Re: Lossless Compression Test (multiple genres + high resolution + multichannel)

Reply #25 – 2024-12-20 09:42:32

@Hakan Abbas : I'll try with pleasure. It'll be a bit difficult for me in the next days but I hope I'll able to it before the end of the year

Re: Lossless Compression Test (multiple genres + high resolution + multichannel)

Reply #26 – 2024-12-20 17:15:27

Thanks for the information guruboolez.

Re: Lossless Compression Test (multiple genres + high resolution + multichannel)

Reply #27 – 2025-01-18 12:27:45

Happy new compressed year.

It is possible to get OptimFROG results out for 96/192 too big files, going by way of raw.
I'd say it is worth it: The frog is in because it is a "high anchor" and not so much because people would do that sort of thing in everyday life.
If I read the spreasheet correctly (that is not a given!), all your 96 and 192 files are 24 bits, and none are mono.

• Gather yourself a folder of the 96 kHz stereo files as flac, and a folder of the 192 kHz stereo files also as flac.
•   flac -d --force-raw-format --endian=little --sign=signed *.flac
  
• OptimFROG command for --preset 0, replace "96000" by "192000" for the other folder:

ofr --encode --preset 0 --md5 --raw --channelconfig STEREO_LR --sampletype SINT24 --rate 96000 *.raw
(I have included the "--md5", which costs 22 bytes it seems.)


As for other codecs:
* Not saying that the following is sufficient to justify the inclusion of MPEG-4 ALS, but:
There is information in encoding by its default and by its -l switch, as the latter turns on the wasted bits feature. If and when -l does  improve, it is an explanation why Monkey's and ALAC don't do better. 
* shorten could have been included as "low anchor" (I actually discovered that shorten has more settings, that might improve compression - but that is for the morbidly curious?)
Suggestion: shorten to the far left as "low anchor", and move the experimental / in development (HALAC for the moment) to the end, as that is what is prone to be updated?
* Again as goes TTA, the bitstreams generated by ffmpeg and reference are the same. Although at this stage you could argue that it is abandonware, I don't see any reason to delete more than one column, at least not if ALS is added to it.

Re: Lossless Compression Test (multiple genres + high resolution + multichannel)

Reply #28 – 2025-01-18 17:44:14

Happy New Year! For encoding in OptimFROG format, I also think that the encoder is indispensable for benchmarking. I could indeed try a few presets again to populate the "high resolution" part, if I can bypass the format bug without too much hassle. I'll try your solution. Finding a solution was not a priority for me last spring, and I admit I wasn't too displeased that the encoder stumbled on these files (it does take a lot of time and energy to encode in this format, much more at high resolution than for Red Book specifications).

Regarding other formats: my comparison work this spring took a lot of time, and I felt it was necessary to do some sorting, otherwise I wouldn't have been able to complete the process. But I'm still keen to explore other formats. Maybe not right away, but the idea of continuing the work started last year appeals to me!

Re: Lossless Compression Test (multiple genres + high resolution + multichannel)

Reply #29 – 2025-01-18 20:27:24

I should add, there is a way to get the frog to accept piping (--incorrectheader, I found out at https://hydrogenaud.io/index.php/topic,122222.msg1009054.html#msg1009054 ). But with my ability to mess up scripts, I would likely have done it the above way if I had drive space enough for both uncompressed and compressed. Looks like you got a mighty 200 GB uncompressed of 192/24 stereo?