Hi folks,

I'm going to add Ahead MPEG-4 AAC test files to SoundExpert rating system, encoded with the following settings:

a1c - MPEG-4 AAC (CBR)
CODER: Ahead MPEG-4 AAC Encoder Plugin for Ahead Nero Burning Rom 5.5.10.0
- Encoder quality - High
- AAC Profile - Low Complexity
- 32 (32.1) kbit/s - 16000 kHz Stereo, Enable PNS - Yes
- 64 (63.7) kbit/s - 32000 kHz Stereo, Enable PNS - No
- 96 (94.8) kbit/s - 32000 kHz Stereo, Enable PNS - No
DECODER: Ahead MPEG-4 AAC Decoder Plugin for Ahead Nero Burning Rom 5.5.10.0

a1v - MPEG-4 AAC (VBR)
CODER: Ahead MPEG-4 AAC Encoder Plugin for Ahead Nero Burning Rom 5.5.10.0
- Encoder quality - High
- AAC Profile - Low Complexity
- Enable PNS - No
- Tape (79.9) kbit/s - 32000 kHz Stereo
- Streaming (104.9) kbit/s - 44100 kHz Stereo
DECODER: Ahead MPEG-4 AAC Decoder Plugin for Ahead Nero Burning Rom 5.5.10.0

Your criticism are wellcome. BTW, does anybody (especialy author) have any information about future improvements of the plug-in? I'm not sure if it is (more or less) final.

Also I would like to add mp3Pro (VBR) to SoundExpert, but first - to listen to your comments:

m3v - mp3PRO (VBR) by Fraunhofer IIS & Coding Technologies
CODER: mp3PRO Encoder Plugin for Ahead Nero Burning Rom 5.5.10.0
- Encoding Quality - Highest
- Allow intensity stereo coding - yes
- Allow downmix - yes
- Padding - ISO
- Low (48.7), Medium (62.5), High (84.9), Highest (104.9) kbit/s - 44100 kHz Stereo
DECODER: THOMSON mp3PRO Audio Player Demo 1.1.0 (Build 0)

Nero mp3 encoder and decoder are buggy and will not decode full file and face some problem in decoding frequency above 16 kHz.

Instead, use mp3PRO encoder in the newest MUSICMATH!



Edit: I'm pro and I'm not sure about this but if you wanna know more about mp3 decoder

http://mp3decoders.mp3-tech.org/decoders.html

Generally a good idea, but...



... please wait until you've read this thread at http://www.audiocoding.com/phorum/read.php...1&i=2219&t=2219 and see what Juan comes up with or if Ivan may want to listen to these samples and change something first. Your list of settings also made me think if the original poster has enabled PNS or another profile than LC perhaps, so I'll have to ask this, too...



Yeah right, no one seems to have tested mp3PRO with variable bitrates yet, so this is a good idea, too.

Well... I need some time to compare them. Thank you anyway.

In general, only thing that should be taken care before testing is that:

1. You have latest nero, 5.5.10.0 or newer

2. Set resampling quality to highest -

... File / ... Preferences / ... Plug-in Lookup

Then, select "Ahead RAW Audio Converter"

Then, switch from "Playback" to "Other Purposes" (combo box)

Set resampling quality to "Sinc" - Quality 100

This will avoid noisy artifacts during resampling.

It would be good if the resampling quality was highest by default...i don't understand why it isn't, compromise quality for speed?

Well.. that's a long story..

Anyway, even Sinc @50 (default for 5.5.10.0) is OK for most of the samples,

But earlier NERO versions had a bug which introduced harsh noise (like PNS - when I heard that for the first time I also thought it was a PNS bug in my encoder). So, download updates often

Thanks for the link. I knew about resample problems in Nero, so I followed Ivan's instructions on setting the highest possible resampling quality in Nero.
The process of codec tuning is almost endless and the question is - when intermediate stops have to be made? It seems to me that it is quite close to its final release now, but I'm not sure still. Perhaps Ivan has the answer....
Juan Avila used the same encoder settings (LC, no-PNS, HQ). PNS slightly improved 32 kbps CBR samples, and I used it in this case.

OK, so then the main problem has been solved already... A few days ago guruboolez also pointed out some problems with Nero AAC in comparison to PsyTEL and uploaded some harpsichord samples to his homepage. I don't know if they are still available and if Ivan has heard them yet, but they suffer from a trembling sound compared to the sharp attacks of the original sample (Bach Prelude), even at very high bitrates. So I guess that resampling has no influence in this case.



That's right, and as you want to use lower bitrates than -normal, the problem mentioned above isn't that important maybe - unless you have a harpsichord in your repertoire.

By the way, I wonder what Nero will do with mp3PRO encodings and a default resample SINC option of 50%, maybe produce similar noise too? Normally mp3PRO won't resample a WAV during encoding, so there shouldn't be any negative influences, but who knows...

For mp3Pro encoding NERO resampling is not used, if I got the code correctly :-) Instead, internal resampler in CT encoder is used, which is good enough.

Yes, one of the nine SoundExpert samples is harpsichord from SQAM disc.

D'oh!

I've just made additional tests with resampling outside Nero. I used nine SoundExpert samples and CEP 1.2, which has very accurate procedure of resampling. The differences are obvious even at 96 kbps. The most critical sample is glockenspiel. I can't say that all previously resampled excerpts sound better (some - yes, some - no), but the fact is they are different.

Tonight I'll upload all nine SoundExpert reference samples to the server (http://www.soundexpert.info/theory.htm, FAQ-3Q) for public use. All who care could investigate more thoroughly what I was talking about. Hope this samples would be helpful in other cases as well.

While the Nero AAC codec is sensible to the way of resampling, it seems logical to test this codec at SoundExpert with Nero resampling mechanism as the most frequent partner of it.

Weird...



You mean the original WAVs or the encoded Nero AAC files done with CEP 1.2? I think these SQAM samples are also available at the MPEG website, aren't they?

Not very strange indeed, couse inaccurate resampling could reduce the amount of information in signal, making it easier to be encoded. A kind of preprocessing...
Original WAVs of course. They are uploaded already.
Yes. some SQAM samples are available there. But not all of SoundExpert samples are from SQAM.

While I was looking at your website today, I noticed one or two things in your suggested MP4 settings:

Is it possible in Nero now to use another resample rate than 32 kHz for 64 kbps CBR? I guess the cutoff is fixed (probably at 10 Khz), so a sample rate of 22 kHz would be enough. If the cutoff is higher (or editable), a sample rate of 24 kHz would be fine, too.



I think you should wait with the VBR samples, until the bug has been fixed that "reads" the CBR setting because of a typo in the source somehow and so affects the average bitrate (and sound) of all VBR settings. (Hi, guruboolez... ) See the Audiocoding forum for more infos.

IIRC it is around 13 kHz at 64 kb/s without PNS, and around 14-14.5 kHz with PNS, so 24 kHz sampling rate won't work.



This bugger was fixed (it surely affected VBR), but I am not sure when the new update will come out exactly, I'll try to make it out a.s.a.p.

Does the new 5.5.10.7 contain updated plugin?

BTW looks like there won't be 5.6 rather 6.0 will be next major release, I guess, (with mpeg-4 video plugin?.. B))

No, 5.5.10.7 does not have updated plug-in, but next version will surely have it

Well, the default setting in PsyTEL with -br 64 is a cutoff at 10 kHz (just checked it again), but of course it can be altered with -c to something else (not like the VBR presets). So does this differ from Nero AAC, i.e. you implemented a fixed and higher cutoff there? As mentioned earlier, I think 64 kbps is the limit where plain AAC (without SBR) loses out too much compared to e.g. mp3PRO or WMA9, so a higher cutoff than 10 kHz would probably make things worse than before, not better. But maybe this is not the case with Nero AAC anymore...



If it's only possible to download the whole Nero package, people would probably prefer to wait a little longer and get an update with more bugfixes rather than a "quickie".

Thanks Ivan and Hans. I confirm the bug too and I’m waiting for the fix. Hope Ivan informs us all as soon as it comes out.

BTW, I could mention similar bug in Nero mp3PRO plugin (being very sorry for the offtopic). SoundExpert sample #8 (Mike Oldfield, "Music From The Balcony"), encoded with the following settings:

CODER: mp3PRO Encoder Plugin for Ahead Nero Burning Rom 5.5.10.0
- Encoding Quality - Highest
- Allow intensity stereo coding - yes
- Allow downmix - yes
- Padding - ISO
- Highest (104.9) kbit/s - 44100 kHz Stereo
- CRC Check - OFF

has a very audible artifact and the decoded stream has a lack of 769 samples. Switching “CRC Check” ON fixes the problem. Cool Edit PRO 2.0 encodes this excerpt correctly both with “CRC Check” ON and OFF. I think this is definitely a bug of Nero mp3PRO plugin. So I would like to ask Ivan (if he considers this important) to inform Ahead stuff who is in concern. The sound excerpt mentioned above (the reference one) could be downloaded here - http://www.soundexpert.info/refsampl/moldfield.rar

And finally, as far as I understand Nero AAC (CBR) operates correctly and the proposed settings are suitable for SoundExpert testing. If so, I’m ready to add the codec to the system tonight.

If you can't change the resample rate or cutoff in Nero for 64 kbps CBR manually, youve got no other choice than accepting the 32 kHz, even if the cutoff is lower than 13 or 14 kHz.

By the way, when the VBR bug is fixed, you could also try out -radio for your samples, because it's a preset where PsyTEL AAC can compete with other codecs in my opinion (~80 kbps). And of course -internet (I hope Nero AAC also resamples at 32 kHz with this setting?), because it should sound better than a CBR 96 kbps setting and averages at ~100 kbps normally. That low average bitrate you got for -streaming is probably caused by the bug, as this preset averages at 128 kbps with PsyTEL and so would be too big for your purposes.

nero55107b.exe came out. Does it contain AAC(VBR) bug fix?

Current version: 5.5.10.7b
Release date: February 03st, 2003



No the current release still has VBR bug.

I'm sorry, but I don't understand this. What CBR setting? Do you mean the quality switch of VBR? What exactly doesn't it do right? I wanted to start encoding using the Nero plug-in with VBR, but this stopped me

No, the chosen value for the constant bitrate influences the overall/average bitrate of the VBR presets, which should not happen of course. You can test this for yourself, if you want, or just believe the developer Ivan Dimkovic.

Ah, yes, now I see. I've just tested it. So, what would the correct CBR setting be for VBR encoding (something like a workaround), because the bitrates from the mp4 files change a lot depending on the CBR bitrate

EDIT: btw, with CBR bitrate set to 16, in_mp4 plug-in for Winamp2 reports "Unsupported file type"

There's no correct answer to this right now, I guess. You could start with 128 kbps for the CBR setting, but a better idea would be to wait for the update, because it probably will be out in a few days (especially if you want to rip your whole CD collection or something like that).



Hmm, hopefully Ivan notices this, too...