IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Bang for the byte and quality of HE-AAC / LC vs MP3, Quality of HE-AAC / LC 44.1 KHz / 22.05 KHz relative to MP3
califauna
post Aug 2 2013, 13:01
Post #1





Group: Members
Posts: 3
Joined: 2-August 13
Member No.: 109425



Hi all,

I have some audio in the following format, according to MediaINfo:

Format : AAC
Format/Info : Advanced Audio Codec
Format profile : HE-AAC / LC
Codec ID : 10
Duration : 3mn 3s
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 44.1 KHz / 22.05 KHz
Compression mode : Lossy

In terms of bitrates what would be the equivalent bitrate in an MP3 file to match the audio quality of that format (assuming all other things equal, such as the best encoders etc) ?

Also wondering how that format compares in file size with MP3. Its 3 minutes 3 seconds and 1.12 Mb so again, knowing the relative quality of the audio in that format Ill know the bang for the byte compared with MP3.

Cheers.
Go to the top of the page
+Quote Post
pdq
post Aug 2 2013, 13:21
Post #2





Group: Members
Posts: 3442
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



If my math is right, that works out to about 49 kbps. An mp3 would have to be significantly higher bitrate to give equivalent quality. Of course, if you are transcoding to mp3 then to even maintain nearly the same quality would require a fairly high bitrate.
Go to the top of the page
+Quote Post
califauna
post Aug 2 2013, 15:34
Post #3





Group: Members
Posts: 3
Joined: 2-August 13
Member No.: 109425



QUOTE (pdq @ Aug 2 2013, 14:21) *
If my math is right, that works out to about 49 kbps. An mp3 would have to be significantly higher bitrate to give equivalent quality. Of course, if you are transcoding to mp3 then to even maintain nearly the same quality would require a fairly high bitrate.



Would you hazard a guess at the bitrate required? Genrally Id like to know the quality comparison between the formats.
Go to the top of the page
+Quote Post
IgorC
post Aug 2 2013, 16:26
Post #4





Group: Members
Posts: 1578
Joined: 3-January 05
From: ARG/RUS
Member No.: 18803



Rough estimation (considering high quality AAC and MP3 encoders).

HE-AAC 48 kbps ~ LC-AAC 64 kbps ~ MP3 96 kbps
Go to the top of the page
+Quote Post
[JAZ]
post Aug 2 2013, 16:35
Post #5





Group: Members
Posts: 1796
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



@califauna, that does not work that way.
You are asking for a linear quality scale, when lossy codecs use different tools that produce different types of quality degradation. (I.e. it is not the same to degrade the stereo image, to remove the highs, or to "simulate" the highs, to name just a few).

As such, it is possible to answer your question, but only with the premise that different people will preffer different degradations.

With that premise, HE-AAC is an extension of LC-AAC which roughly requires half the bitrate. So 64kbps HE-AAC (v1) would roughly be similar to 128kbps LC-AAC, and 48kbps HE-AAC (v1) would roughly be silimar to LC-AAC 96kbps which is roughly similar to 128kbps MP3.

A direct comparison cannot be done, because they produce identificable differences, especially when compared to the source file.



BTW: you should also learn that bitrate = bits per second = number of bits required for one second of content = length of file in bits/length of sound in seconds (1byte=8bits).

That's why pqd has been able to guess the bitrate without you telling about it.
Go to the top of the page
+Quote Post
jensend
post Aug 2 2013, 18:03
Post #6





Group: Members
Posts: 149
Joined: 21-May 05
Member No.: 22191



QUOTE ([JAZ] @ Aug 2 2013, 09:35) *
@califauna, that does not work that way.
You are asking for a linear quality scale, when lossy codecs use different tools that produce different types of quality degradation.
Listening tests are practically always organized around rating samples on single linearly-ordered quality scales. Sure, ratings will vary not only among listeners but among different trials with the same listener, but when he asks for something to "match the audio quality" he's not asking for it to generate identical output or to generate the exact same ratings in every trial from every listener. We can determine with reasonable accuracy where a typical listener will (in repeated trials) rate different encoders/settings, and it's clear something like this is what he's after. Jumping to "correct" him just because he didn't formulate his question in whatever complicated manner would meet with your approval doesn't help anyone.
QUOTE ([JAZ] @ Aug 2 2013, 09:35) *
With that premise, HE-AAC is an extension of LC-AAC which roughly requires half the bitrate. So 64kbps HE-AAC (v1) would roughly be similar to 128kbps LC-AAC, and 48kbps HE-AAC (v1) would roughly be silimar to LC-AAC 96kbps which is roughly similar to 128kbps MP3.
This doesn't agree with listening test results at all. Normally, 48kbps HE-AAC will not hold a candle to 96kbps LC. IgorC's claims (48kbps HE ~ 64kbps LC ~ 96kbps MP3) are much more realistic.
Go to the top of the page
+Quote Post
[JAZ]
post Aug 2 2013, 19:29
Post #7





Group: Members
Posts: 1796
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



@jensend: My way to express it wasn't the best either, as it seems. I just wanted the OP to understand that there is no strict answer to that question.

About the values... Yes, probably my assumption is wrong, now that I tested Winamp's latest fraunhofer codec with the values IgorC mentioned.
Maybe i was more used to HE-AAC v2 (which obviously, has more bandwidth free for the LC part due to parametric stereo), or really LC-AAC has come a long way in low bitrates, because the LC one sounds fine, while the HE-AAC has clearly a problem.

I've uploaded a sample here http://www.hydrogenaudio.org/forums/index....howtopic=102059 , just in case it is a bug in the HE-AAC code, even though it could simply be the expected behaviour.

Concretely, the problem is some warbling that happens after the cymbal, which is clearly visible in the spectogram: the white zones at the 10KHz band. Mmm. well, i'm not sure now if it's that part, or the empty zones between 4 and 8Khz, or simply something that should be looked with other zoom levels.. but it's easy to hear.

This post has been edited by [JAZ]: Aug 2 2013, 19:36
Go to the top of the page
+Quote Post
jensend
post Aug 2 2013, 20:02
Post #8





Group: Members
Posts: 149
Joined: 21-May 05
Member No.: 22191



QUOTE ([JAZ] @ Aug 2 2013, 12:29) *
@jensend: My way to express it wasn't the best either, as it seems. I just wanted the OP to understand that there is no strict answer to that question.
Makes sense, no worries.
QUOTE ([JAZ] @ Aug 2 2013, 12:29) *
About the values... Yes, probably my assumption is wrong, now that I tested Winamp's latest fraunhofer codec with the values IgorC mentioned.
Maybe i was more used to HE-AAC v2, or really LC-AAC has come a long way in low bitrates, because the LC one sounds fine, while the HE-AAC has clearly a problem.
It's certainly true LC encoders have made some progress, though I don't know how the improvement histories of HE and LC compare. But 48kbps HE wouldn't have matched 96kbps LC even at its introduction ten years ago either. This 2003 paper shows 48kbps HE-AAC barely edging out 64kbps LC.

HE-AAC v2 only adds parametric stereo, which is mostly intended for bitrates lower than 40kbps. In HA's 2006 tests at 48kbps HE-AAC v2 actually "lost" to V1 (lower average but not enough ratings to make it statistically significant). I imagine an up-to-date v2-capable HE-AAC encoder will just avoid parametric stereo at this bitrate unless you force its use. Others (esp. C.R.Helmrich) would know a good deal more about this.

That 2006 test does provide a good reality check for your "HE-AAC @48kbps ~ MP3 @128kbps" claim as well- though the AAC encoders have made more progress in the last 7 years than LAME has, it's not by anywhere near enough to make up that kind of margin.

This post has been edited by jensend: Aug 2 2013, 20:05
Go to the top of the page
+Quote Post
C.R.Helmrich
post Aug 2 2013, 20:23
Post #9





Group: Developer
Posts: 690
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



I agree with Igor's rule of thumb. Of course, for some particular items like the one just uploaded, HE@48kbps might sound worse, though.

QUOTE (jensend @ Aug 2 2013, 21:02) *
I imagine an up-to-date v2-capable HE-AAC encoder will just avoid parametric stereo at this bitrate unless you force its use. Others would know a good deal more about this.

Don't know about other providers, but Fraunhofer's encoder does what you tell it to, meaning it will do HE v2 up to 48 kbps or even more. But if you use the slider in Winamp in Automatic mode it will switch to v2 only below 40 kbps, I think.

And yes, the Fraunhofer encoder certainly sounds better at low-bit-rate LC than a few years ago. I recall the quality of the iTunes LC@48-kbps anchor in the 2011 HA public test. 48-kbps LC don't have to sound that bad.

Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
[JAZ]
post Aug 2 2013, 20:56
Post #10





Group: Members
Posts: 1796
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



I've checked now with Nero's encoder, and while the result is a bit worse in both files than fraunhofer's, HE at 48 is still worse than LE 64.
I really don't know why i had a different idea, really.
Go to the top of the page
+Quote Post
califauna
post Aug 3 2013, 13:24
Post #11





Group: Members
Posts: 3
Joined: 2-August 13
Member No.: 109425



Thanks for the rules of thumb above on the enigmatic 'audio quality' factor.

I realize the listening tests they are based on are often subjective but for me they can still help when someone is only familiar with one format, like me with MP3. For example I know the approximate bitrate when cymbals begin to lose fizz , and (all other things being equal) the differentiation threshold between recordings in mbps, and how 'Listener X' uses all these and various other factors to arrive at ratings of this 'audio quality' for MP3 recordings at different bitrates. However, I have little idea what bitrate (in practical terms for me, file size) another format would need to arrive at this same 'audio quality' .

The decision on the format to choose is for me principally a tradeoff between size and quality. So long as the rules of thumb give a fair representation of 'listener X', they help me with this decision.

This post has been edited by califauna: Aug 3 2013, 13:25
Go to the top of the page
+Quote Post
Mach-X
post Apr 5 2014, 01:43
Post #12





Group: Members
Posts: 288
Joined: 29-July 12
From: Windsor, On, Ca
Member No.: 101859



The best subjective answer is to simply encode a couple of files and listen for yourself. Also for simplicity sake I use winamps built in encoder for aac encoding. For portable listening, he-aac v2, which is the lowest setting in winamp produced acceptable sound quality, but far too many imaging errors. The next setting up which is he-aac v1 provides perfectly listenable results for portable use, and really small file sizes, probably on par with lame at 128. This does not imply transparency, as that can only be determined through double blind testing. Since I have poor short term memory, artifacts I may hear during abx sessions I don't hear in casual listening using he-aac v1 so I no longer use abx to determine my portable encoding, I just listen for obvious artifacts.
Go to the top of the page
+Quote Post
Hex144
post Apr 7 2014, 08:57
Post #13





Group: Members
Posts: 33
Joined: 13-February 13
Member No.: 106615



QUOTE (IgorC @ Aug 2 2013, 18:26) *
Rough estimation (considering high quality AAC and MP3 encoders).
HE-AAC 48 kbps ~ LC-AAC 64 kbps ~ MP3 96 kbps

Doesn't this mean that Fh AAC's vbr modes 2 and 3 are about the same quality? (~64 SBR vs. ~96 LC)
Maybe the main difference between vbr3 and vbr2 is just trading file size for encoding & decoding speed / playback time?

This post has been edited by Hex144: Apr 7 2014, 09:06
Go to the top of the page
+Quote Post
nu774
post Apr 7 2014, 09:26
Post #14





Group: Developer
Posts: 562
Joined: 22-November 10
From: Japan
Member No.: 85902



QUOTE (Hex144 @ Apr 7 2014, 16:57) *
QUOTE (IgorC @ Aug 2 2013, 18:26) *
Rough estimation (considering high quality AAC and MP3 encoders).
HE-AAC 48 kbps ~ LC-AAC 64 kbps ~ MP3 96 kbps

Doesn't this mean that Fh AAC's vbr modes 2 and 3 are about the same quality? (~64 SBR vs. ~96 LC)
Maybe the main difference between vbr3 and vbr2 is just trading file size for encoding & decoding speed / playback time?

I don't see why did you take it as HE-AAC 64kbps ~ LC-AAC 96kbps.
Even if you have assumed that the ratio above remains the same for more higher bitrate (which is a very questionable assumption), still 48:64 (=3:4) is not equal to 64:96 (=2:3).
(edit: wording)

This post has been edited by nu774: Apr 7 2014, 09:35
Go to the top of the page
+Quote Post
Hex144
post Apr 7 2014, 10:56
Post #15





Group: Members
Posts: 33
Joined: 13-February 13
Member No.: 106615



this:

QUOTE (jensend @ Aug 2 2013, 22:02) *
This 2003 paper shows 48kbps HE-AAC barely edging out 64kbps LC.

But maybe, like you said, the difference doesn't scale with bitrate, so that's why I was asking.


Just found this (from 2009). LC 80 was actually rated higher than SBR 80. I guess this answers it.

This post has been edited by Hex144: Apr 7 2014, 11:06
Go to the top of the page
+Quote Post
saratoga
post Apr 7 2014, 13:27
Post #16





Group: Members
Posts: 5116
Joined: 2-September 02
Member No.: 3264



SBR becomes a lot less useful above 64kbs, so it definitely does not scale with bitrate.

This post has been edited by saratoga: Apr 7 2014, 13:28
Go to the top of the page
+Quote Post
IgorC
post Apr 7 2014, 15:03
Post #17





Group: Members
Posts: 1578
Joined: 3-January 05
From: ARG/RUS
Member No.: 18803



Some time ago I've experimented with the results of some blind tests (public, personal, verification etc.)
This is a preliminary version of the graph (music samples) . There might some variations due to different encoders but it should reflect performance of formats good enough.



Go to the top of the page
+Quote Post
saratoga
post Apr 7 2014, 15:40
Post #18





Group: Members
Posts: 5116
Joined: 2-September 02
Member No.: 3264



Just curious what encoders and versions were used for opus and USAC?
Go to the top of the page
+Quote Post
IgorC
post Apr 7 2014, 15:46
Post #19





Group: Members
Posts: 1578
Joined: 3-January 05
From: ARG/RUS
Member No.: 18803



USAC RM10. Opus 1.1
Go to the top of the page
+Quote Post
jensend
post Apr 7 2014, 16:34
Post #20





Group: Members
Posts: 149
Joined: 21-May 05
Member No.: 22191



Interesting. Two things one notices right off the bat about "received wisdom": "HE@48~=LC@64~=MP3@96" matches your graph pretty closely, and "USAC @X kbps ~= best of LC/HE/HEv2 at X+8 kbps" matches it very closely from 24kbps up to 96kbps.
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: 24th November 2014 - 08:08