F35/F65 F35: Fixed Pattern Noise?

R

RayZan

Active member
I have been struggling with noise levels in underexposed areas with the F35. So, I did some tests today.

I simply aimed at a black board that was evenly lit and recorded some clips. I started exposing the first clip with the board around 5% Waveform value in the 7Q+ (about 60 IRE, clearly underexposed) and kept on increasing the light amount on the board with each clip. I wanted to know when the noise would stop.
When analyzing the clips, to my surprise, the noise never stops. The same exact pattern of noise “blocks” appears in each consecutive clip, exactly in the same location. It just becomes lighter. I used to see that blocking before only in dark areas of normal subject matters. But I guess it kind of blends in in lighter areas of the scene and I never noticed it.

I pasted some jpegs of the clips. They are magnified by 800%. Notice that the blocking patterns are identical from clip to clip. Just lighter. This has nothing to do with the black board. The camera between clips was nudged and moved slightly. At some point, I even turned off the camera and the 7Q+ altogether (I think after the 4th clip) to do something and then 10 minutes later turned them on again and resumed with the rest.

It seems that there is a fixed pattern of how the sensor’s photocytes respond to the same amount of light, differently, with some giving a slightly higher signal than others. But they are consistent with that. So that at the end there is a fixed pattern of noise that the sensor has that doesn’t change.

Any thoughts?

1_1.8.1.jpg
2_1.9.1.jpg
3_1.10.1.jpg
4_1.11.1.jpg
5_1.12.1.jpg
6_1.13.1.jpg
7_1.14.1.jpg
8_1.15.1.jpg
 
FPN like this is very difficult to see in a single frame. Do you have a video clip?

I do see that one orb or spot, though.
 
IMO, this doesn't look like FPN...just noise (particularly blocks as you mentioned, which can be common).

FPN has clear vertical lines you can see like a grid. (Like this: https://vimeo.com/190958791)

But a clip as James requested would be best.
 
Will post the clips very shortly. They are 10 clips in total, Prores 444. Should I deliver them as one clip from a timeline in dnxhd (I have a pc)? And post them in Dropbox?

The reason I'm saying it's FPN Norbro is that the noise pattern is identical throughout the clips. It is not symmetrical. Unevenly distributed, but still consistent from clip to clip. Those are in slog. When you apply contrast, you will see the pattern much easier.

Will post clip.
 
RayZan said:
Will post the clips very shortly. They are 10 clips in total, Prores 444. Should I deliver them as one clip from a timeline in dnxhd (I have a pc)? And post them in Dropbox?

The reason I'm saying it's FPN Norbro is that the noise pattern is identical throughout the clips. It is not symmetrical. Unevenly distributed, but still consistent from clip to clip. Those are in slog. When you apply contrast, you will see the pattern much easier.

Will post clip.
Click to expand...

That makes sense. It is noise in a fixed pattern.

Though, in at least the CMOS world, people often incorrectly identify FPN.



Individual clips always works for me. Easier to download, organize by name, etc, sometimes. But one long clip works too.
 
Is this 10-bit 422 or 12-but 444 we are looking at? 12-bit should be finer.

I find turning on extended dynamic range increases the noise also.

Once I shot friends wedding on 16mm and it was dimly lit and seeing the footage I worried it was too extreme. After seeing the magic NEAT video can perform, I don’t worry about noise now. I use it a lot with F35 to get rid of that noise “sparkle”
 
Here is the link for individual clips.

Essentially, I taped two foamboards together, one white and one black and then bracketed exposure (done with adjusting light with dimmer and not aperture. Aperture remained on T2).

You will notice the issue more with the black foam board (though it is there in the white one as well). I think the reason is: the black board will tend to surface glare more as light intensifies and that create a bit of micro contrast that "excites" these patterns.

My initial rationale for the test was: If the black board was to be the darkest part of the scene, how would I need to expose it in order to minimize noise in the frame. In other words, it was supposed to be a test for the sensitivity (and DR) of the camera.

Link for 10 clips:

https://www.dropbox.com/sh/3ntjlnog4rv87tm/AAC9PjA5YlX2RA9XekiJpI4Wa?dl=0
 
yachacha said:
Is this 10-bit 422 or 12-but 444 we are looking at? 12-bit should be finer.

I find turning on extended dynamic range increases the noise also.

Once I shot friends wedding on 16mm and it was dimly lit and seeing the footage I worried it was too extreme. After seeing the magic NEAT video can perform, I don’t worry about noise now. I use it a lot with F35 to get rid of that noise “sparkle”
Click to expand...

That was 12 bit 444 Yachada.
I turned off Extended and set it to "Normal" for DR during the test (and in general). Extended has way too much noise.
Unfortunately, I don't think noise reduction can solve this issue. When I use the one in resolve, the mini blocks get bunched together into several giant blocks and the effect is almost worse than the initial noise. I can upload more footage to show that.
I really need to solve this noise issue with the F35 before I use it in my upcoming project. It is the only thing holding me back right now.
 
I don't think any of the files work...unless they're supposed to be 1 second long which would be no better than still frames.
 
NorBro said:
I don't think any of the files work...unless they're supposed to be 1 second long which would be no better than still frames.
Click to expand...

Are you trying to see if the noise pattern changes over time? It does not at all but I could upload an untrimmed version of the files.
 
Just wanted to see in general what it looks like (usually you want to upload at least 5-10 seconds).

But even with the stills...to me it just looks like an old, noisy camera with an unfavorable amount of light; dozens I've seen and used in my life. The original C500 from the same period looked exactly like your stills (blocks of noise). ARRIs get noisy too. Blocky, compressed noise.

Whereas all my Blackmagics had distinct lines in anything higher than their native ISOs like in the video I linked (which has true fixed pattern noise versus a fixed pattern of any other noise).

But I guess regardless if it's true FPN or not, you probably just want a cleaner image so hopefully that gets resolved soon.
 
Below is the type of noise pattern that I was trying to solve. Jpeg with 400% magnification (plus unmagnified image). I really don't think I could. It might be the nature of the camera, or am I pixel peaking (I think on a big screen that noise would be obvious as daylight though) Or, is something wrong with my camera?
Has anyone else experienced this?

15_1.10.1.jpg
16_1.10.2.jpg
 
At glance I am not seeing anything. 3-10 second clips would be best. if I is easier for you, all the clips in one timeline is fine too. sorry
 
RayZan said:
Sure thing.

Here are the new clips at their full length. They are on average 3 to 4 seconds because when i was recording I never thought it would help to make them longer. Sorry.

https://www.dropbox.com/sh/sa1uh1jl5idnocs/AAAo4-sfZJwaFZdlE4rRQr--a?dl=0

I will do more tests later today with hypergamma and test other evf menu options for denoising etc. Will post results.
Click to expand...

I'm sorry, I don't know what I am looking for anymore. The fine grain noise looks fine. It dances around. There is what looks like compression blocking which has a static look, but that changes slightly with each clip, so I have to assume that is just compression artifacting.

Not seeing FNP, at least not able to.
 
That is no compression blocking. It is there in the original 12bit prores recording and it is the same with dpx recording. It is either something wrong with my camera or something inherently wrong with the F35. I was hoping someone may have seen the same thing with their camera.
Can someone just shoot a uniform dark toned wall or board at various exposures and post some clips or stills?
 
I think You are seeing what I call the “honey comb pattern” I’ve noticed it in lots of dimly lit footage from my F35, handheld you can really see it. I found best solution is a fast lens. It’s not issue with your camera, it’s like said above, it’s old camera.

The pattern is always there, you will only ever see it when you are not getting enough light to sensor.

Stop looking at 800% magnification and go start shooting as much footage as you can.

One last note, although this isn’t what is happening to you. I got lots of strange highlighting and artifacts when I upgraded F23 to F35 and turned it on without updating the firmware. Also I have seen cameras without complete firmware update, you have to do all 3 steps. So it’s always a good idea to run the firmware complete when you get a new camera because I do believe mismatch on this level will cause strange images.
 
Last edited:
I never noticed this on my two F35's at the time. But I also never shot a black board. not knocking the test, just mean that I never looked into it too deeply because I never figured out my Fan Max issue.
 
As more and more cameras get 12+ stops DR and 10bit codecs, we are going to see higher and higher expectations from the F35. It is a 12+ stops DR camera, and can be a little bit of a learning curve if coming from newer cinema cameras.

But it is also the only 12+ stops DR camera that I like. It so effortlessly gets the tonal contrast correct. You just have to watch those uncontrolled lights when shooting all dogma 95. But at least the clipping is about as nice as you could hope for in that range. the three channels level out so evenly, they all appear to go white at the same rate. Even the C500ii goes all yellow at the very last moment just before absolute clipping.
 
You must log in or register to reply here.
Back
Top