Here's a mailing list exclusive: the following URL leads to a pre-release version of the MAD plug-in.
http://www.mars.org/home/rob/proj/mpeg/mad-plugin/in_mad-0.12.3b-pre1.zip
This version includes a special "Statistics" tab in the file info dialog for the currently playing song.
Among other things, you'll find there an indication of how often and how much samples are being clipped. Furthermore, you can choose to attenuate the signal on-the-fly based on the amount of clipping.
Attenuation is performed internally on high-precision pre-dithered samples, so it should not affect sound quality except to improve it by eliminating clipped samples.
Further ideas I have in this direction include: - "Automatic clipping attenuation" configuration option - remembering the attenuation setting on a per-file basis
Currently the attenuation setting is automatically reset at the beginning of each song.
I'd like feedback on where this is going. Is this useful? If an automatic attenuation option is made available, would the manual buttons still be useful?
Would there be any benefit in an even more manual configuration like, say, a slider? What about amplification rather than attenuation?
Cheers, -rob
Hi,
On Sun, 3 Dec 2000, Rob Leslie wrote:
I'd like feedback on where this is going. Is this useful?
For me, err... I'm trying to find the good word... No. ;-)
When I listen to clipped mp3s (up to 3dB clipping according to your plug-in), there's nothing that bother me, so...
Now, a little question : Is it too hard to add in the stats something which display the percentage of LR/MS/IS frames, when the mp3 is encoded using JS ?
Bye,
Pierre Darbon
Pierre Darbon wrote:
Hi,
On Sun, 3 Dec 2000, Rob Leslie wrote:
I'd like feedback on where this is going. Is this useful?
For me, err... I'm trying to find the good word... No. ;-)
When I listen to clipped mp3s (up to 3dB clipping according to your plug-in), there's nothing that bother me, so...
Funny, I can hear audible distortion with just 0.3dB clipped samples in some files. I haven't even found any song in my collection with as much as 3dB of clipping, either, so I imagine it would sound even worse. You should try more varied musical selections.
Pierre Darbon licinfo@cict.fr wrote:
Now, a little question : Is it too hard to add in the stats something which display the percentage of LR/MS/IS frames, when the mp3 is encoded using JS ?
I'm thinking about adding something to show this, yes.
I'll also probably change the peak/clip indicator to show the actual peak even when it's not clipped, maybe and/or with a dB value.
Cheers, -rob
Here's a mailing list exclusive: the following URL leads to a pre-release version of the MAD plug-in.
http://www.mars.org/home/rob/proj/mpeg/mad-plugin/in_mad-0.12.3b-pre1.zip
This version includes a special "Statistics" tab in the file info dialog
for
the currently playing song.
This tab confirmed that the file I'm using to test resistance of decoders is a big winner in the "broken mp3" contest. It's really impressive how it's bad:
decoded frames: 13551 clipped samples: 4941 peak clipping: "Attenuate" attenuated the file by -16.7 dB sync errors: 36 crc errors: 1 (but there is NO crc in the file) other errors: 88
If one day you need an highly broken mp3, I've got a good one. Btw several errors in this file can be repaired by hands, using an hex editor. It would be nice to have a decoder able to do the same thing automatically.
Regards,
--
Gabriel Bouvigne - France bouvigne@mp3-tech.org mobile phone: gsm@mp3-tech.org icq: 12138873
MP3' Tech: www.mp3-tech.org
What a cool pre-release! We're not worthy! we're not worthy! :D
Rob Leslie wrote:
Here's a mailing list exclusive: the following URL leads to a pre-release version of the MAD plug-in.
http://www.mars.org/home/rob/proj/mpeg/mad-plugin/in_mad-0.12.3b-pre1.zip
This version includes a special "Statistics" tab in the file info dialog for the currently playing song.
Among other things, you'll find there an indication of how often and how much samples are being clipped. Furthermore, you can choose to attenuate the signal on-the-fly based on the amount of clipping.
It would be useful if it shows the peak clipping number (instead of the graphical bar), or the bar and also the actual number. With this number, the user could then have a general idea of how to manually change the attenuation before playback (the manual slider you mentioned). Right now, you can only attenuate after-the-fact, and thus can't really listen to the samples with and without varying attenuation to see what kind of quality difference there is easily.
Attenuation is performed internally on high-precision pre-dithered samples, so it should not affect sound quality except to improve it by eliminating clipped samples.
Further ideas I have in this direction include:
- "Automatic clipping attenuation" configuration option
- remembering the attenuation setting on a per-file basis
I'd like feedback on where this is going. Is this useful? If an automatic attenuation option is made available, would the manual buttons still be useful?
For automatic clipping attenuation, would this happen only after the highest clipped sample? Or would it read ahead somewhow?
This looks like a really useful feature. I have this one really harsh Christina Aguilera song (most modern pop CD's sound like this though) with bad sibilance. I thought it was my new speakers which had too bright treble, but after I found out the song was clipping very badly and after I attenuated it, it played much smoother. Verry nice. :)
What about amplification rather than attenuation?
Amplification at the frame level during playback sounds like a really, really great feature! However, wouldn't you get into the trouble of not knowing just how much to amplify, thus inducing clipping? So with automatic attenuation and automatic amplification, you'd amplify to clipping, only to attenuate again, losing the peaks of some samples. How can you amplify just shy of clipping at one value for the entire song without going through the the frames before-hand?