I should have mentioned that indeed, I only expect something like this to be visible on a totally unlit (black) section.lidnariq wrote:Fair enough! I tested with Galaxian, and see 3.7 halflives per 1/60th of a second, or a halflife of 4.5ms. This means that the correct constant (with this TV) for the expression I gave above is k=1/13. This is low enough that on the NES, the effect will only ever be visible when transitioning to black pixels. So here's an animated gif, simulating same:It's really subtle, especially without any of the phosphor size blur. You'll also want to enlarge it; I can't see anything when it's at 100dpi.
And after manually capturing 17 frames from FCEUX, here's how I processed it:Note that I'm fixing up the gamma; if I hadn't, it'd be even harder to see.Code: Select all
for i in `seq 0 16`; do pnmarith -maximum $i.ppm previous.ppm > n$i.ppm; ppmtogif n$i.ppm > n$i.gif; pnmgamma -ungamma 2.2 n$i.ppm | ppmdim .077 | pnmgamma 2.2 > previous.ppm; done pnmcat -tb n*.ppm | ppmtogif > all.gif gifsicle --use-colormap all.gif -O3 -V -o anim.gif -d2 n{0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16}.gif
I suspect that standard 24-bit displays are not actually deep enough to show this in a compelling way; after two refreshes (1/13)² even full scale content is just 1 LSB.
As the phosphor is no longer being struck by the electron gun, the "Starting brightness" of the fade is very dim, but the "decay" of the dim trail is very slow.
On mine, I can see the trails last for almost two seconds, but the curve is something like this (please excuse the graph quality):