Upsampling song vs export at higher sample rate

[slightlyprog]

I just want to understand exactly what is happening 'under the hood' on this. I have searched previous topics but haven't seen the answer.

lets say I have a song at 48khz. Lets assume some plugins that generate harmonics can still alias at 48k.

If I convert the whole song to 96k, they no longer alias. If I then render the mix at 96k, still no aliasing, this is totally intuitive.

Now, what happens if instead of converting the song to 96k, I leave it at 48 and simply export at 96 when rendering the mix. Is the Reason engine mixing at 48k, and then converting only the output to 96? Or is it mixing at 96k (and thus avoiding the aliasing that would happen at 48k) ? I'm running version 10, in Windows7 by the way.

(and, I hope it does not have to be said, - please no opinions about whether the aliasing can be heard or whether it's bad etc. etc. - thanks in advance).

[Aquila]

Whenever I export at a higher resolution than my working rate, I notice Reason has to perform additional processing first before the main render. I can only assume that the software is upscaling any samples or audio channels during this stage.

Not been able to confirm this 100% but it's my guess.

[slightlyprog]

Yeah, it shows that it's calculating first, in some form or fashion, but not sure exactly how. I have seen this same question answered for a different DAW (protools) where someone said that PT was still mixing at 48, but I know Reason does some things a bit differently.

[orthodox]

If I convert the whole song to 96k, they no longer alias.

They still alias, just quieter.

Now, what happens if instead of converting the song to 96k, I leave it at 48 and simply export at 96 when rendering the mix. Is the Reason engine mixing at 48k, and then converting only the output to 96? Or is it mixing at 96k (and thus avoiding the aliasing that would happen at 48k) ?

When you specify 96k on export, the whole Reason engine with all the devices will switch to 96k for the duration of the render, so it's equivalent to playing the song at 96k.

[slightlyprog]

If I convert the whole song to 96k, they no longer alias.

They still alias, just quieter.

Now, what happens if instead of converting the song to 96k, I leave it at 48 and simply export at 96 when rendering the mix. Is the Reason engine mixing at 48k, and then converting only the output to 96? Or is it mixing at 96k (and thus avoiding the aliasing that would happen at 48k) ?

When you specify 96k on export, the whole Reason engine with all the devices will switch to 96k for the duration of the render, so it's equivalent to playing the song at 96k.

Thanks very much. They may not alias at all, it is frequency dependent, so if no signal exceeds the nyquist, you won't have any aliasing, and whatever aliasing does occur will begin proportionately higher, and thus may be beyond audible range.

[visheshl]

If I convert the whole song to 96k, they no longer alias.

They still alias, just quieter.

Now, what happens if instead of converting the song to 96k, I leave it at 48 and simply export at 96 when rendering the mix. Is the Reason engine mixing at 48k, and then converting only the output to 96? Or is it mixing at 96k (and thus avoiding the aliasing that would happen at 48k) ?

When you specify 96k on export, the whole Reason engine with all the devices will switch to 96k for the duration of the render, so it's equivalent to playing the song at 96k.

Will it work for RRP too ? Or is it DAW dependent ?
For example i use Ableton as the daw and only use RRP as plugins(both instruments and fx) no other vst in the project not even Ableton plugins.
Will it function the same way while exporting the file ?

[orthodox]

They may not alias at all, it is frequency dependent, so if no signal exceeds the nyquist, you won't have any aliasing, and whatever aliasing does occur will begin proportionately higher, and thus may be beyond audible range.

If the harmonics went beyond the Nyquist at 24k, they may just as well go beyond the higher Nyquist at 48k. It will then take just 1.5 octaves to get back below 24k and into the audible frequency range, which results in only -10 dB reduction of aliasing level for saw/square harmonics at 96k vs 48k sample rate.

[slightlyprog]

They may not alias at all, it is frequency dependent, so if no signal exceeds the nyquist, you won't have any aliasing, and whatever aliasing does occur will begin proportionately higher, and thus may be beyond audible range.

If the harmonics went beyond the Nyquist at 24k, they may just as well go beyond the higher Nyquist at 48k. It will then take just 1.5 octaves to get back below 24k and into the audible frequency range, which results in only -10 dB reduction of aliasing level for saw/square harmonics at 96k vs 48k sample rate.

The thing about that is, with the added headroom, you can low pass filter effectively without hearing the filter itself (well before 48k and yet higher than audible range).

[orthodox]

If the harmonics went beyond the Nyquist at 24k, they may just as well go beyond the higher Nyquist at 48k. It will then take just 1.5 octaves to get back below 24k and into the audible frequency range, which results in only -10 dB reduction of aliasing level for saw/square harmonics at 96k vs 48k sample rate.

The thing about that is, with the added headroom, you can low pass filter effectively without hearing the filter itself (well before 48k and yet higher than audible range).

When plugins you mentioned generate harmonics, they will be already aliased into the audible range on the plugin output and the LP filter can't do anything about them, it will just cut off inaudible frequencies.

[slightlyprog]

The thing about that is, with the added headroom, you can low pass filter effectively without hearing the filter itself (well before 48k and yet higher than audible range).

When plugins you mentioned generate harmonics, they will be already aliased into the audible range on the plugin output and the LP filter can't do anything about them, it will just cut off inaudible frequencies.

No, I'm talking about the fact that at 96k you can filter before going into any plugins that will alias

[ScuzzyEye]

No, I'm talking about the fact that at 96k you can filter before going into any plugins that will alias

Correct. And/or low-pass after the effect creating the harmonics, even if they have aliased down from 48k. As long as the reflected (in)harmonics don't reach 24k, you can remove them pretty effectively.

Or export at 192k (or 176.4k since integer resampling is a bit easier) and use something like r8brain to resample back down to 44.1, and use its really steep low-pass and excellent resampling algorithm to basically brute force 4x oversampling on every device in the rack.

As noted, before exporting at a rate that's different from the working rate. Any recorded audio will be resampled by Reason. Reason's resampling engine is better than it used to be, but still isn't the best out there. If you're trying to reduce artifacts as much as possible, it may be worth bouncing lower rate audio to disk. Resampling it with higher-end software, and then replacing those recordings. You can still work at 44.1/48, and use Reason down-sampled cache audio. It'll use the "native" higher rate audio on export then.