NOS1a and Q5 & W7 vs W8

[christoffe]

What's a bit new is that I'll receive a new library for the driver which in my opinion can solve the stopping of W8 after 10.x minutes, which some people still suffer from. The message : It can happen that W8 forgets to ask for a new (audio) package, so the PC actually doesn't send it and sound stops.
Notice this is a derival from other issues solved in this library and whether it really solves that problem is to see, but it is (hopefully) a clear example for you.

Interesting to know if the driver update solutions also impact sound quality. This is win8 only? I have not experienced any stops after the pagefile/reboot tricks.

Regards, Coen

I do have this problem (stopping after 10. min) when moving the "Playback Drive" to a RAMdisc.

Joachim

[PeterSt]

What's a bit new is that I'll receive a new library for the driver which in my opinion can solve the stopping of W8 after 10.x minutes, which some people still suffer from. The message : It can happen that W8 forgets to ask for a new (audio) package, so the PC actually doesn't send it and sound stops.
Notice this is a derival from other issues solved in this library and whether it really solves that problem is to see, but it is (hopefully) a clear example for you.

Interesting to know if the driver update solutions also impact sound quality. This is win8 only? I have not experienced any stops after the pagefile/reboot tricks.

Regards, Coen

I never have the problem any more (most do not I think), but I (only now) feel it can be related to using the Silverstone. Since then I also have USB2 and USB3 disabled in the BIOS. I am not sure anymore when it stopped to happen though, and what I recall is "since an upgrade". But for me this tells nothing (always using all canges right away); for all of you it will.

Peter

[CoenP]

Something noteworthy about the hump:

Yesterday I went back to win7 again, and I thought wtf!? A clear emphasis on the mid bass compared to what I was hearing before.

BUT:
Then I noticed that I was still in "normal" OS and rebooted to "minimized" OS. And from there everything turned normal again: the nice win7 sound without a noticeable hump.

regards, Coen

[acg]

Surely someone else would have noticed this.

Anthony, then you go claim that both sound the same ?

Otherwise you ask rhetorical questions with the most obvious answer that I don't know (and I am saying that from the start of W8 - there HAS to be a difference and I don't know where it can be, but at least I think I am one step further now).

On a side note (or not) what always puzzled me is that AmirM (former lead developer of Vista Audio Team) was asked the question (WBF through Mani) and he didn't even respond. Why ?

But the most logical is it has to be in "USB" somewhere. At least from that I can reason that a digital loop-back (to check bit perfectness) not necessairly travels the same path as when we go to a DAC. This is all driver stuff with also the notice that W8 is totally different here than W7 (with btw "solutions" in 8.1), and OS/X being even more of a mess.
IOW you don't know what (audio) driver developers all have to do in order to let USB work decently; it's completely tweaked by now (OS/X worse than W8).

What's a bit new is that I'll receive a new library for the driver which in my opinion can solve the stopping of W8 after 10.x minutes, which some people still suffer from. The message : It can happen that W8 forgets to ask for a new (audio) package, so the PC actually doesn't send it and sound stops.
Notice this is a derival from other issues solved in this library and whether it really solves that problem is to see, but it is (hopefully) a clear example for you.

Peter

Hi Peter,

Sorry abut the semi-rhetorical questions.  I guess that I never thought that the frequency response would be anything but flat in any operating system...just too basic an assumption in hindsight obviously...surely not difficult to get right.

Regards,

Anthony

[CoenP]

Hi all,

While we are amidst fact/opinion finding I wondered about the origins of the bass hump in win7. I like to share some thoughts that go left and right on this for those who wonder too.

When we exclude the possibility that nVolts of noise anywhere can cause +2.5dB difference in DAC output, the foremost candidate is the datastream itself. Since the bump is in such a limited frequency range it is IMO impossible that it stems from semi random flipping bits. The phenomenon is therefore a time related issue and then the order of the samples becomes prime suspect.

Now if we have an interface that does not notice dropped (or all 0) or duplicate samples or systematically corrupt samples or swapped samples or whatever sample "error" and this happens at a certain (approximate) interval, then the DACs output will contain a  signal at that frequency. Like in a way noise shaping takes advantage of.

This may seem plausible but:
A for me unexplained factor following the reasoning above is that only the ~90Hz tones seem to be impacted. This should be noticeable on a frequency sweep on the nos1a.
Or are they not alone and do the 90Hz tones show up in any FFT ?!
And what about the hump having the same relative gain for all levels?! I would expect it to be proportional to the "volume" level of the data.
Or what about it not being random at all because otherwise your tracks would have sounded different each time you listen....(or do they...). This consistency points IMHO to software components that repetitively create the same errors.

Another interesting part is how the OS plays a role in all this. Apparently there are differences in the way the OS handles the audio data and maybe there are also differences in interaction between hardware, OS and XX-software. Like a different chipset, #cpu etc requires different drivers with their own influence.
Anyway like Peter said, you may establish a successfull USB transmission, but what is assuring you that the data actually reach the DACs in the intended order? And how sure are we/can we be that the proper data is sent?

Anyway, if we assume that the Hardware and OS together with XX are capable of influencing the extend and frequency of the sample "misbehaviours" then we have an audible and consequent influence on the sound. And also one that is consistent over the same configuration and thus is different for others!

If we in the end cannot find a digital mechanism that causes such a consequent narrow banded anomaly then it must be PC noise still creeping in,   big time.

Thanks for reading so far !

Regards, Coen

[PeterSt]

Hi Coen,

After reading this I get a bit scared. I mean, we now (obviously !) make this a heavy subject, while the means it was detected by me is very very indirect and therefore vague. Too vague to dedicate the very discrete potential causes like you pose them. The least I should do is applying the same measurements (or calibration actually) by means of W8.
It is only that I will run into an at this moment unknown number of installs of software and drivers etc., in order to be able to do it in the first place, so I don't like to do it. But as soon as we are going to talk "too difficult" then that is first, right ?

What is also first is that a sweep you talk about can "almost" be done in the digital domain. With this I mean that best would be to measure at the DAC output and which is nothing strange to begin with and if that does not show the 2.5dB difference for W7 and W8 then what. *This* now is easy, because all it needs is playing a test signal of ~90Hz through W7 vs W8 (can be done by XXHE of course) with an analyser hooked up to the same DAC and look for differences.
So this should be the first thing to do.

Thanks,
Peter

[PeterSt]

But of course I won't believe myself that I am going to see that 2.5dB difference and this is also why I start to be scared;

My story is great of course; calibrate a speaker under W7, listen to it under W8 for judgement to next state that it was tuned to W7 and therefore sounds different in W8 than what was measured from it which was done in W7. Next, we take out one notch which was under suspicion anyway, and now the sound in W8 is the same as in W7 with notch. Well, it is true I'm afraid, but if really really true it should show that 2.5dB difference on the DAC output.

All in other words (and not much different from your ideas about it Coen) : I won't believe that I will see the 2.5dB difference because I see no explanation for it, while on the other hand the audible difference is 1000% clear.

In the mean time I went a step further with this filter suggestion : Re: Your First Weekend DSP Twip.
You can't see it in there, but what I did with that was making more flat which was not. So, one part of it I copied into below screenshot, and with some imagination (or experience) you can see that this part compensates for its left part of something like a 80Hz center frequency (never mind the 104.2Hz center frequency physically used which is related to another filter) to 40Hz. So, correction stops at the mouse pointer.

No Mani, I did not forget about your 40-80Hz remark.

Everybody should realize (and keep on doing that) that we talk about the calibration of a speaker and all we see, and what I myself *can* see, is what it produces for flat line or not. If there is a peak somewhere we try to flatten it, and might there be a dip somewhere we can try to boost it (with DSP). No single person is anticipating "playback software" behaviour, obviously. However, while we now clearly talk about that (Operating System), we should never forget that we stil look through a speaker which requires corrections.

When I was finished with the lot, not only that 90Hz required notch came from it, but also the more broadband 40-80Hz. It is a bit hard to see them separate, but say that the correction is from 40 to 90Hz and this implies a notch. So envision below graph to be inverted and leave out the part to the right of 90Hz.
But this is strange. Well, here we go again, it is theoretically strange because where the driver should start rolling off (~100Hz) it needs more notching. But where are theories when the design of the speaker is just compensating for that roll off to begin with. Thus, what I'm saying is that we just look at "a result" and take it for granted. Not much more to do.

Yes, until you take out that broadbanded notch by means of an "anti-boost" and which is what I did with that latest filter suggestion and ... which now even MORE resembles W7 when used under W8.

So, ... or nothing ?

Let's emphasize : All I did was putting on a blindfold and make a line straight which I can expect to be straight where it for filter results. This is (still) hard to explain, but let's say that if anything, with a rolling off driver beyond 100Hz there should be a compensating boost. I did not even apply that - I applied a straight line. This thus, while Windows 7 showed a "compensating" notch to be necessary.
And as a side note : the real boost begins at 40Hz where "W7 also needs that" (??) and I did not touch that. Do notice though that if I leave it to the people themselves (Mani, VJ, KnB) they all add just over 2dB to the whole line (up to the crossover) first, which I did too (!). So from this you could determine that the whole of W7 shows ~2dB more bass up to ... dunno (but say the crossover).

Still with me ?

For sound results it is up to now quite hard to determine whether this last trick (in the Twip topic) colours the sound or just is for the better overall. And you know what ? I tend to say the latter. It is clear to me that it requires more fine tuning, but without that it looks like I can live with this because what now happens in that low area is 10 times better than it was. And the colouring I digest and seems like needing to get used to (also : I am not sure whether I hear colouring or whether things got nicely profound while a virtual nothing was there before - this is close to 100Hz and might be that necessary 90Hz notch after all, for more or lesser extend).

Sorry for the difficult sentence(s).

Point is and remains : I now listen to W7 but still better. Yesterday, like the day before, I hear no colouring anymore after 30 minutes or so. Instead, all what I play now shows roar and thunder and more than deep down earth. I now can play the more old "ambient" (say 2007 and earlier) with the idea that that stuff has to have it from the lower frequencies and which did not show at all before (under W8). Therefore it was "nothing" because the higher frequencies on their own made no sense (to me).
The normal music is not coloured at all, BUT, implies a fullness which may not be real. And yes, it has something like a cheap speaker (VJ didn't put it like that, but I know what he means). Still, it is a famliar sound and the fullness (filling the room) is something I like.

Notice : This latter is exactly where W8 turned into that other direction to begin with. A sort of less fuzzy in the bass. More clean and a perceived more dark. But maybe something we shouldn't like much, because of too clean.

Lastly a nice other note :
What we have with the two "Twip" appliances, brings us to the limits of Windows 8 were it for the low end and the Orelo MKII speaker. I mean, if more deep down earth comes from this, the vulcano will explode. *Because* of being at this limit, I am sure it is interesting to listen with these DSP settings to Windows 7 again, because all should be a totally clear complete mess.

Peter

[CoenP]

All in other words (and not much different from your ideas about it Coen) : I won't believe that I will see the 2.5dB difference because I see no explanation for it, while on the other hand the audible difference is 1000% clear

Check. It would probably have been noticed anyway in the development of the NOS1.

Now something that causes a percieved difference of a few dB's can't be small or can it? To me this is the real audio mystery and allways raises the question:if we cannot measure it, are we measuring the right stuff then?

Now you seem to have got the bass on win7 level, what about the highs? Can you mimic the gray (in your setup) highs of win 7 in win 8? Not that you gain any direct advantage from that. Yet this would be an ultimate test that OS disturbances primarily cause percieved frequency abberations which is much more desirable than a fine detail robbing noise.

Regards, Coen

[PeterSt]

Now something that causes a percieved difference of a few dB's can't be small or can it? To me this is the real audio mystery and allways raises the question:if we cannot measure it, are we measuring the right stuff then?

Yesterday I started another post but I didn't finalize it because I got stuck in my reasoning (it prooved some adverse effect of what I was attempting to achieve). Still the truth can be in there. Here is that unfinished text :

At writing my last sentence I got an idea of what's also an explanation, but of such different kind that it will justify the 90Hz to have the same output on both W7 and W8 (again, which is what I expect because I can't see how it will be different) :

So W7 is more messy in general. Now think the whole spectrum, with us needing to remember that the highs are grey just the same and actually no accuracy/resolution is there to begin with (all relative to W8).

What I clearly notice for a difference, is that W8 now goes as low as W7 but it is not messy at all. Again hard to explain, but MKII owners will understand. Thus, no matter how deep down earth low, the vibrations of the frequencies remain, while W7 will just show "low". Also to remember : with W7 you do not feel the woofers following what you (think you) hear, while with W8 you do.

I have said before that when a lower frequency distorts for not being straight (this will not be exactly the same as 2nd, 3rd harmonic distortion), it will go more low. This is actually the "warble" and how the distortion repeats at for example each 5 cycles of the wave. Thus, 100Hz becomes 20Hz, although with not as much energy as intended (plus a few more frequencies, thus messy).

If we believe a bit in such story, we can envision that
a. This shows on the analyser for an indeed somewhat lower output of the 100Hz;
b. It could show a 20Hz small peak (FFT) which of course nobody took care to look for (I at least do not);
c. What may show for minor differences on the anlayser may show way more on the "more analogue" output.



Yes. So what I was actually saying is that what's small distortion on the DAC output, not necessarily remains small in a speaker driver. This is the IMD (Intermodulation Distortion) stuff again. So let's think for our comfort (and my story) that any small "warble" signal in the more digital domain (but think DAC output) is totally harmless *there*. It's just electronics (or electricity) and the fact that this springs from the source will NOT be detrimental to the D/A conversion (assumed all is fast enough). Now :

With a speaker driver this is a totally different story. With a speaker driver you could say that *any* two tones in one driver will cause problems. So now envision that 20Hz to be there because the source implies it (W7) and while this 20Hz is now the carrier, the 100Hz needs to ride on that and all sorts of distortion will be our share (Doppler to begin with).

So what I am trying to say is that what the analyser may not show at all (or I don't know where to look, plus the differences are minor), can be audible vastly. Also, just as this is audbile, the measurement for calibration (which is through microphone) will detect it as easily as we perceive it audibly.

Where I got stuck is that my 100Hz example will show a lower level (because of spread to 20Hz) while it should show higher level (remember, the W7 measurement needs notching).
Of course today I can try to say that a false 20Hz is that product what needs notching, but this is not how measurement works. Thus, if a put up a 100Hz signal, the 20Hz is not observed. Or might it work the other way around : when I put up a 20Hz signal and the 100Hz would be a false byproduct, then the 100Hz is not observed - only the 20Hz.

All together, this post was about IMD in a speaker being much more profound than IMD in electronics, but this time with a base of IMD being there in the source. So two (or more) tones (implied by the OS somehow) while one should be there, and these byproducts being so low in level that you can't observe them, while the speaker driver tries to represent it, and inherently shows additional distortion because of that.

Far sought ?

[CoenP]

Of course today I can try to say that a false 20Hz is that product what needs notching, but this is not how measurement works. Thus, if a put up a 100Hz signal, the 20Hz is not observed. Or might it work the other way around : when I put up a 20Hz signal and the 100Hz would be a false byproduct, then the 100Hz is not observed - only the 20Hz.

Why is it not measurable?

Some thoughts and associations:

What I think is problematic with your observations is that the 20Hz in your example is a byproduct of another tone (100Hz+a litllebit of something). So its magnitude will depend on that tone(s) being present, not on legitimate 20Hz tones. No way this can be compensated for with a notch or hump at 20Hz.
Iow the correction would be for the amplitude of real 20Hz tones and not others.

Also we need to establish if the magnitude of the added tones in win 7 is relatively fixed or not: If you always listen at the same level, the correction may just be perfect to correct the byproducts generated at that frequency at that level, but it will not work anymore when playing louder (effective correction becomes to big) or softer (correction becomes to small).
Secondly you really want to know if the unwanted "boost"  is related to the tones themselves or to others in the music. If a byproduct, then different kinds of music would need different corrections. (the modulation will be linked to the general level of the music and as such grow and shrink along with the rest of the spectrum.)

So I really doubt that applying a loud speaker notch for win7 can completely "cure" a boost problem.
Furthermore I think that the boost range is so limited and specific (non-random) that we must think of an analog modulation of the output signal which can be caused by jitter (can we get any lower?), PSU (ultra low noise), grounding (...USB...?) or radiation (...). This modulation can also fog up the rest of the spectrum.

regards, Coen

[CoenP]

I stubbled upon this thread on CA, especially the tekst in the first post is remarkable:

-link removed see post below-

With these figures you would never expect an USB audio device to be any better than a budget CD player. I think you have to take them with a pinch of salt, but it is all very fragile and provides a basis for audibility of OS/hardware and software through the USB line yet I am not shure how exactly that should impact sound.

Peter joins in at the end and that is the point where it gets interesting (technical) again. Will the NOS1b have an 10G ETN optical interface? Or more interesting: will it sound like WIN 8 (I hope not!  )?

regards, Coen

[PeterSt]

Maybe you linked to the wrong thread there Coen. Nothing much interesting in that thread that I know.
I think you wanted to refer to something which quotes Gordon Rankin. Or ?

Regards,
Peter

[CoenP]

Maybe you linked to the wrong thread there Coen. Nothing much interesting in that thread that I know.
I think you wanted to refer to something which quotes Gordon Rankin. Or ?

Regards,
Peter

Yep, wrong thread. Indeed it was the GR quote:

http://www.computeraudiophile.com/f8-general-forum/gordon-rankin-says-im-wrong-about-universal-serial-bus-industry-standard-cables-connectors-and-communications-protocols-between-computers-and-electronic-devices-cable-sound-20814/#post328624

I will delete the link from the previous post.

regards, Coen

[listening]

Hi Coen,

just another aspect is discussed hier: http://www.ecnmag.com/articles/2012/12/differential-signaling-best-practices

Especially the line-to-line impedance imbalance and transmission line lengths seem to be interesting.


Georg

[CoenP]

thanks Georg,

Dispite all the mess we can still enjoy music (if everything works)!

regards, Coen