Blackmagic Pocket Camera 4K Colour Fringing and colour issue

Hi Guys,

I've with pleasure received my BMPC4K.
I've straight away grabbed some shots and started noticing problems.

It seems for some reason the colour of the screen vs reproduced colours as a bit off, that's reasonable given that this is a low segment camera. the worrying thing is the rendering of the color red. this obviously being a hard colour to reproduce for sensors in the past, for example FS700 had the same problem and other older 8bit camera's, but to see it in the pocket camera is something i wouldn't have expected, the color itself has fringes and seems to not be calibrated correctly given extreme contrasty annoying edges.
I have some samples, obviously the color is a bit overexposed, but shouldn't reproduce in this way, it just seems like the whole red channel causes fringes and strange colour casts around highlights.

wetransfer link to grabs:
https://we.tl/t-8ju7ISHAvw

Is this a problem regarding my unit, or is this a wider known problem?

Cheers,
Sico

That is a result of clipping and colorscience version 4 and is pronounced in high ISO gain because there the camera has a reduced dynamic range. But I admit the camera should clip more gracefully here.

If you search web for "Pocket 4K red clipping" you can see a lot of reports about this problem.

Looking to these and other samples provided by other users i guess that red clipping may be caused by not too large enough color gamut used in Color Science v4. Original BMDfilm LOG files are low saturated, if you compare saturation level it appears that BMDflm gamut is even larger than Rec2020. BMDfilm is a native sensor color gamut.
P4K unprocessed LOG samples looks way more saturated, it feels like closer to DCI P3 gamut, or variations of early REDcolor/DRAGONcolor versions slightly larger than sRGB. Extreme saturated light emitted objects like neon lights, lasers, electric sparks, single color LEDs require VERY wide gamut to be captured fully without clipping. My guess that with Color Science v4 BMD decide to unify gamuts between different cameras (as many manufacturers do) and add a in-camera conversion from native sensor gamut to some kind of optimized Color Science v4 gamut. Smaller gamut may produce less global color shifts, it is easer to compress to Rec709 and probably allow more precise sensor color calibration. P4K footage with Color Science v4 looks really well calibrated compare to older cameras. But as side effect it may produce strange color clipping. Hope BM will be able to fix it somewhere in future updates.

There is a fix for this:
- Add Color Space Transform node in Resolve.
- Set in and out to same Gamma/Gamut
- Turn ON Saturation Mapping
- Adjust Saturation clipping settings.

Thanks for the tips.

Red is in fact shifted towards more Purple/Redish colour, is not red anymore after recording.
i remember this problem in early c300 models, they also had fringing in the highlights of certain lights such as Fluorecent lights and other units.

I'll make an official ticket, because most likely this will be fixable in a firmware upgrade with a more narrow calibration, most likely all bright LED and neon like colours will be flagged with this issue, like you posted earlier.

for the price it's neglectable, but would be epic if fixed.

P4K unprocessed LOG samples looks way more saturated, it feels like closer to DCI P3 gamut, or variations of early REDcolor/DRAGONcolor versions slightly larger than sRGB.

Use the cst plugin in resolve and transform from broadcast color v4 (which is said to be for pocket 4K as well) into arri wide gamut and you'll see they're almost exactly the same. So I think the gamut is almost exactly the arri gamut, but the pocket 4K doesn't have anywhere near the arri dynamic range and clips faster. Especially in those low light scenes of cars at night where people are using the higher analog gain setting that has even less dynamic range.

It's similar but reds drift towards magenta and don't seem to have any rolloff at all.
I don't think it's supposed to be like that. Other cameras with the same sensor (and same horizontal noise pattern) don't show this issue.

Interesting.

Is this just a ProRes / high ISO Problem or also present in ProRes/low Iso or RAW?

Cheers

Stumbled upon this.

Seem like a youtuber by the name of “studio sumizoon” have come up with a color compensation filter as a solution. The idea is to use a Hoya filter by the name of CD5000.

This filter is normally used for ccd cameras enhancing the colour distribution in the visible spectrum while eliminating any UV and near IR transmission. More or less like a UV/IR cut filter it may seem, but this one does not cut as fast in the low end of the wavelength spectrum and seem to fall down in a nice curve compare to a normal IR cut filter. Im guessing that’s what causing the nice charger light rolloff that we can see in the video they put out. Take a look about 09:00 – 09:50 into the clip.

What you also see is that it fixes the problem with the 1000-1250 iso switch occurring with no IR cut filter. So I guess this one can replace a normal IR cut filter.

Normal IR cut filter:

img_deposition_filter_4.jpg (16.89 KiB) Viewed 132 times

CD5000 color compensation filter (IR-CUT):

img_graph_7.png (55.67 KiB) Viewed 132 times

Here is more information about the filter.. one problem is that it doesn’t exist like a screw on filter as I can tell. http://www.hoyacandeo.co.jp/english/products/eo/color/07.html

If anyone have some experience with it, than I’m all ears.