Question about sample rate conversion

[Pedro Teixeira]

So I've been giving a bit of thought about this issue. Imagine you're mastering a 24bit 48khz file, since most hot masters nowadays are constantly hovering around -0dBFS (or -0.3/.5 in that case) I've been having doubts when should you convert the sample rate, because if you do when bouncing, you're doing it after the limiter, so, bringing in the case of hot masters, then there is a possibility of inter-sampling peaks ocurring, but if you do it before you're not processing at 48khz but at 44.1khz which would defeat the whole purpose in the first place...

Any thoughts on this matter?

 

[jipchen]

Personally, I'd convert the sample rate after mastering because otherwise the higher sample rate in the first place would be useless.
On the other hand mastering a 48 or 44.1 file makes probably no real difference anyway.

SSL offers a free plug-in to check inter-sample peaks, I guess it could help you: http://www.solid-state-logic.com/music/X-ISM/index.asp

 

[gabriel g.]

1. There is nothing higher then 0dBFS. so no interpeak sample would make you audio any louder then it is

2. With quality converters, there is NO need to use higher samples rates then needed (audio cd 44.1khz)

3. I would change the samplerate before mastering.

 

[gabriel g.]

Personally, I'd convert the sample rate after mastering because otherwise the higher sample rate in the first place would be useless.
On the other hand mastering a 48 or 44.1 file makes probably no real difference anyway.

SSL offers a free plug-in to check inter-sample peaks, I guess it could help you: http://www.solid-state-logic.com/music/X-ISM/index.asp

most of the time you do more harm then good to your audio with tracking at 48khz for an audio cd.
Because the dither math gets complicated and you get rounding errors

 

[Pedro Teixeira]

Yes I know there's nothing further than 0dBFS, since it's the full scale, but what I mean is, by having it hovering at -0.3dB it may peak to -0dB since it's being done AFTER the limiter.
I haven't had a problem with this, just something I've been thinking of.

So, you're saying the best thing to do is, before mastering, dither the file and then after the processing bounce it to 16bit without dithering since it has already been dithered... what is the exact advantage in that? Isn't it exactly the same thing? The file is getting truncated anyway

 

[gabriel g.]

The thing is, you dont dither anything with sample rate conversation. Dithering only change the bit-depth not the sample rate.
Dither makes 24->16bit.

And dither wont make any interpeak samples because it only realy works with the information near the noisefloor....

Thats why I wrote (btw edited, because I was an idiot) that I would change the SAMPLERATE before mastering

 

[Pedro Teixeira]

I know it doesn't, when I asked right now about dithering I was asking about bitrate, I wasn't very clear, sorry about that.

I guess that would make sense, so you would dither 24bit to 16bit, process and then export at 16bit? I'm asking about this, because it seems redundant since it's going to be done anyway so being before or after it seems the same to me, but it probably isn't, I don't know.

 

[jipchen]

most of the time you do more harm then good to your audio with tracking at 48khz for an audio cd.
Because the dither math gets complicated and you get rounding errors

Yeah, I see it the same way.. I always record in 44.1 kHz.
But given this situation I'd convert it afterwards because the harm (if any) will be done anyways and at least you get the small profit of the higher samplerate for the mastering process.

 

[wishtheend]

Depending on the converter, you can have higher peaks when going from 48khz to 44.1khz. In the case of mastering, I did all processing I wanted for the mastering chaing MINUS the loudness portion (so eq, mb comp, glue comp, etc) so that I have some headroom, did the conversion to 44.1/24bit and then just brought that up to the needed loudness and exported a 16bit version.

I use FG-X for loudness, so it's a little easier to do everything BUT loudness, but I guess if you have a more complicated mastering chain it might require a little trial and error.

 

[gabriel g.]

I know it doesn't, when I asked right now about dithering I was asking about bitrate, I wasn't very clear, sorry about that.

I guess that would make sense, so you would dither 24bit to 16bit, process and then export at 16bit? I'm asking about this, because it seems redundant since it's going to be done anyway so being before or after it seems the same to me, but it probably isn't, I don't know.

ok good,

With changing the sample rate you can get interpeak samples, so I do that before mastering. I import 48khz into a 44.1khz session and let my sequencer of choice (logic, protools, cubase will all do a good job) make the sample rate change.
Then I master.
I use any dither I think sounds best, sometimes the fg-x, ozone, or the build in from logic.

The build in will work at the last process. The plugin ones work before the mix down.

I realy never had any problems with both ways.

I normaly use the apogee 16bit dither in logic.
So it is included in the pcm after loudness maxing.

then I check the files before burning in waveburner and it shows me the same level as in the mastering project.

So no loudness differences.

Hope this helps

 

[gabriel g.]

and the small headroom was the same too (I always use -0.4dBFS)

 

[timislegend]

- inter sample peaks never exceed 1dB (-1.0dbfs will be perfect for src).

- src first (izotope rx advanced / weiss saracon)

- limit / dither second

people talk about dithering and src mathematics as if the exist in the same domain... but they do not.

to avoid rippling and time smearing artifacts in the passband, it is best to convert from higher sample rates.

 

[gabriel g.]

- inter sample peaks never exceed 1dB (-1.0dbfs will be perfect for src).

- src first (izotope rx advanced / weiss saracon)

- limit / dither second

people talk about dithering and src mathematics as if the exist in the same domain... but they do not.

to avoid rippling and time smearing artifacts in the passband, it is best to convert from higher sample rates.

ideal would be 88.2khz....

 

[timislegend]

ideal would be 88.2khz....

word!

i was going to add... to your comment, about even math.

the most common src used is 48 to 44.1 (which is not even math) however 88.2 to 44.1 is even and should be the most common (which is what i use... and love)

 

[John_C]

2. With quality converters, there is NO need to use higher samples rates then needed (audio cd 44.1khz)

not true.

I've long since given up on the idea that 44.1kHz sounds better on its own, as you mentioned: a high quality converter can happily give you a pretty good reconstruction of the intended signal at 44.1

However, aliasing is a real problem with digital audio processing. Oversampling helps, but it introduces it's own problems.

Example: people often dislike the high shelves in eq plugins. You're starting to have to work with signals that are close to the limit of the sample rate and you run into problems. Working at 88.2kHz you can get the same eq quality that you used to get at 10kHz all the way up to 20kHz.

Basically, if you're ever been able to hear the difference between a plugin with and without oversampling then you've heard the advantage of higher sample rates.

Also, you can get away with much greater time stretching at higher sample rates.

 

[ForHerDeadEyes]

Just throwing this in:
http://src.infinitewave.ca/

 

[timislegend]

Just throwing this in:
http://src.infinitewave.ca/

i actually started a thread about this a while ago and most people didn't seem to understand it.

it's definitely useful ...and from time-to-time i will go back a mess around with it.

 

[ForHerDeadEyes]

i actually started a thread about this a while ago and most people didn't seem to understand it.

it's definitely useful ...and from time-to-time i will go back a mess around with it.

Must've missed that..

I have no idea how to use it practically, other than using the ones which has the best overall conversion..
Ozone / Awave studio / R8Brain.. Those are the ones that seems to be the best..