Digika wrote: ↑29 Oct 2020, 19:56
Chief Blur Buster wrote: ↑25 Oct 2020, 19:11
I also want manufacturers to use overdrive sliders (100 settings) rather than 2 or 3 or 4 Overdrive settings.
I think the reason they dont do that is to reduce the amount of testing, if you have set up of 20/60/80/100 you just have main 4 passes, but with step of 1, 1-100scale that becomes 100 main passes. Takes crapload of time.
<Programmer POV>
Since you're a software developer I think....
The Overdrive Algorithm
How overdrive is generated is that they run photodiode oscilloscope on many passes to find correct overshoot colors. For example using grey value 220 to speed up a transition from grey 50 to grey 200. That's an voltage-level overshoot of +20 for a specific color combo without showing a vision-level overshoot ripple on an oscilloscope. That's the "strongest visually overshootless overdrive" you tune for -- the biggest voltage overshoot you aim for that doesn't produce a visible/measurable photon-level overshoot. There's 3 separate monochrome pixels per color pixel, so overdrive is processed 3 times independently for each color channel, since overdrive only cares about the monochrome transparency of the pixel in the LCD layer, without regards to the color filter layer. So you have one LUT that can be used for all color channels.
The Overdrive Lookup Table
You measure this for all grey combos to create an
Overdrive Lookup Table of A(B) = C where A is source grey, B is destination grey, and C is the grey to use to accelerate transition from A to B. There are 65536 values in a 256x256 OD LUT on an 8-bit panel, though most manufacturers use a subset such as 17x17 OD LUT and use bilinear interpolation to generate the rest of the LUT (256x256) to save scaler memory. Ideal OD LUTs is generated on a per-Hz basis, calibrated at 20C room temperature. Also, reduced-resolution OD LUT is a big problem in the 240Hz+ and 360Hz+ refreh rate race because of stronger localized hotspots in the GtG heatmapping, which would benefit from finer-resolution original OD LUTs. VA panels, in particular, benefits significantly from larger-sized OD LUTs, combined with an OD Gain, to help better mitigate the dim-colors VA ghosting.
The Overdrive Gain
Overdrive Gain is a simple multiplier to the delta (difference of B and C in the A(B)=C formula) in the OD LUT. The middle setting is usually a 1.0 multiplier. So the overshoot is a +20 there. Now if you slide the slider lower, the multiplier may become 0.6. So your overshoot becomes only +12 in that case. Now if you slide the slider higher, the multiplier may become 2.0. In this case the overshoot becomes +40.
Deviating from neutral (visually overshootless) overdrive, will produce more ghosting or more coronas. And if temperature changes, the calibration may require a slight overdrive gain shift to compensate for temperature or panel lottery effects. Also, values outside the gamut is clipped (e.g. clipped to 0-255). Also, Overdrive LUTs can be multiple refresh depth if need be (e.g. needing to clamp a pixel to color value 0 for two or three refresh cycles to speed a transition from greyscale 50 to greyscale 5, which is unusually slow on a VA panel during cold temperatures, so multiple-refresh-cycle overdrive can be a big benefit for such panels. But it's more expensive to engineer for. However, Overdrive Gain is a simple engineering move, and it is frankly, it is surprising that manufacturers have been resistant to letting users adjust Overdrive Gain.
Overdrive Gain is absurdly simple mathematics, and that's why an overdrive slider is simple Math 101 stuff.
</Programmer POV>