Apart from the discussed gate driver details, which are just O.K. in my view, the circuit in post #35 misses any means for effective PWM operation with a battery load, particularly a storage inductor and a diode.
As shown so far, it's using only the batteries internal resistance to translate duty cycle into average current.
As I understand it, lighty is charging up the boost regulator's output C to a high value, turning off the pwm, then dumping the C into the battery. This has the effect of "burp" charging the battery. During the resting interval, excess charge in freshly charged sites in the battery is allowed to drift away from the conduction areas into other less-conductive areas of the cell, increasing the likelihood that those areas will be able to be charged. In many ways this is like cleaning a dried spot of food on the counter: if you put a little water on it, and then come back and rub, you accomplish the cleaning with a lot less abrasion to the surrounding area.
I think FvM reffered to the Wizpic circuit as shown in the post #35.
As for my intentions you're correct. That is exactly what I'm doing. You also forgot to add, that fast, high-current discharges help desulfation of the plates. ;-)
Well, the concept is nothing new. I simply want to make it automated as much as possible and more safe.
BTW - I finally decided to use external ADC (22-bit DeltaSigma, single-ended). It should provide me a with a very high resolution and since I'll be measuring voltage of battery between discharge pulses I will just add some low pass filter to prevent possible ringing after the discharge. Averaging is something I also consider adding into code to provide me with stable values as well as omitting 6 LSB bits. It's a slow going process but I'm doing it in my spare time (I don't have much of that) so it is to be expected.
Good plan. You can never have too much data. With resolution that high, you might be able to detect the incremental increase in cell activity after each pulse. Definitely worth running a bunch of batteries and plotting the deltaV per pulse.