Modding the Kodak Reels 8mm Film Digitizer (Firmware Hack)

[omega]

Thanks for the answer, I have to use version C. And I have a problem with the white balance on this one. It goes a lot towards the blue shade.

 

[0dan0]

Thanks for the answer, I have to use version C. And I have a problem with the white balance on this one. It goes a lot towards the blue shade.

White balance is now adjustable in the 6.6 version. Just go to the where you adjust sharpness and there is a white balance control now.

 

[0dan0]

New update.
The Tri X film is a nasty one with a lot of visible grain and the development in Caffenol accentuate it.
My attention was captured by the high bitrate that the previous clip had eq. 46.7 Mbit/s, well above the 35 Mbit/s 0dan0 set.
View attachment 22945

So i tried to to reduce sharpness more, at -1.5 and -2. At -2 I hit the sweet spot of 34.6, loosing in definition but removing the jitter effect quite completely.
The clip now is usable, a little muddier but usable. Maybe there will be some software improvements in the future.
View attachment 22946
I also removed the ghost images converting the clips to monocromatic in Davinci Resolve.

I digitized again a color film, (an Ektachrome 40 developed in 1975), with a bit ate of 22.3 Mb/s, sharpness -2. The image suffers from jittering in the upper part of the frame and is also visible the ghosting.
I'll do some more tets.
View attachment 22949

If there is a relationship between compression and the jitter, it would only be compression size of the first few frames. As I've noticed that once the jitter starts is it always there, which which why I found setting the short shutter helpful (like it moves the jitter out of frame.) You can test the compression theory by starting the encoding on blank leader. The compression is adjustable. I set a start QP factory and a maximum QP. I think is starts around 24 and slowly increases to 18 or 19. If there is a buffer overflow warning, I drop the QP back to 24, it does this continuously for a crude rate control. I'm not trying for 35Mb/s, but rather the high quality while keep the capture reliable. For the high grain source these might both be too high. So rather than changing the sharpness, change the QP limits. There is some documentation in the binaries.

 

[omega]

White balance is now adjustable in the 6.6 version. Just go to the where you adjust sharpness and there is a white balance control now.

I know, but even at 2 it's not enough for some films.
Otherwise the picture is nice and would probably be even better if I could check the sharpness. But I can't get the lens cap off and I don't want to take the whole device apart just yet.

 

[domb84]

Have you found this for Reels scanners?

I can't say I've tried to be honest as I do not have one. From my comparisons on the other similar scanner firmwares it seems to be a common file.

 

[ThePhage]

Under the picture setting "Exposure" is replaced with "White Bal", for a manual, but fixed, white balance control. You should not need to reflash FW to change the white balance any more.

White Bal. values to their RGB gains

White balance	Red gain	Green Gain	Blue Gain
2.0	1.75	1.25	1.0
1.5	1.5	1.125	1.0
1.0	1.375	1.0	1.125
0.5	1.25	1.0	1.25
0.0	1.0	1.0	1.0
-0.5	1.0	1.0625	1.125
-1.0	1.0	1.125	1.25
-1.5	1.0	1.1875	1.5
-2.0	1.0	1.25	1.75

I'm finally starting some test scans utilizing the v6.6 Manual User Controlled White Balance setting to determine the best setting for my old films (8mm from 1950s & 1960s. I have a couple of observations that relate to @omega's statement above about +2 not being enough:

With 3 brief tests (WB: 1.0, 1.5, 2.0) this footage requires varying amounts of post color corrections. For the purposes of evaluation, the color correction was applied using an Auto-Color feature in Adobe Camera Raw that adjusts Temperature and Tint, and provided the following read-outs:

WB 1.0 required increasing the Blue/Orange Temperature slider to +50 (out of -100 to +100)
WB 1.5 required increasing the Blue/Orange Temperature slider to +29 (out of -100 to +100)
WB 2.0 required increasing the Blue/Orange Temperature slider to +12 (out of -100 to +100)

All 3 (WB 1.0, 1.5, and 2.0) required increasing the (Green/Magenta Tint slider to +25 (out of -100 to +100)

For visual comparison, here are the 3 test scans (with no post color adjustments), and I also included a previous scan using Mac84's firmware (with the stock firmware's Auto White Balance).

I know almost nothing about color science, except that it's not straightforward. And I imagine that old and faded (and amatuer!) film will have a wide range of color issues. That being said @0dan0 I'm wondering if your RGB gains could use further adjustments. 2 specific suggestions based on my very unsofisticated test:

1. Proportionally increased Red gains
2. Decreased Green gains

These would apply to at least the WB 1.0, 1.5, 2.0 options I tested. Unsure what might be helpful for WB below those.

Thankful for all the work you've done!

 

[0dan0]

I never encountered film that blue, and I was guessing what WB values might be useful. I'm away from any scanners at the moment. And the table of histogram values are compiled C code, making remote debugging more tricky, but I think I found the RGB gain for WB 2.0.

0x1c0 - 1.75
0x140 - 1.25
0x100 - 1.00

As these gains results in too much green, try changing the "40 01" to "00 01". If that is too little green, try "10 01". This is hard to get perfect with a single axis WB control, would be improved if it was separate kelvin and tint controls. I could hijack the the "tint"/saturation control. I will think on that.