From a technical point of view , I think that we need to consider perception of SQ especially when there are a lot of of individuals that hear a positive correlation .
Of course we know since the 1950 that THD+N (after a certain threshold ) is in no way correlated to SQ . So why should we look at this information since we know its useless ?
What is SQ ? We don't know at this point and no machines ( I use AP every day) will tell us.
In the search of better sound , we must listen to what people (many) say .
I think that a good path...technically speaking .
Of course we know since the 1950 that THD+N (after a certain threshold ) is in no way correlated to SQ . So why should we look at this information since we know its useless ?
What is SQ ? We don't know at this point and no machines ( I use AP every day) will tell us.
In the search of better sound , we must listen to what people (many) say .
I think that a good path...technically speaking .
A lot of people believe in horoscopes and crystal healing. Doesn't make them technically correct. If someone wants to spend 100s or 1000s on placebos to make them feel better about their system fine, but encouraging others to do the same needs to be questioned and the claims put under a microscope.
Data arriving on ethernet is not a perfect square wave. It has edges , some small sine wave ( and some RF fast edges). Do you think it has no effect on SW CPU ? Zero modulation ..?
As for the noise on ethernet lines (after the isolator magnetic) just use any scope and check... lots of noise depending on bandwidth .
So we did establish there is noise on the " galvanically isolated ethernet "
How is that impacting SQ. Frankly I don't know the mechanism , I dont even know (me or anyone on this thread) what SQ is..
What we do know...is a majority report increase in SQ .
" We cannot measure thus its not real "
Maybe you are right .. but math say bigger probability is that you are wrong and we just have not found , yet , the test and mechanism .
You appear to be taking issues from SPDIF with an embedded clock and assuming ethernet must have this. The data on ethernet is buffered before sending onto DAC (by whatever method). Ethernet runs at 10x a 24/192 bit rate and the CPU is running at GHz clock speeds. I cannot see how clock edges affects anything.
For my world, I cant see what tempted you to add this post. What was your intention? I find many of your posts quite vague.. like you don't really have a firm opinion but rather seek to see if you can somehow alter the discussion... for no obvious reason...
I'm with chris719 here...
//
I'm with chris719 here...
//
Basically planet10 just pointed out that the argument of data integrity is insufficient.
Being convinced that an exchange of a "whatever" (like an ethernet switch in this case) can not lead to an audible difference is fine, but generally it helps to think about the premises that have to be true so that the conclusion is correct.
Plausible is not a synonym for correct; the list of interface protocols (and implementations) that never could lead to audible differences, but nevertheless did (or better to say where a physical possible reason for audible differences was found after through analyses), is quite long and the data integrity wasn't the reason in each case afair. It usually are second or third order effects.
If you remember the recent argument about the "data transport from server to server"; it illustrates where the models used for the argument differs from reality.
Data integrity is a necessity for any interface but it is insufficient for the mentioned conclusion.
(Disclaimer: That is just a general remark, I do not develop or sell ethernet switches nor have I done sensory tests on this kind of effects)
There were claims a few years ago that the power supply used on the device for doing rips of CD's affected the sound of the ripped files (ie regardless of playback device, the files ripped with the good power supply were superior to those ripped with the normal power supply). This is along the same vein as this thread.
There is a lot about digital transfer of information that people just don't seem to understand. The world would be in big trouble if the transfer of digital information was affected by the quality of the network devices being used!!
Tony.
Post 12 is about digital data transmission . Good post .
This is not what we are discussing , the question is can any type of noise on digital bit perfect bus affect the SQ ?
Yes ethernet is buffered but that means it will stop all packet communication ? Will that mean that noise on ethernet lines is suddenly eliminated ?
Now for clock edges RF .. of course the energy of (some) of those edges will modulate the Gnd and propagate into the CPU . Any output of that CPU will have some of that RF noise on the bus
I can say this...on one of my DACs (MCLK) I had a perfect almost jitter free square wave.on both 44/45Mhz Yet I could hear a difference between the 2 sample clocks
Upon looking further with spectrum analizer I could see that some very small RF was present on one clock output (1000Mhz bandwidth) in the orders of 10-20uV multiple spikes .
So I did what I had to do...and removed the noise. SQ clearly improved . THD+N stayed the same
Yeap noise on the bus is important not only jitter and bit perfect.
This is not what we are discussing , the question is can any type of noise on digital bit perfect bus affect the SQ ?
Yes ethernet is buffered but that means it will stop all packet communication ? Will that mean that noise on ethernet lines is suddenly eliminated ?
Now for clock edges RF .. of course the energy of (some) of those edges will modulate the Gnd and propagate into the CPU . Any output of that CPU will have some of that RF noise on the bus
I can say this...on one of my DACs (MCLK) I had a perfect almost jitter free square wave.on both 44/45Mhz Yet I could hear a difference between the 2 sample clocks
Upon looking further with spectrum analizer I could see that some very small RF was present on one clock output (1000Mhz bandwidth) in the orders of 10-20uV multiple spikes .
So I did what I had to do...and removed the noise. SQ clearly improved . THD+N stayed the same
Yeap noise on the bus is important not only jitter and bit perfect.
Post 12 is exactly what we are talking about, that data through a ethernet switch cannot possibly change the quality of an audio signal.
Data packets are buffered to allow out of order frames to arrive before they can be passed up through the OSI layer.
But now you have moved the goal posts to CPU noise.
Data packets are buffered to allow out of order frames to arrive before they can be passed up through the OSI layer.
But now you have moved the goal posts to CPU noise.
I am not moving goal post , but in DA conversion , data and bus noise are linked as far as SQ
Post 12 is about "getting 100% perfect digital data transmission" and you get it on DA silicon ...along with noise of that bus .
In my experience , it has an influence on SQ and even THD ( harmonics change with hi levels of RF noise)
Post 12 is about "getting 100% perfect digital data transmission" and you get it on DA silicon ...along with noise of that bus .
In my experience , it has an influence on SQ and even THD ( harmonics change with hi levels of RF noise)
100BaseTx is around 31MHz clock. CPU bus clock on your computer is? To take an example PCIe lanes start at 2.5Gb/s LVDS. The ethernet clock is laughably slow compared to all the other RF going on in there.
This is the DAC master clock which is one place where noise and jitter CAN be an issue. Possibly the only place.
Finally, some truths !
Thanks Dave. Finally, some sense/insight:
" USB & Ethernet as transport methods are still developing and getting better as the little gotchas that many do not even know exist are worked thru."
pj
Thanks Dave. Finally, some sense/insight:
" USB & Ethernet as transport methods are still developing and getting better as the little gotchas that many do not even know exist are worked thru."
pj
Precisely.
We naturally want to put things in boxes when we think we understand them. It is a great strategy to help us cope with a very complicated world. But in audio this can limit our thinking. We lose context.
An ethernet pulse transformer is labelled "galvanically isolated" so there can be a tendency to overlook it's influence on other parts of the circuit. The truth goes much deeper than that: it is isolated at DC (to allow good digital signal integrity by eliminating ground loops) but not at AC of course. It has capacitance, leakage inductance, crosstalk etc. In short is not perfect. Exactly how blameless it is can only be determined if you understand every interconnection and interaction with the rest of the circuit and therefore the system (I do not profess to know that!).
Another example of box thinking: Us audio types stick anything labelled "ground" in the zero ohms at all frequencies box because that is how we can rationalise it. A RF engineer will tell you that there is no such thing as "ground", it is simply not possible to achieve zero ohms at all frequencies, the higher you go, the more impedance depends on circuit materials and geometry and the more unpredictable things get. We all can hear the detrimental effects of uncontrolled RF demodulating in semiconductor junctions to bring us splashy gritty highs, even though as cdsgames says, it often makes no dent on a static THD figure. If you can't hear that then I envy you
Last edited:
Still waiting for some evidence.........have you got any because I don't expect any from Dave?
That's pretty much misplaced trust. People hear different things depending on time of day, venue, friends around, peer pressure, fatigued or not, price, color, shape, reputation of equipment, the list goes on and on.
It is very common here on this forum that people have radically different opinions on SQ of something. Half say it's great, other half say it sucks. They can't all be right at the same time, so that is solid proof that about half of them are wrong. You just don't know which half.
Interestingly, my AP shows exactly the same numbers today as it did yesterday. Even when I am fatigued, stressed, etc. ;-)
Now before you see a chance to derail this discussion with 'we don't listen to numbers', I know that. And I am not claiming that.
Jan
Last edited:
This is a pretty intense and worthwhile conversation, please keep it civil so it can continue.
Supporting something that is BS doesn't un-BS it. But it does give us a better insight into your logical processes, such as they are ;-)
Edit: Kevin if you think this is too personal feel free to eradicate it.
Jan
Last edited:
It is not rational to question a very expertly constructed data transfer control protocol. If it were problematic, all computers that communicate this way all over the world would be processing the wrong data somehow.
In addition, switching devices have a very simple task.They are connecting at least two different node and address them.
Also data does not care which kind of ethernet cable is used for transmission (unless the length is a problem).
In addition, switching devices have a very simple task.They are connecting at least two different node and address them.
Also data does not care which kind of ethernet cable is used for transmission (unless the length is a problem).
- Status
- Not open for further replies.