[4.6K] Need other opinions, is this normal FPN ?

Hey everyone,

Loving my camera so far, 4.0 is a blast. But I think I'm noticing FPN at ISO1600 on my 4.6k.
(it was there before the update btw)

I know this camera needs light (..."blabla"..."not a low light camera"..."not an a7s"). No need to say it all again.

But this is my first experience with a camera outside the dslr/mirrorless world, and I just want to make sure this is normal or not, since the 4.6k is supposed to be better than the 4k on the low light side, and what I'm seeing is just ugly.

edit: check the wetransfer video below

Is this normal ? Maybe someone from blackmagic here can tell me if something's wrong or not ?

Thank you in advance

EDIT: did another video, original 4.6k 2.40 ProRes 422 here: https://we.tl/HF0fIOvkB1
EDIT²: a couple DNG's here: https://we.tl/arm0YMC7Qt

For what it's worth, mine is about the same, if not worse. I can see FPN at ISO 800 if I go looking for it but of course that's without proper lighting. I wish there was a BMD-recommended test we could perform, but that's opening up a can of worms I guess. But some have said they can't provoke FPN even if they try, I find that pretty strange. Difficult to have a meaningful discussion about that I guess unless there's testing method everyone can agree on.

For the final answer, contact support and send them some DNG:s.

For what it's worth, mine is about the same, if not worse. I can see FPN at ISO 800 if I go looking for it but of course that's without proper lighting. I wish there was a BMD-recommended test we could perform, but that's opening up a can of worms I guess. But some have said they can't provoke FPN even if they try, I find that pretty strange.

For the final answer, contact support and send them some DNG:s.

What you described is about the ursa mini I have. I think it's normal behavior, and maybe some are a tad cleaner, I don't know...

Sent an email to the BM Support, I'll keep you guys updated.

For what it's worth, mine is about the same, if not worse. I can see FPN at ISO 800 if I go looking for it but of course that's without proper lighting. I wish there was a BMD-recommended test we could perform, but that's opening up a can of worms I guess. But some have said they can't provoke FPN even if they try, I find that pretty strange.

For the final answer, contact support and send them some DNG:s.

What you described is about the ursa mini I have. I think it's normal behavior, and maybe some are a tad cleaner, I don't know...

Yeah, unless I see some real issues in "normally lit" low light I'm not too worried.

Actually, what I'm seeing in your files looks a lot like the scaling issues that have been haunting some of us. See this thread for more:

https://forum.blackmagicdesign.com/viewtopic.php?f=2&t=50175

I see that too on this file with VLC, but not in Resolve with the settings that work best for my monitoring.

Scaling issue shows a grid pattern in videoplayers or NLE (I see it in Media Player Classic when not fullscreen). I don't have that in Premiere Pro, and it's not the problem here.

It's definitly FPN, you can check the DNG's in Photoshop but it's more noticeable in video.
Also, I can see it on the camera screen directly. Just want to know if it's normal or not.

Yeah, unless I see some real issues in "normally lit" low light I'm not too worried.

Normally lit or not, that FPN pattern should never appear in darker parts in my opinion.
Grain noise, of course, but not straight lines like that :/

Just looked at the clip, at first I had the scaling problem like Jamie mentioned, so I resized the Quicktime window and that went away.

I can defiantly see some fixed pattern noise too though, will be interested to hear if BM class that as normal!

Valentin from what I can see the RAWs look fine in ACR... Quicktime has a much higher gamma and looks washed out I've noticed which only amplifies them problem noise as well as artifacting and band banding created by the export... they dont seem to be on the RAWs afaics...

Here's a comparison of Quicktime vs Windows 10 in-built video player

Mov vs WMP.jpg (982.77 KiB) Viewed 5041 times

Just checked your .dng's. Definitely FPN, but with a S curve most of it goes away.

Do you also see it at 800? I just did a quick test and mine is about the same, maybe ''worse''. I'm interested in BMD's reply back as well.

Yep, mine looks the same but generally I only see FPN on ProRess files.
I'd also be very interested to know if this is considered normal.

Just got my answer from BM, apparently it's totally normal.

Hello Valentin,

thanks for your email,

for the lighting condition that you have been using this is considered normal. The camera requires more light to return a less noisy picture and you should consider using a ETTR (expose to the right) method to then achieve your look in a colour grade post process.

I hope this has helped.

Hum, well... Okay... Thought the camera would perform better than that.

heres a low light video I made... its at 1600 iso... u can see some noise...

i unfortunately have to return the camera because slot b doesn't read cards... and I'm nervous that Ill get a worse sensor? a lot of ppl seem to have fun problems

also take this into consideration... what card are you using? I noticed that going past what your card can handle introduces noise also... I can't shoot process 444HQ on my card w/o it looking worse that prores 444

also take this into consideration... what card are you using? I noticed that going past what your card can handle introduces noise also... I can't shoot process 444HQ on my card w/o it looking worse that prores 444

Lexar 3400x 256Gb, didn't experience any issue with them.

@FrankApollonio: I saw your video before buying the 4.6k, looking for lowlight footage.
Honestly it looks good imo !

what settings are you shooting at just curious?

In the videos of my first post, 23.976fps, shutter 180°, Sigma 18-35 @ f1.7
4.6k 2.40 ProRes 422 for the WeTransfer video

Hi Valentin, I know u're in Belgium. Just curious why do you shoot 23.976fps?


Sent from my iPhone using Tapatalk

I shoot only for the web

Right


Sent from my iPhone using Tapatalk

I've yet to do a definitive test, but i believe most of my FPN i get with the 4.6K Mini goes away when i "reboot" my sensor, but turning the camera off-then on.

Something about sensor calibration.

I've yet to do a definitive test, but i believe most of my FPN i get with the 4.6K Mini goes away when i "reboot" my sensor, but turning the camera off-then on.

Something about sensor calibration.

Interesting indeed


Sent from my iPhone using Tapatalk

Hm.. besides the FPN (which isn't too bad imho) the DNGs definitely show this weird "maze" (bayer?!) pattern at 1:1 zoom. This would only become a "scaling issue" depending on which scaling algorithm/filter you use (or your viewer/app uses) as it becomes even more pronounced then, but it's definitely there before any kind of scaling is involved!
Checked with other debayers (Adobe etc).

Others have reported it before as well. Personally I've only seen it a few times with the Pocket when there was a flaring high brightness light source in view (which is also the case here)....

So is this some kind of sensor issue in general with this sCMOS family?
Or because of this particular combination of no LPF whatsoever + it's microlens/bayer array + certain lighting situations?

Here is a little 1:1 crop.. 3x magnified and exposure/contrast up to make it more visible, but it's quite visible already without upping exposure that much. Would be nice to get an explanation on whats going on there...

I wonder if this is what the 0.6K is for. When downscaling to UHD or 4K a lot of the maze artefact should be eliminated. Of course, that means there's less incentive to use windowed HD for example.

Blackmagic has obviously chosen this debayering algorithm over several other algorithms and perhaps they're after the sharpness it provides compared to algorithms that are completely maze-free. Just speculation on my part.

There's an app called CornerFix that can alter the BayerGreenSplit parameter in CDNG files. By using that I was able to get a (ever so slightly softer) maze-free image. However, DaVinci Resolve doesn't read the BayerGreenSplit parameter for some reason. A few preview apps I tried do, but I haven't tried the Adobe apps.

There's an app called CornerFix that can alter the BayerGreenSplit parameter in CDNG files. By using that I was able to get a (ever so slightly softer) maze-free image. However, DaVinci Resolve doesn't read the BayerGreenSplit parameter for some reason. A few preview apps I tried do, but I haven't tried the Adobe apps.

Yeah I remember that parameter from my own RAW "experiments".. it was a very important variable for debayering. I think the Adobe apps are honoring it. By the way the enlarged screenshot above was taken from Adobe Camera RAW. So the maze is not exclusive to Resolves Debayer... it has something to do with the 4.6K sensor or at least the DNG or it's attributes...