Why hasn't AAC (LC or HE) become the standard instead of AC3, DTS for multi-channel?

[JakFrost]

I'm curious why AAC in LC (http://en.wikipedia.org/wiki/Advanced_Audio_Coding) or HE profile (http://en.wikipedia.org/wiki/HE-AAC_v2) hasn't become the defacto popular standard used by people for multi-channel audio yet? (I know AAC is the recommended MPEG standard for future multi-channel audio but that's a technical standard. This is a bit of a rant about audio codecs on my part.)

Everywhere I look I see AC-3 or DTS being used for multi-channel audio and nowhere do I see AAC in LC or HE profile being used. I feel like everyone just keeps the AC-3 tracks in the movies that they encode for themselves with a better video codec and nobody thinks of the audio.

I look at AC-3 multi-channel files with bitrates of 448 Kbps and 640 Kbps and huge DTS files with 1,536 Kbps rates. Those are just HUGE! These two older standards remind me of MP2 audio of the old days and the huge bitrates make me think that they are inefficient.

I'm just getting into H.264 (MPEG-4 AVC) encoding because it is the new 3rd generation standard after DVD MPEG-2. I very happy with MeGUI and x264 encoder since I can shrink down my Blu-rays from 25-40 GB to 8 GB each and keep them on my 1 TB hard drive instead of swapping discs from my folder.

XviD (MPEG-4 ASP) and MP3 are getting long in the tooth these days and I feel the same about AC-3 and more so about DTS. I would have thought that as H.264 overtakes XviD that AAC-HE multi-channel would overtake old MP3 2-channel.

I purchased a whole new Samsung PN50A550 Plasma HDTV (http://www.samsung.com/us/consumer/detail/spec.do?group=televisions&type=televisions&subtype=plasmatv&model_cd=PN50A550S1FXZA), 60-foot HDMI cables to hook-up my computer, and Pioneer VSX-1018AH-K AV receiver (http://www.pioneerelectronics.com/PUSA/Products/HomeEntertainment/AV-Receivers/PioneerReceivers/VSX-1018AH-K) with AC-3, DTS, TrueHD, and Master Audio support and I have a 5.1 surround system. I want to enjoy high-definition video and multi-channel audio movies now and in the future.

I looked at the audio test quality for 64 Kbps (http://www.rjamorim.com/test/64test/results.html) and another for 48 Kbps (http://www.listening-tests.info/mf-48-1/results.htm) files and the results clearly point to AAC-HE as the preferred codec. Also the fairly recent July 2007 test of 64 Kbps (http://www.listening-tests.info/mf-64-1/results.htm) shows good results for AAC-HE.

I even did my own encoding tests with MeGUI to determine what bitrate for ABR and or quality setting I should use for VBR. I've read that VBR is much more preferred for audio encoding so below are some of the files that I ended up with based on different quality settings. I used the Nero codec with profile set to Auto and the quality settings set like below.


403 MiB AC-3 448 Kbps 6 ch 00001 PID 1100 3_2ch 448Kbps DELAY 0ms.ac3

75 MiB AAC-HE 82 Kbps 6 ch Q=0.1.mp4
99 MiB AAC-HE 110 Kbps 6 ch Q=0.15.mp4
148 MiB AAC-HE 164 Kbps 6 ch Q=0.2.mp4
201 MiB AAC-HE 223 Kbps 6 ch Q=0.25.mp4
253 MiB AAC-LC 280 Kbps 6 ch Q=0.3.mp4
305 MiB AAC-LC 337 Kbps 6 ch Q=0.35.mp4
374 MiB AAC-LC 414 Kbps 6 ch Q=0.4.mp4
449 MiB AAC-LC 497 Kbps 6 ch Q=0.45.mp4
517 MiB AAC-LC 576 Kbps 6 ch Q=0.5.mp4


I haven't had a chance to listen through the files yet. However size wise I would hope that if Q=0.2 at 164 Kbps for 156 MB or even Q=0.25 at 223 Kbps for 211 MB would be almost transparent in quality to AC-3 at 448 Kbps then that would be a good trade off for 3x or 2x filesize shrink.

I'm just a little upset that there is no clear popular standard for compressed audio that has yet emerged. DVD for MPEG-2 for was good, XviD for MPEG-4 ASP was better, now x264 for MPEG-4 AVC (H.264) is great. But for audio it was old MP2, LAME for MP3 is good, and now nothing more for multi-channel except for huge AC-3 and DTS files. What the heck!

[squid_80]

Simple: because most multi-channel decoder hardware already out there handles AC3/DTS fine, but can't decode AAC.

[smok3]

dolby digital for example is also a full solution, example;

quote1

Midnight mode allows low-volume listening with high-volume benefits, reducing the volume on just the loud effects of a program, increasing the volume on quiet sounds, and maintaining dialogue at a consistent level. A Dolby Digital feature applies dynamic range compression that preserves low-level sounds, prevents dramatic passages from getting too loud, and keeps dialogue intelligible during lower-level listening.

The amount of compression is not arbitrary, but is decided in advance by the soundtrack's producers and coded right onto the soundtrack.

quote2

Dialogue normalization automatically adjusts the volume when you change Dolby Digital programs so that the level of the dialogue remains constant. It does not alter the dynamic range, only the overall playback level, based on data about the dialogue that is coded into the program signal.

from http://www.dolby.com/consumer/technology/dolby_faq_1.html

so anyway, it is mucho more than just low-bitrate to look for, which makes your rant really look silly...

[tebasuna51]

Simple: because most multi-channel decoder hardware already out there handles AC3/DTS fine, but can't decode AAC.

I agree with that.

@JakFrost
Of course aac is more eficient codec than ac3 or dts, but if we use until 8 GB for a 100 min. HD movie we can spend until 240 MB for audio (320 Kb/s LC) instead 120 MB (160 Kb/s HE).
In this comparative (http://www.ebu.ch/CMSimages/en/tec_doc_t3324-2007_tcm6-53801.pdf) you can see aac 320 Kb/s is near 640 Kb/s ac3 or 1536 Kb/s dts.
Talking in quality terms q 0.35 more or less. Less than 0.3 may be is for streaming pourpose, and not for backup. And more than 0.4 need lossless source to be useful.

@smok3, true but obsolete solutions

quote 1 Dynamic Range Compression

When hardware players was to slow to calculate the DRC values on the fly, have the values pre-calculated (at encode step) in the same stream was a great advantage.

Now is easy apply DRC on the fly (see DirectShow Ac3 Filter for PC) and must be applied not only for ac3/dts but also for lpcm, lossless formats (flac, ...), mp3, aac, ...

quote 2 Dialog Normalization

Also a good idea to try normalize the sound volume. But obsolete because not all audio streams use this method, only ac3.

In ac3 the dialog volume was at -31 dB, very slow compared with TV commercials, CD Audio, ReplayGained mp3, ...

How many questions in these forums begin with: "I have an ac3 with low volume ..."?

[smok3]

"I have an ac3 with low volume ..."?
i hope the answer is, adjust your volume knob?

[tebasuna51]

i hope the answer is, adjust your volume knob?
Now we can use:

eac3to input.ac3 output.ac3

to remove the DialNorm without re-encode.

[smok3]

or use a compliant player? :)
ok, i'am gonna stop now...

[JakFrost]

I agree with that.

@JakFrost
Of course aac is more eficient codec than ac3 or dts, but if we use until 8 GB for a 100 min. HD movie we can spend until 240 MB for audio (320 Kb/s LC) instead 120 MB (160 Kb/s HE).
In this comparative (http://www.ebu.ch/CMSimages/en/tec_doc_t3324-2007_tcm6-53801.pdf) you can see aac 320 Kb/s is near 640 Kb/s ac3 or 1536 Kb/s dts.
Talking in quality terms q 0.35 more or less. Less than 0.3 may be is for streaming pourpose, and not for backup. And more than 0.4 need lossless source to be useful.

Thank you very much for the link to this EBU Evaluations of Multichannel Audio Codecs test results (http://www.ebu.ch/CMSimages/en/tec_doc_t3324-2007_tcm6-53801.pdf), it was very interesting and exactly what I have been looking for. I read and skimmed it for info regarding the codecs and the results. It is very interesting especially the results in quality between DTS 1.5 Mbps, DD+ (AC3) 448 Kbps, and AAC-HE 128, 160, 192 along with AAC-LC 320 Kbps.

On page 13 they list the encoders used and unfortunatelly they used the CT AAC codec. I wish they used the Nero codec since in the other tests that I found online, Nero AAC is rated as the preferred codec. But nonetheless the results are still good, keeping in mind that the Nero AAC codec (http://www.nero.com/eng/nero-aac-codec.html) would bring the results up. Also they did not mention what encoding mode they used but it seem that it wasn't VBR based and it was probably CBR since there is no mention of ABR, but then again I don't know what the default is for CT AAC or what settings they changed.

Looking at the Phase 1 results on Page 21 and seeing the score for original at ~94, DTS 1.5 Mbps at ~93, and DD+ (AC3) 448 Kbps at ~92, is very educational. However, the real interesting results for me are for Phase 2 results on Page 22 showing the original at ~97, and AAC-HE 160 Kbps at ~88, and AAC-LC 320 Kbps at ~89, along with DD (AC3) 448Kbps at ~87.

I'm glad to see AAC-HE holding it's own with AAC-LC using half the bit rate and outdoing AC3 at 2.8x bitrate. (I have always been a secret proponent for new audio encoding technologies so I personally favor AAC-HE because of the geek-factor.)

I wonder if the tests would have not been better for AAC-HE if the Nero AAC encoder would have been used. I heard one person mention that the "new" (don't know what version he was referring to) had some enhancements put in especially for multi-channel efficiency but I don't remember where I read this. I would really appreciate if someone had a reference to this.

But regardless a ~87 result for AAC-HE at 160 Kbps is a very acceptable score in my opinion when compared to the results of the other codecs used and especially comparing their bitrates to AAC-HE.

One thing that is still a little unclear to me is the whole Q=0.35 setting that you mentioned regarding the quality to use for encoding in VBR mode. When I look at the test results and see that there is practically no difference between AAC-HE at 160 and 192 then I look at the Q=0.2 value that I used that produced a 164 Kbps file at 2.7x smaller size, I would think that this is the Q=0.2 setting to use for encoding.

I might be wrong on this but I think that maybe the Q=0.35 setting that you recommend and Q=0.5 that others mention, that I have seen recommended on HydrogenAudio forums is more geared towards stereo 2-channel sound and is not proper for surround 5.1-channel. I remember reading in another thread that the "new" (once again I don't remember the reference) Nero AAC codec improved multi-channel efficiency to let either all the channels, or just the Left & Right channel pairs, to share data to improve the bitrate.

When I use Q=0.2 on the Fifth Element AC3 448 Kbps audio track I get AAC-HE at 164 Kbps in VBR at 152 MB and 2.7x size decrease. When I use Q=0.35 like you recommend I estimate that the file would be AAC-LC at ~345 Kbps and ~340 MB and that is just too close to the AC3 at 413 MB size making re-coding worthless.

That just does not sound right at all when you consider that in the test results on Page 22 AAC-LC at 320 Kbps only got a result 1 point higher than AAC-HE at 160 Kbps.

So now, wouldn't Q=0.2 be the proper setting to get VBR AAC-HE at ~164 Kbps at half the size of 320 Kbps AAC-LC?

On the other hand, I do agree with you and see that when you deal with 8 GB of available space, 413 MB for the original AC3 audio track is not really a big consideration. Also since all the audio decoders out there support AC3 and the software decoders just pass the signal through S/PDIF it makes sense to leave the audio AC3 untouched. I have been doing just that for the Blu-ray movies that I have been encoding to AVC, I just demux the AC3 track and keep the original without recoding.

My original question and rant is probably more focused towards portable 1 CD (700 MB) encodings using AVC (H.264) that still use MP3 2-channel instead of AAC-HE at ~128 Kbps minimum or preferribly at ~160 Kbps, even when using ABR instead of VBR. Since MP3 isn't directly supported that AV hardware just like AAC-HE, then why stick around with MP3 2-channel MP3 audio when you could do 5.1 channel with AAC-HE at the same bitrate? Who do I talk to or hit on the head to make people realize this? Or am I missing something in this picture? Perhaps a soft spot for MP3 audio in people's hearts and unwillingness to move on to newer technology?

Basically, I'm just curious why AVC + AAC-HE hasn't become the standard for smaller encodings, just like Xvid + MP3 did?

PS: I only saw DD+ (AC3) at 448 Kbps and there was no 640 Kbps tested like you mentioned, unless you mean that DD+ 448 is equal to DD 640.

[smok3]

imho AAC-HE just came out a bit late (it will certainly get its big share for web video in the very near future thought), with late i mean/guess:
- hardware support (standalone players)
- hd space is cheap, so why bother
- who on earth is still doing CD versions of the rips?
- for 5.1 you basically need a device/computer with analog outputs
- also for radio, bandwidth is cheap enough and i don't know a lot of radios that would broadcast 6 channels
- i'am sure i could find more

[Blue_MiSfit]

If it's on the PC, AAC can be easily transcoded on the fly to 640kbps AC3 using AC3Filter, but obviously this is a no-go for hardware boxes (at least most of them). Dolby wants to keep making money off licensing, so they have pretty strict policies that allow transcoding of BluRay / HD-DVD audio formats (like DD+ or TrueHD) to AC3 for S/PDIF output, but they certainly won't license anyone to take AAC and go to AC3!

~MiSfit

[shon3i]

@tebasuna51, thanks for PDF, i looked into and seem like MPEG Surround are tested, full surround @ 64kbps give amazing results, i realy can't wait to hear myself.


On page 13 they list the encoders used and unfortunatelly they used the CT AAC codec. I wish they used the Nero codec since in the other tests that I found online, Nero AAC is rated as the preferred codec.Well from my internal tests CT AAC always show more respect (quality) than Nero at same bitrate, in multichannel encoding. It's aslo valid for lastest Nero 1.3.3.0 encoder, the big difference are noticable in surround channels, and LFE, but only on HE-AAC mode. So CT AAC maybe fail by hair in stereo encoding, but more because we have only one ratecontrol (ABR) comparing to Nero's VBR which is great.

probably CBR since there is no mention of ABRCT AAC uses ABR, but everywhere they print that uses CBR.

[JakFrost]

imho AAC-HE just came out a bit late (it will certainly get its big share for web video in the very near future thought), with late i mean/guess:
- hardware support (standalone players)
- hd space is cheap, so why bother
- who on earth is still doing CD versions of the rips?
- for 5.1 you basically need a device/computer with analog outputs
- also for radio, bandwidth is cheap enough and i don't know a lot of radios that would broadcast 6 channels
- i'am sure i could find more

Yes, I feel the same about AAC-HE being a little late out of the gate. AC3 just seems to be supported by all hardware AV receivers and it is becoming more supported on the PC side with AC3Filter and FFDshow codecs.

Hard drive space is cheap and plentiful but I still feel like many others that having a 40 GB Blu-ray movie release that's encoded in MPEG-2 is a waste of space when you can do the same with AVC at 8 GB with full resolution and quality. Even the 22 GB Blu-ray releases in VC-1 and AVC seem a little crazy when a 8 GB version would give you almost the same quality at 3x the space savings. There's always a technological push towards more efficient encodings and bitrates no matter how cheap space is because everyone always finds that no matter how much capacity you buy, you always run out sooner than you think. My 1 TB hard drive that I just bought is already a third full of Blu-ray movies and encodings and I have a feeling that once more stuff that is worthwhile to keep is released I will be running out of space quickly.

Just browsing around I see that CD sized releases are still very popular. It seems that the standard is XviD and MP3 2ch at 112 Kbps, or sometimes 128 Kbps. I'm curious why the AVC (H.264) and AAC-HE 6ch at 128 or 160 Kbps hasn't cought up as a new standard and that's one of the reasons for this thread.

I don't get the 5.1 being analog output thing you mentioned. One of the reasons I got into this whole multi-channel thing and came to learn about the AC3 and AAC-HE along with S/PDIF output is because I was hooking up my onboard soundcard's S/PDIF output to the video card input to get S/PDIF digital audio over the HDMI cable and into my AV receiver in the living room. My other option would be to run S/PDIF coax or optical cable into the AV receiver if I couldn't get it over HDMI but it worked once I replaced my video card with a nVidia GTX 260 model. I did it this way to keep the signal all digital and also to keep it on one cable. So I don't get what you mean by 5.1 analog outputs thing. (But I'm getting off topic here regarding the audio cabling on this thread.)

[JakFrost]

Well from my internal tests CT AAC always show more respect (quality) than Nero at same bitrate, in multichannel encoding. It's aslo valid for lastest Nero 1.3.3.0 encoder, the big difference are noticable in surround channels, and LFE, but only on HE-AAC mode. So CT AAC maybe fail by hair in stereo encoding, but more because we have only one ratecontrol (ABR) comparing to Nero's VBR which is great.

CT AAC uses ABR, but everywhere they print that uses CBR.

Hmm, that's interesting to know because everywhere I've read and seen old codec tests for 2-channel Nero seems to be ahead of CT. I would have thought that the same would hold true for multi-channel but I haven't seen a test yet between CT, Nero, and iTunes for AAC-HE 128 or 160 Kbps for multi-channel. I would really like to see this type of a test. But regardless of that I like AAC-HE no matter what encoder is used overall.

One thing that I'm not so clear about is that with ABR/CBR you get nice predictable bitrate for your encoding. In MeGUI the audio profiles let you select NDAAC-HE-MultiChannel-128Kbps, 160Kbps, or HQ-192Kbps. All of these profiles are ABR to make the bitrate consistent.

I know from the MP3 days and LAME encoder settings for old -V2..., --alt-preset standard, and new --preset standard, that VBR is it and that's what's recommended. In my first post I did my encoding size tests for quality and came up with Q=0.2 to give me 160 Kbps for AAC-HE. But when I tried the same quality setting a few weeks back on other AC3 tracks I get bitrate values that vary widely by many dozens of kilobits. I'm going to do some re-coding again to see how much the Q=0.2 varries.

I'm still groping around the dark here and I'm not quite clear is Q=0.2 is good to use or is a higher quality necessary?

[JakFrost]

I took the Terminator 2 soundtrack and also did a VBR encoding on it and came out with the same bitrates for the Q= settings. Basically Q=0.2 gives me the closest setting of 164 Kbps for both soundtracks so I think that's the one to stick with as per the test results that show AAC-HE doing well at this rate and only a slight 1 point improvement at 192 Kbps that is not worth the bits and the results for 128 Kbps show a lower result on the tests.


Terminator 2 Soundtrack

Size Format Bitrate Channels Filename

625 AC3 640 6 00008 PID 1100 3_2ch 640Kbps DELAY 0ms.ac3

79 AAC-HE 81 6 Q=0.1.mp4
106 AAC-HE 108 6 Q=0.15.mp4
161 AAC-HE 164 6 Q=0.2.mp4
217 AAC-HE 221 6 Q=0.25.mp4
275 AAC-LC 280 6 Q=0.3.mp4
332 AAC-LC 338 6 Q=0.35.mp4

[lchiu7]

If it's on the PC, AAC can be easily transcoded on the fly to 640kbps AC3 using AC3Filter, but obviously this is a no-go for hardware boxes (at least most of them). Dolby wants to keep making money off licensing, so they have pretty strict policies that allow transcoding of BluRay / HD-DVD audio formats (like DD+ or TrueHD) to AC3 for S/PDIF output, but they certainly won't license anyone to take AAC and go to AC3!

~MiSfit

Some broadcasters going to DVB-T for their HD broadcasts are using AAC-HE with LOAS/LATM. That is a real pain to process with there being only one (that I know of) open source DS filter to handle it and no open source filter that can extract the audio from the ts streams. dgavcindex can demux the audio which is a start though but I have yet to find (possibly ffmpeg?) a program that can read this format and output wav or AC3

[Guest]

I have yet to find (possibly ffmpeg?) a program that can read this format and output wav or AC3 WinAmp with its file writer plugin.

[lchiu7]

I knew that but it's not exactly something you can do in batch (well I haven't checked). What I have been doing is demuxing with your program, playing the AAC file in Winamp and outputting wav and then encoding wav to AC3 with besweet. Works but does require manual effort.

[J_Darnley]

Enqueue lots of files, hit play and wait. I've forgotten what the default directory is but once you have done my suggestion, you will have a directory full of wav files. [EDIT] Sorry, you are right in that it it would be hard to use from a batch file because Winamp would not close once it has finished playing the files.

[lchiu7]

I didn't even know winamp could be run from batch. Having it still up at the end is no issue. What I would like to do is have one batch file that demuxes the audio from the ts file to create the AAC file, play Winamp to produce the wav file and then besweet to produce the AC3 version. Seems as if I can do that but just have Winamp still up which isn't a problem. Problem at the moment is getting the batch file to work with the really long names that my PVR software creates and having the paths correct

[Blue_MiSfit]

Why not just use eac3to? It can read audio from a TS and transcode to AC3 on the fly.

Hmm.. on second thought not sure if it supports reading AAC.

~MiSfit

[frogman]

Regarding Topic: cause the real reason is that they don't want to support Open Source! NO licenses fee's, NO $$$ in it for the studio's or Manufactures of AV equipment. Simply economics 101.

[dom_b]

Hard drive space is cheap and plentiful but I still feel like many others that having a 40 GB Blu-ray movie release that's encoded in MPEG-2 is a waste of space when you can do the same with AVC at 8 GB with full resolution and quality. Even the 22 GB Blu-ray releases in VC-1 and AVC seem a little crazy when a 8 GB version would give you almost the same quality at 3x the space savings. There's always a technological push towards more efficient encodings and bitrates no matter how cheap space is because everyone always finds that no matter how much capacity you buy, you always run out sooner than you think.[/I]

Not to do with audio but was thinking the opposite of this point... If I buy a Blu-Ray disc for £20 I don't want them saving space and releasing a single layer disc encoded at a lower bit-rate to save space regardless of codec. I personally would want them to use the full 50GB available on a dual-layer blu-ray disc, mostly for the feature as well. Ive got a lot of Blu-Rays where the feature is around 12GB where they could have at least doubled the bitrate on it. I realise the space issue but if you want to save space for your own rips you should just re-encode it. At the end of the day Blu-Ray releases are not designed for ripping off and storing on your media center... although a lot of us do :)

[Gokumon]

Regarding Topic: cause the real reason is that they don't want to support Open Source! NO licenses fee's, NO $$$ in it for the studio's or Manufactures of AV equipment. Simply economics 101.

What exactly does open source have to do with the adoption of AAC? Your post makes absolutely zero sense. And yes there are plenty of license fees for the people who own patents in the AAC standard just like any other patent pool managed by the MPEG-LA.

[SeeMoreDigital]

Thankfully, the HDMI connection has given high compression multi-channel audio formats such as AAC and WMA Pro a new lease of life!

If your stand-alone player supports AAC audio (within .MP4 or .TS), the multi-channel AAC stream will be converted to a multi-channel PCM stream, which in-turn can be passed via HDMI and decoded by your surround-sound amplifier....