It's not about data loss. We agree, the bits arrive as they should, otherwise there would be a dropout or massive distortion. The explanation is on the UpTone Audio website, from an expert in silicon chip design and a leader in the field of network audio, which you are clearly not.
cpurdy
: "It boggles my mind that people could be using "dedicated streamers" that have so little buffer space that they are converting UDP packets in real time into audio. Seriously, 1990 called and wants its archaic technology back."
"
Q. Some music renderer endpoints have software or hardware buffers, as do the USB or Ethernet interfaces of some DACs. Do the effects of upstream phase-noise and leakage still impact these?
A. Buffers by themselves do not block phase-noise overlay. As long as there is input data (such as in most USB endpoint buffers) the ground-plane noise from the data still enters the DAC’s ground-plane whether there is a buffer there or not. In addition, this phase-noise overlay also occurs inside the buffer itself, plus the input clocks on both the input and output side are subject to the threshold affects of noise on the ground-plane AND on the ground wires inside the buffer chip. So small buffers with data going in and out don't really make much of a difference, and if not done just right can actually make things worse. USB systems usually have a very small buffer, and since there is always data coming in while data is going out, phase-noise overlay from incoming data is still a factor. Even if the buffer after the USB receiver is large, there are always a fair amount of USB packets still on the bus. The only way that goes away is to completely close down the USB connection."
https://cdn.shopify.com/s/files/1/0...enson_EtherREGEN_white_paper.pdf?v=1583429386
You're taking this a bit too personally, but simultaneously you are assuming that a company that is trying to make money selling ridiculously over-priced antique electronic equipment using (literally) unbelievable claims and imaginary technical words is somehow completely objective
However, to keep this as objective as possible: It's been over 30 years since I have done any silicon chip design work, and I am not a leader in the field of network audio. You have successfully called me out, and I am every bit the fraud that you assumed that I was. Other than the part where I didn't actually claim to be designing silicon chips or leading the field of network audio.
Of course, outside of this particular thread, I've never even seen the name UpTone Audio, so my guess is that they are also similarly fraudulent, in that they are neither designing silicon chips nor leading the field of network audio.
Now let's get into the Q&A:
Buffers by themselves do not block phase-noise overlay.
I'm not sure what phase noise overlay is. I know what phase noise is, and I suppose that the use of the term "overlay" here refers to the possibility of more than one phase noise source, which is not an unreasonable use of the term if you are trying to describe the visual output on an oscilloscope to a child, for example.
As long as there is input data (such as in most USB endpoint buffers) the ground-plane noise from the data still enters the DAC’s ground-plane whether there is a buffer there or not.
This is a soup of words, but I think what they're trying to say here is that "sending information over cables doesn't eliminate ground plane noise". That's not an unreasonable statement. If your ground plane does not have any noise, then you are living in some alternative universe. Ground plane noise is quite literally unavoidable, which is why people go to university for more than a few years to learn how to design electronic circuits.
In addition, this phase-noise overlay also occurs inside the buffer itself, plus the input clocks on both the input and output side are subject to the threshold affects of noise on the ground-plane AND on the ground wires inside the buffer chip.
This meaningless jumble of words is probably where the story falls apart. I'm not even sure where to start dissecting it, but let's go through it:
*
"this phase-noise overlay also occurs inside the buffer itself" So what? That has no effect on the data in the buffer.
* "the input clocks on both the input and output side are subject to the threshold affects of noise on the ground-plane" So what? The clocks are oscillators which (a) aren't impacted by that noise (which is why we use crystal oscillators etc. in the first place) and (b) do not generate phase noise (materials are selected for oscillators based on this specific property, i.e. a single dominant phase). See for example: Quartz.
No, really, look it up!
*
"the input clocks on both the input and output side are subject to the threshold affects of noise [..] on the ground wires inside the buffer chip." That's the same ground. Perhaps they should have listed a few dozen other places where that ground is used?
That is what a ground is!
(But note that ADCs and DACs will generally use two separate grounds, i.e. electronically isolated within a larger system, with one on the digital side and one on the analog side. See the TI links at the end of this response for some visual examples.)
So small buffers with data going in and out don't really make much of a difference, and if not done just right can actually make things worse.
It is true that we used small buffers up until the 1980s and even early 1990s. Back when you could see transistors with the naked eye.
As I said previously, the 1990s called and want their antiquated technology back.
As far as "
can actually make things worse", that statement is meaningless without additional information. The reason that buffers are and were required is that without them, there is no data.
Years ago, we could get by with tiny buffers (as small as 2 bits per line) because (drum roll) we had extremely dependable oscillating clocks. USB and Ethernet need much larger buffers, because of (i) packetization and (ii) support for mismatched clock rates.
USB systems usually have a very small buffer, and since there is always data coming in while data is going out, phase-noise overlay from incoming data is still a factor. Even if the buffer after the USB receiver is large, there are always a fair amount of USB packets still on the bus. The only way that goes away is to completely close down the USB connection."
There's no such thing as a USB system. A USB transceiver chip may have a relatively small buffer (perhaps in the low kilobits size range). For example, TI documents it thusly:
"The receiver contains an elastic buffer used to compensate for differences in frequencies between bit rates at the two ends of a link. The elastic buffer must be capable of holding enough symbols to handle worst case differences in frequency and worst case intervals between SKP ordered sets. A SKP order set is a set of symbols transmitted as a group. The SKP ordered sets allows the receiver to adjust the data stream being received to prevent the elastic buffer from either overflowing or under-flowing due to any clock tolerance differences"
That particular chip (the 1310 model) is 15 years old now, so we're not talking about some new invention.
But let's cut to the chase: The claim from the
"expert in silicon chip design and a leader in the field of network audio" is that noise can come over a wire, even USB or network, and that noise will come out as part of the analog signal from the DAC. Because (
hear me out, because this part is bleeding obvious) the electrical engineers designing the DAC had absolutely no idea that electrical noise in the system was even a remote possibility, because they have never heard of USB or clocks or networks or ground planes.
Yeah, I find that all a little hard to believe, too.
I don't work for TI, but they're nice enough to have published a few little papers on this topic (which I hadn't seen before, but they look pretty easy to digest for a lay-person, which obviously I am, being a fraud
):
https://www.ti.com/lit/an/slyt499/slyt499.pdf (part 1) and
https://www.ti.com/lit/an/slyt512/slyt512.pdf (part 2)
Just to be 100% clear though: I'm not claiming that it's impossible for a DAC to have noise in its output that comes from its ground plane. I think most well designed DACs of the past decade will not have
any perceivable noise in their output that comes from their ground plane (indicated by the noise floor measurement, subject to the explanation of measurement noise here:
https://www.analog.com/media/en/training-seminars/tutorials/MT-001.pdf). But there will always be noise, and it will always be measurable with precise measurement devices, until Congress repeals the laws of thermodynamics.
