Hi there,
I'm using a Lundahl pulse transformer LL1572 (1:1) on the SPDIF unbalanced (coax) output stage of my Goldmund Mimesis 36+ CD transport, and I rather like it : smooth and natural sounding, much better than the stock Coilcraft WB2010.
However, some people seem to prefer Murata DA100 series, Scientific Conversion SC939 or Newava S37211...
Could somebody help me ? Some experience on the topic ? What criteria or rating is important to choose a 1:1 pulse transformer for SPDIF ?
I'm using a Lundahl pulse transformer LL1572 (1:1) on the SPDIF unbalanced (coax) output stage of my Goldmund Mimesis 36+ CD transport, and I rather like it : smooth and natural sounding, much better than the stock Coilcraft WB2010.
However, some people seem to prefer Murata DA100 series, Scientific Conversion SC939 or Newava S37211...
Could somebody help me ? Some experience on the topic ? What criteria or rating is important to choose a 1:1 pulse transformer for SPDIF ?
Can I ask how a pulse transformer on a digital signal can effect the sound! If the analogue was going through the transformer I could understand there being some effect, but a digital signal!
The end device has to pull a clock out of SPDIF to use for playback, and the less damaged the SPDIF signal is before it gets there, the better the clock recovery is going to be.
Think of SPDIF as an RF signal with sharp edges, and your goal is to preserve those edges. Making sure the interface you come up with doesn't ring and get extra edges, suffer from reflections that can bash subsequent edges, or take too much slope off the edges.
As for transformers... my usual metric for a transformer is a return loss versus frequency plot, you want it to be as low as possible at SPDIF frequencies. My usual transformer for SPDIF is a Mini-Circuits ADT1-6T, chosen for this reason. It's a 50 ohm transformer, and I match a 75 or 110 ohm input to 50 ohms when I use it. Never really used anything else.
Thank you gmarsh for your comment ! Anybody else who tried one of these transformers (Murata, SC, Lundahl, Newava) ??
I am fully conversant with high speed digital, high speed digital layout and signal integrity.
Is there some empirical measurements and data that we can relate to, to say why a pulse transformer can sound different. If the waveform is bad before the pulse transformer then its gonna be bad coming out of the pulse transformer, and if you wave is so bad then there are signal integrity issues that require addressing.
The best way would be collating the digital waveforms before and after the pulse transformer, and also the analogue output of the DAC.
As stated signal integrity issues have to be considered and the only way of knowing what is going on is to capture the waveforms, so that the any problems such as excessive ringing etc can be observered. Though ringing that apears on the top of the wave is nowhere as critical as non monotonicity of the rising (or falling) edge, where the switching threshold is.
So what I am trying to determine is empirical data to support and determine what is happening to the square wave, that will affect the sound, as digital is very tolerant of noise etc. Or is the data just from listening...
Is there some empirical measurements and data that we can relate to, to say why a pulse transformer can sound different. If the waveform is bad before the pulse transformer then its gonna be bad coming out of the pulse transformer, and if you wave is so bad then there are signal integrity issues that require addressing.
The best way would be collating the digital waveforms before and after the pulse transformer, and also the analogue output of the DAC.
As stated signal integrity issues have to be considered and the only way of knowing what is going on is to capture the waveforms, so that the any problems such as excessive ringing etc can be observered. Though ringing that apears on the top of the wave is nowhere as critical as non monotonicity of the rising (or falling) edge, where the switching threshold is.
So what I am trying to determine is empirical data to support and determine what is happening to the square wave, that will affect the sound, as digital is very tolerant of noise etc. Or is the data just from listening...
It's the way the clock is encoded with the data that makes spdif so susceptible to the interface. There are several papers written by Julian dunn on how bandwidth limitation of the transmitter medium results in data modulated jitter.
Another factor is also that a poor transformer will have quite high leakage inductance making it harder to terminate properly and a poorly terminated interface gives rise to reflections which will also degrade the recovered clock by causing level uncertainties.
However with some care it is possible to approach the limits of what it is possible to measure with 16bit data and the "Jtest" method for measuring data modulated jitter as introduced by dunn.
Its a form of manchester encoding that is used quite widely for data transmission, though the differential method is prefferable.
With digital it is either there or it isn't, SPDIF is quite a slow interface these days, and should with modern devices and what is known about transmitting signals, shouldn't present much of a problem, there are numerous digital interfaces that work at higher speeds, that work quite succesfully. We use custom made pulse transformers for a variet of interfaces as well as planar transformers built into PCB's.
What I am after is the explanation and data on how analogue terminology for the effect on sound can be ascribed to a digital interface, how you ascribe one sounding smooth etc. This would imply that a regular change to the analogue output is being achieved by the transformer! somthing usualy only achieved by DSP in the digital domain.
The artifacts that effect digital signals and the problems of noise are covered quite extensively in a few good tomes, Howard Johnson, High speed digital design for the Signal Integrity and Henry Ott for EMC problems...
With digital it is either there or it isn't, SPDIF is quite a slow interface these days, and should with modern devices and what is known about transmitting signals, shouldn't present much of a problem, there are numerous digital interfaces that work at higher speeds, that work quite succesfully. We use custom made pulse transformers for a variet of interfaces as well as planar transformers built into PCB's.
What I am after is the explanation and data on how analogue terminology for the effect on sound can be ascribed to a digital interface, how you ascribe one sounding smooth etc. This would imply that a regular change to the analogue output is being achieved by the transformer! somthing usualy only achieved by DSP in the digital domain.
The artifacts that effect digital signals and the problems of noise are covered quite extensively in a few good tomes, Howard Johnson, High speed digital design for the Signal Integrity and Henry Ott for EMC problems...
I tried Lundahl LL1572 on S\PDIF output of Musiland 01 USD,
after a day I replaced with Newava. Much better IMHO.
after a day I replaced with Newava. Much better IMHO.
Scientific Conversion! Not as well known as the others mentioned, but IMHO the best out there. Do not even hesitate. What those folks don't know about pulse transformers just is not known. Regards
Hmm. Not according to this thread, this proved a big education for me. I strongly suggest anyone interested in actual analysis of what is best and how to acheive it read the whole thread.
The SB3/SPDIF output thread.......
Regards,
Rob.
I would suggest that those using SC xformers know their limitations and have designed their circuits accordingly. If they are not taking this into consideration, perhaps you should consider another designer's units. Regards
While the point of the thread in the link I posted is, amongst other things, that ALL transformers need to be optomised in circuit, one of the clear messages also was that the SC transformers art tested were more compromised than the Naweva. He certainly had the means and the knowledge to optimise both; with both optimised the Naweva was the preferred option. SC may have improved their designs since, I don't know.
Marce,
Your previous post was inaccurate - yes, you will have crap coming out of the transformer if crap is going in. However, S/PDIF is a RF transmission line. The moment you stick in a transformer (most usually for isolation), you've screwed up the transmission line. So it then boils down to picking one that has the least negative impact and like Robert F said, optimizing the transformer circuit properly so that the impedances match, thus generating the least loss and reflection. Since you know high speed digital, you'll appreciate how tricky it is for a board fab house to create a stackup that'll meet various SE/Diff transmission line impedances.
The OP's post was more of a subjective question - which one sounds better and how to pick one. In that case, there are many answers. From the subjective perspective, go for one that has the most hype - e.g. Scientific Conversion. If you made the mod yourself and perform a sighted listening test (especially without a duplicate un-modded unit for A/B comparison), then you may well hear a positive difference. Gmarsh recommended Mini-Circuits. I use Newava simply because Digikey carries them.
Your previous post was inaccurate - yes, you will have crap coming out of the transformer if crap is going in. However, S/PDIF is a RF transmission line. The moment you stick in a transformer (most usually for isolation), you've screwed up the transmission line. So it then boils down to picking one that has the least negative impact and like Robert F said, optimizing the transformer circuit properly so that the impedances match, thus generating the least loss and reflection. Since you know high speed digital, you'll appreciate how tricky it is for a board fab house to create a stackup that'll meet various SE/Diff transmission line impedances.
The OP's post was more of a subjective question - which one sounds better and how to pick one. In that case, there are many answers. From the subjective perspective, go for one that has the most hype - e.g. Scientific Conversion. If you made the mod yourself and perform a sighted listening test (especially without a duplicate un-modded unit for A/B comparison), then you may well hear a positive difference. Gmarsh recommended Mini-Circuits. I use Newava simply because Digikey carries them.
How is it innaccurate:
It is digital not rf, it is a form of manchester endocing, that is a form of digital signal transmission used extensively, that combines the clock and the dtat so that there is a transition every clock cycle, otherwise it is very hard to transmit a series of 1's or 0's as they are effectively a DC voltage.
Again I am asking for an explanation of how a digital transformer can have a SOUND.
It is digital not rf, it is a form of manchester endocing, that is a form of digital signal transmission used extensively, that combines the clock and the dtat so that there is a transition every clock cycle, otherwise it is very hard to transmit a series of 1's or 0's as they are effectively a DC voltage.
Again I am asking for an explanation of how a digital transformer can have a SOUND.
Had a quick loook at the thread, the statement this is DIY sod EMI says it all, EMI and signal integrity are two sides of the same coin, signal integrity is EMI that effects the internal working of the unit, EMI is external (This is based on Keith Armstrongs' introduction to these problems). So if we start of a thread by saying SOD EMI, then were not going to be following good engineering practice.
'Digital' only exists in theory and in computational space. The moment you put a digital signal on a wire, cable, PCB trace... it becomes analog. And transmission line (RF) effects start to rear their ugly head.
The nice 'digital' square wave generated in the playback device, once it's sent through the SPDIF interface and its associated cables, transformers, connectors, passives, etc.. can end up coming out awfully mangled on the other side. And the choice of any of these devices, including the transformer, can affect this. Pulling correct bit values out of the 'mangled' received signal is straightforward, but that's only half the problem with SPDIF, and by far the easier problem.
SPDIF embeds the clock in the data stream, and the receiving end has to use a PLL to extract it and provide the playback clock. It does this by detecting the edges in the SPDIF stream and trying to line up the edges of a local oscillator with it. The more the edges get pushed around in time by the 'mangling' of the interface, the harder the PLL has to work, and the quality of the recovered playback clock degrades. Jitter happens. And it can be audible.
So yes, a transformer can affect the sound of SPDIF.
NO signal integrity and good engineering can avoid the problems you mention.
High speed digital design and layout is a seperate skill set from Rf layout, I do both, I also use Signal Integrity Verify tools that are heavily based on transmission lines. The fact that high speed digital uses very high frequencies, does not make it rf as such. RF is analogue technigues where as for high speed digital you use different techniques.
Bad engineering and choosing the wrong transformewr can mangle the waveform, but as stated previously, digital is very tollerant of noise and other problems that would destroy an analogue waveform, hence its use for data transmission these days.
A badly engineered interface will mess up the signals, that just bad engineering, but poetic language is being used to say this transformer sounds smooth, this one harsh etc etc, where is the data, waveforms both digital and analogue to explain and show what is happening.
Again digital is not analogue, you only have 2 states, end of story... Digital signal tranmission is transmiting an occasionly square wave made up of numerous harmonics, the highest harmonics present being determined by the signal rise time... and at SPDIF frequencies these days it is not hard to design and layout a path that does not destroy the wave. If you want to see how bad a 'square' wave can look have a look at some of the waveforms for DDR memory, as long as the switching intervalk is constant, even quite severe ringing on the top of the wave will not effect the end result.
High speed digital design and layout is a seperate skill set from Rf layout, I do both, I also use Signal Integrity Verify tools that are heavily based on transmission lines. The fact that high speed digital uses very high frequencies, does not make it rf as such. RF is analogue technigues where as for high speed digital you use different techniques.
Bad engineering and choosing the wrong transformewr can mangle the waveform, but as stated previously, digital is very tollerant of noise and other problems that would destroy an analogue waveform, hence its use for data transmission these days.
A badly engineered interface will mess up the signals, that just bad engineering, but poetic language is being used to say this transformer sounds smooth, this one harsh etc etc, where is the data, waveforms both digital and analogue to explain and show what is happening.
Again digital is not analogue, you only have 2 states, end of story... Digital signal tranmission is transmiting an occasionly square wave made up of numerous harmonics, the highest harmonics present being determined by the signal rise time... and at SPDIF frequencies these days it is not hard to design and layout a path that does not destroy the wave. If you want to see how bad a 'square' wave can look have a look at some of the waveforms for DDR memory, as long as the switching intervalk is constant, even quite severe ringing on the top of the wave will not effect the end result.
OK Marce, since you brought up DDR, please explain why DDR doesn't work if you put it through a cable. Why it has to be done with short, controlled impedance traces on a pcb. Why the memory controller typically supports no more than 3 loads (DIMMs) on a signal trace. As you figure that out, you'll know why a high speed digital signal behaves like a RF signal.
Good luck.
Good luck.
- Status
- Not open for further replies.