Sample Rate Conversion and bit reduction - when and how with software?

[MattGray]

Originally Posted by adamlloyd83

But Izotope outputs a 32-bit float file.

Are you talking about iZotope RX Advanced or AE's WaveEditor or SampleManager?

Quite frankly I don't know why it wasn't made to just return the file to the source bit depth after resampling, that's annoying to me, unless I'm missing something. It isn't a problem for downsampling, because we usually dither at the last step anyway, but upsampling is a different story.

What Alexey (one of the very talented programmers for iZotope) was trying to explain to you is that iZotope's SRC does it's DSP calculations at 64 bit floating point internally. All SRC's (and all native plug-ins in general) are done at either 32bit floating point (most common) or 64bit floating point so you can not avoid your audio increasing in wordlength if you want to upsample to 96kHz.

So..."Are there any (high quality) SRC programs that will upsample 24 bit files without converting them to 32 bit float files that need dithering back to 24 bits?

No... refer to the info above. However a lot of Native DAW's will import & playback 32bit floating point files directly (with the exception of Pro Tools) which would save you dithering & requantizing to 24bit fixed point again. What DAW are you using?

Mind you one truncation from 64bit or 32bit float to 24bit isn't likely to be audible as it's truncation errors/distortion will be below the noise floor of even the best D/A converters unless you heavily process or raise the gain dramatically post the initial truncation.

Matt

[adamlloyd]

Originally Posted by MattGray

Are you talking about iZotope RX Advanced or AE's WaveEditor or SampleManager?



What Alexey (one of the very talented programmers for iZotope) was trying to explain to you is that iZotope's SRC does it's DSP calculations at 64 bit floating point internally. All SRC's (and all native plug-ins in general) are done at either 32bit floating point (most common) or 64bit floating point so you can not avoid your audio increasing in wordlength if you want to upsample to 96kHz.



No... refer to the info above. However a lot of Native DAW's will import & playback 32bit floating point files directly (with the exception of Pro Tools) which would save you dithering & requantizing to 24bit fixed point again. What DAW are you using?

Mind you one truncation from 64bit or 32bit float to 24bit isn't likely to be audible as it's truncation errors/distortion will be below the noise floor of even the best D/A converters unless you heavily process or raise the gain dramatically post the initial truncation.

Matt

Hey Matt..

Ahh, yes I tried using RX Advanced. I tried truncating, and the tracks sounded fine at first when importing them to a 96k session. But after mixing and mastering the song, the end result just didn't sound as smooth to the ears (perhaps my imagination, I'm not sure). I assumed it was the truncation.

I am a PTLE guy (for now... we'll see for how long), so unfortunately I need 24-bit files. It sounds like Wave Editor might do the trick because it works at 64-bits and can return the file to it's original bit depth without problems (from what I understand). I would have to move all the files to my mac (production sys is on PC) to upsample, but that's not a big deal.

Do you know where the distortion occurs when "truncating" the 64-bit process back to a 24-bit file? I'm assuming it's way way down there. I've actually never thought of 64-bit process > 24-bit as truncation in the classic, nasty sense...

Thanks!

[Audiop]

Has anyone used SoX on a Mac? I'm scratching my head trying to make it work on a MBP.

I have been able to use it with a wrapper (SoX Wrap) but it's not updated since 2005 and there are some limitations in settings of dither and the SRC parameters.


/Peter

[jdg]

Originally Posted by Audiop

Has anyone used SoX on a Mac? I'm scratching my head trying to make it work on a MBP.

I have been able to use it with a wrapper (SoX Wrap) but it's not updated since 2005 and there are some limitations in settings of dither and the SRC parameters.


/Peter

yah, what do you need it to do?
i can make a simple front end for one or two of its many "tricks"

just sample rate convert?

[Audiop]

@jdg

I want to SRC 24bit fixed and 32bit float @ 88.2-192kS/s down to 44.1kS/s with same wordlength as input and also the option to dither down to 16bit with TPDF and the shaped dither called "low-shibata" according to SoX.

This I want to do with linear phase VHQ.


If I can only have one batch-file or similar I would like LP VHQ down to 44.1kS/s, no dither.

Where do I send the check? :-)

Thanks a lot man for even consider this. thumbsup


/Peter

[speerchucker]

Originally Posted by Bob Olhsson

NEVER ASSUME audio software developers or audio manufacturers know what they are doing! It simply doesn't pay enough to attract the best engineering talent.

Ha! So true.

Take for example the pathetic SRC in Protools.

[Noooway]

I made my first script ever, a simple script for using SoX for SRC under Mac OSX.
I've used the the VHQ Linear Phase parameters in this example converting from whatever to 44.1 kHz.

Download latest SoX and copy the main sox file to your root/bin directory (hidden).
Copy and paste the script (without the ---) into e.g. BBEdit and save the file named e.g. "soxsrc44.command" (without the " of course).
Copy this .command-file to the directory where you have the files you want to convert and run it! (I guess you also can put the main sox file here too if you did not put it in your root/bin dir)
The script does not delete any original files. Just adds "_44kHz" to the file name.

Note: The Script does not seem to handle directory path names with spaces. If you have a dir named "Audio Files" you got to name it "AudioFiles".

It should be easy editing the script to do what you want, e.g. dithering etc., simply by changing or adding a few parameters in the:

"sox "$i" "$val\_44kHz.wav" rate -v -s 44.1k"

To convert to 48 kHz VHQ Minimum Phase, simply change to:

sox "$i" "$val\_48kHz.wav" rate -v -s -M 48k

Hope it is useful :-)

----------------------
#!/bin/bash

cd `dirname $0`

for i in *.wav; do
val=${i%.wav}

echo "** Converting wav to 44.1 kHz (using VHQ Linear Phase)"

sox "$i" "$val\_44kHz.wav" rate -v -s 44.1k

echo "** $i conversion complete."
done
----------------------

[Cellotron]

Originally Posted by speerchucker

Ha! So true.

Take for example the pathetic SRC in Protools.

Dunno - the most recent version actually tests out pretty decently in the comparisons up at http://scr.infinitewave.ca (which also confirms it's the same exact algorithm used in both HD 7.2 and LE 8.03). One that tests out surprisingly poorly is SADIE - hopefully the upcoming version 6 improves on this.

Best regards,
Steve Berson

[CARLOSNAHY]

some body has used the alesis masterlink... ?? some opinion?? maybe just as src copying the 24bit file exported from your daw ???... im thinkig that i can export from protools hd to desktop, do a cd with the data and use the alesis as src, doing another cd with the 16-44 and do the master in jam...
sorry for my english...

[dcollins]

Originally Posted by CARLOSNAHY

some body has used the alesis masterlink... ?? some opinion?? maybe just as src copying the 24bit file exported from your daw ???... im thinkig that i can export from protools hd to desktop, do a cd with the data and use the alesis as src, doing another cd with the 16-44 and do the master in jam...
sorry for my english...

I would avoid any processing in the Masterlink. That thing has ten foot pole marks all over it.


DC

[DuX]

All I know is... to keep it simple, I keep everything in 96k or 88.2k/32bit FP format. It just feels so safe and the backup space is really not a problem these days, especially if you use Wavpack to compress such files. It's so easy to use them any time, edit it, whatever, without worrying about SRC or dither. You do what you have to do to it, SRC it to whatever the frequency you need and then dither/quantise it to whatever the bitdepth you need. Simple. And the main backup is always here to use it again...

Cheers!