comparing almost identical files

So, I’ve been using audacity to compare the quality of several audio files which are vinyl rips with various bitrates.
I’m aware of the procedure to check for identical files (invert one>mix and render), I have used it with success (I mean the result was silence), and other times it just didn’t do it, the rendered version’s spectrogram was like an inflated version of either one of the original files, so I knew the audio files were definitely different.

but recently, I have had another case in which the spectrograms look visually very similar (like 95%+, strikingly more similar than with audio files that yield an inflated spectrogram), the audio is almost indistinguishable, but the bitrates are different and when I do the abovementioned procedure, the result is neither silence nor an inflated spectrogram, but a very diminished spectrogram that plays only a certain type of high frequencies.

I’m no sound engineer, but my guess is both files are linked, they likely originate from the same vinyl, but one of them has known some conversion process and maybe a tiny bit of editing, as in a soft noise reduction procedure.

My question is the following : in this case, which file should I consider the better one ? is this a purely subjective question (it amounts to what I prefer), or is there a way audacity can show me an objectively measured quality difference ?

I’m not sure I could only look at bitrates and decide, since in my understanding, it is totally possible that a modification process would have yielded either a lower or higher bitrate depending on how the person that modified the file configured their audio processing program.
I suspect the higher bitrate file is the modified version since it is constant, contrary to the other file I compare it too (which comes in different bitrates for each track), but as I said, in my understanding, the opposite could be true.
I guess I could also ask the person whom I downloaded the files from, but in this day and age, I fear they might not be the person who modified the original file, so another method is what I’m looking for…
typical example of rendered tracks with diminished spectrogram.png

If I had to guess, I’d say that the bigger file size version would be the better quality one.

A common practice is to make a high quality “archive” version using a format that is capable of very high quality (such as 24-bit WAV), and then make more user friendly versions that are smaller in size (and probably slightly lower sound quality). However, there’s no guarantee that has been done, so you may need to rely on a subjective listening test.

A clue that you may be able to see in the track spectrogram is the upper frequency limit. You will need to zoom out vertically to see this ( In this example (below), the upper track is the original high quality version, and the lower track is a copy that has, at some point in its life, been down-sampled to 22050 Hz, and then resample back up to 44100.
If the files are for your own listening pleasure only, then the file that sounds best is the one that you want.

Thank you very much for your reply.
I have done the vertical zooming and found again very similar spectrograms, that both clearly have a comparable higher frequency limit (19k). All the files that show the reduced spectrogram when mixed and rendered are mp3 though, but that might not be relevant.

However when zooming really close (using the little magnifying glass pointer) I see that the colors of one track are slightly more saturated towards the top and the peaks even seem to get a little higher (not by much though, 0.01k range).

So apparently the more saturated one is indeed also the larger file - I guess if there’s no other method, I’ll simply repeat this procedure and subjectively assess the audio when in doubt…

Thanks again for the quick advice !

Can quickly toggle between the two versions …
then judge by sound alone.

yeah toggling is one of the things I have already been doing.
In the specific case where I get a diminished spectrogram after reverse>mix and render, the audio difference is barely noticeable and has to do (for all I can guess, because it is so close it’s really tough to be certain) with the higher frequencies, i.e. a parasitic noise that occurs in the same spot (which kinda proves the original audio source is the same) will be a slightly muted in a version, but the rest of the track is identical.

with the other comparisons I do, where I get a broader spectrogram after mixing and rendering, toggling generally is way more efficient, indeed.

All the files … are mp3 though

MP3 files are a little magic. Tiny, efficient MP3 files are made by very carefully rearranging musical tones and leaving some of them out. If you compare the data before and after, they will be significantly different, but even expert listening tests will insist before and after are the same.

There is one very important feature of MP3. You can’t make an MP3 from an MP3 without causing damage. It will try to do all of its tricks again and this time the damage will be audible.