OagMachine Problem

[boombastic]

I tried several times to encode an AC3 into an MP4 with the OagMachine.Thisi is the log file

BeSweet v1.5b23 by DSPguru.
--------------------------
Using azid.dll v1.8 (b825) by Midas (midas@egon.gyaloglo.hu).
Using Shibatch.dll v0.24 by Naoki Shibata & DSPguru (shibatch.sourceforge.net).
Using bsn.dll v0.2 by DPeshev,Richard,DSPguru (http://DSPguru.Doom9.org).

Logging start : 11/20/03 , 16:35:03.

BeSweet.exe -core( -input D:\My Movie\VIDEO_TS\My Movie AC3 T01 3_2ch 448Kbps DELAY 40ms.ac3 -output D:\My Movie\VIDEO_TS\My Movie AC3 T01 3_2ch 448Kbps DELAY 40ms.mp4 -logfilea E:\Programmi\Gordian Knot\BeSweet.log ) -azid( -g max -L -3db -c normal -s stereo ) -ota( -d 40 ) -bsn( -2ch -config ) -profile( The OagMachine v0.1 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : D:\My Movie\VIDEO_TS\My Movie AC3 T01 3_2ch 448Kbps DELAY 40ms.ac3
[00:00:00:000] | Output: D:\My Movie\VIDEO_TS\My Movie AC3 T01 3_2ch 448Kbps DELAY 40ms.mp4
[00:00:00:000] | Floating-Point Process: No
[00:00:00:040] +-------- AZID -------
[00:00:00:040] | Input Channels Mode: 3/2, Bitrate: 448kbps
[00:00:00:040] | Output Stereo mode: Stereo
[00:00:00:040] | Total Gain: 12.319dB, Compression: Normal
[00:00:00:040] | LFE levels: To LR -3.0dB, To LFE 0.0dB
[00:00:00:040] | Center mix level: BSI
[00:00:00:040] | Surround mix level: BSI
[00:00:00:040] | Dialog normalization: No
[00:00:00:040] | Rear channels filtering: No
[00:00:00:040] | Source Sample-Rate: 48.0KHz
[00:00:00:040] +---------------------
[01:52:27:048] Conversion Completed !
[00:22:10:000] <-- Transcoding Duration

Logging ends : 11/20/03 , 16:57:13.

The DLLs are form Nero 6.0.0.23

Then i would like to merge it to an existing .MKV with MKVmerge 0.75 but i always get desynchronization between audio and video.Where is my error?

[boombastic]

I thought it was a problem of wrong delay reported in the Stream Information file so i tried with two other my DVDs but the problem remains.Any suggestion?Thanks

[Tuning]

Hello boombastic and Welcome to Doom9 forum,

I think you have not re-sampled your source to 44.1khz, that may be the problem. Re-sampling to 44.1 Khz is needed if you are using nero dlls(>=60015), so as your's is 60023, I think that is the problem.

There is a problem if your source is 48khz, the header of Mp4 file be written with 44 instead of 48 if you use newer dlls.

For more on that look to BSN=BeSweet+Nero (http://forum.doom9.org/showthread.php?s=&threadid=62672&highlight=BSN)
Its long thread, but the problem is explained there...

[filewalker]

Strange is that if I encode to stereo(surround2 profile)with 48 kHz it's desynchronized.
If you encode to -bsn( -6chnew ) with 48 KHz output it's synchronized
If you encode to -bsn( -6chnew ) with 44 KHz output it's synchronized


So it's safer if we always set the sample rate to 44 KHz for all profiles.

Cu filewalker

[Tuning]

If you are using older nero plugin (60011), then whatever sample rate and channel preset you select the output is not having any kind of problem.

Because of this and the speed attained( because there is no resampling), I'm stick to older dll.

-Tuning-

[SeeMoreDigital]

Doom9 has just posted a link for a revised OagMachine guide.

It's available via his 'News' (http://www.doom9.org/) page (dated 23Nov03) which you may find useful.



Cheers

[filewalker]

stupid question: :rolleyes:
for 2.0 (stereo) encoding I always checked the "LFE to LR Channels" and used -3db if I had a 6ch AC3.

But for 5.1 AAC encoding: is it still necessary to check it?

Thanks in advance!

Cu filewalker

[Hobojobo]

I am just curious.
The problem which turns up with Nero's dlls (>=6.00.11) and BeSweet,
is it a "bug" of BeSweet or Nero?
Downsampling means loss of quality, doesn't it?
Does a upcoming update probably fix this problem?

BTW, OagMachine is a great tools, have I already metioned that?

[bobsc]

Originally posted by filewalker
for 2.0 (stereo) encoding I always checked the "LFE to LR Channels" and used -3db if I had a 6ch AC3.

But for 5.1 AAC encoding: is it still necessary to check it?
No it's not.

[Tuning]

Originally posted by Hobojobo
I am just curious.
The problem which turns up with Nero's dlls (>=6.00.11) and BeSweet,
is it a "bug" of BeSweet or Nero?
Downsampling means loss of quality, doesn't it?
Does a upcoming update probably fix this problem?


Menno (Autor of Nero AAC plugins), once told the fact. Whilst using latest Nero dll (>=60015), the application (Here BeSweet+BSN) must provide 44.1 khz sampled audio to the encoding plug-in.
This implies Nero's plugins are optimized for 44.1 sampled sound and we need to re-sample the 48 khz -> 44.1 before feeding the encoder.

So this is not a bug. If BeSweet, by default re-samples to 44.1Khz, then no one could notice the problem. ;)

-Tuning-

[tiki4]

Originally posted by filewalker
Strange is that if I encode to stereo(surround2 profile)with 48 kHz it's desynchronized.
If you encode to -bsn( -6chnew ) with 48 KHz output it's synchronized
If you encode to -bsn( -6chnew ) with 44 KHz output it's synchronized


So it's safer if we always set the sample rate to 44 KHz for all profiles.

Cu filewalker

That is quite normal behaviour that you can try for yourself. Instead of direct transcoding you may decode the AC3 to WAV / AIFF and try with them what Nero does.

1. case: stereo 48 kHz -> Nero will resample to 44.1 kHz
2. case: stereo 44.1 kHz -> Nero won't change sample rate.

3. case: 6ch AIFF (44.1/48) -> Nero won't chane sample rate.

By the way: It depends on which preset you are using if Nero tries to resample or not. In case 2 the Nero plugin asks the input application for resampling if the rate != 44.1 kHz. BeSweet doesn't handle that request, thus you get asynchronous audio. I'd say it's more safe for now to resample. Otherwise someone with good hearing and the required setup should try to ABX two samples (44.1 vs. 48 at the same bitrate).

tiki4

[Doom9]

I've updated the guide to include the "always downsample" bit for AAC output. I hope that in time, we can relax this restriction, because for me that's a major showstopper.. I hate downsampling. Most devices today (soundcards, DVD players) operate mostly with 48KHz so it makes no sense of not using it (with soundcards, not using 48KHz often requires card internal upsampling because the card can only output 48Khz).

[ChristianHJW]

Originally posted by Tuning If you are using older nero plugin (60011), then whatever sample rate and channel preset you select the output is not having any kind of problem.

Because of this and the speed attained( because there is no resampling), I'm stick to older dll.

You are aware that, even if this DLL works with 48 KHz, it will sound like complete shit because the Nero AAC encoder has not been optimized for 48 Khz yet ?

The same is valid for Vorbis and most other audio encoders, simply because of the fact that all of them are targeted towards CD encoding, and music CDs have 44.1 KHz by definition.

If you ever take the time to talk to a developer of an audio encoder, he will make clear to you how difficult it is to tune a psy model so it fits both sampling rates .... for this reason, most existing encoders will perform much better for 44.1 KHz material, and the same is valid for Nero DLLs .....

[tiki4]

While you are of course right, ChristianHJW, I still think, that the problem lies a little bit different here:

I think certain bitrates are better suited for certain sampling rates. For example LAME starts resampling the audio if your average bitrate is smaller than 128 kBit/s (AFAIK 32 kHz for --alt-preset 96). The same do other audio codecs. I think oggenc and mppenc also have features to resample, maybe not automatically.

The problems is that Nero tries to resample but can't when the input comes from BeSweet (or Nencode). Thus a 44.1 kHz header is written to a 48 kHz file.

Anyway, I think drawbacks from resampling aren't so problematic (apart from the speed issue). Most modern sound cards will resample at playback time to 48 kHz. Whether or not a person can hear the difference pretty much depends on the sound hardware and personal ability.

For my part I can honestly tell that I can't hear the difference (using one of the said problematic Audigy cards).

Regards,

tiki4

[boombastic]

My question now is:LAME downsample when you go under a certain bitrate;is better to downsample also with AAC if you use a very low bitrate?For example if i use 40-50 Kbs should i downsample to 22050 or 38000?

[tiki4]

I could only dig up this (http://www.hydrogenaudio.org/index.php?showtopic=12623&hl=) for the moment. Maybe you want to take a look here (http://forum.doom9.org/showthread.php?s=&threadid=64441&highlight=resampling) also.

[bobsc]

Originally posted by boombastic
My question now is:LAME downsample when you go under a certain bitrate;is better to downsample also with AAC if you use a very low bitrate?For example if i use 40-50 Kbs should i downsample to 22050 or 38000?
It sure would be nice if Menno or Ivan would clear this up.

[Tuning]

Originally posted by ChristianHJW
You are aware that, even if this DLL works with 48 KHz, it will sound like complete shit because the Nero AAC encoder has not been optimized for 48 Khz yet ?

The same is valid for Vorbis and most other audio encoders, simply because of the fact that all of them are targeted towards CD encoding, and music CDs have 44.1 KHz by definition.

If you ever take the time to talk to a developer of an audio encoder, he will make clear to you how difficult it is to tune a psy model so it fits both sampling rates .... for this reason, most existing encoders will perform much better for 44.1 KHz material, and the same is valid for Nero DLLs .....

I'm just new to programming world and really don't know how difficult is making an encoder working in both sample rates. That I admit and respect your stand for delivering information. Thanks.:)

If you notice my other post, I have clearly stated what menno told and you can see, I have wriiten Nero is optimized for 44.1Khz too. The former statements were my experience on using BSN. ( Started testing nero plugins since DSPGuru released BSN v0.1 till 60023)

I have both plugins in Besweet folder in-order to test again at anytime. So what I was telling is older dll did not had problem on encoding in 48Khz/44.1 Khz at what preset you choose ( implies what channel preset -- for replying filewalker's post). Which is from the experience that tests given to me. Now also I can assure that. So its only my opinion and nothing more, as i said I still use older(60011) dlls.

Regarding sampling rate optimization for hearing, I think ( and could not notice ) there will be almost zero " hearable " difference between 44.1khz and 48 khz encoded AAC as already explained by tiki4.:)

...even if this DLL works with 48 KHz, it will sound like complete shit because the Nero AAC encoder has not been optimized for 48 Khz yet ?

Do anyone who used older dll (60011) has similar *shi**y hearing experience ? :D

With later dlls(>=60015) its some shitty experience.:p

To me it was as good as at 44.1khz.;)

Thanks for replying. :)

BTW, I respect every developers work. (Just to make it clear)

[Tuning]

Originally posted by boombastic
..is better to downsample also with AAC if you use a very low bitrate?For example if i use 40-50 Kbs should i downsample to 22050 or 38000?

You don't need to resample if you are using HE-AAC, it will do downsapling to 22050, as it is build up on SBR technology.[http://www.audiocoding.com].

I think Nero encoder automatically alters sampling rates according to your bitrate.
Menno's reply in BSN=BeSweet+Nero (http://forum.doom9.org/showthread.php?s=&threadid=62672&perpage=40&highlight=BSN&pagenumber=2)"If the bitrate gets lowered the samplerate will also get lowered, to give the best quality result."

[DSPguru]

Originally posted by tiki4
The problems is that Nero tries to resample but can't when the input comes from BeSweet (or Nencode). Thus a 44.1 kHz header is written to a 48 kHz file.small correction : 44.1khz aac stream not only differs from 48khz aac stream in headers, eg, you cannot fix this issue by just changing few bits in the stream's header.

[bobsc]

Originally posted by Tuning
I think Nero encoder automatically alters sampling rates according to your bitrate.
Menno's reply in BSN=BeSweet+Nero (http://forum.doom9.org/showthread.php?s=&threadid=62672&perpage=40&highlight=BSN&pagenumber=2)
Your missing the point here, since BeSweet doesn't automatically resample you have to do it yourself.

[Tuning]

bobsc, you are right. I just tested this to confirm your opinion.
Thanks.

BTW, Nero Wave Editor and Burning rom (Encode files) resamples according to bitrates chosen. LC Stereo tests results is given below.
Test file was a LAME mp3 file created from ac3 test file, 100kps and 48Khz.
The latest nero encoder plugin(60023) and nero wave editor was used.

-------------------------------------------------------------
CBR
LC-CBR-16kbps-8Khz
LC-CBR-24kbps-11Khz
LC-CBR-32kbps-16khz
LC-CBR-40kbps-22Khz
LC-CBR-48kbps-22Khz
LC-CBR-56kbps-24Khz
LC-CBR-64kbps-32Khz
LC-CBR-80kbps-32Khz
LC-CBR-96kbps-32Khz
LC-CBR-112kbps-44Khz
LC-CBR-128kbps and above -44Khz ( Never reach 48Khz because of optimization integration towards 44khz )
-------------------------------------------------------------
VBR
LC-VBR-Tape:Lowest - 24Khz
LC-VBR-Radio:Low - 32Khz
LC-VBR-Internet:Medium - 48Khz (I think if 44Khz source was used, the sampling rate of ouput would be 44khz from here onwards)
LC-VBR-Normal:High - 48khz
Above this 48Khz.
--------------------------------------------------------------

In VBR profiles no 44.1 Khz file is produced, which i think either optimization towards 44khz is not working or there is some other reason. (Can any one tell me ? :) )

Hope someone find this useful....:)

-Tuning-

[tiki4]

@DSPGuru: Thanks for correction.

@Tuning: Thanks for performing these tests. I wanted to do them myselves but didn't find time yet. However I think 'internet' and 'streaming' VBR profiles should resample (at least in my experience). I will try to test this with a plain WAV source tonight.

tiki4

[Hobojobo]

@all: thanks for the provided information.

As long as I have to downsample 48kHz audio to use Nero aac, I stick to other codecs/encoder.
What is the point in downsampling, as Doom9 said, and losing quality in order to use a superior audio codec ?

For music in 44.1 KHz Nero aac is a good choice.

For 48 kHz stuff, wait and have a cup of tea....:)

[filewalker]

Hi,

I have another question.

Which gain should I use for AAC encoding?
I read in this audioforum something about Hybrid gain and AAC...now I searched for this thread but I can't find it anymore.

Is the hybridgain tag decoded, if I mux it in MKV and decoded by CoreAAC? or have I use another gaín option in the Oagmachine?

Thanks in advance.


Cu filewalker

[DSPguru]

no directshow filter currently supports aac's hybridgain tag, so you should use it with vorbis/mp2/mp3, but not with aac.

[Tuning]

Thanks DSPGuru for this info.

I think BeSweet has Lwinggain implementation - so considering there will be a future DS filter(May be Nero Digital audio decoder) capable of post gain features, may I use hybrid/post gain in AAC ?

Thanks.

[filewalker]

Thanks DSPGuru for this info,too.!:)

I always used Hybridgain :rolleyes: ...so which setting (or Cl) is recommended to get equivalent results in BeSweet?

Cu filewalker

[DSPguru]

yes, hybridgain is already implemented in BeSweet for mp4 streams, and it will probably be supported in the future by some ds filters, but until that happens, i cannot advise using it.
so just use the alternative "normalize to 100%" feature.


btw, tuning - loved your signature!

[Tuning]

@DSPguru, Thanks as I have got the answer.:D

I have a feature request too. Sorry if this idea is dumb or if already implemented. Similarly like adding pregain, can a equalizer function be added ?. That is increasing signals of different frequency on-the-fly transcoding. (For eg, to increase treble of sound file )

Thanks.:)

[DSPguru]

exists :D !
read doom9's OagMachine guide.. ;)

[Tuning]

Thanks again. Sorry for wasting your time. :stupid: . Just missed the doom9 guide. :thanks:

Infact, doom9's guide pointed to supereq guide, that pointed to shibatch tools page and then to BeSweet CLI page. These all guides were unknown to me. I was looking for the same CLI pages for months.

Thanks again.

[filewalker]

To follow Tuning's signature...:D

THANKS, DSPGuru for the info!:)

Cu filewalker

[Hobojobo]

Nero 6.00.28 is out. (ftp://ftp.nero.com/nero60028.exe)
Including new versions of aac.dll, aacenc32.dll and NeroIPP.dll.
:)
Have not tested it yet. So I do not know whether that brings us something new.

[hubereevez]

Nero 6.00.28 is out.

It seems that i have bigger files with he 50-70.....
Or should I just stop eating bad mushrooms?

bye

hubhub

[mfluder]

Originally posted by Hobojobo
As long as I have to downsample 48kHz audio to use Nero aac, I stick to other codecs/encoder.
What is the point in downsampling, as Doom9 said, and losing quality in order to use a superior audio codec ?

For music in 44.1 KHz Nero aac is a good choice.

For 48 kHz stuff, wait and have a cup of tea....:)
I already posted this but I'm in a good mood so I'll repeat it. You DON'T have to resample as long as you keep 'aac.dll' file from the first Nero release. Then you just use 'aacenc32.dll' from new Nero releases which is actually encoder and it is _the_ file that influences quality. So when you do that you don't have to resample, you will have perfect 48KHz file with all new quality updates. If you do this just remember to use '-6chold' switch.

So people, please stop spreading misinformation by saying that resampling is required. As you can see that's not the case, you just have to pay attention and read more carefully. Also, I remember Menno saying that this will be fixed in December release but it isn't.

mfluder

[mfluder]

Originally posted by hubereevez
It seems that i have bigger files with he 50-70.....
Or should I just stop eating bad mushrooms?
No, there is no need to stop eating bad mushrooms :D New encoder does increase filesize considerably. But I really hope that filesize isn't the only thing that increased.

mfluder

[Tuning]

@hubereevez
It seems that i have bigger files with he 50-70.....
Or should I just stop eating bad mushrooms?

No, only one way left : Abandon eating mushrooms (not only bad) and eat anything else. :p

Back to increased file size scenario : This is because the maximum bitrate now attained in VBR mode is increased considerably. So still if you want smaller size then choose a smaller VBR profile. I think as AAC codec's version increases the max VBR bitrate is getting higher.

@mfluder

I remember Menno saying that this will be fixed in December release but it isn't.
Mfluder, this is not the case. The problem is happening due to incompatibility of BeSweet/BSN to cop up with the bitrate used. i.e, the encoding app has to feed the aac API with appropriate sampling rate. Instead of using BeSweet if you use Nero then that app re-samples 48khz sound to 44.1 and encodes it @ 44. In the same time BeSweet doesn't resample automatically to 44.1 and we get wrong 48khz stream causing all the trouble. So you have to manually resample to 44 in BeSweet.
(note that this 44.1 optimization was brought from later versions (>=60015)
Some explanation is posted by DSP guru and tiki4 above in this thread. Some test results is also posted, (done by me with 60023)

Hope that helps.. :)

[mfluder]

@Tuning
I know all this. I'm using Nero's AAC since the first release of Nero 6. First with AC3 plugin and now with BeSweet. Take a look at that BSN thread and you'll see that you and I actually discussed some things there. I'm also well aware of the resampling problem and why does it happen. The thing is, this happens _always_ no matter what preset you choose. It doesn't depend on preset, at least when BeSweet is used. If what you say is true then when using Transparent preset Nero's dlls shouldn't resample the output. Because I mainly use this preset I don't like the idea of resampling and that is why I use older 'aac.dll'.

And actually you were the one who replied to someone saying that Menno said this will be fixed in next release. Here is a quote:

From Menno's reply,the other day,he had said the next version of AAC encoder will have the 48khz problem solved.
I don't know where you read that as I haven't found it, but I thought I could trust your word ;)

mfluder

[Tuning]

Sorry mfluder, that was a mistake and DSPGuru and tiki4 cleared all in this thread.
The man who cleared my confusion is actually bobsc, he quoted and I tested, so you can see,
"Your missing the point here, since BeSweet doesn't automatically resample you have to do it yourself."
This established that bsn is not doing resampling in my simple tests. So finally I could understand that was the problem.

Sorry for this mistake. :stupid:
Hope you will understand that... :)
Btw, I'm also using older aac.dll
Thanks.

EDIT: Here is Menno's replies to our current topic:
Originally posted by tiki4
"the new aac.dll from 6.0.0.19 has issues with 48 kHz input".

Menno's reply1:
It does not. The application you use has trouble correctly feeding data to the plugin (no correct resampling).
Menno's reply2:
sometimes the AAC configuration can determine that the audio data needs to be resampled. The application using the plugins should handle this by resampling the data.

Bobsc's reply:
Sometimes? I thought you should always resample to 44.1 kHz with current Nero.
Menno's reply1:
Depends on the rest of the configuration.
Menno's reply2:
If the bitrate gets lowered the samplerate will also get lowered, to give the best quality result.
-------------------------------------------
Menno's reply to calinb in He-AAC playback (or encoding) problem,... (http://forum.doom9.org/showthread.php?s=&threadid=62570&highlight=AAC)
quote:44.1kHz vs. 48kHz

44.1kHz is the "recommended" frequency that the plugin uses when encoding. Only problem is that NeroMix doesn't do the resampling, so the plugin gets 48kHz data while it is expecting 44.1kHz. Bug has been reported by me some time ago, and should be fixed in newest versions or next version. In Nero it works fine.
edit: Most apps using the Nero plugin (Nencode, BeSweet, foobar2000) do not handle this resampling properly. (I think it's fixed in foobar now).
-----------------------------------------------------------
Tiki4's question in the same thread:
Now I've got a simple question: What's the difference in handling 48 kHz input in aac.dll from Nero 6.0.0.11 vs. 6.0.0.19? Some problems were reported in that other thread and the question is why we get this behaviour.

If I remember correctly, one could encode a 48 kHz input file into 'streaming' profile and still a 48 kHz AAC file is produced, while the later versions seem to encode at 48 kHz but write a 44.1 kHz header to the file.

Menno's Reply:
The presets have obviously changed in the latest version of the plugin, this is why it now encodes at 44.1kHz, this probably has to do with the fact that the 44.1kHz mode is much better tuned than the 48kHz mode.

The info on NeroMix was that I considered to be the 48fix!(:o hahaha!)
:stupid:

I think now you will get the idea.
Bye.

[hubereevez]

This is because the maximum bitrate now attained in VBR mode is increased considerably. (...) I think as AAC codec's version increases the max VBR bitrate is getting higher.

So nero does not telling the truth when displaying HE-AAC 50-70 ?
Strange
++
hubhub

[tiki4]

@Tuning:

Now we have it all together, thanks. Maybe some mod should add this stuff to the BeSweet FAQ. It gets already difficult to inform everybody over and over again about those things.

@hubereevez:

Nero's aacenc32.dll was considerably tuned for better quality from release to release. You can follow that development on HA.org. However when tuning the VBR profiles for higher quality also the bitrate increased for certain sources. Maybe the bitrate ranges shown in the presets are no longer very good, but I more like Ivan Dimkovic tuning the quality of the encoder than fiddling with the GUI for it. I guess someday when he finds the time he will update that, too.

AAC is about high quality audio and not only on low bitrates IMHO.

tiki4