Skip to main content
Topic: Pre-Test thread (Read 44391 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Pre-Test thread

Reply #125
I would suggest to use the nero profile Portable 50-70kbps HE-AAC HQ instead of Streaming for the test. Average Bitrate is always arround 64kbps, mostly even lower, and the quality is surely better than 64kbps cbr (at least for my ears)

Pre-Test thread

Reply #126
Quote
I would suggest to use the nero profile Portable 50-70kbps HE-AAC HQ instead of Streaming for the test.

Dude, where the heck did you see a "Portable" profile in Nero? 

Pre-Test thread

Reply #127
Argh, now i got it! Preset "VBR/Stereo - Portable, 50-70 Kb/s (HE-AAC) is identical with the standard option (in this case "streaming"). Sorry, my mistake^^;

Pre-Test thread

Reply #128
Oh, well. The developer I talked about declined participating in this test. The codec suite and settings will be the ones I mentioned sooner. Test starts next wednesday (09-10).

Looking forward to your participation.

Roberto.

Pre-Test thread

Reply #129
OK, the sample packages are ready, except for Vorbis, that, hopefully, will be released until tomorrow night (I can't wait more than that)

Unfortunately, I can't postpone this test anymore. It would mean the test would start at 09-17 and end at 09-28, and I'm going to travel on that sunday.

So, let's hope Xiph releases Vorbis 1.0.1 soon.

On another issue, could someone please clarify what is the overhead introduced by the Ogg container? I'm trying to find out the container overhead for the formats I'm featuring and so far I came up with this:

-MP3: Containerless
-Real Audio: Does it even use a container?
-WMA: ~6Kbps
-MP4: Negative (!), since the ADTS headers are ripped prior to multiplexing
-Ogg: ?

Thanks for any info;

Roberto.

Pre-Test thread

Reply #130
[spam removed by moderation]

Pre-Test thread

Reply #131
@danielperez

Do me and the rest of the community a favor by NEVER EVER POSTING HERE AGAIN.  Thanx.       
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Pre-Test thread

Reply #132
Jesus Jumping Christ 

Pre-Test thread

Reply #133
PLEASE tell me that my eyes are playing tricks on me, and that a few posts before mine is NOT an 80-page pyramid scheme?  My, how times have changed at HA.

Pre-Test thread

Reply #134
Quote
Jesus Jumping Christ 

Indeed.
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Pre-Test thread

Reply #135
why, oh, why in MY test thread? >_<

Now people won't notice the info request I posted at my last post (just before the 80-page spam)

Pre-Test thread

Reply #136
Back, not quite on but slightly closer to the topic for a secound.  How could the MP4 container size be negative Roberto?  I can see it being negligibly small, but how could it be possible for it to shrink the filesize?
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Pre-Test thread

Reply #137
Quote
Back, not quite on but slightly closer to the topic for a secound.  How could the MP4 container size be negative Roberto?  I can see it being negligibly small, but how could it be possible for it to shrink the filesize?

Because it strips all the headers from the AAC files prior to multiplexing them.

There's a header (either ADIF or ADTS) in every AAC stream with information about sampling rate, bit depth, number of channels, AAC profile, etc.

It removes all these headers (making the file a RAW AAC), multiplexes it in the MP4 container and stores the stream information in the MP4 header. That's why it's slightly smaller.

Edit: Do the test for yourself: Encode a track with Psytel AACenc or FAAC, wrap it in MP4 using MP4ui or MP4creator, and compare file sizes.

Regards;

Roberto.

Pre-Test thread

Reply #138
Quote
PLEASE tell me that my eyes are playing tricks on me, and that a few posts before mine is NOT an 80-page pyramid scheme?   My, how times have changed at HA.

It's not even a well-written pyramid scheme 

Quote
PLEASE MAKE SURE you write every name & address ACURETLY. This is critical to YOUR success.


This is just sad 

Pre-Test thread

Reply #139
*yawn* spam deleted, account banned, have a nice and productive day.

Pre-Test thread

Reply #140
Quote
Because it strips all the headers from the AAC files prior to multiplexing them.

There's a header (either ADIF or ADTS) in every AAC stream with information about sampling rate, bit depth, number of channels, AAC profile, etc.

It removes all these headers (making the file a RAW AAC), multiplexes it in the MP4 container and stores the stream information in the MP4 header. That's why it's slightly smaller.

Edit: Do the test for yourself: Encode a track with Psytel AACenc or FAAC, wrap it in MP4 using MP4ui or MP4creator, and compare file sizes.

Hmm - someone should implement converting of ADIF AAC files to MP4 in MP4Creator (if this is not done already)

Encoding with ADIF header (psytel, faac) won't generate frame-headers and still there will be enough information to build the MP4 file, and therefore there would be no undercoding/bit reservoir issues.

Pre-Test thread

Reply #141
Quote
OK, the sample packages are ready, except for Vorbis, that, hopefully, will be released until tomorrow night (I can't wait more than that)

Unfortunately, I can't postpone this test anymore. It would mean the test would start at 09-17 and end at 09-28, and I'm going to travel on that sunday.

So, let's hope Xiph releases Vorbis 1.0.1 soon.

On another issue, could someone please clarify what is the overhead introduced by the Ogg container? I'm trying to find out the container overhead for the formats I'm featuring and so far I came up with this:

-MP3: Containerless
-Real Audio: Does it even use a container?
-WMA: ~6Kbps
-MP4: Negative (!), since the ADTS headers are ripped prior to multiplexing
-Ogg: ?

Thanks for any info;

Roberto.

If you have a recent CVS snapshot (i.e. from after September 2), that is 1.0.1 as far as encoding quality is concerned. AFAIK, the 1.0.1 release is currently only being held back due to Win32 build problems.

Regarding the Ogg container overhead, http://www.xiph.org/ogg/vorbis/doc/framing.html says that the overhead shouldn't be more than 2% of the bandwidth. If this is correct (but I don't know if it is), that would give you 1.3kbps overhead on a 64kbps stream.

Hope this helps,

Carsten Haese
Ogg Traffic Editor, Xiph.org Foundation

Pre-Test thread

Reply #142
John33 shall build new Vorbis binaries then...
;)
"To understand me, you'll have to swallow a world." Or maybe your words.

Pre-Test thread

Reply #143
Quote
Hmm - someone should implement converting of ADIF AAC files to MP4 in MP4Creator (if this is not done already)

Encoding with ADIF header (psytel, faac) won't generate frame-headers and still there will be enough information to build the MP4 file, and therefore there would be no undercoding/bit reservoir issues.

Well, one can always use ADIF2MP4 in that case... (given you know where to find it, of course  )

Quote
If you have a recent CVS snapshot (i.e. from after September 2), that is 1.0.1 as far as encoding quality is concerned. AFAIK, the 1.0.1 release is currently only being held back due to Win32 build problems.

Regarding the Ogg container overhead, http://www.xiph.org/ogg/vorbis/doc/framing.html says that the overhead shouldn't be more than 2% of the bandwidth. If this is correct (but I don't know if it is), that would give you 1.3kbps overhead on a 64kbps stream.


Interesting. Thanks a lot for the info.

Quote
John33 shall build new Vorbis binaries then...


Well, I believe these ones are already up-to-date then:
http://www.hydrogenaudio.org/forums/index....topic=13019&hl=
(Dated September 2...)

Regards;

Roberto.

 

Pre-Test thread

Reply #144
OK. This discussion is now officially closed. The sample packages are already done and uploaded to a temporary server (where Spoon will grab them and move to the definitive server - thanks Spoon  ) and the batch files are ready as well as the test setup files.

Vorbis 1.0.1 didn't make it, it only made me lose one week. Shows you shouldn't rely on Xiph for deadlines.

One bad news is that the size of the sample packages now sum to 126Mb. Sorry about that, blame on the formats that force me to distribute them in lossless for lacking a command line decoder (WMA, MP3pro and Real Audio). Shouldn't be a hassle for broadband users, but people on dial-up might need to let the files downloading overnight. The average package size is 10Mb.

Thanks for all the support you guys gave me in this thread. It was hugely valuable.

The test starts tomorrow sometime by the (brazilian) afternoon.

Looking forward to your participation.

Best regards;

Roberto.

 
SimplePortal 1.0.0 RC1 © 2008-2019