GLCD overclocking (and awful contrast consistency)

nick12ab

Senior Member
Ages ago in my thread Serial Controller for KS0108B GLCD I mentioned the GLCD flickering with the (supposedly) inferior choice of resistor. Overclocking is performed by changing the resistor. I finally decided to attempt overclocking of the GLCD and found that there are good reasons why the manufacturer selected the slower speed.

Why would the manufacturer choose the slow speed when the faster speed would eliminate the flicker? Here's why:
  1. Slightly less power consumption
  2. Magnitude of contrast voltage required reduced
  3. Contrast consistency better
That last one is the biggie. The faster the controller is running, the more awful the contrast consistency is. The contrast consistency of the GLCD at stock speed isn't great so it's best not to make it worse.

By 'contrast consistency', I'm referring to the following:
  • How little/much the contrast level 'changes itself' according to the amount of lit pixels on the display.
  • The horizontal and vertical ghosting effect caused by many lit pixels on the same line.
Overclocking the GLCD makes both these worse. So some slight flicker when all the pixels are lit is better than this. Picture (GLCD sold by Rev-Ed) shows all.

View attachment 14107

What the picture does not show is that if the display has most of its pixels lit the unlit pixels are also lit due to the severe contrast inconsistency. This would mean that if a display running at a high speed was used it would require constant adjustment of the contrast control. The viewing angle was also marginally worse as a result.

When the display was massively underclocked, the contrast consistency was brilliant but only part of the display was lit at any time and it had a scanning effect so was useless. The chosen resistor value must be a compromise between the problems caused by lack of speed and too much speed.

So if you have bought a GLCD and found that it had unusually poor contrast consistency even after correcting the viewing direction and fine tuning the contrast voltage, it might just be that the cheap Chinese manufacturer decided to use a lower value of resistor without investigating the consequences.

Incidentally I had similar results with the LCD when overclocking a Gameboy (where LCD refresh rate is controlled by the same clock signal as the CPU and for that reason I don't think its LCD will work with PICAXE). This shows that the problem is not just limited to KS0108-based GLCDs and can apply to GLCDs using other controllers too. I also tried it with an AMPY 2001-11 character LCD and the effect was there but much less significant; probably because of the much better duty cycle in character LCDs (1/16 instead of 1/64).
 
Top