Testing a few digital cables

[manisandher]

Did I hear the same/similar differences in both the files (your example track and mine) or do I suffer from amnesia ?

Yep! This is what you said about my original files:

I don't know what the gag or pitfall is, but with the first version I thought "oh boy, finally cymbals to compare but now all lack dynamics. It really disturbed me. Also the colour of the cymbals was not right. The tick in the beginning on them is too loud compared to the too thin remainder.
I noticed that especially at the end things were not right at all (was it hitting on the rim ? very close by that on the skin ?).

With the second there was much more body to the cymbals but the punch (sort of) in the drums was so much lacking that when the bass started I seemed to have missed the lower keyed drums (compared to the first version). Colour seemed too dirty of everything and halfway the piano started to disturb me (buzzes a bit through the room).

With the third all was suddenly OK ? Dynamics were there, cymbals had the right colour, the bass seemed to be more life-like and the end sounded dynamic as should.

So, although these have not been analyzed, we can assume that they are indeed identical too.

Mani.

[PeterSt]

I assume you supplied the latest incarnation of the files, right ?

Anyway, I am pretty sure they did not compare the files as should. And you know what ? I have massive problems with doing it myself this time. The first time this is so !

Working on it ...

[manisandher]

I'm going to redo the test, but this time through an DAC and then ADC. The path will be:

microRendu (no DSP) or Mach II (no DSP) -> Chord 2Qute DAC -> Prism AD124 ADC -> Tascam spdif digital in

In this case, even though the mR and the Mach II are producing bit-perfect outputs (as proven by Ray on ASR), we certainly should expect to have a difference in SQ. We'll see...

Oh and BTW, this time I am not going to reveal which captured file is which! Let's get any listening biases out of the way.

Here we go!

1. Original CD rip (converted to wav, for those not using XX):

https://drive.google.com/open?id=0B0PU5LO5jVjfTUdrMGtRRkZuZkk

2. Digital transport A (microRendu or Mach II):

https://drive.google.com/open?id=0B0PU5LO5jVjfM0paUHFfRVppZHM

https://drive.google.com/open?id=0B0PU5LO5jVjfQ1JYdk9mVWtEbHM

3. Digital transport B (microRendu or Mach II):

https://drive.google.com/open?id=0B0PU5LO5jVjfSGxOVE9BWmR2aWs

https://drive.google.com/open?id=0B0PU5LO5jVjfMEF3eURHVkU1ZWM

[Edit: Actually there was no need to redo these - they were fine in the first place. I've added new links.]

Of course, these 3 files will not sound identical. The original file (#1) is going through the Chord 2Qute DAC and then through a Pass Labs X1 preamp (to convert the Chord's SE output to balanced) and then through the Prism AD124 ADC (balanced input only). But nevertheless, we should be able to determine which digital transport, the mircroRendu or the Mach II, is doing the least harm.

So, which file, #2 or #3, sounds closest to #1?

Mani.

[manisandher]

I assume you supplied the latest incarnation of the files, right ?

Yes, the files I supplied were of 'your' Konbanwa track.

Here is are the null results: http://audiosciencereview.com/forum/index.php?threads/sonore-microrendu-hardware-tear-down-review.770/ (scroll down towards bottom).

Mani.

[PeterSt]

Yes, the files I supplied were of 'your' Konbanwa track.

Of which you supplied two versions ...

I have seen that thread over there. This is why I say that they did not compare as should. Ok ?

[manisandher]

I wanted to be totally consistent and upload 3 files of the same format to ASR, and not 1 flac + 2 wav. Not everyone uses XX, and I didn't know whether other software players would treat flac and wav identically (as I know XX does).

So, I converted your flac to wav with dBpoweramp Converter. I used this wav file to make the digital recordings, #2 and #3. And this is somehow invalid? Again, enlighten me...

Mani.

[PeterSt]

And this is somehow invalid?

Don't make yourself crazy. Did I say or even imply somewhere that something is invalid ?

Don't you see that I am comparing myself and that I thus need to know what you supplied those guys ?
Btw, I could have downloaded it from there of course, but I didn't (I already had the lot). And, aligning the files takes an hour or so.

Anyway, start putting yout Tascam, mR or Stealth to eBay. Or all of it. Something is not right and nothing is equal anywhere.

Satisfied ?
I am.
Btw I am not going to listen to the latest test. I have fun as it is.

But seriously ... for you, start thinking what can be wrong with the latest (is 2nd) version of my track you put up. And notice it can even be FLAC. The differences are always 1 or 2 decimal (on 65536 possible values) - more than 8 million of them.



What I like you to do is make a second take of the mR output. So, just 100% the very same as you did last time with my track and also with the very same source file (if you converted it to WAV first, then again do that os use the WAV if you still have it). Thus, changing nothing compared to the previsous time !
If that resulting one is different from the one I have here (lead-in and -out differences trimmed) then it has to be your Tascam.

Peter

PS: I thought you had learned by now that you shouldn't give random people some files to compare. Because remember, those people are the same who prove that XXHighEnd sounds the same as JRiver. They just don't know how to do it (apparently).

PPS: No I am not deaf. I also can't be placeboed. But you can ! hahaha

[PeterSt]

I am not sure what the options are, but let's start with these :

1. FLAC encoding/decoding is not lossless. And keep in mind, I just used the literal XXHighEnd method for the decoding (I grabbed the UnicodeTrack file coming from it).

2. The Tascam fails on us.

3. The digital output of the two means (mR and Stealth) are not bit perfect. Or at least one of them.

4. The digital output of XXHighEnd is not bit perfect. Btw, think of it ... you can avoid upsampling all right, but I don't think you can avoid the bit depth going from 16 bit to 32 bit. And then what ? then something must be rounding, because the files are 16 bits all right.
Ha !

Please add other possibilities (anyone is invited).

I did not take the #3 file yet, because I thought that the tick in the beginning could bother me. And it still can, as the other file(s) has been trimmed already and possibly it needs more of that to avoid (skip over) the tick.
Anyway, interpret 4. above carefully because I imply that XXHighEnd's output can not be compared ... but I did not yet. But with the mR the same can be going on.

Peter

[manisandher]

Something is not right and nothing is equal anywhere.

I'm sure you picked this up, but I just wanted to be absolutely certain...

There are clicks/ticks on file #3, but not on #2. Their frequency varies with varying SFS. Ray has referred to them as 'stuffed zero samples' - there seem to be 7 in total over the 4min 36 sec track. These have increased the length of #3 by 7 samples over #2, and obviously need to be taken into account in the nulling process.

Sorry if this is totally obvious to you, but I wanted to be sure.

Mani.

[manisandher]

But seriously ... for you, start thinking what can be wrong with the latest (is 2nd) version of my track you put up. And notice it can even be FLAC. The differences are always 1 or 2 decimal (on 65536 possible values) - more than 8 million of them.

I have no idea what the difference between your flac and my wav files of #1 could be. More than 8 million of them? How has this happened in the flac-to-wav conversion?

What I like you to do is make a second take of the mR output. So, just 100% the very same as you did last time with my track and also with the very same source file (if you converted it to WAV first, then again do that os use the WAV if you still have it). Thus, changing nothing compared to the previsous time !
If that resulting one is different from the one I have here (lead-in and -out differences trimmed) then it has to be your Tascam.

Will do.

Mani.

[PeterSt]

OK. I don't think it will be really needed to re-do the mR recording as I suggested; this was before I found the rounding probable problem (though in the same post).

I have been looking at it again, and while I now found other means for better proof to myself, I could also find back the differences I had a hard time with (as mentioned, for the first time).

Thinking further ... it is hard to dictate this all to rounding errors, with also the knowldege that expaning to 32 bits without upsampling, should just add (padd with) zeroes. Also I am not even sure any more how XXHighEnd deals with this these days (or better : 5 years ago, because it must have been that long that I looked into that for the last time).
I also come to this "conclusion" because looking at it in the way I want(ed all the time today) I am reminded by the dithering XP always did and it just looks like that. So ... maybe you forgot to switch off Dithering in HQP ?? (I suppose it can do that at 16/44.1 just the same)
XXHighEnd does not dither.

Below you see the difference file of it. This is always 1 decimal of difference. And yes, earlier on I said "1 or 2" but I think I misread the hexa-decimal data somewhere.

Peter

[PeterSt]

I will now do the #3 file as well; I think this will be useful for conclusions, especially because I'm on the path of Dither now.

[manisandher]

So ... maybe you forgot to switch off Dithering in HQP ?? (I suppose it can do that at 16/44.1 just the same).

No definitely no dithering in HQPlayer.

Mani.

[manisandher]

Thinking further ... it is hard to dictate this all to rounding errors, with also the knowldege that expaning to 32 bits without upsampling, should just add (padd with) zeroes. Also I am not even sure any more how XXHighEnd deals with this these days (or better : 5 years ago, because it must have been that long that I looked into that for the last time).

I set DAC to 16 bits in XX. This is the only setting that would work with my 16/44.1-only capable Legato USB-to-spdif converter. The 'DAC is 24 bits' and 'DAC is 32 bits' buttons became grayed out once I selected 'DAC is 16 bits'. I therefore assume there was no padding going on.

Mani.

[PeterSt]

Mani, correct. But I just forgot whether the NOS1 plays at that setting. Uhm ... which you did not use of course.
So indeed, with that setting it is and remains 16 bits. So, good.

Peter