Can you hear 16 times the jitter?

[manisandher]

... Mani, try with the Arc Prediction filter only (not any of the Custom Filters).

 

Well, I do know the outcome, but tried this nevertheless (in my office system, with a NOS1a DAC). And yes, the difference between the two 'original' files is more pronounced - 'the studio cave' file now has a slightly larger cave

 

Over here, if I load the full original FA track (5min 56.04sec) into Audacity and truncate it down to 30s, the sound (of the first 30s, obviously) changes. The truncated file sounds indistinguishable from one of those Dennis loaded (not the 'cave' file, but the other 'original'). My full original meanwhile remains markedly superior.

 

Why this should be the case, I have no idea. I don't want it to be the case. It's a PITA. But it's what I'm hearing. I could so easily have just voted and left it at that (anyone can see how I did BTW once Dennis releases the results). But in all the tests I've gone a step further and actually described what I'm hearing. Over here it's real, and it's repeatable. I don't know what else to say...

 

Mani.

[esldude]

Well, I do know the outcome, but tried this nevertheless (in my office system, with a NOS1a DAC). And yes, the difference between the two 'original' files is more pronounced - 'the studio cave' file now has a slightly larger cave

 

Over here, if I load the full original FA track (5min 56.04sec) into Audacity and truncate it down to 30s, the sound (of the first 30s, obviously) changes. The truncated file sounds indistinguishable from one of those Dennis loaded (not the 'cave' file, but the other 'original'). My full original meanwhile remains markedly superior.

 

Why this should be the case, I have no idea. I don't want it to be the case. It's a PITA. But it's what I'm hearing. I could so easily have just voted and left it at that (anyone can see how I did BTW once Dennis releases the results). But in all the tests I've gone a step further and actually described what I'm hearing. Over here it's real, and it's repeatable. I don't know what else to say...

 

Mani.

The thing is both my 'original' files were truncated. Neither was exactly the first 30 seconds. I can check later when I am at home and set how much the difference.

[mansr]

The thing is both my 'original' files were truncated. Neither was exactly the first 30 seconds. I can check later when I am at home and set how much the difference.

 

The Fiona Apple originals are offset by 207 samples, Hyperion Knight by 149 samples. I don't have the full tracks, so I can't tell the absolute starting point.

[manisandher]

The thing is both my 'original' files were truncated. Neither was exactly the first 30 seconds. I can check later when I am at home and set how much the difference.

 

That might be interesting.

 

I could so easily simply say that both your 'original' files sound different to my 30s file - I mean, I know that you shifted both. But that's not how I hear it. Only one of your shifted files (the 'cave' file) sounds very different here. The other sounds pretty much indistinguishable from my 30s file.

 

Mani.

[manisandher]

The Fiona Apple originals are offset by 207 samples...

 

They should have different offsets, giving different checksums, no?

 

Mani.

[YashN]

New generic Chinese 12AT7 installed a couple of weeks ago and yesterday I though the sound in my system improved again + last time I did a listening test with Mani's files, our perception here seemed to be at odds with others' so take this with a grain of salt:

 

Listening today:

 

Man, this is what is wrong with a lot of gear and gear discussions and gear listening: a slow, jazzy as in not much timbre variation and minimalist to top it all - makes it very difficult to latch on some characteristics I'd like to listen closely to like rapidity of resolution of attack transients being one. Here, there's precisely nothing to allow me to judge that.

 

How about a New Order - Blue Monday or a Kraftwerk - The Robots, where the determination can be done right away a few seconds into the track for eff's sake?

 

This said, here are my thoughts based on listening impressions this morning, and the SET Tube amp is on for 20 mins max yet, I probably listened to each track 7 times, so I haven't spent a lot of time on the task (Then again, GIGO as mentioned):

 

Y and Z: the voice sounds more spacious, there's a slight buzz on a bass attack which sounds like the instrument normally buzzing if the strings are pinched on the fret imperfecly.

 

Z: the voice sounds flatter and smaller, the slight buzz above on the bass attack sounds like a glitch instead of a normal performance imperfection.

 

So, in my system, and in a pissed mood at the choice of music:

 

In my opinion, X is the worse-sounding.

[mansr]

They should have different offsets, giving different checksums, no?

 

Mani.

 

The offsets of the originals differ by the numbers I gave, but I can't tell how far into the full track they start.

[manisandher]

Y and Z: the voice sounds more spacious, there's a slight buzz on a bass attack which sounds like the instrument normally buzzing if the strings are pinched on the fret imperfecly.

 

Z: the voice sounds flatter and smaller, the slight buzz above on the bass attack sounds like a glitch instead of a normal performance imperfection.

 

So, in my system, and in a pissed mood at the choice of music:

 

In my opinion, X is the worse-sounding.

 

With the risk of pissing you off even further, do you want to sort your X, Y and Zs out?

 

Mani.

[manisandher]

The offsets of the originals differ by the numbers I gave, but I can't tell how far into the full track they start.

 

OK, let's wait for Dennis to get back to us.

 

Mani.

[PeterSt]

Well, I do know the outcome, but tried this nevertheless (in my office system, with a NOS1a DAC). And yes, the difference between the two 'original' files is more pronounced - 'the studio cave' file now has a slightly larger cave

 

OK, wrong answer; my theory is that they should sound the same now; apparently they don't. Then this has to be crucial in this situation :

 

Over here, if I load the full original FA track (5min 56.04sec) into Audacity and truncate it down to 30s, the sound (of the first 30s, obviously) changes.

 

This too can be explained but possibly only to those who know which XXHighEnd's dials may vastly change sound and how they operate (or just believe what's perceived of that). But this is speculation at this time, so for another day.

 

 

Dennis, what I was hinting at is the way how digital filters operate;

Not talking about ringing per se, it would be true that the audio output lags a little, caused by the filter. This is because of the output result of one sample is determined by the previous samples (and the "how" largely depends on the filter). If we simply think averaging over 100 samples then 99 empty (silent) samples contribute to the first non-empty sample and the output will be the level of the current sample divided by 100 (I'm sure you get the gist). This means that with the upcoming samples all showing the same level, only at the 100th sample the level is what the samples contain for real. Thus a massive lag. This lag continues throughout the file (track) and we can say that the last sample will still contain the information of the first samples (think sliding average now, which most often is an underlying method of digital filters). Now :

 

That chopping 3 samples off the back end of a 1.3 million sample string changes the character of the two files?

 

This is not entirely correct because a. it is more (and unimportant) but b. you did similar at the start (and again more). Now the above applies, BUT as I said, depending on the DAC, with that meaning : depending on the filter.

 

So by not starting a track at the beginning of it, but randomly in the middle, it will already sound different. It *is* different, because the filter output is different.

 

Mentioned Arc Prediction filter does not apply any averaging of any kind (it is genuinely interpolating) so with such a filter what I described just can not happen. Sadly that (my own little test of this towards Mani) did not work out. This certainly does not mean that Mani is hearing ghosts, because the first time that we both will disagree about such tests is yet to happen (if my memory serves me well). And so I easily take it for granted that something else is going on as well.

 

Of course I should have listened myself, but the test, say, backfired on me; I saw Mani judge three different ones while only two different should be in there, blindly trusted that result, and thus looked into the files instead of listening to some "set up" which could be spoofed. The setup was NOT spoofed. But in the end the obfuscation now backfired on the test itself (still thinking "depending on the DAC").

[PeterSt]

They should have different offsets, giving different checksums, no?

 

Mani.

 

(I only looked at Fiona set of files)

 

Ah, wait. I only now recall the real reason why I dove into the files instead of listening what I surely planned;

I saw that the size wasn't even equal. So how could they be the same.

 

But hey, actually this is smart obfuscation. So two of them really *are* the same (I checked that). But first chop off those lead-ins and -outs.

[PeterSt]

Especially for Mani :

 

Fiona Apple Y1.wav

Fiona Apple Y2.wav

 

(rightclick - Save As to save)

 

Please describe the difference, if any.

[goldsdad]

The ultimate audiophile tweak will surely be spawned by this thread: remove/insert some samples from/to the start of each file to liberate the music from the edgy cave.

Of course, one will need to hunt for the optimum offset adjustment for each file, which will vary for each DAC. And whenever you change your DAC, the process will need to be repeated.

 

Audio store staff will just love it when they hear, "Hmm... sounds a tad cavernous and edgy. Can I listen to that track again, but without the first 173 samples?".

[YashN]

With the risk of pissing you off even further, do you want to sort your X, Y and Zs out?

 

Haha, no no prob, but I grouped Y and Z because I didn't think there were perceivable differences at least by me alone, in my system.

[kdubious]

Assume:

A filter that is influenced by the prior 100 samples, a 100000 sample track ("Track 1") that starts at sample 1 (and ends at sample 100000), and a second track ("Track 2") that starts at sample 50 (relative to Track 1) and ends at sample 100000 (relative to Track 1, for a total length of 999950).

 

Once you get to sample 150, the preceding 100 samples are the same in each track.

 

So you'd have slightly different sound between the point where Track 2 starts in Track 1, and the actual intro in Track 2.

 

But the outros, and the entire remainder of the tracks, (again assuming a filter that is only influenced by the preceding 100 samples) would sound the same.

 

Maybe you left out something in the explanation of your filter?

[esldude]

Assume:

A filter that is influenced by the prior 100 samples, a 100000 sample track ("Track 1") that starts at sample 1 (and ends at sample 100000), and a second track ("Track 2") that starts at sample 50 (relative to Track 1) and ends at sample 100000 (relative to Track 1, for a total length of 999950).

 

Once you get to sample 150, the preceding 100 samples are the same in each track.

 

So you'd have slightly different sound between the point where Track 2 starts in Track 1, and the actual intro in Track 2.

 

But the outros, and the entire remainder of the tracks, (again assuming a filter that is only influenced by the preceding 100 samples) would sound the same.

 

Maybe you left out something in the explanation of your filter?

 

What you are asking is what I don't get. I understand how the filter results will roll around and around. The result even many thousands later could be a touch different. But how much is that difference at that point. I can't put it together in a way it would be anywhere close to an audible difference.

 

I do have a couple ideas to test this and see what the actual real world difference would be. Might not get to them today. I honestly think I am wasting my time, but will see what happens.

[esldude]

The ultimate audiophile tweak will surely be spawned by this thread: remove/insert some samples from/to the start of each file to liberate the music from the edgy cave.

Of course, one will need to hunt for the optimum offset adjustment for each file, which will vary for each DAC. And whenever you change your DAC, the process will need to be repeated.

 

Audio store staff will just love it when they hear, "Hmm... sounds a tad cavernous and edgy. Can I listen to that track again, but without the first 173 samples?".

 

No kidding. A mastering guy is doing final edits. He cuts 100 samples this rather than that. One version is an open, wonderful beautiful sounding track, and the other is dull, lifeless, boring and sounds in a cave. Oh, and yes, as you point out, the exact parameters for that will vary depending on the DAC in use and the filtering it has.

 

I find my credibility stretched beyond the horizon on this idea.

 

Of course applying Occam's razer is what I should do.

 

EDIT because I forgot: Beside the above scenario, remember an 8th generation copy can be confused for either of these. Just by accident, happenstance and luck. Despite clock drift and jitter.

[mansr]

What you are asking is what I don't get. I understand how the filter results will roll around and around. The result even many thousands later could be a touch different. But how much is that difference at that point. I can't put it together in a way it would be anywhere close to an audible difference.

 

I do have a couple ideas to test this and see what the actual real world difference would be. Might not get to them today. I honestly think I am wasting my time, but will see what happens.

 

A FIR filter won't be affected like that but an IIR filter might.

[esldude]

The beginning of the Fiona Apple track was chopped by 523 samples. Then the other one was 207 samples more offset than that.

[manisandher]

I remember having an argument with one of my lecturers at Uni. He said something like “CD sound is perfect”, and I made it clear that based on my experience at that time that I didn’t think so at all. He pretty much took an hour lecturing me (can you believe?) about how the N-S theorem says so.

 

That was back in 1986.

 

Over and out.

 

Mani.

[goldsdad]

No kidding. A mastering guy is doing final edits. He cuts 100 samples this rather than that. One version is an open, wonderful beautiful sounding track, and the other is dull, lifeless, boring and sounds in a cave. Oh, and yes, as you point out, the exact parameters for that will vary depending on the DAC in use and the filtering it has.

 

I find my credibility stretched beyond the horizon on this idea.

 

Of course applying Occam's razer is what I should do.

 

EDIT because I forgot: Beside the above scenario, remember an 8th generation copy can be confused for either of these. Just by accident, happenstance and luck. Despite clock drift and jitter.

 

Not only is the 8th gen file distorted by multiple generations of clock drift and jitter, the encoded waveform exhibits significant group delay relative to the original files. Image (click it to enlarge) showing the 8th gen aligned with original:

 

 

 

Yes, it is amazing that Mani’s system makes the 8th gen file sound only subtly different to one of the originals, while the originals, which differ only in offset, are made to sound grossly different to each other.

 

 

Also, if offsetting a file’s samples greatly alters its sound quality, it is astonishing that you accidentally created just the right offset in one file to make it sound indistinguishable to Mani’s own un-offset 30s file when played by his software and DAC.

 

 

 

[sdolezalek]

Yes, it is amazing that Mani’s system makes the 8th gen file sound only subtly different to one of the originals, while the originals, which differ only in offset, are made to sound grossly different to each other.

 

Also, if offsetting a file’s samples greatly alters its sound quality, it is astonishing that you accidentally created just the right offset in one file to make it sound indistinguishable to Mani’s own un-offset 30s file when played by his software and DAC.

 

I love it! Now we have a case where MEASUREMENTS indicate real differences that most of us can barely hear. And the same people are hearing significant differences between files that measure as "same."

 

Dennis: Were you intentionally stirring up this much controversy? It sounds as though you yourself are a bit confused by these results...

[sandyk]

Now we have a case where MEASUREMENTS indicate real differences that most of us can barely hear. And the same people are hearing significant differences between files that measure as "same."

 

Surprise, surprise !

The usual suspects simply refuse to accept that modifying .wav files (NOT normal Data files) and saving them again, results in a degree of degradation as PCs aren't fully electrically quiet. In this exercise there have been several lots of processing in audio editors and saving the files again.

As I have said previously we are more often than not given files of differing degrees of mediocrity instead of the original higher quality original.wav files.

G.I.G.O.

[kumakuma]

The usual suspects simply refuse to accept that modifying .wav files (NOT normal Data files) and saving them again, results in a degree of degradation as PCs aren't fully electrically quiet. In this exercise there have been several lots of processing in audio editors and saving the files again..

 

Perhaps you should actually read what people are saying. No one here is saying that the simple act of saving an unmodified file changes its sound.

[sandyk]

Perhaps you should actually read what people are saying. No one here is saying that the simple act of saving an unmodified file changes its sound.

 

Well I am saying that the simple act of saving an unmodified track to a different storage location as a COPY of the retained original does degrade it's sound ,the actual amount depending on how electrically quiet the PC is, as well as the actual storage location, especially if it's on USB, and I don't give a damn whether you believe me or not.

All of the files provided have been modified to some extent, as they are excerpts from the original track.

As Dennis is worried about clear violations of copyright, perhaps he could have provided links to the COMPLETE tracks on request for personal use ONLY, without naming the files directly.

 

I will not be entering into further discussions with you on this subject.