Thanks for the plotting tips. I was wondering how to do that.
The red and blue curves in your plot are as you describe, only they were timed exposures. The latest exposure was done with a PPM-2 using the standard dose determined using your method.
Looking back on my previous difficulties linearizing, I realize now that exposure differences were probably a big factor in failures to converge. In my recent tests with the PPM-2, I found as much as 1/6 stop difference between time-based exposures. That much difference could flip a smoothed curve from convex downward (overexposed) to convex upward (underexposed). I saw that that pattern a lot, and I blamed QTR-Linearize-quad, but QTR-Linearize-quad was probably doing a good job. But when I printed a target with the linearized curve and then (say) underexposed it, the curve flipped over the diagonal. The effect of exposure overwhelmed the effect of linearization.
It would be useful to model (mathematically) the effect of fractional stops of over- and under-exposure on a straight line. It’s just a gamma function, I think.
So maybe the conclusion is: Linearization is extremely sensitive to variations in exposure. If you’re linearizing, you must use a dose meter.
I made a new target negative with master-lin. I will print a target tonight and measure it tomorrow morning.