I started a thread under MP3 general, where I was trying to find suggestion to just slightly reduce the size of an mp3 I was creating. One of the suggestions was to try using a low pass cut off to reduce the file size and I was given the suggestion of using the command line -V 4 --lowpass 15 After making that change, I found the file size had greatly grown instead of being reduced. I also noticed something in the lame dos window during the encoding that looked odd. There was something about resampling to 32 kHz which I don't understand. Below are screen captures from the 2 encodings. The first is just using the standard "-V 4". The second is using "-V 4 --lowpass 15". Does anyone have an idea why its doing a resample to 32khz? Is that what is increasing the file size. You can clearly see below that the blocks are getting created with a higher average bps. with the low pass filter????

Thanks,

Bob

LAME version 3.96.1 (http://lame.sourceforge.net/)
CPU features: MMX (ASM used), 3DNow! (ASM used)
Using polyphase lowpass filter, transition band: 17960 Hz - 18494 Hz
Encoding C:\Download\Jtmp91--6.wav to C:\Download\Jtmp91--6.mp3
Encoding as 44.1 kHz VBR(q=4) j-stereo MPEG-1 Layer III (ca. 10x) qval=3
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
5250/5288 (99%)| 2:37/ 2:38| 2:37/ 2:38| 0.8734x| 0:01
32 [ 0]
40 [ 0]
48 [ 0]
56 [ 0]
64 [ 0]
80 [ 28] *
96 [ 471] ****************
112 [1685] ********************************************************
128 [1982] *********************************************************
160 [1079] ************************************
192 [ 5] *
224 [ 0]
256 [ 0]
320 [ 0]

LAME version 3.96.1 (http://lame.sourceforge.net/)
CPU features: MMX (ASM used), 3DNow! (ASM used)
Resampling: input 44.1 kHz output 32 kHz
Using polyphase lowpass filter, transition band: 15097 Hz - 15484 Hz
Encoding C:\Download\Jtmp)!3--.wav to C:\Download\Jtmp)!3--.mp3
Encoding as 32 kHz VBR(q=4) j-stereo MPEG-1 Layer III (ca. 10x) qval=3
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
3800/3837 (99%)| 1:57/ 1:59| 1:58/ 1:59| 1.1595x| 0:01
32 [ 0]
40 [ 0]
48 [ 0]
56 [ 0]
64 [ 0]
80 [ 0]
96 [ 0]
112 [ 0]
128 [ 2] *
160 [ 161] *****
192 [2573]*********************************************************
224 [1063] ****************************
256 [ 1] *
320 [ 0]

It resamples because of this:


However, I am not sure why you encountered such a bitrate increase.

Without looking at the code it appears to me that there may be a bug or 2 involved here. I played around with different lowpass values and found that if the lowpass was set to 15.65 or higher, I got the normal(lower) bitrates. If I set the lowpass to 15.64 or lower, I got the inflated bitrates.

Located below are abbreviated screen shot this time. At these lowpass settings the encoding remains at 44 kHz in both examples. Also notice the transitioning band range for the first example doesn't seem to be correct, even though thats the example that correctly keeps the lower bitrates.

Example 1 "-V 4 --lowpass 15.65"

LAME version 3.96.1 (http://lame.sourceforge.net/)
Using polyphase lowpass filter, transition band: 15826 Hz - 16360 Hz

96 [ 74] *******
112 [ 367] ********************************
128 [ 758] ***********************************************************
160 [ 738] **********************************************************
192 [ 13] **

Example 2 "-V 4 --lowpass 15.64"

LAME version 3.96.1 (http://lame.sourceforge.net/)
Using polyphase lowpass filter, transition band: 15115 Hz - 15648 Hz
128 [ 3] *
160 [ 279] **************
192 [1384] *********************************************************
224 [ 274] **************
256 [ 10] *

You might have a buggy exe - or a very odd track. I cant reproduce this behaviour on any samples, and your readouts show all M/S blocks dispite m-j stereo used.
The lowpass and resampling details are normal, you cant set lowpass exactly how you like, it rounds to the nearest available interval that the lowpass algorithm ~finds most convientient.

It seems the bitrate inflation may be tied to the switch in sampling frequency... Is it possible to force it to use 44.1khz, and if so, does it still have the same problem?

In the last examples of lowpass filitering at 16.64 and 16.65, the sampling frequency was kept at 44.1 khz in both examples because the lowpass was high enough to avoid the shift down to 32 khz(at least the lame ms-dos box stated both were recording at 44.1 khz. The source .wav file may have something to do with the problem. It was extracted from a VHS tape of an old 70's movie and may not have been stereo to begin with. I only have a free email account that limits me to 2048k email attachments, but if someone knows of a place,(perhaps an ftp site), where I can place the wav file for other's to look at, I'd be happy to do that. I also tried this in the latest 3.97 a8 release and the same thing happens, so I think I'd rule out a buggy 3.96.1 .exe compile at my location.

Bob

In the last examples of lowpass filitering at 16.64 and 16.65, the sampling frequency was kept at 44.1 khz in both examples because the lowpass was high enough to avoid the shift down to 32 khz(at least the lame ms-dos box stated both were recording at 44.1 khz. The source .wav file may have something to do with the problem. It was extracted from a VHS tape of an old 70's movie and may not have been stereo to begin with. I only have a free email account that limits me to 2048k email attachments, but if someone knows of a place,(perhaps an ftp site), where I can place the wav file for other's to look at, I'd be happy to do that. I also tried this in the latest 3.97 a8 release and the same thing happens, so I think I'd rule out a buggy 3.96.1 .exe compile at my location.

Actually, I can use a friends email account to send a larger file, but I know many email servers limit the size of incomming attachments so if someone is interested in me emailing the file, just make sure they can accept large attachments. I guess I could also trim the .wav down to what ever size someone requests. It doesn't need to be 30 megabytes to recreate the problem, I could simply send a 2 megabyte clip.

Bob

Can't you upload to the Uploads forum?

Maybe the bitrate increase after 32kHz sampling is due to the sfb21 limitation ?

Then maybe you have a TV pilot signal on your source that is causing masking of slighlty lower freqs.
The lowpass, removing this pilot, would also remove the strong masker.

I'm not sure how to check for that pilot signal, I was going to try to upload a small sample (9meg) of the wav file to the upload forum as someone else suggested but it wouldn't work. Perhaps you need special permissions to upload to that forum. It let me browse to the file on my hard drive, but when I click on the attach file button, I end up at a "page can not be displayed..." microsoft error page.

Bob