IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Real Lossless - Efficiency
Defsac
post Jun 23 2005, 10:15
Post #1





Group: Members
Posts: 347
Joined: 17-May 05
Member No.: 22107



Real's page claims that Real Lossless offers:
QUOTE
Flawless reproduction of audio at less than half the size of the original audio file


As we had no efficiency data in the wiki for Real Lossless, I decided to test some sample material to see whether they did indeed offer <50% compression, which would put them up with the best of the lossless formats. The first sample I examined was Tomahawk's 101 North. The source file is 52.8mb. I loaded it up in RealPlayer, and after going through all sorts of arcane procedures to actually convert a file, managed to get to the conversion screen.



Amazingly, both the bit rate and the file size indicate a near exact 50% size reduction. After conversion, I looked at my media file to discover it was indeed 705kbps.



Actually, it isn't. The file is not 26mb as originally indicated, it is 35.2mb.



The library doesn't list file sizes, but clicking the file info option reveals it's true size inside RealPlayer.



The problem is, the bit rate is still listed as 705kbps. For a file of that size the average bit rate should be around 920kbps. It seems to me Real is being deliberately deceptive concerning the efficiency of their format. I'll update this when I have determined the average efficiency of the codec, but I wanted to let people know Real doesn't seem to be being entirely honest with us.

This post has been edited by Defsac: Jun 23 2005, 10:19
Go to the top of the page
+Quote Post
Liisachan
post Jun 23 2005, 10:32
Post #2





Group: Members (Donating)
Posts: 119
Joined: 9-July 04
Member No.: 15225



Interesting...
Go to the top of the page
+Quote Post
johny5
post Jun 23 2005, 10:37
Post #3





Group: Members
Posts: 110
Joined: 27-May 04
Member No.: 14369



After my first experience with Real i never touched it again. Apperently my insticts were right, and Real isnt the sort of program i would like on my computer. Thanks for this info.
Go to the top of the page
+Quote Post
Defsac
post Jun 23 2005, 11:29
Post #4





Group: Members
Posts: 347
Joined: 17-May 05
Member No.: 22107



I have analysed 20 samples and they have an average efficiency of 69.8%. The details can be found here. I'll try to do some more tomorrow, however it had trouble with some samples and continually crashed with a Visual C++ error trying to convert them so right now I've just about had it with RealPlayer.

This post has been edited by Defsac: Jun 23 2005, 11:29
Go to the top of the page
+Quote Post
precisionist
post Jun 23 2005, 11:52
Post #5





Group: Members
Posts: 739
Joined: 16-January 04
From: Germany
Member No.: 11279



Real is always worse than Monkey in its "normal" setting! I'd consider it very inefficient.
It would be nice to include replaygain values in the list, cause compression size is mostly strictly related to loudness. Codecs may compress with different efficiencies at different loudnesses; I noticed that LA compresses much better than Monkey at very low volumes.


--------------------
I know that I know nothing. But how can I then know that ?
Go to the top of the page
+Quote Post
Defsac
post Jun 23 2005, 12:17
Post #6





Group: Members
Posts: 347
Joined: 17-May 05
Member No.: 22107



QUOTE (precisionist @ Jun 23 2005, 08:52 PM)
It would be nice to include replaygain values in the list, cause compression size is mostly strictly related to loudness. Codecs may compress with different efficiencies at different loudnesses; I noticed that LA compresses much better than Monkey at very low volumes.
*

I've updated the list, you might need to refresh it in your browser to see the updated version.

This post has been edited by Defsac: Jun 23 2005, 12:32
Go to the top of the page
+Quote Post
johny5
post Jun 23 2005, 12:28
Post #7





Group: Members
Posts: 110
Joined: 27-May 04
Member No.: 14369



QUOTE (Defsac @ Jun 23 2005, 12:17 PM)
QUOTE (precisionist @ Jun 23 2005, 08:52 PM)
It would be nice to include replaygain values in the list, cause compression size is mostly strictly related to loudness. Codecs may compress with different efficiencies at different loudnesses; I noticed that LA compresses much better than Monkey at very low volumes.
*

I've updated the list, you might need to refresh it in your browser to see the updated version.
*



The average compression isnt even close to 50%. Its more like 70. Its really interesting to see a company like real make a "mistake" like this. Somehow im not supriced that Real is doing something like this. They didnt get a bad rep by doing nothing.

This post has been edited by johny5: Jun 23 2005, 12:33
Go to the top of the page
+Quote Post
Defsac
post Jun 23 2005, 12:32
Post #8





Group: Members
Posts: 347
Joined: 17-May 05
Member No.: 22107



QUOTE (johny5 @ Jun 23 2005, 09:28 PM)
The average compression isnt even close to 50%. Its more like 70.  Its really interesting to see a company like real make a "mistake" like this. Somehow im not supriced its Real doing something like this. They got a rep.
*

The problems I see with the format so far are:

a) It's not clear at the start of the conversion process that the values that are given are estimates. It says "Required: 20MB", which gives the impression the output file will be that size when in reality it is around 20% larger.

b) It's doesn't use any kind of prediction to get these values, it simply halves the bit rate and size. If they're simply going to use a constant for estimates it should be closer to the format's real efficiency which is almost 70%.

c) The bit rate is always detected as 705kbps, even after the conversion process. This is simply half the PCM bit rate. It wouldn't be hard to perform a basic calculation to get the actual average bit rate, this just seems to be an attempt to decieve the user into believing the <50% efficiency claim on their web site.

This post has been edited by Defsac: Jun 23 2005, 12:33
Go to the top of the page
+Quote Post
rjamorim
post Jun 23 2005, 15:37
Post #9


Rarewares admin


Group: Members
Posts: 7515
Joined: 30-September 01
From: Brazil
Member No.: 81



According to my calculations (somebody please shout if I'm wrong), the average compression ratio in RAL, according to Defsac's test, is 69,80%.

I gotta admit these values seem a bit odd to me. They make Real Lossless by far the worst codec out there (compared to Hans Heijden's and Speek's values), even worse than RAR.

Maybe the choice of samples is too un-representative? Or maybe I'm that hopeless with arithmetic :B

Big thanks to Defsac for his test.

This post has been edited by rjamorim: Jun 23 2005, 15:42


--------------------
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org
Go to the top of the page
+Quote Post
Busemann
post Jun 23 2005, 15:43
Post #10





Group: Members
Posts: 730
Joined: 5-January 04
Member No.: 10970



QUOTE (precisionist @ Jun 23 2005, 02:52 AM)
Codecs may compress with different efficiencies at different loudnesses; I noticed that LA compresses much better than Monkey at very low volumes.
*


I've noticed that with ALAC as well. I have a friend who rips albums to wav-> normalize -1dB in audio editor->then convert to alac, and the resulting files are often 50kbps smaller each than if just directly importing.

(He insists that normal replaygaining (or soundcheck in iTunes) only affects the playback volume and does nothing to reduce clipping).

This post has been edited by Busemann: Jun 23 2005, 15:45
Go to the top of the page
+Quote Post
johny5
post Jun 23 2005, 15:54
Post #11





Group: Members
Posts: 110
Joined: 27-May 04
Member No.: 14369



QUOTE (rjamorim @ Jun 23 2005, 03:37 PM)
According to my calculations (somebody please shout if I'm wrong), the average compression ratio in RAL, according to Defsac's test, is 69,80%.

I gotta admit these values seem a bit odd to me. They make Real Lossless by far the worst codec out there (compared to Hans Heijden's and Speek's values), even worse than RAR.

Maybe the choice of samples is too un-representative? Or maybe I'm that hopeless with arithmetic :B

Big thanks to Defsac for his test.
*


I think you are wrong, but it depends how you calculte the average.
I just summed up everything and the input was 824,819 MB and the output 538,510MB
This means the average compressed file is 538,510 / 824,819 = 0,652882632 of the original size so 65%.

EDIT if you sum all the percenteges and divide it by 20 you do get 69.8015, so you did calculate it correctly wink.gif. Its just the method used thats different.

This post has been edited by johny5: Jun 23 2005, 16:11
Go to the top of the page
+Quote Post
rjamorim
post Jun 23 2005, 16:29
Post #12


Rarewares admin


Group: Members
Posts: 7515
Joined: 30-September 01
From: Brazil
Member No.: 81



QUOTE (johny5 @ Jun 23 2005, 11:54 AM)
EDIT if you sum all the percenteges and divide it by 20 you do get 69.8015, so you did calculate it correctly  wink.gif. Its just the method used  thats different.
*


What a mess smile.gif


--------------------
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org
Go to the top of the page
+Quote Post
tgoose
post Jun 23 2005, 19:59
Post #13





Group: Members
Posts: 407
Joined: 12-April 05
Member No.: 21399



QUOTE (johny5 @ Jun 23 2005, 11:54 AM)
EDIT if you sum all the percenteges and divide it by 20 you do get 69.8015, so you did calculate it correctly  wink.gif. Its just the method used  thats different.
*

If I understand correctly, however right the calculation is done, it's not the right calculation to use, unless all files were of equal size. I may be misunderstanding, though.

This post has been edited by tgoose: Jun 23 2005, 20:00
Go to the top of the page
+Quote Post
rjamorim
post Jun 23 2005, 20:21
Post #14


Rarewares admin


Group: Members
Posts: 7515
Joined: 30-September 01
From: Brazil
Member No.: 81



QUOTE (tgoose @ Jun 23 2005, 03:59 PM)
If I understand correctly, however right the calculation is done, it's not the right calculation to use, unless all files were of equal size. I may be misunderstanding, though.
*


No, you are correct. I should have calculated the other way.

Still, 65% is a painfully bad compression ratio. That sounds very weird for a modern codec.


--------------------
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org
Go to the top of the page
+Quote Post
Jojo
post Jun 23 2005, 21:09
Post #15





Group: Members
Posts: 1361
Joined: 25-November 02
Member No.: 3873



QUOTE (rjamorim @ Jun 23 2005, 11:21 AM)
QUOTE (tgoose @ Jun 23 2005, 03:59 PM)
If I understand correctly, however right the calculation is done, it's not the right calculation to use, unless all files were of equal size. I may be misunderstanding, though.
*


No, you are correct. I should have calculated the other way.

Still, 65% is a painfully bad compression ratio. That sounds very weird for a modern codec.
*


65% is not that bad. I have seen albums that average at 69% @ APE extra high. I just did a quick test with 40 random files using FLAC and the average is 63%.


--------------------
--alt-presets are there for a reason! These other switches DO NOT work better than it, trust me on this.
LAME + Joint Stereo doesn't destroy 'Stereo'
Go to the top of the page
+Quote Post
moozooh
post Jun 23 2005, 21:39
Post #16





Group: Members
Posts: 357
Joined: 22-September 04
From: Moscow
Member No.: 17192



It depends on the music and its loudness, anyway.
So, the results obtained are hardly representative for the overall efficiency of RAL.


--------------------
Infrasonic Quartet + Sennheiser HD650 + Microlab Solo 2 mk3. 
Go to the top of the page
+Quote Post
Defsac
post Jun 24 2005, 05:38
Post #17





Group: Members
Posts: 347
Joined: 17-May 05
Member No.: 22107



QUOTE (Mo0zOoH @ Jun 24 2005, 06:39 AM)
So, the results obtained are hardly representative for the overall efficiency of RAL.
*

What is? Unless you're planning to test every peice of music ever released the best you can achieve is an estimation. The reason I have left the efficiency value in the wiki blank is that I don't consider 20 samples statstically significant enough to accurately calculate an average. I hope to test at least seven times this amount.
Go to the top of the page
+Quote Post
rjamorim
post Jun 24 2005, 10:54
Post #18


Rarewares admin


Group: Members
Posts: 7515
Joined: 30-September 01
From: Brazil
Member No.: 81



QUOTE (Defsac @ Jun 24 2005, 01:38 AM)
The reason I have left the efficiency value in the wiki blank is that I don't consider 20 samples statstically significant enough to accurately calculate an average. I hope to test at least seven times this amount.
*


I agree. The ideal would be to test whole albums, like Speek and Heijden did.


--------------------
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org
Go to the top of the page
+Quote Post
xmixahlx
post Jun 24 2005, 18:25
Post #19





Group: Members
Posts: 1394
Joined: 20-December 01
From: seattle
Member No.: 693



QUOTE (rjamorim @ Jun 24 2005, 02:54 AM)
I agree. The ideal would be to test whole albums, like Speek and Heijden did.

who will bribe them to add it to their charts? ...i mean, RKAU is on both charts, and someone out there *could* actually use Real Lossless.


later


--------------------
RareWares/Debian :: http://www.rarewares.org/debian.html
Go to the top of the page
+Quote Post
Polar
post Jun 29 2005, 23:44
Post #20





Group: Members
Posts: 266
Joined: 12-February 04
Member No.: 11970



QUOTE (xmixahlx @ Jun 24 2005, 17:25 UTC)
QUOTE (rjamorim @ Jun 24 2005, 09:54 UTC)
I agree. The ideal would be to test whole albums, like Speek and Heijden did.
who will bribe them to add it to their charts? ...i mean, RKAU is on both charts, and someone out there *could* actually use Real Lossless.
I've tried on a couple of occasions, and not just the ones reported here and here.
Go to the top of the page
+Quote Post
rjamorim
post Jun 29 2005, 23:54
Post #21


Rarewares admin


Group: Members
Posts: 7515
Joined: 30-September 01
From: Brazil
Member No.: 81



QUOTE (xmixahlx @ Jun 24 2005, 02:25 PM)
who will bribe them to add it to their charts?
*


Heijden tried adding it once, but he ran into non-losslessness problems.

Hopefully he'll give it a try again whenever he has the time.


--------------------
Get up-to-date binaries of Lame, AAC, Vorbis and much more at RareWares:
http://www.rarewares.org
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 30th August 2014 - 09:44