sample rates

[starfinger]

I know that it's desirable to maintain a high bit-depth throughout the recording process, so I've been rendering my mixes at 24-bit, mastering them in sound forge, and finally dithering them to 16-bit.

What about the sample rate? Is there an analogous benefit for keeping files at 96khz throughout the process, if you know it will eventually have to be 44khz?

In other words, will it sound better to apply my mastering plugins before or after resampling?
Or should I just render my mix at 44k/24-bit and master/dither that?

I might be answering my own question here, but I guess I should see if my softsynths sound better rendered to 96khz and then resampled to 44k, or just rendered to 44k directly.

-craig

[Southwest_Statistic]

Before.

[Adam!]

Because 99% of the effects and processing (including simple digital mixing like gain and EQ) reduce the effective bit depth it can be very beneficial to work at 24 or even 32 bits/sample. Since most digital audio environments have internal mixers that operate at 32+ bits-per-sample you probably won’t lose any CPU performance by working at a higher bit depth (I know you already agree, just summarizing).

Unlike bit depth, I have found very few reasons to work at higher sampling rates than 48 kHz. In fact the only reason I work at 48 kHz instead of 44.1 kHz is because my soundcard has latency issues otherwise. 99.9% of effects and processing do not reduce the sampling rate, so loss of fidelity during mixing should not be a concern. Although recording at 96 kHz will allow you to capture the 20-40 kHz range, I highly doubt your equipment produces or preserves much of this information. Even if it does, unless you are drastically pitching down your recordings, or you’re primary audience is dogs and bats, no one will ever know the difference between 48 kHz and 96 kHz. Also, most audio environments let you set the sample rate they operate at (instead of up-sampling everything like they do for bit depth). This means that if you are working at 96 kHz you’ll probably have to do twice as much processing than you would normally do for every second of audio. That’s bad.

One upside is that many soundcards measure their buffers by samples-per-second, so by using a higher sample rate you can reduce by half the theoretical minimum latency of your setup. Again, this is one of those things where you ask yourself “Will anyone be able to tell the difference? Will I be able to tell the difference?” Also, there’s the there-are-bigger-things-to-worry-about argument: for instance, you could be recording at 64-bit/192 kHz, but if everything you record has the fan whir from your pushed-to-the-limits PC your mix will probably sound worse than before.


EDIT: Oh yeah, to answer your question... mastering before down-sampling can never sound worse, will probably sound the same, and could sound better.

[deshead]

What about the sample rate? Is there an analogous benefit for keeping files at 96khz throughout the process, if you know it will eventually have to be 44khz?

The consensus I've read is that you want to minimize the amount of resampling required (because aliasing introduces noise,) so it's generally best to keep everything at the sample rate you'll end up with.

In other words, will it sound better to apply my mastering plugins before or after resampling?

Theoretically, before is better, since the higher sample rate allows the plugins to work with any ultrasonic frequencies that might be present. But like Puce said, you'll probably never hear the difference.

Here are a couple of articles on the subject:

http://www.digitalprosound.com/Htm/Soap ... Apogee.htm

http://www.prorec.com/prorec/articles.n ... B20022F4CA

[starfinger]

a ha, thanks guys.

The ultrasonics comment was closest to what I was concerned about, but that now seems pretty irrelevant.

-craig

[toddlans]

the Nyquist theorum states that you only need a sample rate of twice of the highest frequency you'll be recording. so for our puposes since not many of us can hear beyond 20k (and actually i bet a lot of us can't hear above 18 given the fact we've all probably grown up blasting headphones and such) 44.1 is definitely high enough. i read a lot about recording and about once a month this debate ends up somewhere on an audio forum. some people will argue that theoretically a higher sample rate would give you increased detail, etc, and other people tell them they're full of shit or that for practical purposes its not needed. the point is if a group of pro engineers can't decide if theres a benefit to 96k or higher sample rates then 99% of people couldn't tell the difference even if there is in fact one. You do want to work with 24bit though, especially while recording tracks. Taking a sample thats 16bit and making it 24 bit won't do so much, but if you have the ablility to record tracks at 24bit, do so as you will get higher resolution and don't have to worry as much about getting just below 0db for a good sound as you do with 16bit. I record at 24-bit/44.1k with Cubase set at 32bit float.

[starfinger]

i sent (basically) this same question to izotope, with regards to using ozone for dithering. I thought I'd share their response with you guys.

Hi Craig,

The best time to downsample from 96 to 44 is after mastering, because you'll get better sound in applying your audio effects at the higher sample rate and bit depth. Downsampling after dithering will ruin it, so a good order for your process could be:

1. Master and apply Ozone (without Dithering) at 96/24
2. Downsample to 44/24
3. Add Ozone Dithering at 44/24
4. Immediately convert 44/24 to 44/16

I hope this helps, and let me know if it works out.

Best,

Dave