I am comparing rips stored on the Melco N10 which have originated on Unitiserve and Melco D100 respectively, if that is your question?
Have you made the comparison, or are you still doing the comparison?
The only way I can see Audacity showing identical ripped data differently is if the files still have additional metadata for volume levelling/normalisation etc
I the two files are bit for bit identical audacity should show them as such.
1 Like
Agreed. It really would be good to have some repeat of the tests - and ideally with multiple people: even better with more than one piece of music.
Thatās correct, but the electromagnetic noise doesnāt get saved into the files, only the binary data.
The electromagnetic noise is created in real time by the playback equipment and not read from the file.
Jitter is a type of noise - itās random variance of the signal (i.e.ānoiseā) in the time domain.
But as what alteration of what specific type of material or energetic substrate is the binary data actually coded in each medium (HDD, SSD, pen drive, etc.)
Until you examine that scale you will never know whether there is something different about the way the data are āscoredā or āwrittenā into that actual medium that cannot be seen by staring at the data.
Indeed for HDD, there can be sub level data left on the platten.
However the drive electronics filters this out and the only thing transmitted to the disk interface is the true binary data at full digital levels. All sub bit level information is destroyed in the process of disk reading and never gets out of the disk drive itself (this equally applies to both SLC and MLC SSDs).
Ok well:
- as originally stated there may be no audible difference in a double blind test anyway.
Or, if there is an audible difference: - Then there would have to be a method of transferring that difference through the process you describe. One can imagine possible ways this could be done - e.g. if there were some noise-created extra data in the sub level data that is left behind, that somehow effects the way the intended audio is extracted (e.g. slows down its retrieval) and thus audibly affects the way the audio data is read and/or reconstructed.
Otherwise any audible difference - if there is one - would be magic or parapsychology!
2 could happen at playback, but there is no mechanism by which sub level data can be transmitted into the drive at the data write time (the sub level data are present from data previously written to the disk and now being overwritten by new data), so the sub level data on an HDD or SSD cannot be correlated to the mechanism used to supply the data to be written (i.e. an optical drive if ripping a CD or the internet connection for downloads).
OK, well Iāve reached the end of my speculative tether now.
If youāre right then there cannot be an audible difference.
Which makes the results of any double blind test interesting.
I understand the programming aspect, the digital electronics used and even the underlying physics of both the electronic devices and the electronic and magnetic storage media involved.
The nail head is hit!
First thing to do, which absolutely requires blind testing, is to establish for certain that there is a difference. There was a telling quote of couple of days ago on the strangely disappeared (I hope only temporary) ethernet +switch madness thread, where someone couldnt tell the difference between two things in blind testing, then suddenly could when he or she could see again which was playing!
The test that the opening post of this thread cited is the only occasion of which I am aware where the metadata has been stripped from two rips of the same music on different mavhines, files said to have been compared and verified as bit-identical, then compared by playing and reported as sounding different (+ the reported Audacity difference). If verifiable, then the original questions still stand.
Yes indeed! So the question is, hasanyone has yet done blind testing (leave aside double-blind) of rips of the same music done with the Melco ripper and, say, dBPoweramp and found them to sound different?
And If they have, and found a consistent difference, then the next obvious step is to see if the files are bit-identical - and that is what triggered this very thread, where someone decided to remove the metadata where variables might occur, and checking whether the files are then bit-identical. At that point, if identical, digital computing tells us they cannot sound different. If they donāt sound different, then the thing to do is go back a step to see what differences in the metadata, or in the way the metadata is stored by the two rippers, causes the sound difference. But if they do sound different and are bit-identical, as was claimed (which needs verifying, then we are into the inexplicable, which needs investigating to understand and find the best solution for different situations ā¦unless you believe in magic!
This is not actually true though. Software is written in blocks which are often walled off from each other. If I download MS Outlook and there is a single binary error then maybe one element wonāt work sometimes but the rest will be fine. As regards banking there is no bit based error that could produce that kind of error.
A sound file is very different. Thereās no walled off element as such and the file will play (despite assertions about error correction preventing this).
There is software available to prove this. Turn your audio file back into code; alter one thing and reconvert. Itāll play just fine.
Ok I have some input on this. A Melco owner on Roons forum has shared a track ripped from an album via the Melco drive and Melco and via his old Unitiserve. I did not know which file was from what ripper. On receiving the files I put them straight into my Roon cores music storage and proceeded to listen to them on my Headphone rig. Roon presents them to me as being the same, same name, same length, same analysis waveform and DR data. On listening to them they do however present themselves differently which shocked me. Itās not huge but its different. At first I thought they must be from a differnet master as one appeared to have more air around it especially the vocals and guitar seemed a little clearer and lighter.
I decided to check the files themselves after this in Audacity. There was an offset between the files with them having silence padded at start and end. So to compare them I would need to line them up accurately. I used the time slip tool in the app to get them accurately lined up each other and compared the waveforms of each one. They looked identical, zero difference. Zoomed in massively no difference. Playing each one with seamless switching in real-time between them as you can easily do this in Audacity without introducing any clicks or stutters and I could not hear a difference at all was like the same track throughout. This was also using the same DAC and headphones as my previous test with Roon. Listening on same computer as audacity using Roon and the same kit I could notice a difference when playing onr then playing the other.
This does not make sense. If the waveforms are the same the underlying data is the same any difference would show up in the waveform. Also why would one app sound the same and one not?
So another test to check if the audio data is identical is to perform a differnce test. So in Audacity I Inverted the phase of one of the tracks and then mixed the two tracks down together. With this principal any audio that is the same would cancel each other out like it will if speakers are out of phase and you would get nothing but silence, if not the same then you would hear the bits of the track that differ and get silence for those that are. Just got silence. They are the same in an audio editing and mixing application.
Somebody else on Roons forum also got the same files and he confirms the files are identical in their music data. He removed all metadata, removed the padding and did a CRC hash check, file size check and what ever else can be done and it all confirmed they where bit identical. Heās not fed back on if he could hear any difference though.
For me itās all puzzling, as the real imperical evidence shows they are the same, Audacity shows me this and listening via it they sound the same when they are played in sync with each other and seamlessly switch between them.
So I feel that the difference I am hearing which is subtle and not always as evident the more you listen to one then the other is the result of the cognitive process of trying to compare sound from memory which is what we are doing. It appears we are flawed in this capacity or at least I am. Using Audacity I hear no change in tone, timing anything because there is none. The difference in the comparison is that this one is in real-time with no delay or memory involved itās a continuous process so Iād it was different you would hear it. Our brains are easily fooled it seems or is there something else in play?
I now plan to compare the files without metadata and with the padding removed and try and see if this also has the same effect when listening in Roon.
1 Like
Brilliant description CG.
Just try listening to the two versions back to back but blind.
Ideally double blind, where you donāt know which is which and thereās not someone else in the room who does know either.
Double blind mean you donāt know what the change is.
Double blind means that any info which can influence either the subject or the experimenter is withheld - so yes, that would include knowing what to look for.
But to do that CG would have to set up and expt with various other people which is not likely to happen.
And even then youād end up asking the subjects whether the 2 tracks sound different.
1 Like
Interesting. One variation on the ear test might be to hear a random sequence of the two recordings (an independent person tosses a coin to choose which to play).
One thing that might not have been considered is the effect of noise and whether there is some memory left in the system from what it previously played. There may be differences in time intervals between the playing of the two recordings. Hysteresis type effects that are temperature dependent for instance.
Phil