Best Cable Option: Streamer to DAC


I was recently told that the inherent limitation of SPDIF connection is PCM 192Hz. I didn't know that. Many new streamers spec 384KHz and I am also told that to achieve higher sample rates (and presumably the full capabilities of the new units) I should use USB rather than SPDIF.  So it made me wonder what actually is the best connection between streamer and DAC:  USB, COAX BNC-SPDIF, AES/EBU or something else?   From a practical standpoint, is there any audible difference from the higher sample rates?  If so, my system should be able to reproduce it.  I'm just looking for help, not trying to start any arguments on here.
papafrgog
You won’t get any definite consensus here....choose a streamer / DAC that does not limits your connectivity. This way you can try and use the protocol that sounds best to your ears.

“I was recently told that the inherent limitation of SPDIF connection is PCM 192Hz.” - True, but I wouldn’t look at this as a limitation since current streaming services tops at 24bit/192kHz resolution. Jitter is no longer an issue with most modern DAC’s or streamers that now has internal clocks to combat jitter.

“is there any audible difference from the higher sample rates?” - Yes, as long as you have a system that allows you to appreciate the difference. You do need DSD files to appreciate the uptick in SQ. In this case USB is the obvious choice. 
S/PDIF is also limited to 24 bits while USB can pass 32 bits (in case you are interested).

Is there an audible difference? Now that’s a loaded question!

My feeling is that, all things being equal, yes there is an audible difference.

I’m a big proponent of using a DDC when passing hi-res files (DSD) through USB to a DAC. I connect my media player to a DDC (Singxer SU-1) then on to my DAC (Gustard x20 Pro) via I2S. 
My media player is an Asus VivoPC (a fan-less, small form factor pc) so to remove all noise from the USB, I rely on the DDC. At lower resolutions it may not make a noticeable difference though.


It’s a very good question.  Unfortunately, the answer isn’t  nearly as good: It depends.

It depends on your particular components.  In newer units, and only in general, it would appear to most forum participants that USB tends to sound better than SPDIF (and AES EBU which is essentially an SPDIF over a balanced cable) as it is higher bandwidth and enables the DAC to be the clock coordinating the two.  There are those who disagree, of course.  And, many successfully argue that USB is a terrible audio interface as it’s really optimized for computer convenience.  It can pick up electrical noise and be finicky.

Nevertheless, in my own units, USB is superior to SPDIF.   It’s pretty clear that there are very good older DACs for which the USB interface was an afterthought and falls far short of SPDIF.  If you use SPDIF, be sure to use a 1.5M true digital cable with a 75 ohm impedance. (Worth some google time as to why if you’re wanting to use and SPDIF).

Optical solves an electrical noise problem but it tends to be “jittery” and I haven’t seen many people claiming it’s the equal of USB or SPDIF.

Some new units also accept Ethernet and some are using IS2. I am using the latter between my Auralic Aries G2 and Auralic Vega G2.1. It’s at least the equal of the USB (probably slightly better) but also allows additional communication and resource sharing.  

In sum, the only way to really know is to try the various options with your own gear.  You may conclude something entirely different than what I’ve suggested above, or you may find that your results are somewhat consistent with the above.  Unfortunately, there’s no substitute for trying it yourself.  But, i do think your DAC manufacturer will have some guidance as to what they suggest will work best.  Perhaps start there.

Best,

Nyquist applies only to continuous waves, so intermittent (music) 22kHz signal would have sidebands that may encroach on the 1/2 sampling frequency limit. It is remotely possible, but amplitudes of folded signals will be extremely small (as are 22kHz harmonics). 96kHz sampling would fix it for sure, but some people claim they need 192kHz. Same goes for number of bits. 16 bits means about 96dB dynamic range requiring 96dB of system total S/N to hear it. 24dB provides about 144dB dynamic range. In order to hear lowest bits you need 144dB S/N. Nobody has that, but you want 32 bits (192dB), claiming it sounds better?
On one hand it is insanity, but on the other - if it helps (placebo effect) it helps, and there is nothing wrong with it. ;)
Other than injected noise (computer cable is connected) asyn USB should be better, since it delivers only data (timing is not involved). S/Pdif has timing encoded with data and that can produce timing jitter (DAC clock is based on it). There are many factors involved in selecting best interface (Ethernet, USB, Toslink, Coax) and nobody can know for sure what is the best solution in your home. Trying at home and comparing seems to be the best course of action.
"Jitter is no longer an issue with most modern DAC’s or streamers that now has internal clocks to combat jitter."

This is totally incorrect as virtually any DAC will provide better SQ if it is fed a cleaner digital signal.

There are many fine renderers or digital bridges that will aid in providing a better digital stream to the DAC than the one coming out of a PC, for example. Some of them terminate with a s/pdif, some with USB and others with I2S. Users have found quality sound with each of these types. The price ranges for these are quite large. Some of the highest value devices use s/pdif as output.

My DAC will accept an I2S input and, theoretically, it should be best. But I did some research on the issue to find, to my satisfaction, that any such device at a reasonable price did not perform as well as the unit I finally ended up with that has a USB output.
kijanki
Nyquist applies only to continuous waves, so intermittent (music) 22kHz signal would have sidebands ...
It isn't clear what you mean here. Music is a "continuous" wave.
This is great information and I appreciate the thoughtful responses. I have experimented with my system connections, but never tried USB, thinking it would be inherently noisy. Noted definite improvement of SPDIF over Toslink. I use Transparent Reference or Ref XL cabling exclusively. I am getting Aesthetix Romulus DAC upgraded from Signature to Eclipse at this moment so I want to extract everything it can offer when I get it back. 
Intrigued by the idea of DDC between streamer & DAC. Will be the subject of more research. The obvious consensus is USB, and my takeaway is to get one that is well made and shielded. The notion that SPDIF has timing encoded with data was new to me vs USB only data. Never really thought of it in those terms. 
This is how the forum is supposed to work. Thank you all
99% of the back catalogue is redbook. Presently the best connection is mainly dependent on the quality of the interface implementation between streamer and DAC. Increasingly, the engineers spend their efforts predominantly on USB. There are however still instances where specific proprietary solutuons (e.g. I2S) outperform.
@cleeds It is likely that harmonics of 22kHz come from percussion instruments, that are not continuous by nature.  100Hz (drum roll) will give you (modulated) 22kHz with bunch of sidebands spaced 100Hz apart.  First two sidebands (21.9kHz and 22kHz) will have already very small amplitude and the rest of them can be ignored.  Sidebands appear because signal is modulated (not continuous).  The danger here is that anything above half of sampling frequency will, in D/A process, fold into 0Hz and up.  With 44kHz sampling 22.1kHz will become 100Hz signal and there is no way to remove it.  We might not be able to hear 22.1kHz, but will definitely hear 100Hz.  I stated that even with 44.1kHz it is very remote possibility and folded-over frequencies will have extremely small amplitudes.  96kHz would eliminate any possibility of this happening while 192kHz is a waste of space on HD, IMHO.  I have few plain redbook CDs with such breathtaking quality, that makes me believe that 16/44.1 format is not the limiting factor.  192kHz sampling or 24bit resolution might be important in studio during mixing, to avoid loss of quality, but final product in 16/44.1 is fine with me.

 
papafrgog,  rate of incoming signal and rate of D/A conversion have to be synchronized.  With async USB computer sends bunch of samples (frames) at certain rate (usually 1 kHz).  DAC receives such frame and places it into buffer.  Dac signals back to computer when buffer is over or underflowed and computer adjusts the size of next frame.  That way rate of D/A conversion is based on DACs internal stable clock only and no data sample is lost.
@cleeds instead "First two sidebands (21.9kHz and 22kHz) will have already very small amplitude and the rest of them can be ignored."
it should say:
"First two sidebands (21.9kHz and 22.1kHz) will have already very small amplitude and the rest of them can be ignored."
kijanki
It is likely that harmonics of 22kHz come from percussion instruments, that are not continuous by nature.
Any percussion instrument is an "continuous" as any violin or other instrument. Consistent with the Fourier Transform, this wave can be recorded.
... 100Hz (drum roll) will give you (modulated) 22kHz with bunch of sidebands spaced 100Hz apart ... Sidebands appear because signal is modulated (not continuous).
The signal is analog, it is continuous.
The danger here is that anything above half of sampling frequency will, in D/A process, fold into 0Hz and up ...
Anything exceeding half the sampling rate is filtered out as part of the A-D conversion process. It doesn't "fold" into anything. You can easily prove this with measurements.
Any percussion instrument is an "continuous"
No, it is not. If you record drum roll it will show individual bursts with silence in between.
The signal is analog, it is continuous.
Analog and continuous are two different terms. When you strike drum once signal is not continuous. There is a silence before and after. Any modulated frequency will show on FFT as a root frequency and sidebands. It applies not only to amplitude modulation, but to any modulation, including time jitter of digital signal.
Anything exceeding half the sampling rate is filtered out as part of the A-D conversion process. It doesn’t "fold" into anything. You can easily prove this with measurements.
There are analog filters before A/D process but they cannot be very sharp. By definition they have to be even group delay (Bessel), and those are very hard to make in analog domain. Bessel filter characteristic show practically the same, weak attenuation within 2x Fc, no matter how many poles you use. The whole idea of single bit converters (Delta-Sigma) was to avoid sharp filters by pushing quantization noise higher. In any A/D and D/A process there is ALWAYS something that folds over. The only issue is the scale (amplitude) and all I stated is that Nyquist applies only to non-interrupted (continuous) frequencies, but artifacts of this violation are likely non-audible with redbook CD and definitely non-audible with 96kHz sampling.



kijanki
... If you record drum roll it will show individual bursts with silence in between ... When you strike drum once signal is not continuous. There is a silence before and after ...
It isn’t clear why you think percussion instruments are not "continuous." They're as continuous as a trumpet - every trumpet player runs out of breath eventually. There’s nothing especially unique about the analog wave produced by percussion instruments.
There are analog filters before A/D process but they cannot be very sharp.
That something is not perfect does not make it useless. Filters used in A/D conversion can be measured to show essentially perfect rejection of frequencies that would violate Nyquist.
... Nyquist applies only to non-interrupted (continuous) frequencies ...
Nyquist and Fourier Transform apply equally to percussion, trumpet, violin and other sounds. Consider that Nyquist is not a theory - it’s a theorem.
Again, any signal that is amplitude modulated will have sidebands.  Organ note played will show single frequency on FFT, but interrupted 10 times a second will show root frequency and sidebands spaced 10Hz apart.   

I have to agree with you, that filters in front of A/D will be essentially perfect.  I forgot that they record at 192kHz (not 44.1kHz) and 10 pole Bessel will have, at 10x Fc, attenuation above 100dB (below one bit of resolution).