ledmatrix: gamma correction for smoother brightness ramps#99
ledmatrix: gamma correction for smoother brightness ramps#99cbiffle wants to merge 3 commits intoFrameworkComputer:mainfrom
Conversation
|
FWIW, I'm not at all sure that I hooked the gamma correction at the right points in the code --- definitely open to guidance if you'd like it moved! |
|
It's also worth calling out that gamma correction makes all mid-range brightnesses look dimmer (by design!). So it might be good to increase the default brightness multiplier (currently 51 on PVT looks like?) to something higher. I'm happy to find a value that matches the approximate intensity of the original firmware if you'd like. |
|
This change reminds me of what FastLED does. Great idea and work. Can I use this code in my fork of the repo? |
|
Would appreciate progress on this change! I also noticed this issue. |
This costs ~256 bytes of flash (once referenced). I've tried to keep the generated code readable.
The brightness values sent to the LED controller actually control a PWM duty cycle. LEDs emit light roughly in proportion to the PWM duty cycle, but human vision perceives brightness on an exponential curve -- generally it takes 2x the physical luminous flux for the eye to perceive something as one step brighter. As a result, brightness ramps (like the one generated by --all-brightnesses) were rapidly brightening to what appeared to be max, and then flattening. This change adds gamma correction, which maps the linear input brightness values to a non-linear function of PWM duty cycles, causing the ramp to look far more ramp-y. The gamma value I've chosen here is arguably a preference, I messed with output on a PVT panel until I found something that looked approximately right.
With gamma scaling, 51 is very dim, need to increase the default brightness to be able to see the bootup animation. Signed-off-by: Daniel Schaefer <[email protected]>
JohnAZoidberg
force-pushed
the
ledmatrix-gamma
branch
from
0e2480b to
4dbdac6
Compare
|
I rebased on the main branch and I tried it out. I noticed that the default brighness was way too low with this change, so I increased it a bit.
On my system everything below 37 is off. Everything below 51 is basically off, the LEDs have some variance, some are off, some are extremley dim. |
4 participants
The brightness values sent to the LED controller actually control a PWM
duty cycle. LEDs emit light roughly in proportion to the PWM duty cycle,
but human vision perceives brightness on an exponential curve --
generally it takes 2x the physical luminous flux for the eye to perceive
something as one step brighter.
As a result, brightness ramps (like the one generated by
--all-brightnesses) were rapidly brightening to what appeared to be max,
and then flattening.
This change adds gamma correction, which maps the linear input
brightness values to a non-linear function of PWM duty cycles, causing
the ramp to look far more ramp-y.
The gamma value I've chosen here is arguably a preference, I messed with
output on a PVT panel until I found something that looked approximately
right.
Photo of a panel with the firmware from
main(left) vs this change (right):