Topic: upsampling for post-processing??

[Chris K]

a taper pal of mine, that is not on ts.com (yet), says that he sometimes upsamples his d7 dat tapes from 16bit to 24bit in order to add post processing compression, normailzing, etc.

he says that it allows for a smoother transition from the raw data to the bounced data

he then down samples back to 16 bit for cd burning, shn and flac distribution

does this make sense?

anyone else do this?

[Sean Gallemore]

your description makes since

[Chanher]

I "upsample" to 32-bit in CEP whenever I do something extraordinary.  For example I've finally gotten around to converting some of my very first tapes and almost half have significant clipping.  so I convert to 32-bit, run clip restoration, then back to 16-bit.

nice name btw

[Chris K]

what i am wondering is, will the higher bit rate make an audible difference assuming you do the same post processing in 16bit?

what are the advantages?

this is quite new to me, and i find it very interesting!

[Sean Gallemore]

a lot of people will record in 24bit and 48ktz when possible, even if the end result will be a CD (16/44.1).  Maybe it allows for more percision

my take is that you can't get more from less.  16 > 24 > 16 in my limited knowledge wouldn't make it better.  A horrible comparison is using mp3s, going 128 > 192.  You just can't get the data that isn't there.  However, maybe the smooting out theory is right......fuck I dunno, I'll shut up

[scb]

first: it's not upsampling.  bit rate change has nothing to do with changing the number of samples

more bits = more precise numbers.  doing calculations in the 24 bit realm allows for more precision

[Craig T]

doing any kind of editing in 16bit, especially DSP (eq, etc), kills the quality.  take identical copies of a recording, DSP one in 16bit.  the other covert to 32bit-float (or 24bit) then DSP and dither back to 16bit.  that should give you your answer.

[Chris K]

first: it's not upsampling.  bit rate change has nothing to do with changing the number of samples

more bits = more precise numbers.  doing calculations in the 24 bit realm allows for more precision

thanks...that actually makes alot of sense...more precise numbers should equal better calculations which should equal better sound...in theory

my inqury has been satisfied...PLAY ON!

[Kelso]

well it's stupid to use 48 kHz if the final product is a cd. For a very non-significant gain in quality in recording you will lose more at the final conversion. 48 to 44.1 conversion is a very complex conversion and it's destructive.  The proper way to go is 24/44.1 or 24/88.2 for cd mastering.

[zowie]

well it's stupid to use 48 kHz if the final product is a cd. For a very non-significant gain in quality in recording you will lose more at the final conversion. 48 to 44.1 conversion is a very complex conversion and it's destructive.  The proper way to go is 24/44.1 or 24/88.2 for cd mastering.

That's a myth.   Math is math, it's not destructive if done properly, or even very complex, just large numbers.  Unless you're writing the code, what do you care?

[jpschust]

well it's stupid to use 48 kHz if the final product is a cd. For a very non-significant gain in quality in recording you will lose more at the final conversion. 48 to 44.1 conversion is a very complex conversion and it's destructive.  The proper way to go is 24/44.1 or 24/88.2 for cd mastering.

no, the proper way is the highest possible bit and sample rate possible with the best possible conversion tools.  ideally id like to be working in 24/192, but just not enough out there that is up to that par yet.  yet.

[Kelso]

Are you sure it's not destructive? Every pro audio software Iv' used propose different flavour of quality versus speed which would mean it is lossy (there is no "perfect" quality). Just thinking of what sampling is, I find it almost comparable to d/a a/d conversion, but maybe I'm missing something.
 Even if it's a very non-significant loss what's the point for 44.1 to 48 upsampling? If you use 96 right, but 48 it means losing space and processing time for almost nothing. Count more than realtime on a fast computer.