Because its high speed digital, thats why, I reguraly lay out DDR interfaces, and we use high speed digital techniques, NOT rf. When I lay out a radio or transmitting ciruit that is rf I use rf layout techniques. As to trace impedance, at the speeds DDR works at it is critical to match the impedance of the whole signal path to avoid reflections, we work to trying to achieve as near to 50 ohm as possible. The longer the trace the more of the high frequency element of the signal is lost, and the more chance of problems with any reflectione etc, capacitive losses become more critical at higher frequencies, crosstalk is more problematic.
At the end of the day DDR and interfaces work, there is a whole wealth of guidlines that guide one through laying them out, and there are the simulation tools to verify that the layout will work (use transmission line theory of course), so if we can get them working at silly frequencies, tha same engineering disciplines can be used for SPDIF to get the interface right. I always thought the idea behind a lot of DIY auio was to get the best possible sound, so if SPDIFF interface is such a big problem then engineer the interface and connecting cables etc to solve the problems, but then that would not be as much fun as discussing the sound of transformers!
There are problems and techniques common to both, but it is still hiigh speed digital, it is not (intentionaly) going to be transmitted, and if you wanna nit pick IBIS is high frequency analogue used to simulate a high speed digital interface... But at the end of the day the signal you are transmitting is either a 1 or a 0.
You forgot to mention that DDR also requires the various data lanes, clocks, address signals etc to be length matched, due to timing issues.
High speed digital is a discipline in its own right, and everyone I talk to involved in this sort of design refer to high speed digital design as high speed digital design and when discussing rf refer to rf design techniques, both involve getting high frequency signals from a to b, so there is some commonality, but they both employ different layout techniques.
PCB Layout Examples - Printed Circuit Boards
At the end of the day DDR and interfaces work, there is a whole wealth of guidlines that guide one through laying them out, and there are the simulation tools to verify that the layout will work (use transmission line theory of course), so if we can get them working at silly frequencies, tha same engineering disciplines can be used for SPDIF to get the interface right. I always thought the idea behind a lot of DIY auio was to get the best possible sound, so if SPDIFF interface is such a big problem then engineer the interface and connecting cables etc to solve the problems, but then that would not be as much fun as discussing the sound of transformers!
There are problems and techniques common to both, but it is still hiigh speed digital, it is not (intentionaly) going to be transmitted, and if you wanna nit pick IBIS is high frequency analogue used to simulate a high speed digital interface... But at the end of the day the signal you are transmitting is either a 1 or a 0.
You forgot to mention that DDR also requires the various data lanes, clocks, address signals etc to be length matched, due to timing issues.
High speed digital is a discipline in its own right, and everyone I talk to involved in this sort of design refer to high speed digital design as high speed digital design and when discussing rf refer to rf design techniques, both involve getting high frequency signals from a to b, so there is some commonality, but they both employ different layout techniques.
PCB Layout Examples - Printed Circuit Boards
And I'll re-iterate: getting the bits through a SPDIF interface correctly (picking out individual ones and zeros) is the easy part. Deriving a low jitter clock from the interface is the hard part.
Since you're a signal integrity guy, fire this into your tools...
Suppose you have a serial, pseudorandom data signal switching at frequency F, with a rise time of 0.1/F. Now run it through an improperly terminated trace with an electrical delay of around 0.48/F - lets say the return loss on the sending end is 3dB, return loss on the receiving end is 10dB.
Ignore the data validity window. What happens to the edges?
Since you're a signal integrity guy, fire this into your tools...
Suppose you have a serial, pseudorandom data signal switching at frequency F, with a rise time of 0.1/F. Now run it through an improperly terminated trace with an electrical delay of around 0.48/F - lets say the return loss on the sending end is 3dB, return loss on the receiving end is 10dB.
Ignore the data validity window. What happens to the edges?
Ah, another thread to go the way of a pointless circulating argument. You do know Marce, that by adopting this half sarcastic, half argumentative approach in your posts, you are alienating many and actually depriving many of us a reasonable discussion of these things.
Lighten up, and if you don't agree, state so once, in one post and leave it there.
Fran
Lighten up, and if you don't agree, state so once, in one post and leave it there.
Fran
Hi, I think Marce has some valid points.
FWIW I used the Newava S22083 and a Mini Circuits type but the latter are hard to find here. The Newava is not expensive and way better than most. Let's say that when I find one in a device it just stays there untouched and when there is something else I like to replace it for the Newava. To be honest I regret having to find excellent transformers for a lousy interface (OK let's call it a less-than-optimal interface, I use it every day too) and I am surpised no one invented a better interface yet in all those years. I understood an external I2S variant was designed but never heard about it again.
The Audio Note SPDIF transformers are a joke. They remind me of coils you can find in light dimmers. I wish someone would measure those with the right equipment to verify objective results. Toroids are not the way to go and the wires should be thinner. But I think the brand name is enough to sell those like many other items in this field.
FWIW I used the Newava S22083 and a Mini Circuits type but the latter are hard to find here. The Newava is not expensive and way better than most. Let's say that when I find one in a device it just stays there untouched and when there is something else I like to replace it for the Newava. To be honest I regret having to find excellent transformers for a lousy interface (OK let's call it a less-than-optimal interface, I use it every day too) and I am surpised no one invented a better interface yet in all those years. I understood an external I2S variant was designed but never heard about it again.
The Audio Note SPDIF transformers are a joke. They remind me of coils you can find in light dimmers. I wish someone would measure those with the right equipment to verify objective results. Toroids are not the way to go and the wires should be thinner. But I think the brand name is enough to sell those like many other items in this field.
Last edited:
As a futher note concerning the thread on Audio Circle. RF engineering is not DIGITAL engineering. I trained many years ago in RF engineering. I am not trained in DIGITAL engineering, but I do know enough to differentiate between the two. The gentlemen who designed the Anedio DAC which I use (and has an SC xformer) is a noted digital engineer. That makes all the difference. Digital signals are not RF. Regards
I replaced a Pulse PE-65612 with a Murata Da101c and noticed an improvement. The Pulse was an improvement on the original fitment, which I binned.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
In the past I did make my own transformers for SPDIF with small toroids. Couldn't make them as good as ready made ones despite winding very tight. I also recall that I wound them bifilar and that smaller toroids were better than large ones. This was very long ago so details are lost (like which cores I used). One thing I clearly remember that it cost me a lot of time with marginal results.
IMO the leakage inductance/capacitance could be the reasons but I am a DIYer and by no means an expert in any way on this subject. I did replace ready made "audiphile" SPDIF toroids for normal industrial ones with good results. But if you are who I think you are you are an expert in this field. Please give your professional opinion.
IMO the leakage inductance/capacitance could be the reasons but I am a DIYer and by no means an expert in any way on this subject. I did replace ready made "audiphile" SPDIF toroids for normal industrial ones with good results. But if you are who I think you are you are an expert in this field. Please give your professional opinion.
Last edited:
Toroids are renowned for having the lowest leakage inductance - that comes from the almost perfectly constrained magnetic path. In practice they do leak around the lead-outs. I'm of the (minority) view that common-mode noise rather than jitter is the primary reason for audible differences between transformers, so I'd go for the ones with the lowest interwinding capacitance. That I believe means SC, unless there are others touting this feature since I last checked. Increased jitter caused by higher leakage inductance can be cleaned up later, CM noise is harder to deal with.
Jean Paul,
Yes I am who you think I am, but to be honest, it is some 10 years ago that I experimented with SPDIF transformers.
You may well remember I had a DAC at the time, and initially I was using the transformers used by the PCM63 Elektuur design. These transformers were using (when I remember correctly) socalled Amidon cores (core material was some iron powder variant).
When I got more into transformers I tried an SPDIF transformer wound on the smallest cobalt based amorphous strip toroidal core available (some 15mm diameter).
I wound (abraxalito) with the lowest possible interwinding capacitance by putting one section opposite the other (so not bifilarly) because my instinct told me that transfer should be by core, not by capacitance. The used core material is very well able to do this transfer.
When looking at the SPDIF signal on the scope it was apparent that my transformer showed a much better signal than the Amidon transformers in that the signal looked much more like a square wave with steeper slopes (which the SPDIF signal is supposed to be when I am right); this measurement confirmed the quality of the core for this application.
Sonic evaluation easily confirmed the better quality of this transformer.
Since that time however I did not compare with commercially available transformers like Scientific Technology for instance (what's in a name....).
I have a bunch of these toroids still here; when interested I could send you some for your evaluation.
(
Yes I am who you think I am, but to be honest, it is some 10 years ago that I experimented with SPDIF transformers.
You may well remember I had a DAC at the time, and initially I was using the transformers used by the PCM63 Elektuur design. These transformers were using (when I remember correctly) socalled Amidon cores (core material was some iron powder variant).
When I got more into transformers I tried an SPDIF transformer wound on the smallest cobalt based amorphous strip toroidal core available (some 15mm diameter).
I wound (abraxalito) with the lowest possible interwinding capacitance by putting one section opposite the other (so not bifilarly) because my instinct told me that transfer should be by core, not by capacitance. The used core material is very well able to do this transfer.
When looking at the SPDIF signal on the scope it was apparent that my transformer showed a much better signal than the Amidon transformers in that the signal looked much more like a square wave with steeper slopes (which the SPDIF signal is supposed to be when I am right); this measurement confirmed the quality of the core for this application.
Sonic evaluation easily confirmed the better quality of this transformer.
Since that time however I did not compare with commercially available transformers like Scientific Technology for instance (what's in a name....).
I have a bunch of these toroids still here; when interested I could send you some for your evaluation.
(
Hi, I am always willing to try new stuff but S22083 is my standard type nowadays or no transformer at all (oops, feels like admitting that you drank a beer on a AA meeting). If you like you can send them but I can not promise when I will actually try them out 😉
Yes I know the Amidon cores. The PCM63 DAC was a very good DAC. I recall comparing it to a AN DAC with E88CC outputstage and transformers etc and we could not hear a difference. Also the Elektor PCM1710 DAC I made then had self wound cores.
BTW I guess that if you start winding sym-asym trasnformers for the newer DAC chips with symmetrical outputs you will definitely have demand for them. Like CS4397/9, the PCM voltage out types, ESS types etc. The Neutrik and Haufe transformers I tried sound way better than opamp stages. Easy to implement too. Hifimediy is developing a CS4398 DAC specifically without output stage so transformers can be used. It seems output transformers in DACs are back on track again.
Yes I know the Amidon cores. The PCM63 DAC was a very good DAC. I recall comparing it to a AN DAC with E88CC outputstage and transformers etc and we could not hear a difference. Also the Elektor PCM1710 DAC I made then had self wound cores.
BTW I guess that if you start winding sym-asym trasnformers for the newer DAC chips with symmetrical outputs you will definitely have demand for them. Like CS4397/9, the PCM voltage out types, ESS types etc. The Neutrik and Haufe transformers I tried sound way better than opamp stages. Easy to implement too. Hifimediy is developing a CS4398 DAC specifically without output stage so transformers can be used. It seems output transformers in DACs are back on track again.
Last edited:
Jean-Paul,
Maybe I was ahead of times back in 1999; the DAC I sold had PCM63 converters with 100 ohms I/V resistors and 1:5 step-up transformers sporting a "healthy" 700 mVRMS... output.
Output stage was a nice jfet source follower, but for obvious (.....) reasons the tube guys bypassed the fet output stage.
Yes, I also notice that DIY DAC's are pretty hot nowadays.
However not all DAC chips are the same; the CS43.. types are differential voltage output with do not seem to need additional amplification. The differential current output types will work great with a transformer no doubt, but that might be a different transformer because of the required amplification.
The ESS types seem to be somewhere between current and voltage output so again a different situation.
When you have a good design I could wind a prototype for your evaluation, but we are OT in the meantime....
Maybe I was ahead of times back in 1999; the DAC I sold had PCM63 converters with 100 ohms I/V resistors and 1:5 step-up transformers sporting a "healthy" 700 mVRMS... output.
Output stage was a nice jfet source follower, but for obvious (.....) reasons the tube guys bypassed the fet output stage.
Yes, I also notice that DIY DAC's are pretty hot nowadays.
However not all DAC chips are the same; the CS43.. types are differential voltage output with do not seem to need additional amplification. The differential current output types will work great with a transformer no doubt, but that might be a different transformer because of the required amplification.
The ESS types seem to be somewhere between current and voltage output so again a different situation.
When you have a good design I could wind a prototype for your evaluation, but we are OT in the meantime....
Gmarsh - done a bit of basic sim this morning, havn't got all the parametrs in, but looks bad and gets worse... Adding mismatched transformers should be even more interesting. The attached was cobbled together rather quickly, I would like to do a proper SPDIF interface using devices that are generally used, so any pointers to suggested devices would be helpful.
Woodturner Fran- sorry if you think I am being argumentative and sarcastic, dont read my posts😉, but at the end of the day I think I am asking valid questions...
So you understand my point, SPDIF is by todays standards should not present a problem in getting a clean signal form A to B, compared with the ridicoulous speed a lot of todays digital interfaces run at with crazy rise times and 100Mhz plus clocks. We should be able to engineer the interface so that the signals arrive cleanly, this can be done using current engineering techniques, such as replacing nasty RCA connectors with BNC etc. That is why I am questening the effect transformers have, and looking for data, so that I can understand more of the problem.
At the end of the day we are debating a subject that we all hold differing views on, that is the beauty of debate and some threads, would you rather it be a sycophantic discussion where we all agree with some current trend or idea, I dont, I like to ask questions and occasionaly get answers.
As to digital and RF engineering, we do both where I work (tactical communications and Surveillance) and do differentiate between them both as seperate engineering practices, its just that high speed digital involves frequencies that are also used for RF...
Agree on the Audionote transformers, again it is selling somthing on belief not engineering practice, IMO.
Even if we dont all agree, theres no need to get upset, always question your beliefs, as they may not always be true😀
Woodturner Fran- sorry if you think I am being argumentative and sarcastic, dont read my posts😉, but at the end of the day I think I am asking valid questions...
So you understand my point, SPDIF is by todays standards should not present a problem in getting a clean signal form A to B, compared with the ridicoulous speed a lot of todays digital interfaces run at with crazy rise times and 100Mhz plus clocks. We should be able to engineer the interface so that the signals arrive cleanly, this can be done using current engineering techniques, such as replacing nasty RCA connectors with BNC etc. That is why I am questening the effect transformers have, and looking for data, so that I can understand more of the problem.
At the end of the day we are debating a subject that we all hold differing views on, that is the beauty of debate and some threads, would you rather it be a sycophantic discussion where we all agree with some current trend or idea, I dont, I like to ask questions and occasionaly get answers.
As to digital and RF engineering, we do both where I work (tactical communications and Surveillance) and do differentiate between them both as seperate engineering practices, its just that high speed digital involves frequencies that are also used for RF...
Agree on the Audionote transformers, again it is selling somthing on belief not engineering practice, IMO.
Even if we dont all agree, theres no need to get upset, always question your beliefs, as they may not always be true😀
Attachments
Marce: well put indeed. Perhaps (more than perhaps! obviously!) I misunderstood your questioning.... just that I have seen many threads go south due to someone endlessly stating their case and then picking fights so they can restate it again!! A friend was banned from here as a result of being caught and probably baited by others with this kind of carry on..... which I think was more of a loss to the community here than to him. Anyway, I guess its easy sometimes to not "get" the tone/clues that you would pick up if we were actually talking - apologies for the accusation!
***********************************
I did do some listening way back comparing newava and pulse tx and the newava was clearly better, but not as good as none at all (shigaclone to buffalo dac). Never got to get some SC ones, and as for the audionote ones.... they look like I might as well wind it around my finger! I do remember seeing scope shots of a signal through the various brands of Tx. A search might turn it up maybe? You could clearly see ringing with some Tx, and much less with others.
Fran
***********************************
I did do some listening way back comparing newava and pulse tx and the newava was clearly better, but not as good as none at all (shigaclone to buffalo dac). Never got to get some SC ones, and as for the audionote ones.... they look like I might as well wind it around my finger! I do remember seeing scope shots of a signal through the various brands of Tx. A search might turn it up maybe? You could clearly see ringing with some Tx, and much less with others.
Fran
But does the ringing really matter? John Kenny's SPDIF unit had some of the worst ringing I've seen, yet it worked flawlessly, sounded fine, and the measurements from the analog side (which is all that really matters for the sound, which I *think* was part of marce's point) were as good as any of the other units I've tried.
FWIW, I'm using a TX102, but don't claim any sort of superiority- my DAC locks and plays the music without incident.
FWIW, I'm using a TX102, but don't claim any sort of superiority- my DAC locks and plays the music without incident.
Oh, you measured that? I think I missed that in the thread that was running at the time (subsequently closed IIRC).
Fran
Fran
The ringing? Yes, I measured it, with and without attenuators (no difference). Also on two other units I had on hand, and a homebuilt SPDIF-to-AES/EBU converter using the TX102. The one with the least ringing had a slower rise time than the others, but worked just fine. I think I ended up where abraxalito is- any audible differences are more likely due to noise than to waveform fidelity.
I'm still shaking my head about the Audio Note units. I'm afraid to ask what they're charging for those...
I'm still shaking my head about the Audio Note units. I'm afraid to ask what they're charging for those...
Did you compare results with different transformers on one and the same DAC ? That is what counts IMO. They all work but some work better than others and some receivers are more susceptible than others too.
A better signal form will pay off in jitter terms with the older Crystal recievers. In fact sometimes best results can be obtained by omitting the transformer with CS types. I wouldn't know with WM8805 for instance how much a transformer influences matters. My newer DACs have DIR9001 and WM8805 and no transformers anymore as I got really confused when the transmitting side has a different transformer than the receiving side would get. The very scientic approach was that 2 diiferent ones wouldn't improve matters. Since I already had galvanic isolation I just didn't bother to use transformers at the receiving side.
Last edited:
No, I didn't. But that wasn't the point I was trying to make- what I did do was compare interfaces with very different amounts of ringing and rise-time and found that, in practice, it made little to no difference in the analog signal as long as the transmission wasn't pathological and as long as there was SOME sort of transformer in there. There could be noise coupling differences, but they didn't show up when I looked for any effects from them; I wouldn't be surprised if that could be different in different setups.
If someone has controlled (ears-only) listening tests showing that, within reason, the details of the shape of SPDIF waveforms have an effect on the audio, I'd be interested in that!
- Status
- Not open for further replies.