linearizing PWM'd LED output for greyscales

Does anyone have a reference/suggestion for a curve to linearise the apparent brightness of a LED
with respect to its PWM drive signal. i.e. to map 256 'grey' levels to 4096 PWM ratios such that the
'grey' scale appears linear to the eye (and so things like antialiasing work well).

As the drive is PWM, at constant current, I don't think the LED characteristic itself is especially
relevant, as the instantaneous intensity doesn't change, so I'm guessing this is essentially an
optical perception question. Is there a major difference between colours (Orange, ~600nm in my case)
? I would imagine that white at least might be different to a single-wavelength source.

I've messed with some log based curves in Excel, which certainly work way better than a linear
mapping, but I'd like to try to get things closer for an app that uses antialiasing to smooth motion
on a fairly low-res bar-graph type display, and I think it can probably be less jerky than I'm
seeing at present.


.

Relevant Pages

  • Re: strength of chess computer programs vs. time
    ... computer chess "technology curve" ... whatever) mean a straight increase in play strength. ... obvious it was a curve and not linear. ... There was a 'deminishing returns' to the growth of the machine power / ...
    (rec.games.chess.computer)
  • Quantum Gravity 166.3: Comparing dy/dt = ky and y = exp(kt) As Linear vs Exponential
    ... a bounded linear operator or bounded linear transformation T has ... rather Birkhoff Causation which approximates Probable ... To try to piece together how they relate, look at the curve y = exp ... is not only a Singularity but an Infinity! ...
    (sci.physics)
  • Re: best linear fit to a curve
    ... So you wish to find a piecewise linear ... approximation to your curve. ... then its a free knot spline problem. ...
    (comp.soft-sys.matlab)
  • Re: Exposure function
    ... I scrapped the old function I had, because the slope of the linear part ... again that in semilog-x scale the curve looks quite a bit like sigmoid. ... I probably need a third parameter for the curvature of the ...
    (rec.photo.digital)
  • Re: contrast in p&s digital cameras vs film
    ... obtained on 35mm film. ... The ones taken with digital cameras usually appear to have less contrast ... curve and digital cameras are linear. ... characteristic curve comes when you develop it. ...
    (rec.photo.digital)