Available July 26, 2006

New and improved IntuitQ and two very special additions. "intuitMatch" and a Harmonic Cursor which shows you the fundamental frequency along with the corresponding 9 harmonic frequencies.. This version is off the hook!!

There's only one intuitQ button, but now there are two extra cursor tools. The intuitQ and intuitMatch tool allows you click and drag over a frequency range you want processed and a harmonic tool which is a data inspection tool

The intuitM button allows you to match your track to any reference you load with just a push of a button. There are a number of things this is useful for. Especially when you have a nice kick drum that you want to match or even a vocal that was done in a difference studio by the same vocalist or any instrument for the matter. You can match part of the track or the entire spectrum of the track to your favorite commercial track. This addition was added by popular worldwide demand.

The previous implementation simply used frequency domain convolution (filtering) to smooth the spectrum and come up with the desired shape, then took the difference between actual and desired and smoothed that to get the har-bal filter frequency response. The problem with that approach is that simple smoothing comes up with a reference shape that can be markedly different to the original such that the original doesn't overlay it (hence it is biased).

The new approach is built on the same process of building a smoothed reference from the original spectrum but in this case it is realised using a cubic spline of limited resolution (ie. few node points used to cover the spectrum) that is fitted to the original by minimising the mean square error between it and the original. That in itself provides some smoothing but it doesn't get rid of the wide valleys or broad hills that should be reduced to reduce masking.

The way that is handled is by adding a curvature penalty to the least square fit. What that basically means is that it finds the best fit cubic spline with least bends in it. That is what fills in the holes and shaves the tops of the mountains. Because it is a least squares fit you'll almost always find that when you look at the frequency response it realises the boosting and cutting is evenly spread across the entire frequency range. By that we mean that the frequency response across the spectrum hangs around 0dB. That is what makes it sound tonally neutral with regard to production intent.

The other respect that this version differs from the original is the smoothing applied to the difference. In the original we used smoothing based on a 2 octave band coverage (ie. low Q or roughly equivalent to a 5 band graphic EQ) whereas this version uses 1 octave band resolution (ie. higher Q or roughly equivalent to a 10 band graphic EQ). It seems counter intuitive going on what Pro mastering engineers say (ie. only use low Q parametric filter changes) but the higher Q 1 octave band resolution sounds superior to the lower Q case in the testing we've done.

A lot tighter and truer to the original. It seems the whole problem with higher Q is not that it is bad but that it is very difficult for people to set the correct parameters thorough listening.

We've prototyped the intuit Match feature and the algorithm worked as expected! In fact it works exceptionally well.

Cheers,
Earle and Paavo