I jumped on the 96khz bandwagon a long time ago for a variety of reasons commonly cited by those in the pro audio community. One of those reasons was the absence of aliasing in the audio band when using higher sample rates. Recently the topic came up again and the scientist in me wondered how much of a problem aliasing really is at 44.1kHz, particularly when it comes to Mixbus's built in tape saturation.
Aliasing is commonly demonstrated by introducing signals of higher frequency than nyquist, either directly or by a high frequency below nyquist run into a clipper or saturator. As the frequency is increased, you can watch the alias component "walk" across to the left on a spectrum analyzer. The problem with this sort of demonstration is that the level of the original high frequency signal is much higher than what would be typically encountered in real use (anything that is not a guitar amp sim).
For example, if you look at the spectrum of any reasonably balanced mix, the signal level drops as frequency is increased. Running this mix through the tape saturation in mixbus, its the lower frequency energy acting on the saturation stage. Without the low frequency content, there is not enough energy in the high end to even get an indication on the saturation meter.
An example:
For this experiment I chose to use "Peg" by Steely Dan. Partly because of the clarity of the track and it had been a long time since I've listened to it. I loaded this track into Mixbus and routed the output to a Mixbus only. I left the tape saturation at its default setting of -10. I adjusted the track fader so that the saturation meter averaged a"straight up" indication. I then loaded a spectrum analyzer on the output of the mixbus.
I created another track and loaded a sine generator set to a high frequency (<11kHz) that was certain to generate an alias component when clipped by the saturator. This track was fed to the same mixbus as my reference music track. The level of this track was adjusted that it just poked into the spectrum of the music track while playing. The test signal was now at a real world level.
Muting the music track, there was no alias component. Theoretically its there, but its below the noise floor. increasing the tape saturation from -10 to +10 caused the alias component to appear, but down 72dB from -18dB(0VU)! -72VU can be audible in a very quiet room but given the amount of saturation required, you'll never hear the alias in the context of music.
Heres why: Say you had a mix playing and the saturation meter is straight up as in the example above. Increasing the tape saturation by 20dB would then drive the signal into oblivion, something you would only do for some sort of effect. The alias that exists at -72VU wouldn't even be discernable as the original signal would be barely recognizable anyway!
In real life, audible saturation will occur at low frequencies long before the high frequencies. Unless you like to crank up the saturation on your high-hat tracks, then you would probably hear aliasing.
Did the same test with the limited set of plugins I like to work with. Some had no problem at all, the worst had an alias down -60dB when pushed with HQ set to off.
For me, the scale has tipped back the other way in favor of working at 44.1kHz (less disk i/o and CPU).
Aliasing is commonly demonstrated by introducing signals of higher frequency than nyquist, either directly or by a high frequency below nyquist run into a clipper or saturator. As the frequency is increased, you can watch the alias component "walk" across to the left on a spectrum analyzer. The problem with this sort of demonstration is that the level of the original high frequency signal is much higher than what would be typically encountered in real use (anything that is not a guitar amp sim).
For example, if you look at the spectrum of any reasonably balanced mix, the signal level drops as frequency is increased. Running this mix through the tape saturation in mixbus, its the lower frequency energy acting on the saturation stage. Without the low frequency content, there is not enough energy in the high end to even get an indication on the saturation meter.
An example:
For this experiment I chose to use "Peg" by Steely Dan. Partly because of the clarity of the track and it had been a long time since I've listened to it. I loaded this track into Mixbus and routed the output to a Mixbus only. I left the tape saturation at its default setting of -10. I adjusted the track fader so that the saturation meter averaged a"straight up" indication. I then loaded a spectrum analyzer on the output of the mixbus.
I created another track and loaded a sine generator set to a high frequency (<11kHz) that was certain to generate an alias component when clipped by the saturator. This track was fed to the same mixbus as my reference music track. The level of this track was adjusted that it just poked into the spectrum of the music track while playing. The test signal was now at a real world level.
Muting the music track, there was no alias component. Theoretically its there, but its below the noise floor. increasing the tape saturation from -10 to +10 caused the alias component to appear, but down 72dB from -18dB(0VU)! -72VU can be audible in a very quiet room but given the amount of saturation required, you'll never hear the alias in the context of music.
Heres why: Say you had a mix playing and the saturation meter is straight up as in the example above. Increasing the tape saturation by 20dB would then drive the signal into oblivion, something you would only do for some sort of effect. The alias that exists at -72VU wouldn't even be discernable as the original signal would be barely recognizable anyway!
In real life, audible saturation will occur at low frequencies long before the high frequencies. Unless you like to crank up the saturation on your high-hat tracks, then you would probably hear aliasing.
Did the same test with the limited set of plugins I like to work with. Some had no problem at all, the worst had an alias down -60dB when pushed with HQ set to off.
For me, the scale has tipped back the other way in favor of working at 44.1kHz (less disk i/o and CPU).