Super saw analysis ideas

I found a post on Matrixsynth today that lead me to a second paper analysing the JP8000 super saw. 

The post: https://www.matrixsynth.com/2021/07/korg-nts-1-nutekt-digital-synthesizer.html

The project: https://github.com/GrahamJamesKeane/UberSaw

The project page: https://korgnts1beginnersguide.wordpress.com/

Adam Szabo: How to Emulate the Super Saw: https://korgnts1beginnersguide.files.wordpress.com/2021/07/szabo_adam_10131.pdf

Alex Shore - An Analysis of Roland’s Super Saw Oscillator and its Relation to Pads within Trance Music: https://korgnts1beginnersguide.files.wordpress.com/2021/07/ananalysisofrolandssupersawoscillatoranditsrelationtopadswithintrancemusic-researchproject-a.shore_.pdf

This got me thinking - I still think the detune response looks like it consists of three linear components instead of a 11th order polynom like Adam Szabo shows in his paper (I did in 2014 as well when making my analog 7-saw module).

I really want to test my theory in detail, but that requires some automation. Here's my idea:

- Control the JP through webMidi

- Loop through all midi values for detune

- Use a js spectrum analyzer, for example https://github.com/hvianna/audioMotion-analyzer to find the frequency components. We should be able to find 7 distinct maximums (the 7 oscillators). 

- For each maximum, record the pitch and amplitude, possibly as an average over several samples

- Build tables for each oscillator/detune setting and calculate spread and pot response to see what the actual response is.

Edit: I've been looking at the Web Audio API lately, and using it directly could be a good option. It works by chaining nodes, one of them being a FFT node. I assume that can be used for finding the maximums.