🚀 go-pugleaf

Hello

When we linearise a set of speakers, we measure a sound pressure, which is power in micro Pascals. We try to make the speakers flat, so that all tones produce the same sound pressure at the measure device. The power in Watts going into the speakers is much greater for 50hz signals than for 1000hz signals, but it produces the same pressure on the measuring device.

Similarly with microphones, we measure them by determining what sound pressure reaching the diaphragm, produces a particular voltage output. We try to make this a linear as possible.

Now ours ears are nonlinear, but we do not give the speakers or the microphones a arbitrary preemphasis to allow for the nonlinearlty
of our ears, so why give a images such a arbirary inverse gamma.

We should linearise the monitors/speakers, we hope our mics/scanners are linear and we should then apply any necessary correction to the images/audio to allow for the nonlinearities in the system including our eyes/ears.

As far as I know all digital data manipulation happens in a linear domain apart from imaging. This is a left over from the early days of
television, when a inverse gamma had to applied to a TV signal, so that it was viewable on a TV. It was too expensive to put a inverse gammma device into a every TV, TVs being nonlinear devices.

In radio a pre-emphasis is added to improve S/N, but is removed before the audio reaches the speakers. In Dobly noise reduction, the audio is given a high frequency boost before recording, but it is removed on replay before mixing and equalisation.

Doing data processing in a linear domain is not a complicated concept.
Perhaps things will never change, because of the degree in investment in the present way of doing things, but it is not correct.

Mike Engles