soundfreely
Max Output Level: -78 dBFS
- Total Posts : 625
- Joined: 2003/12/29 19:17:54
- Location: NJ, USA
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/23 00:30:50
(permalink)
ORIGINAL: The Scar
I've encountered this - I was working on a friend's system and used his L2 with a 16 bit dither, and it still said the file was 24 bit.... so we figured we had no idea how to work the L2 properly. I'd be very curious to hear from folks who know the L2 better if what Chris says is correct: that it does dither it down but SF still reads it as 16 bit. ('Cause I looked like a bit of a fool that day - I recommended using the L2 to dither and then it didn't seem to work.)
Hence the reason it is a small pet peeve of mine when people say they are "dithering down to..." As already mentioned, dither is noise used to mask the audible problems that would otherwise be introduced by truncation. In your friend's case, the L2 made the audio effectively 16 bits but the file is still a 24 bit file becuase it hasn't yet been truncated. It's essentially a 24 bit file with 16 bits of "good" information until it is truncated down to a 16 bit file. -Erik
|
TheFingers 
Max Output Level: -58.5 dBFS
- Total Posts : 1669
- Joined: 2005/10/28 18:42:44
- Location: A warm canal.
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/23 01:00:50
(permalink)
I think I confirmed it last night. After rendering the L2 on a 24 bit file. SF still has a 24 bit file but if you look at the file with a hex editor (I opened it byte by byte in Matlab). You will se that the 8 LSBs are all zeros. L2 does the dithering on the 16th bit and zeros the remaining LSBs. You then have to have SF change the bit depth, but since L2 has acted already SF is mearely discarding zeros! Just don't double dither! Ozone3 works this way too, when dithering, the host software should be set to discard the zeroed bits. For example, if you dither with Ozone3 and Sonar, say 24 to 16, Sonar needs to be set to 16 on export with the plugin (Ozone3) running. I have SF, but don't use it to dither so I don't know if a feature of SF is to automatically toss the zeroed-by-dither bits.
1973 "A" neck. I'd rather be playing Bass:
|
Mr. Ease
Max Output Level: -71 dBFS
- Total Posts : 960
- Joined: 2003/11/24 18:44:01
- Location: West Sussex, UK
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/23 06:39:24
(permalink)
Just something that does not seem to have been mentioned before. If your target media is CD only, then why do you not use 88.2 kHz sampling? This will not produce the same aliasing problems going to 44.1 kHz as 96 kHz does, so at least then you will only be worrying about dither. Of course if you want to target 48 kHz media as well this is a moot point.
post edited by Mr. Ease - 2006/03/23 06:45:50
|
bthompson
Max Output Level: -86 dBFS
- Total Posts : 227
- Joined: 2005/03/28 11:04:56
- Location: Rochester, NY USA
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/23 08:00:58
(permalink)
ORIGINAL: The Scar
that's totally counter intuitive (at least to my brain) but now that I know how it works I see the logic.
I suppose it is but there's a reason for it. L2 is a plugin. Plugins have to output the same sample size they are fed as input. --Bill
|
fcarosone
Max Output Level: -86 dBFS
- Total Posts : 241
- Joined: 2005/03/04 11:26:28
- Location: Italy
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/24 08:36:02
(permalink)
I join this query perhaps too late, I've been out for some days. I also confirm that Wave Ultramaximizer does dither but does not truncate, essentially because it is a plug-in as someone else wrote. If you have a bit-meter you can see it clearly (I have Metric Halo Spectrafoo on a Protools Mac TDM system at job, and I have Ozone on a Sonar PC at home, they both do).
I would like to add one more info to the query, I haven't seen now: before truncating a bit depth I always wonder if I'm missing something in the low-bits. So I always amplify / apply a gain to the file so that it is close to 0 dB Full Scale (if it is for burning, it can be normalized at -0.4 dB, if you plan to do some mastering you'd better leave some room). I mean, applying a gain to the 32-bit file you "move the bits up" as much as you can, so that when you apply dither and truncate you have lost the minimum from your file.
For instance, I use to work in 24 bit and I have the Adobe audition collection of loops, recorded in 32-bits. I am not interested in such a big dynamics, so (within Audition) I truncate them down to 24-bits before entering Sonar. In order to do so, I first analyze a group of them to grasp their max level (in Audition) and then apply a uniform gain to all of the loops. Only after I dither and convert to 24-bits.
|
bthompson
Max Output Level: -86 dBFS
- Total Posts : 227
- Joined: 2005/03/28 11:04:56
- Location: Rochester, NY USA
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/24 09:13:03
(permalink)
ORIGINAL: soundfreely
Hence the reason it is a small pet peeve of mine when people say they are "dithering down to..." As already mentioned, dither is noise used to mask the audible problems that would otherwise be introduced by truncation.
That's not correct. Dither *does not* mask distortion products. It *prevents* the formation of distortion products. Exactly the right "shape" and amplitude of noise added to the high resolution signal allows the subsequent rounding to a lower bit depth to be done with the only error being a low level random noise floor. There's no distortion at all, just noise, and not much of it. If you just add noise to an already truncated signal, you'll quickly see the difference. The noise won't mask the distortion at all.
|
bandasound
Max Output Level: -85 dBFS
- Total Posts : 277
- Joined: 2004/03/03 22:54:04
- Location: Knoxville, TN
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/29 01:59:16
(permalink)
thanks for the info on Waves L3...Im not familiar with the hexidecimal mathematics to confirm what you said but it does make sense....I really like the results i get even when i save as Mp3 at 256kps.
You know it would be nice if there was a 24bit MP3 option....
hmm....True R&B is like deep fried chicken, greens, and buttered corn bread...
|
soundfreely
Max Output Level: -78 dBFS
- Total Posts : 625
- Joined: 2003/12/29 19:17:54
- Location: NJ, USA
- Status: offline
RE: Converting from 32bits to other formats -> very bad results.
2006/03/29 03:58:09
(permalink)
ORIGINAL: bthompson
ORIGINAL: soundfreely
Hence the reason it is a small pet peeve of mine when people say they are "dithering down to..." As already mentioned, dither is noise used to mask the audible problems that would otherwise be introduced by truncation.
That's not correct. Dither *does not* mask distortion products. It *prevents* the formation of distortion products. Exactly the right "shape" and amplitude of noise added to the high resolution signal allows the subsequent rounding to a lower bit depth to be done with the only error being a low level random noise floor. There's no distortion at all, just noise, and not much of it.
If you just add noise to an already truncated signal, you'll quickly see the difference. The noise won't mask the distortion at all.
How ironic?  No disagreement here. "Mask" was a poor word choice. I was thinking "mask" because of how the dither functions. Summed with a higher bit signal, it essentially "shakes (couldn't think of a better word)" the audio before the truncation thereby preventing any quantization distortion due to brute truncation. However, I was implying that there is a change in the audio after the dither is applied. Thanks for clearing that one up. -Erik
post edited by soundfreely - 2006/03/29 04:07:08
|