More importantly, the question I asked is really very clear:
Is audio quality affected by the hardware you run Roon core on assuming the chain past the RAAT stage is unchanged?
Answers like:
“It cannot matter and that why I never tried”
Are not what I am looking for.
I have tried a bunch of cases as I detailed in my original post and found no dependance. I am asking if anyone has.
Using one’s ears in lieu of a measurement instrument is not the way to go.
Well, both Toslink and coax are S/PDIF; the only difference is the physical line. Not sure what you mean by “recover the lock from the source”, but DACs can’t rely on the bus clock for D/A conversion. They use PLL to adjust their internal clocks to pick up the slack and make sure internal buffers don’t overflow or underflow. Again, it’s the same for both types of cables.
Actually, maybe you really need to listen! And then determine what it is your measurements are not telling you. That’s the scientific method, not the other way around.
Consider amplifier THD measurements. They are done with a 2V input (max input that is). In such measurements, some amps look great, yet sound terrible. How? They have crossover distortion.
What is wrong with this measurement? The signal you’re using probes unimportant regions much more than important regions: a musical signal spends most of its time around low level signals not max level. The time spent in each level interval looks like this:
Where the red is music, the yellow a sine wave.
The point here is that measurements SHOULD spell it ALL out, if they are COMPLETE. But the measurements people do today are simplistic and inadequate.
If you want to use listening as a scientific method, you need double-blind ABX kind of testing. That’s kind of hard to achieve while switching your own cables.
What is the source of this graph, and what are the coordinates?
Recover the CLOCK from the source:
In S/PDIF, the clock is recovered from the source. This means a PLL circuit is following the level transitions and locking into those as a source for the clock. Depending on how crossing (ie the point at which there’s a transition) is done, this crossing time can be affected by drifting voltages, a blurred signal (they are never square of course, that would require infinite bandwidth), drift from the source clock, etc etc etc (yes I do use etceteras).
So even if I recover the same exact digits, because I am slaved to the source clock, my timing can drift around. If for example the source clock effectively drifts between f1 and f2 and back to f1 1,000 times a second, it will create jitter sidebands for a tone of frequency F at F+1000Hz and F-1000Hz.
Some “cavaliere” S/PDIF implementations use a buffer and reclock the signal, but those always run the risk of over or under runs.
USB effectively allows for reclocking as it is able to tell the source to slow down or speed up, guaranteeing no buffer issues.
The horizontal coordinates are amplitudes, the vertical is a histogram in each amplitude bucket.
For a sine wave, the amount of time in a bucket ds is:
(this is the yellow line)
For music, it is like collapsing a waveform onto the signal axis:
and what you get is something like the red line.
That’s nothing cavalier about this. No good DAC implementation relies on the bus clock. Just because there are ways to screw it up doesn’t mean it’s a problem that hasn’t been solved.
You have no idea what you’re talking about. Most DACs do in the case of S/PDIF.
No, they’re usually done over the whole input/output range, just like with pre amps and DACs - you’ve been criticizing ASR…
… but Amir does measure over the whole range, actually, since he knows very well, that distortion profile can vary with frequency and volume.
One thing he doesn’t do is measure a line level device’s output with high and low load impedances, because that can also make an audible difference…
Ok, correct for amps. I take that back. I don’t think I have ever seen this for DACs though.
Plus to be clear, I said ASR measurements are incomplete, I did not say Amir only measures at max input, that was a reference to how most THD measurements are reported.
Check out DAC measurements at ASR…
Sure, so I would assume. However, they have a different transmission protocol as well as different cables
Same here.
Yes but with the limitation only to those products that I have actually used in my main system.
Ok understood. I have heard difference from USB cables and various computers running Roon and Audirvana over USB.
I have never heard differences in Roon core machines when running over RAAT (ie over the network) to the same streaming DAC.
How the DAC treats USB input can indeed make a difference. USB-powered DACS, for instance, may be susceptible to power fluctuations from the computer. For independently powered DACs with a competent USB implementation, there should be no real difference. We have discussed this ad nauseum before.
So those would be Roon Ready DACs with what is essentially a little bit of Roon inside them. Again, not surprised.
(I assume s-bar is s/A)
Something is fishy here. First, if delta-s is fixed (i.e. finite), then delta-t can’t be approximated using the first derivative. If you do that, then any signal that has a local extreme at some moment in time (i.e. where the first derivative is zero) would result in an infinite amount of time spent around that local extreme value, since the derivative is at the denominator. Obviously, all audio signals have local extremes, but the red line in the graph is finite throughout the range. I’m sure the music histogram was derived through other means, then deceivingly compared to the “theoretical” sine wave histogram.
Do the math sweetie. Nothing is fishy. Obviously I approximate ds by s’ dt. Then ds, the bucket, is fixed. This is really trivial stuff.
Drop the sweetie please.
Yes, it’s trivial. dt = ds/s’, so if you have a peak, s’ = 0 and dt is infinite. That would result in vertical asymptotes at each peak in the signal. Like the ones you see for the sine at -1 and +1.
The calc is for a sine wave! And yes, for a very small ds bucket the density goes to infinity. As it should. Look at the yellow line in the waveform plot! I thought you were a physicist?
