Quality analysis / Tag editor / Cache

[LordYanaek]

Hi. Recently read a review about Similarity, can't remeber where, and decided to give it a try. I like it
Well, i didn't post just to tell you i like your app. I have a few questions, especially regarding the unusual but highly appreciated quality analysis feature, the built-in tag editor and the "cache".

Quality analysis

• How is clipping calculated? I have some audio tracks with very high clipping value (up to 211119) even thought i ripped them myself from an original CD, using DBPowerAMP CD ripper (often considdered a very good and secure ripper), verified them against AccurateRip, and encoded in flac. All my tracks are scanned using ReplayGain but as they are flac, the normalization is stored in tags and the file is not modified. Is it possible that the clipping detecor might have troubles with replay-gain or are those tracks poor victimes of the loudness war
• I havn't seen anythin about "click" in the analysis article on this site? How many clicks are considered "harmless" I have a track with 1081 clicks, but that one is a downloaded mp3 (perfectly legal as i it comes directly from the band) Actually all of my high clicks tracks are mp3.
• Are there some values that give best results most of the time for the fft settings in spectro and sonogram? I understand the spectros and sonograms, but not being a mathematician, i have no idea how the fft works, and so i can only try different values and see what happens. So far i havn't seen anything different testing different values. One might think that higher values will be better, but that's not always true

Tag editor

• Is there a way to record the analysis values inside tags using the tag editor. I don't see it appear when i'm on the analysis tab.

Cache

• Currently, i must empty the cache if i want to change groups settings. Is it the intended behaviour or a bug? It's annoying cause each time i must restart the scan or the analysis. Is there a way to backup/restore the cache?

Misc : Is it possible to run the analysis only on some tracks, and pass those tracks from another application (using command line, or exporting/importing a txt or csv file or something)? I would like to store data about tracks quality in tags, but then if i ever add new tracks, i would like to be able to simply search for tracks without quality tags using foobar2000 and scan them using similarity, rather than rescann everything.

[Admin]

QA:
Clip finding algorithm very simple, it increase counter if 3 consecutive points of sound data has very small delta difference near maximum. Example abs(SOUND_MAX - abs(x)) <= epsilon for all 3 points.
Clicks is heuristic algorithm for finding vinyl disks clicks, we update articles soon and add such information.
Sonogram - if you set high fft you loose time resolution, but have very detailed frequency resolution, and vice versa. For spectrogram it's more like averaging, high is better (4096). overlap is how amount of data already used in one block of fft, go to next block, 0 -no overlaping, 90% - 9/10 of old data.

TagEditor: I don't understand question ?

Cache: why do you need clear cache ? Similarity must work with diff. folder groups with old cache. Do you have any problems with caching, can you be more specific ?

You can only analyze folder, without scan for duplicates do it via toolbar 'Play' button menu or main menu. And you can easily export all you data, just open desired tab and select File->Export in main menu (CSV available).

[LordYanaek]

QA:
Clip finding algorithm very simple, it increase counter if 3 consecutive points of sound data has very small delta difference near maximum. Example abs(SOUND_MAX - abs(x)) <= epsilon for all 3 points.

And is it expected to get such high values (211119) for tracks i ripped without any sort of destructive DSP, and thus should be as they are on an original CD?

TagEditor: I don't understand question ?

Well, i would like to be able to record the analysis results inside tags. So i would have a "clipping" tag inside my music file and i could see the clipping value without the need to re-scan. Ideally this would be done automatically to avoid manually tagging 10 000 files.

Cache: why do you need clear cache ? Similarity must work with diff. folder groups with old cache. Do you have any problems with caching, can you be more specific ?

Because if i scan without groups to check duplicates in folder "Music", and later want to use groups because i only want to check duplicates in "Music" (group1) and "Newly ripped"(group2) or watever other temporary folder, i can't. The groups column will be emtpy and i can't auto-mark by groups unless i clear the cache.

And you can easily export all you data, just open desired tab and select File->Export in main menu (CSV available).

Yep, my data is already in a LibreOffice spreadsheet. I was more interesting in being able to import a list of tracks to analyse so i don't re-analyse everything. I guess there is no such feature.

[Admin]

1. Please send this file to us via email or some sharing host, we analyze it more detailed, maybe we change algorithm in Similarity in future.
2. As I understand you want to store clipping information in the audio file itself, in some tag (comment example), am i right ?
3. What version do you use ? This is behavior fixed in 1.8.2, yes we had such bug in 1.8.1, group id restored always from cache.
4. Ok, maybe we implement analysis by use list file in next versions.

[LordYanaek]

1. Please send this file to us via email or some sharing host, we analyze it more detailed, maybe we change algorithm in Similarity in future.

I take it that this result is unusual then. I'll make a couple more tests on my own and if i don't find a reason i'll send you the file through support mail to avoid leaving links to copyrighted material in a forum.

2. As I understand you want to store clipping information in the audio file itself, in some tag (comment example), am i right ?

Yes. Possibly other values too. I store all my files as FLAC and the vorbis comment format allows for much more custom tags than mp3. Not sure why i would need analysis values in the future, but in case i need them, it would be easier to just read a tag than scann everything again or search a cvs file.

3. What version do you use ? This is behavior fixed in 1.8.2, yes we had such bug in 1.8.1, group id restored always from cache.

My bad, i didn't notice a new version was released recently. I've updated to 1.8.2 (and lost my registration doing so, i think i'll mail support anyway) Why didn't PayPro send a mail with the registration code as every other online store i've used does

[ashman5]

"write analysis results to tags"

I just posted another thread with this request.

Similiarity would need to create its on tag labels (clipping, mean(abs.), step,...etc) for each value.