Hi Pierre,
there are several reasons for your observations:
- a change in the threshold calculation for the partition bands
(AACS3 is now activated by default, psymodel.c)
- a change in the quality control for VBR, because of
the change in the threshold calculation
(look for lower_masking_db, quantize.c)
- and the side channel reduction is off, when switching RH_masking on.
The amount of bits needed for the side channel is calculated
by the bits used for the mid channel. RH_masking turns this
behaviour off. Lame calculates then the amount of bits needed
for the side channel, just as for the other channels (R,L,M).
It treats M/S channels like L/R channels.
This needs more time and sometimes more bits,
but I think it produces mp3s of a higher quality.
You may read the "--noath option" thread again.
If you really want to try my RH_masking, I use the following
options:
lame -V6 -B320 -b32 -h -k -m j --lowpass 24 file.wav
-B320 -b32 => allows Lame to select out of the
full range of bitrates
-h -k -m j => to produce high quality streams
without sfb21 cutoff
with M/S stereo enabled
--lowpass 24 => to disable the lowpass filtering for -V6
Robert
Pierre wrote:
> Hello,
>
> Once again, it's a newbie who is talking...
> First, a big "Thanks !" for everyone who replies last time.
>
> I've done some tests with versions 3.59 and 3.60, using VBR.
> Could somebody explain to me (with everyday's words) these results and such a
> difference ?
>
> Command line looks like : ./lame -h -v -V 0 test2.wav
> Actually,test2.wav is 1 minute and 10 seconds (the beginning) of "Enjoy
> the silence" from Depeche Mode (Violator album)
>
> 1) Lame 3.59
>
> LAME version 3.59 (www.sulaco.org/mp3)
> GPSYCHO: GPL psycho-acoustic model version 0.75.
> Encoding testcase.wav to testcase.new.mp3
> Encoding as 44.1 kHz 128 kbps j-stereo MPEG1 LayerIII file (11.0x)
> Frame | CPU/estimated | time/estimated | play/CPU | ETA
> 6/ 6(100%)| 0:00:00/ 0:00:00| 0:00:00/ 0:00:00| 0.6830| 0:00:00
> cmp -l testcase.new.mp3 testcase.mp3 | wc
> 2152 6456 32280
>
> LAME version 3.59 (www.sulaco.org/mp3)
> GPSYCHO: GPL psycho-acoustic model version 0.75.
> Encoding /mnt/disque_d/dae/test2.wav to /mnt/disque_d/dae/test2.wav.mp3
> Encoding as 44.1kHz VBR(q=0) stereo MPEG1 LayerIII file
> Frame | CPU/estimated | time/estimated | play/CPU | ETA
> 2689/ 2689(100%)| 0:04:07/ 0:04:07| 0:04:08/ 0:04:08| 0.2849| 0:00:00
> 32:*
> 40:*
> 48:
> 56:*
> 64:
> 80:
> 96:
> 112:*
> 128:*
> 160:*
> 192:*
> 224:*
> 256:****
> 320:**************************************************
> average: 314 kbs
>
>
> 2) Lame 3.60 without RH_masking
>
> LAME version 3.60 (www.sulaco.org/mp3)
> GPSYCHO: GPL psycho-acoustic model version 0.75.
> Encoding testcase.wav to testcase.new.mp3
> Encoding as 44.1 kHz 128 kbps j-stereo MPEG1 LayerIII file (11.0x)
> Frame | CPU/estimated | time/estimated | play/CPU | ETA
> 6/ 6(100%)| 0:00:00/ 0:00:00| 0:00:00/ 0:00:00| 0.5818| 0:00:00
> cmp -l testcase.new.mp3 testcase.mp3 | wc
> 0 0 0
>
> LAME version 3.60 (www.sulaco.org/mp3)
> GPSYCHO: GPL psycho-acoustic model version 0.75.
> Encoding /mnt/disque_d/dae/test2.wav to /mnt/disque_d/dae/test2.wav.mp3
> Encoding as 44.1kHz VBR(q=0) stereo MPEG1 LayerIII file
> Frame | CPU/estimated | time/estimated | play/CPU | ETA
> 2689/ 2689(100%)| 0:06:02/ 0:06:02| 0:06:03/ 0:06:03| 0.1946| 0:00:00
> 32:*
> 40:*
> 48:
> 56:*
> 64:
> 80:
> 96:
> 112:*
> 128:*
> 160:*
> 192:*
> 224:*****************************
> 256:**************************************************
> 320:*****************************
> average: 264 kbs
>
>
> 3) Lame 3.60 with RH_masking
>
> LAME version 3.60 (www.sulaco.org/mp3)
> GPSYCHO: GPL psycho-acoustic model version 0.75.
> Encoding testcase.wav to testcase.new.mp3
> Encoding as 44.1 kHz 128 kbps j-stereo MPEG1 LayerIII file (11.0x)
> Frame | CPU/estimated | time/estimated | play/CPU | ETA
> 6/ 6(100%)| 0:00:00/ 0:00:00| 0:00:00/ 0:00:00| 0.5610| 0:00:00
> cmp -l testcase.new.mp3 testcase.mp3 | wc
> 2693 8079 40395
>
> LAME version 3.60 (www.sulaco.org/mp3)
> GPSYCHO: GPL psycho-acoustic model version 0.75.
> Encoding /mnt/disque_d/dae/test2.wav to /mnt/disque_d/dae/test2.wav.mp3
> Encoding as 44.1kHz VBR(q=0) stereo MPEG1 LayerIII file
> Frame | CPU/estimated | time/estimated | play/CPU | ETA
> 2689/ 2689(100%)| 0:06:07/ 0:06:08| 0:06:08/ 0:06:08| 0.1916| 0:00:00
> 32:*
> 40:*
> 48:
> 56:*
> 64:
> 80:
> 96:
> 112:*
> 128:*
> 160:*
> 192:*
> 224:***********************
> 256:**************************************************
> 320:*********************************************
> average: 273 kbs
>
>
>
> Lame 3.59 was compiled using :
> CC_OPTS = -O9 -fomit-frame-pointer -march=pentiumpro \
> -ffast-math -funroll-loops
>
> Lame 3.60 :
> FEATURES = -DAACS3 -DAAC_TMN_NMT [-DRH_masking]
> CC_OPTS = $(FEATURES) -Wall -O9 -fomit-frame-pointer -march=pentiumpro \
> -fno-strength-reduce -finline-functions \
> -ffast-math -malign-double -mfancy-math-387 \
> -funroll-loops
>
> Redhat 6.0 on PPro 233, kernel 2.2.10, gcc --version says : egcs-2.91.66
>
> Finally, does it mean that theses AAC optimizations will always make
> improvements in quality when encoding to the 128 kbits "standard" ?
>
> Again, thanks for your replies.
>
> Bye,
>
>
> Pierre Darbon
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
