FFdshow is no subject now ;)

But the provided ffmpeg.exe there, means the recent one:
http://ffdshow.faireal.net/mirror/ff....2006.04.19.7z should have been compiled with the patch included as it was applied to the ffmpeg CVS already one Week or more ago ;)

sure! sorry ^_^

@Inc

Thanks, now I got it. :)

Cheers,

Inc's ffmpeg build works great.
 

@ALL

I do confirm that Inc's ffmpeg build works a treat!!! I just tried it out with a WAV PCM 2.0 which was treated with DVDAudio from Wavegenix Deluxe Mastering Suite. The transcoding was performed without issues and I did listen that volume of audiostream transcoded doesn't lower after all.

Well done, Inc.

I will test new Celtic Druid's build in order to see if the patch was applied.

Cheers,

Thats not a patch applied by me. Its a patch from a member of the ffmpeg-developers community.
I only did .... compile the sources.

Is the volume out of celticdruids build different?

@Inc

From my ears, the audios output for both builds (your and from Celtic Druid) seems the same (with regard the volume).

See ya,

Someone with good ears(mine are not!!) listen this please. :D

WAV audio:
http://rapidshare.de/files/18584346/...GINAL.wav.html

AC3 SOFTENCODE[448 kbps]:
http://rapidshare.de/files/18585005/...CODE_.ac3.html

AC3 BESWEET 1.5B31[448 kbps]:
http://rapidshare.de/files/18585442/....5B31.ac3.html

AC3 FFMPEG CELTIC DRUID 19-APRIL-2006[448 kbps]:
http://rapidshare.de/files/18585857/...DRUID.ac3.html

if you can download it and rate, please post your opinion :D

There have been already a tests made related to Softencode vs. FFmpeg/besweet and tmpegEncs AC3. Do look here in the forum.
http://www.kvcd.net/forum/viewtopic.php?t=15987

AC3 from ffmpeg (was) too low but freqs where excellent. Same like Softencode ... and better than TmpgEnc which CUTs at approx. 15Khz
http://www.edv-tipp.de/dvd/047_dvd_sound06.htm
http://www.edv-tipp.de/dvd/047_dvd_sound08.htm

Its german, so u have to use an online translator

@sparkster

No surprise that SFSE have provide the 'better' result: SFSE let you normalize volume audio using DRC and you did this. For my ears, ffmpeg by Celtic Druid returned a ac3 audio file with 'same' volume level as original. Besweet returned a ac3 audio file with a lower volume.

A better test should be you get the wave audiostream and into it apply some sort of normalization (the suite that I've mentioned does this using a one-click button :D ) and with this already normalized wave audiostream you should do the trancoding to ac3 format. Or so you should do apply the same normalization values with ffmpeg / besweet as used with SFSE.

My R$ 0,02 ... ;)

See ya,

@incredible
thanks for the links! That 15khz cut-off is simply surprising as Dialhot said! 8O

@danpos
thank you :D

:lol:

I normalized it in Soundforge 8:
http://www.digitalfaq.com/archives/i.../2006/04/2.png

I used these(wav->ac3) setting for FFMPEG:
And these(wav->ac3) on besweet:
http://www.digitalfaq.com/archives/i.../2006/04/3.png

I did convert ac3 -> wav using these settings on besweet:
http://www.digitalfaq.com/archives/error.gif

Result for BESWEET 1.5B31:
http://www.digitalfaq.com/archives/error.gif

Result for FFMPEG 19-APRIL-2006 CELTIC DRUID:
http://www.digitalfaq.com/archives/i.../2006/04/4.png

the files (transcoded from the normalized wav):

http://rapidshare.de/files/18621551/...LIZED.wav.html
http://rapidshare.de/files/18621245/...SWEET.ac3.html
http://rapidshare.de/files/18621265/...DRUID.ac3.html

@sparkster

Well done! :D

See ya,

Forget about besweet/belight in this state now. It uses a very old ac3enc.dll which was build from old ffmpeg sources where the volume-bug-fix still wasnt applied. I did send a PM to DspGuru at doom9 that he should recompile the ac3enc.dll.

DRC is no normalization, but a function which affects the dynamic range compression. DRC is smart where normalizing parses the whole file for the max value. SO for a comparison switch DRC in SFSE off.

My tests:

A comparison between the orig PCM Waveform vs. SoftEncode and FFmpeg:

First the freqs of the PCM waveform:
http://www.digitalfaq.com/archives/i.../2006/04/1.gif

Second the freqs of the Softencode output:
http://www.digitalfaq.com/archives/i.../2006/04/2.gif

Third the freqs of the FFmpeg output:
http://www.digitalfaq.com/archives/i.../2006/04/3.gif

Now lets overlay the output-graph of all three:
http://www.digitalfaq.com/archives/i.../2006/04/4.gif



Also very interesting the phase stablity of both re-le channels.

pcm:
http://www.digitalfaq.com/archives/i.../2006/04/5.gif

Softencode:
http://www.digitalfaq.com/archives/i.../2006/04/6.gif

ffmpeg:
http://www.digitalfaq.com/archives/i.../2006/04/7.gif

remark: the thicker/uneasier the purple-blue line gets the more incorrect is the phase.


Now we know that the ac3 encoder engine of ffmpeg is a very good qaulity engine, but someone who reported issues with the more anxient releases of this core on his SAP should test the output of the recent ffmpeg verrsion - IIRC Zyphon had a SAP which was very sensible.

@incredible
pretty amazing :!: :!: :D :D :D
What soft (and version x.xx ?) are you using to plot these frequency and phase analysis stuff ?

Adobe Audition (aka cooledit)

@Inc

Very good presentation indeed. So, your and our tests seems to show that ffmpeg (new build with applied patch) is the way to go in treating of free ac3 encoder ... :)

Cheers,

Inc,

have you had the chance to try any more recent commercial AC3 encoder? My friend has the Sony Vegas 6 suite which contains an AC3 encoder - if you cannot do the test encode, I'll ask if he would do so.

Basically I see both encoders the same, except that ffmpeg cuts of a little higher (about 200Khz), which is good. Phase differences are almost the same in all :)

-kwag

A Remark: I saw that I did encode that stereo Waveform using 448kbits in each encoder! Which is not real condition like!

Now the test shows how those encoders do behave if we do use on that stereo Waveform a 192kbit encode and a 128kbit encode.
192kbit CBR is fully sufficient for a stereo signal at 48000Hz, 16bit seen on a technical point of view.


192kbit - Softencode:
http://www.digitalfaq.com/archives/error.gif

A clean full up to 20.3KHz freq Range.



192kbit - FFmpeg:
http://www.digitalfaq.com/archives/error.gif

Also here up to 20.8Hz freq Range but a phase error which rises from 13Hz. BUT this is minimal -- check if you can recognise it!



128kbit - Softencode:
http://www.digitalfaq.com/archives/error.gif

As shown in the SFSE encoder config settings at 128kbit the encoder forces the Freq being cut at 13.55Khz!



128kbit - FFmpeg:
http://www.digitalfaq.com/archives/error.gif

Full Hz range but a phase error starting at an even earlier Hz point compared to 192!



What do these diagrams say? Very easy:
If you want "no" compromise in stereo AC3 audioencoding then you should go for Softencode at 192kbit as a full very phase-error-free freq range will be provided after decoding the AC3.
If you go for 128kbit AC3 then you have to choose between freq fullrange but minimal phase-errors (ffmpeg) or a limited freq range incl. no errors (SFSE).

Now you should make tests using your ears especiall on the phase errors .... ;) And then you decide what will be your choice.

A point to freq range: dont be affected by reading specs/values! "Normal" human ears wont recognise in a "normal attention" freqs above 16Khz, if so then congrats to excellent ears and for shure mega congrats to an EXCELLENT playback/speaker combination! Audiophile systems which do provide signals in "an accurate db state" so that you can recognise these high freq are very expensive. Shure technical specs of some speakers/Receivers do often mention high freq outputs about 18-22Khz but thats theory and beside this it depends WHAT these high freqs are made of and for shure if they can be provided at a needed db ratio.
BTW: A regular very good TV broadcast in PAL wont exceed 16Khz!

No sorry, I dont have any other choice. When doing tests be shure NO DRC is set as this will provide an almost untouched range.

BUMP! :lol:

Sorry guys but I really can't tell what I'm doing wrong here :(.
Could you please post a step-by-step tutor on how to transcode AC3 with this tool?
BTW screenshots would work wonderfully for me :D.

Cheers

Update: new ac3 encoder based on FFMPEG lib.
 

@Incredible

Take an eye at this topic.

The author (Justin Ruggles) is the same guy that did propose the patch to correct the 'low volume bug' on ffmpeg-devel list ...

JFYI. :)

Regards,