[LAU] Log F-scale for JAPA

[Fons Adriaensen]

Hello all,

I've been asked (in a private mail) if I would provide 'upstream'
support for the logarithmic F-scale patches for JAAA which have
been discussed recently on this list.

The answer is maybe, but certainly not in their current form.

1. As a first and sufficient reason, I don't want to develop
   the current codebase any further. It's a glorious mess, and
   in order to add some other features I want to add it needs
   to be redesigned from scratch or at least completely refac-
   tored. 

2. The current log X-axis grid and annotations are a no-go.
   I pride myself on getting such things (including slider and
   meter scales etc.) right, even if that often takes more work
   than the actual DSP code. Very few tools get this right, even
   Gnuplot's log scale is useless in many cases (try zooming in).
   A decent log F-axis needs a lot of ad hoc code (which I mostly
   have, from  other projects). It needs to support arbitrary
   zooming, and both a technical (i.e. based on a 1, 2, 5, 10
   sequence) and a musical (based on octaves, as in JAPA) format.

3. JAAA's original purpose was to measure organ pipe spectra
   (it was presented at the same LAC as Aeolus IIRC). More
   generally it is a technical tool, meant for for electronic
   or acoustic engineers, not for sound engineers doing their
   normal job of recording or mixing music. As a technical
   tool it must provide results that are both as accurate as
   possible and well-defined. Well-defined means results that
   correspond to quantities, equations, etc. that are used in
   electronics, acoustic and DSP theory, not necessariy the
   most relevant ones in a musical context. For such technical
   uses the most convenient presentation is on a linear F-scale,
   for the simple reason that the analysis performed by JAAA
   (essentially a DFT with some extensions) is defined in terms
   of a linear frequency domain.

4. In those cases where a logaritmic frequency axis presentation
   would make sense (the classical 'marketing style' frequency
   response plots), JAAA is simply not the right tool. It does 
   not provide enough resolution at LF, or tons of irrelevant
   and misleading detail at HF. And anyway, you can't measure
   speakers, mics, etc, with continuous signals (as expected by
   JAAA) unless you have a large anechoic room at your disposal.
   And if you really want to do such measurements, the tool the
   use ATM is JAPA with either a slow logarithmic sine sweep or
   pink noise. There *is* a lack of good Linux based tools for
   this sort of thing, but the cure is not a log F-scale for
   JAAA but some extra features in e.g. Aliki.

But of course I'm looking at this from my perspective. If there
are any people out there using JAAA _and_ wanting a log F-scale
I'd like to know what exactly they are doing. It may not result
in a log F-scale for JAAA, but maybe in something better.

Ciao,

-- 
FA

A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)