CS8412?
Or are we talking about a transformer somewhere else. Elso's helpful circuit diagram shows the transistor he mentions actually driving the 75 cable - rather than at the receiving end.
Is that what Kasra was meaning? Is that what we have been referring to? (i.e a transfomer to drive the 75 ohm cable)
(I was even more puzzled about the cats)
Or are we talking about a transformer somewhere else. Elso's helpful circuit diagram shows the transistor he mentions actually driving the 75 cable - rather than at the receiving end.
Is that what Kasra was meaning? Is that what we have been referring to? (i.e a transfomer to drive the 75 ohm cable)
(I was even more puzzled about the cats)
Weird hats aside, some of us have done a lot of research into this. We strongly feel that using isolation transformers function better that direct connections. We have also done a lot of research into what transformers work, and why.
And the bottom line is to use a transformer, and one with as little leakage inductance as possible.
And yes, all the loop filter mods and other tricks are still a bandaid for a poorly conceived interface.
Incidentally, the weird hat is the result of staring at TDRs and network analyzers for too many years. What happens to chemists that sniff too many unpronouncable chemicals? Just curious.
Jocko
And the bottom line is to use a transformer, and one with as little leakage inductance as possible.
And yes, all the loop filter mods and other tricks are still a bandaid for a poorly conceived interface.
Incidentally, the weird hat is the result of staring at TDRs and network analyzers for too many years. What happens to chemists that sniff too many unpronouncable chemicals? Just curious.
Jocko
well..
Im going to build an interface between the cd and my homemade dac. Well the CD has already some cheap a¤¤ that need to be upgraded. It really leans to Pulse transformers, which seems to be really easy to use.
I have talked with the president of Scientifi Conversions, by specs their values seem to be the better ones on the market when it comes to transformers.
They provided me with everything i needed to know, even measuremeants which are far away from my knowledeges.
They have many to choose from depending on the wanted interface, and they are all optimized for different samplingrates, according to the papers i have read.
Jocko / Elso / Harry / Others, what do you think about using transformers, and perhaps the ones from SC?
/kasra
Im going to build an interface between the cd and my homemade dac. Well the CD has already some cheap a¤¤ that need to be upgraded. It really leans to Pulse transformers, which seems to be really easy to use.
I have talked with the president of Scientifi Conversions, by specs their values seem to be the better ones on the market when it comes to transformers.
They provided me with everything i needed to know, even measuremeants which are far away from my knowledeges.
They have many to choose from depending on the wanted interface, and they are all optimized for different samplingrates, according to the papers i have read.
Jocko / Elso / Harry / Others, what do you think about using transformers, and perhaps the ones from SC?
/kasra
Can't think of any reason to have so many flavors. As long as you can pass the 7th harmonic.
A good place to start is the Schott 67125450 that Digi-Key sells. I believe Crystal recommends it, but if they do it is for the wrong reason. Like everything else in their diatribe on transmission lines.
Jocko
A good place to start is the Schott 67125450 that Digi-Key sells. I believe Crystal recommends it, but if they do it is for the wrong reason. Like everything else in their diatribe on transmission lines.
Jocko
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.