Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Testing of Lossless audio formats (Read 5680 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Testing of Lossless audio formats

During the last few days I have been working on a comparison of all the different lossless audio formats out there (AFAIK).

I used the following tracks joining them in one large wav file:
Quote
Beatles   -   Tomorrow Never Knows   -   02:57
Beatles   -   Yellow Submarine   -   02:40
Beatles   -   Sgt. Pepper's Lonely Hearts Club Band   -   02:02
Avril Lavigne   -   Complicated   -   04:06
Bach   -   Prelude And Fugue No. 1 In C Major   -   04:29
Jimi Hendrix   -   Midnight Lightning   -   06:23
Bob Marley   -   You Can't Blame The Youth   -   04:02
Andrew Lloyd Webber   -   Angel of Music   -   02:20
Andrew Lloyd Webber   -   The Phantom of the Opera   -   05:02
Andrew Lloyd Webber   -   The Music of the Night   -   05:41
Supertramp   -   Give a little bit   -   04:09
Supertramp   -   Breakfast in America   -   02:40
Debussy, Claude  -   Douze Etudes pour piano//Premier Livre//I Pour les «cinq do...   -   02:56
Mozart       -   Haffner Serenade -- Menuetto   -   04:59
Harrison   -   my sweet lord   -   04:43

I realize that I could have used track from more genres but since I don't have any techno or whatever I might have used I just used what I had on my HD. Hopefully this will not make a significant difference.

For your pleasure I drew some graphs:
Compression Ratio
Encoding time
Decoding time
[Compression Ratio]/[Decoding time]

These were the settings used:
Code: [Select]
FLAC (1) - 1.1.0:               flac -5 [INPUT]
FLAC (2) - 1.1.0:               flac -8 [INPUT]
LA (1) - 0.4 beta:              la [INPUT] [OUTPUT] 
LA (2)    - 0.4 beta:             la -high [INPUT] [OUTPUT] 
Monkey's Audio (1) - 3.96b5:    mac [INPUT] [OUTPUT] -c2000 
Monkey's Audio (2) - 3.96b5:    mac [INPUT] [OUTPUT] -c5000 
Shorten - 3.5.1:                shorten [INPUT] [OUTPUT] 
LPAC (1) - 1.40/3.08:           lpac -3 [INPUT] [OUTPUT]
LPAC (2) - 1.40/3.08:           lpac -5 [INPUT] [OUTPUT]
WavPack (1) - 3.97:             wavpack [INPUT] [OUTPUT]
WavPack (2) - 3.97:             wavpack -h [INPUT] [OUTPUT]
OptimFROG (1) - 4.5alpha:    ofr45 --encode --speedup 4x --optimize normal [INPUT] --output [OUTPUT]
OptimFROG (2) - 4.5alpha:    ofr45 --encode --speedup 4x --optimize best [INPUT] --output [OUTPUT]
RKAU (1) - 1.07:                rkau -l2 [INPUT] [OUTPUT]
RKAU (2) - 1.07:                rkau -l3 [INPUT] [OUTPUT]
WMA - dBpowerAMP:               WMA lossless

You can get the raw results here: results

I hope it will be of some use.

Testing of Lossless audio formats

Reply #1 –
I'm interested by the results, especially for classical discs (big lack in the very good Speek's test). Unfortunately, I can't read the detailed results : I haven't Office. I think that Office can export a sheet in html format. Can someone convert this for me ?
BTW, I learned than « insane » profile (-c5000) exists in MAC... 

Thank you
Wavpack Hybrid: one encoder for all scenarios
WavPack -c4.5hx6 (44100Hz & 48000Hz) ≈ 390 kbps + correction file
WavPack -c4hx6 (96000Hz) ≈ 768 kbps + correction file
WavPack -h (SACD & DSD) ≈ 2400 kbps at 2.8224 MHz

Testing of Lossless audio formats

Reply #3 –
In addition to the lossless codecs that you tested, there is also Bonk.  Not nearly as popular as many of the others, but it is out there.

Also, could you label the vertical axes on your second and third graphs?

Thanks for posting your test results --  this is interesting stuff.

Testing of Lossless audio formats

Reply #4 –
I'll test bonk as soon as possible.

"label the vertical axes" -- what sence would that make? The speed is relative to the computer of course.

Testing of Lossless audio formats

Reply #5 –
Thank you. I hoped separate results for each discs, as I saw it on Speek's website (removed now - I can't find them...). Lossless is very nice on classical discs : approximately 40% of the raw size. I'm interested by the behaviour of the different codec/settings with 'low-complexity' (correct ?) signals.
For exemple, on a very quiet piano track :

Code: [Select]
Original   = 32253 Kb
MAC -c5000 = 6182 Kb  (19,17 %)
MAC -c3000 = 6336 Kb  (19,64 %)
MAC -c1000 = 6600 Kb  (20.46 %)
Flac -best = 7264 Kb  (22,52 %)
Flac -5    = 7314 Kb  (22,68 %)


MAC 3.96b6
FLAC 1.10

On this track, FLAC -best is >15% bigger than MAC -high (3000).
Wavpack Hybrid: one encoder for all scenarios
WavPack -c4.5hx6 (44100Hz & 48000Hz) ≈ 390 kbps + correction file
WavPack -c4hx6 (96000Hz) ≈ 768 kbps + correction file
WavPack -h (SACD & DSD) ≈ 2400 kbps at 2.8224 MHz

Testing of Lossless audio formats

Reply #6 –
Jan S,

Labelling the time access would give us a sense of proportion (i.e. if the range of the x axis is 50.0 seconds to 51.5 seconds it doesn't matter much, but if it ranges from 50.0 seconds to 400.0 seconds then it matters)...

Testing of Lossless audio formats

Reply #7 –
In the comp_ratio-dec_time.png is the flac decoding slower than la decoding or just x axis is reversed?
--
pozdr.
yq

Testing of Lossless audio formats

Reply #8 –
Quote
Unfortunately, I can't read the detailed results : I haven't Office. I think that Office can export a sheet in html format. Can someone convert this for me ?


Thank you

have you tried OpenOffice?

Testing of Lossless audio formats

Reply #9 –
JanS, what's new, you find something lacking on my website or you just liked to do this for fun? 
As I said already, good album suggestions are still welcome...

Testing of Lossless audio formats

Reply #10 –
Quote
JanS, what's new, you find something lacking on my website or you just liked to do this for fun? 
As I said already, good album suggestions are still welcome...

Well frankly....I had forgot all about your website till I had already put too much work in it :-/ so I figured I could just as well finish it and tell ppl the outcome.
I'd wish you had seperated some of the info into seperate graphs though. Like I did.
And as mrosscook pointet out we both missed bonk.

yq: no. LA is the slowest decoder.

Testing of Lossless audio formats

Reply #11 –
I don't really see expressing the results in another forms would be useful. Speed, compression, and the relation between them, they are all visible in one graph now. I prefer to keep it to the point like that. That's also why I leave all the many numerical values and calculations "behind the graphs" away from the website!

mrosscook, thanks for the link, I couldn't find a binary of Bonk in the past. If all goes well I'll include it with the next update 

Testing of Lossless audio formats

Reply #12 –
Quote
Thank you. I hoped separate results for each discs, as I saw it on Speek's website (removed now - I can't find them...). Lossless is very nice on classical discs : approximately 40% of the raw size. I'm interested by the behaviour of the different codec/settings with 'low-complexity' (correct ?) signals.
For exemple, on a very quiet piano track ... On this track, FLAC -best is >15% bigger than MAC -high (3000).


From my experience developing La I've found that quiet piano tracks etc often show the biggest variation strangely enough. My test set for example (http://lossless-audio.com/comparison.htm) includes one such track - it compresses in fact to around 15% of the original size!

I should also note that La 0.4 improved a lot on this quiet piano track over 0.3 (after I established that Optimfrog compressed this track better than La 0.3 I realised there must be some room for improvement), and for this reason the comparison shows La0.4's default mode having better compression than La 0.3, whereas on more typical music (i.e. a standard rock/pop song) La0.4 default is actually slightly worse. So by including the piano track my results are in one way slightly skewed, while on the other hand they represent a wider range of genres .

Testing of Lossless audio formats

Reply #13 –
Sorry for the basic question but why did you use MAC 3.96b5 for the test instead of the latest ie. 3.97 (8 months old) ?