As if there was some kind of "inverted differential" mechanism trimming a part of the "common sounds" from the 2 channels.... Just an image of course...
Apart from Jud's great resemblance of what happens (at least here), I think Alain put out the best description of what I perceive. If you read carefully into that little text, you see synthesis, fake and "processing" the least. But also look at this from Jud :
The precision of the rendering may also very possibly be good enough that imperfectly done stereo effects which would wind up in the middle of the soundstage with slightly fuzzier reproduction come out as one effect on the left, and its near-but-not-quite twin on the right. I heard some of this, intentional or not, with the electronic effects on "Cherokee."
I would say we are on to the merits of what is happening, though still not knowing how it happens. So, true; but is it really a twin as how you describe it ? I perceive the exact same sense of it (and indeed not from all music) BUT (and now hop back to Alain's observation), like the timing between left and right is not equal. So, our twins, to me, seem not to be able to form like this, not even in electronic music. Oh, it can be done, but it is not a real thing which is applied. Think like this :
Set up a synthesizer voice with some sawtooth. On that super impose another "VCO" (anothert voice) but have it time delayed for its frequency compared to the first and then such that the second has its tops where the other has its dips (envision the wave). Now output each to its own (L/R) channel. And hey, have the tops and dips in eachother's middle for any frequency ! (which makes the delay a variablee subject to the frequency).
... I don't think that can be done in the first place, and it certainly (??) can't be done with output to separate channels.
When the delay is fixed in time, a rather similar effect may happen, but maybe not so profound; it would depend on the frequency (hence music). But, I am fine if we look at it like this (is what is happening).
Next, it would be fairly easy for the OS to make a few mistakes and output to the left channel vs the right channel in a "lagged" fashion. Say that the samples for L and R sure go to the proper channel, but that someone is smart (??) and collects 10 samples for L and output them and next 10 for R and output them. It would be more efficient ...
... and exhibit this effect.
... and I am a conspiracy thinker of course ...
When a thing like this happens, it would not make the lot bit perfect. And well, I have been looking for reports about this, but all I ran into is that this is OK for W8. But is it really ?
When a thing like this really would be happening, it can happen right in advance of the digital output. This means that any internal loop-back capture would miss it; it will report bit perfect. But take e.g. SPDIF out and loop that back with a cable, and it can't be missed.
This latter is how it should be done and how I always did it. Possibly I saw nobody doing it like this and already the fact that my method needs trimming of the start of the result in order to align it for the remainder is a pain and possibly why people don't do it like that. Somehow I always see people making "null" (diff) files and it seems to easy; as if this trimming is not in order, which possibly is not in order indeed with an internal (software) loopback.
Add to it that in the NOS1 driver this "loop-back" provision is there just the same and that after this a ton of OS kernel code is still in charge.
My sort of point : I don't own a capture device anymore so I can't check.
My sort of other (conspiracy) point : One of the NOS1 users noticed that the system volume could be used with Kernel Streaming. Now THAT is odd. I believe it was so that this only happened with WASAPI active (so, Normal OS Mode or let WASAPI active). Bert ?? But anyway, this looks totally 100% impossible to me, knowing my own driver. This IS a bit tricky, because the W8 part (new in the 1.03 version) of that was not made by myself, but the part which does or does not "subscribe" to Windows Volume events is in my own code, and that code was not changed. However, whatever is in my code can be in kernel code after that and overrule it.
What I also read - and none of it was checked by me in my own system - is that Shared Mode can now be bit perfect. This was not so. So, things changed here for sure, but don't ask me whether it is related to my conspiracy subject.
The fact though that Shared Mode can be bit perfect while Exclusive Mode clearly is not once the Windows volume is in between all, tells me that at least there is a fair possibility that we don't listen to bit perfect at all, and maybe not only because of some influencing volume which should not be there in the first place.
For me it is of utmost importance that I can't get a concistent view of what is happening and that exactly never happened before. This can be looked at as similar to fairly inconsistent judgements of people, which also never happened before much. Somehow, suddenly, all is subject to PC's, chains and speakers. Maybe we all talk about the same thing and it just needs the proper "technical" judgement, like from Alain, Jud and me for that matter.
If someone is willing to grab SPDIF out (can be from the MoBo) and loop that back into a recording device with SPDIF in, record a full track like that ... then I will be happy to spend the time and look for bit perfectness. Better not try to do this latter yourself because when it is bit perfect it can take you an hour already to find the common part at the start of the both files (recording and original) and many hours more before you give up at finding that part (because it just is not there). I have done it many times before so it is far more easy for me.
If this is a NOS1 user doing this, don't forget to deactivate Is NOS1 and play 16/44100 natively (no upsampling).
And *if* someone is going to do this, please announce it in this topic, so we are not all doing the same thing. Another one may do it just the same of course - and which would be good for some safety or sanity check, but it is not really required. One should be enough.
So, hoping with can get some conspiracy thinking out of the way ...
Peter
