I'm going to address a few points here... because there are some nuances... which I will try to mention.
This list is somewhat organized.... but only somewhat...
HOW DIGITAL AUDIO WORKS
1.
A digital audio
FILE is a list of numbers stored in a file.
And, yes, as long as those numbers are all correct and unchanged, that is quite literally the end of
that discussion.
1a.
The way an analog signal is converted into digital content is by sampling.
The value of the analog signal is measured periodically... and the result of each measurement is stored as "a sample".
The timing for these measurements is set by "a clock"... "every time the clock ticks we take a sample".
We assume that this clock is perfectly smooth and consistent.
However, since this happens in the ADC, when the digital audio file is created, we have no control over it (unless we are doing the digitizing).
1b.
In most modern ADCs the process is more complicated, and may involve things like "oversampling".
HOWEVER, since this all occurs in the ADC, the end result is the same:
BY THE TIME WE RECEIVE THE DIGITAL AUDIO FILE IT IS JUST A LIST OF NUMBERS.2.
In order to convert that list of numbers back into an analog audio signal we use a DAC.
The DAC basically reads the list of numbers, converts each number into an analog voltage, and outputs that analog voltage.
The timing for this process is also set by "a clock"... "every time the clock ticks we read a number and output a voltage".
2a.
THIS CLOCK IS NOT PART OF THE DIGITAL AUDIO FILE... Therefore this clock must be "recreated" when we wish to convert the numbers back into analog audio.
And, of course, the clock and data must be made to line up.
(We cannot "let the data overflow" or "run out of data when we need it".
2b.
There are basically two choices about how things happen at this point.
One option is that the source sends the data at the rate at which it is intended to be played.
The DAC then "locks onto the signal and uses the electronic equivalent of a flywheel to smooth it and make a clock from it".
The other option is that the DAC actually simply "creates a new clock from scratch".
And, since the data must be made to line up with the clock, we can only do this if the DAC can control the speed at which the data is sent to it.
2c.
With old-style USB connections the rate at which the data was sent was controlled by the computer.
The DAC then "locked onto the packets" and "regenerated the clock" - usually using one or PLLs (phase-locked loops).
The catch is that this process is never quite perfect... and so the clock you end up with is never quite smooth.
2d.
With a modern asynchronous USB connection the new clock is
CREATED BY THE DAC.
The DAC then uses a control signal to throttle how fast the computer sends the data to ensure that it doesn't overflow or run out.
(Therefore, when we do it this way, at least in theory, as long as the data is "clean", nothing "upstream" matters in terms of clocks and signal speed.)
2e.
With Toslink and Coax connections the situation is more like old-style USB.
The speed at which the data numbers are sent over the connection is determined by the original clock.
And, at the receiving end, that same original clock is used for the data.
(In technical terms "we use an embedded clock".)
Because of this the clock of the sending device, and even the cable itself,
CAN affect the clock rate.
2f.
When we talk about things like Roon, and Ethernet, and DLNA, this all pretty much goes out the window.
Ethernet packets
DO NOT travel consistently or smoothly.
In fact, with Ethernet, packets can be lost, sent out of order, or even sent later if they turn up missing (Ethernet and TCP/IP will take of all of those messy details).
Therefore you can forget about the source clock, and the signal quality, and noise and such.
Ethernet packets are simply a way to get that list of numbers from the source to the destination.
It is up to the player to put the list back together and make up a new clock to go with it.
NOW TO THE FUN PARTS
3.
As we've only suggested up until this point.
OUR GOAL IS TO HAVE PERFECT NUMBERS AND A PERFECT CLOCK.AS LONG AS WE DO IN FACT HAVE PERFECT NUMBERS AND A PERFECT CLOCK THERE IS NOTHING LEFT TO DISCUSS.Period! End! Full Stop!
3a.
However (#1)
Anything that actually prevents the list of numbers from arriving intact can produce audible errors.
So, for example, if there is so much noise or jitter that the signal is "lost" or "interrupted".
3b.
However (#2)
We don't live in a perfect world.
Therefore it is possible that things like noise and interference can "creep into the analog circuitry".
So, for example, a lot of noise
could actually cause the numbers to be garbled...
Or noise could get into the analog circuitry in the DAC and find its way to the analog output as audible noise...
Although, obviously, this will depend on the actual circuitry in the DAC.
So,
IF YOUR DAC WAS PERFECT, they wouldn't matter... but your DAC isn't perfect.
3b(2).
Now we get into the details...
For example, an asynchronous USB connection is pretty robust...
As long as the packets arrive in readable condition it is immune to noise and jitter (because the DAC creates its own clock).
And,
as long as the DAC is well designed, it should be relatively insensitive to noise on the incoming ground and power lines as well.
BUT, since a Coax or Optical connection
DOES use the clock from the incoming signal...
It
CAN be affected by jitter or excessive noise on that signal... or by a "bad cable" that makes those worse...
4.
Filters, and reclocking, and jitter... Oh My!
4a.
Since an asynchronous USB connection uses a clock created by the DAC
MOST of this shouldn't matter.
(Now you know why I consider that my preferred choice.)
4b.
MANY DACs these days include some sort of re-clocking which they apply to all of their inputs.
(Some DACs do this with a separate ASRC chip; Sabre DAC chips have a built-in ASRC function; there are other alternatives.)
This is actually redundant if you're using an asynchronous USB input but can be significant if you're using other types of inputs.
4c.
There are also a variety of
NOISE FILTERS, fancy wires, replacement power supplies, and such.
The short answer there is that, in most cases, they do nothing... but, in
SOME cases, with
SOME DACs, they may make a difference.
One thing I've found is that some USB-powered DACs are sensitive to noise on their USB power lines... and so benefit from actual galvanic isolation.
(But, while I can't rule it out, I have never seen that be the case with an AC-powered USB DAC.)
5.
"Extra" sample-rate conversion.
(Or "What if your signal is not bit-perfect after all?")
5a.
MOST processors and AVRs employ some sort of sample rate conversion under at least some situations.
In
most cases this renders them relatively insensitive to both jitter and noise on their digital inputs.
(And, since in most their analog circuitry shares the cabinet with digital circuitry, they also tend to have good immunity to analog noise sources.)
5b.
In many situations your digital audio signal will be up-sampled or down-sampled at the source.
To be quite frank, while
SOME "high-resolution" files sound quite different than their "CD quality" counterparts...
In my experience, in most cases, there is no audible difference that can
specifically be attributed to sample rate.
(There are often differences but we can rarely rule out other possible causes that may account for them.)
5c.
Some broad generalizations:
- up-sampling is less likely to be audibly bad than down-sampling
(so, if you have to pick a "default sample rate" you're better off picking 24/96k than 16/44k)
- sample rates above 96k are unlikely to be audibly better
(you may be able to hear a difference between 44k and 96k but I
really doubt you can hear a difference between 96k and 192k)
-
EVERY conversion can potentially introduce slight differences... so it's best to avoid conversions when practical
(and, since there is no practical way to create new data, up-sampling is unlikely to ever improve anything... unless your particular DAC "just likes a certain sample rate")
5d.
Again... since I would be remiss to leave it out...
IN MOST CASES THE QUALITY OF THE ORIGINAL SOURCE MATERIAL IS BY FAR THE DOMINANT FACTOR.(For example, it's silly to worry about the difference between 96k and 192k if the original was mastered on analog tape, which is far more limited than either.)
I'm going to close with an interesting quote... about DSD.
This is a paraphrase of a quote from a white paper published by Weiss (which has now become mysteriously difficult to locate online) ...
(Weiss makes Saracon - the most expensive, and arguably still the most respected, studio DSD-PCM conversion software.)
"DSD has no technical benefits over PCM. However, since it offers another option, which some of your customers may desire, it makes sense to offer that option as a product choice."
(Note that the conversion between DSD and PCM is not bit-perfect... so there will always be a tiny difference - audible or not - when converting in either direction.)
Keith said it BEST.
'Bit Perfect'.....
You could send your data to the MOON and back and if it were bit-identical to what you sent? You're Golden....
Is there enough support for beyond 24/96 IN THE RECORDING STUDIO? Or are existing files manipulated?
Who records and ships in such formats?
Are you asking if people buy or listen to 24/192 or DSD? As for bit-identical, are you saying that all that matters is a bit-identical transfer without regard for EMI and RFI considerations? Just trying to understand your point. Thanks.
