Test track with distortion which isn't there...

[BertD]

....weird name for a topic but still it is true. I found a small piece at the start of a track which is recorded very softly which apparently seems to be a hard case for players. I heard this go wrong under XX Engine #2 Doubled under XP and also under Vista using doubling and/or upsampling.

Without upsampling and playing at 44k1 this sounded okay without distortion. Today I tried the track again under Vista #3 (XX version 0.9) with doubling and with upsampling and now it sounds okay no matter what I do with the player.

Something somewhere sometimes can go wrong and probably not just under XX or using upsampling or not. Perhaps nice to try at home to hear if you recognise a piano playing with a small bell "singing" continiously. If not then there is something wrong with your playback...

Play it loud cause it is a rather low level part of the recording, it will end before it gets loud so no worries (unless you have another track selected after that one!).

Bert

[PeterSt]

Bert ! You must be ACrazy person ...

[Bert pointed this out a few weeks ago, when he didn't have #3 available yet]

For those who are interested ... this is my personal explanation :

First of all, those with horn speakers ... dive with one of your ears into the mid/high horn. The right one is best, but left will do also.
Btw, set XX to a Delay (start time) at a convenient number of seconds, so you have the time to get to the speaker without hearing yourself breathing only. And, indeed set your volume to the max, or whatever it needs to hear something anyway. As Bert implied, do not add another track in the Playlist or your windows might go out.
(you could also compare with other players of course).

Now, what's happening here, is that the volume is so much "compressed", that with the 44K1/16bits this track is about, there's very very few headroom to express the variations in volume level. I didn't look at the data, but it would well be 1 % only where all plays, of the 100% available.

The above for me implies that if the smallest thing goes wrong, all goes wrong for the perceived result.

When I listen to this with bit perfect playback the Fireface MME drivers imply, all sounds normal. Assuming you all don't have such a driver, start with normal 44K1 under XP, which is not bit perfect. #1 or #2 does not matter much in this case (tracklength is short enough to use #1 if you want).
To me the bit perfect drivers are slightly better than the not bit perfect drivers. However :

That they do differ, is noticeable by means of doubling. I say : Doubling doubles the error.
So, with #2 and Doubling, you can hear that it's not as clean or pure anymore.

Now we go to #1 or #2 under Vista, just 44K1.
Already, there's no way we can recognize the "original" (remember, which is already not much original under XP with the not bit perfect drivers).
I say : the original is about a piano and small twinkling bells, and suddenly now the piano sounds like a bell.
Explanation : the resampling Vista performs, even if you selected 44K1 as what your soundcard (DAC) can do (Advanced tab in the Device options). SO it resamples from 44K1 to ??? and back.

Now set your Device (we're in Vista) to anything else (like 96K) and it gets worse. There's only bells now ... No piano anymore.

Lastly, select Engine #3. This is your best option, surely when you don't have bit perfect drivers under XP.
Start with 44K1 ... all is okay.
Then Double it ... all remains okay (meaning : we're not "doubling" errors here).
Then Upsample it ... still all okay (meaning : XX' Upsampling is okay).


Of course this is one example only, of very special caliber. I don't know whether it's representative for everything, but at least - to me - it gives the idea to be on the right track. *Also*, obviously, it gives the idea that not bit perfect XP is worse. For this case anyway. Possibly for all.
So, no matter XX may sound way better than anything else ... here's the proof that it really needed Vista to outclass ... well, itself.

If I try to be scientific, I say that this is pure harmonic distortion. I mean, when a piano just sounds like a bell, things are mixed up heavily, and of course we know how many harmonics a bell has. From this situation (again, I already knew it) I started to learn to "hear" harmonic distortion, and which really made me hear the "clipped" data from the example here.

Keep in mind : sound data not being "bit perfect" is just the tyniest form of distortion, of which about the whole world would say it's inaudible. It already was not (to me and others) because of the way you perceive it. Yeah right. Here though, it can be listened to as an absolute phenomenon.
When this is upsampled the way Vista does it internally, it will be *or* very heavily upsampled in order to get to the common denominator first (and you know what I think of that, removing all the squarish waves which are present in the original), *or* it is performed in a dumn wrong way (avoiding the heavy upsampling then). In any case in the end dither must be added, which comes down to not being bit perfect by already that means only.

Bert, I think it is great that you found this example. 
Peter