96k vs 48k .... Aliasing Question

[BenIfin]

This is a bit of a follow-up to something I posted about a month ago - and am seeking further education.

Say I am recording a Digital Guitar Amp Sim Plugin -or- playing that same Digital Guitar Amp Sim Plugin live, and not employing any Oversampling in either case

=> if doing it at 96k, does that mean that Aliasing folding back / artifacts will only be present above 48k [1/2 the Sample Rate] (?)

=> equally if doing it at 48k, does that mean that Aliasing folding back / artifacts will only be present above 24k [1/2 the Sample Rate] (?)

=> or in either case, does the Aliasing still foldback down / artifact below 1/2 the Sample Rate chosen (?)

Thx.

 

[jay mitchell]

=> if doing it at 96k, does that mean that Aliasing folding back / artifacts will only be present above 48k [1/2 the Sample Rate] (?)

Aliasing artifacts - the effect - are always present below Nyquist (half the sample rate). Frequencies that cause aliasing to occur are those that lie above Nyquist. Those frequencies are reflected back to frequencies within the Nyquist band based on how far they lie above Nyquist. For example, if the SR is 96 kHz (IOW, Nyquist = 48kHz) and a nonlinear process generates a signal 1kHz above Nyquist (49kHz), the result of the aliasing will be a signal 1kHz below Nyquist (47kHz). In order for aliasing products to reach the audible band - 20kHz or less - with a 96kHz SR, the frequencies that get aliased would have to exceed Nyquist by 28kHz or more: 48kHz + 28kHz = 76kHz. For example, 77kHz would be reflected back (aliased) to 19kHz.

The arithmetic is the same no matter the sample rate.

 

[BenIfin]

Aliasing artifacts - the effect - are always present below Nyquist (half the sample rate). Frequencies that cause aliasing to occur are those that lie above Nyquist. Those frequencies are reflected back to frequencies within the Nyquist band based on how far they lie above Nyquist. For example, if the SR is 96 kHz (IOW, Nyquist = 48kHz) and a nonlinear process generates a signal 1kHz above Nyquist (49kHz), the result of the aliasing will be a signal 1kHz below Nyquist (47kHz). In order for aliasing products to reach the audible band - 20kHz or less - with a 96kHz SR, the frequencies that get aliased would have to exceed Nyquist by 28kHz or more: 48kHz + 28kHz = 76kHz. For example, 77kHz would be reflected back (aliased) to 19kHz.

The arithmetic is the same no matter the sample rate.

Thanks - for a change I think I actually understand ^this^ - which for me is an achievement :)

In crude unscientific shorthand - would it be correct to say - regardless of the SR chosen - its always there, but at a 48k SR, there is a little bit more in the audile range than at a 96k SR ?

 

[Orvillain]

Even at high sample rates, aliasing can appear anywhere in the audible spectrum depending on how high the distortion-generated harmonics go.

Increasing the sample-rate:
- pushes Nyquist up.
- gives nonlinear processes more headroom before they generate audible aliasing.
- but does not remove aliasing. It only shifts where the reflections land.

Nonlinearities generate (effectively) infinite harmonic bandwidth. You still need anti-aliasing filters, even at higher sample-rates. Signals need to be low-passed in order to avoid aliasing.

 

[BenIfin]

Even at high sample rates, aliasing can appear anywhere in the audible spectrum depending on how high the distortion-generated harmonics go.

Increasing the sample-rate:
- pushes Nyquist up.
- gives nonlinear processes more headroom before they generate audible aliasing.
- but does not remove aliasing. It only shifts where the reflections land.

Nonlinearities generate (effectively) infinite harmonic bandwidth. You still need anti-aliasing filters, even at higher sample-rates. Signals need to be low-passed in order to avoid aliasing.

Thanks hugely :)

 

[BenIfin]

Even at high sample rates, aliasing can appear anywhere in the audible spectrum depending on how high the distortion-generated harmonics go.

Increasing the sample-rate:
- pushes Nyquist up.
- gives nonlinear processes more headroom before they generate audible aliasing.
- but does not remove aliasing. It only shifts where the reflections land.

Nonlinearities generate (effectively) infinite harmonic bandwidth. You still need anti-aliasing filters, even at higher sample-rates. Signals need to be low-passed in order to avoid aliasing.

Just a bit more education :)

If I'm using NAM Captures in real-time live - not recording anything - will running Oversampling in the NAM Host [ie:- Genome] reduce the NAM Capture aliasing (?) or is that un-changeably-baked-in to the NAM Capture regardless of Oversampling applied when playing live through them (?)

Thx again.

 

[Orvillain]

Just a bit more education :)

If I'm using NAM Captures in real-time live - not recording anything - will running Oversampling in the NAM Host [ie:- Genome] reduce the NAM Capture aliasing (?) or is that un-changeably-baked-in to the NAM Capture regardless of Oversampling applied when playing live through them (?)

Thx again.

I actually don't know. In theory, if it is all programmed correctly, oversampling should reduce aliasing. But I actually don't know how that interacts with neural models.