AF101E Video Noise on Black parts of the image - looks like snow

J

jourman

New member
Did anyone encounter this issue.
I recorded on the AF101E in various ISO settings, and various lighting situations, and various gamma settings - all this after setting the Black balance of the camera.
I got very noisy blacks, regardless of the settings i used, even in broad daylight i had noise on the black parts of the image.

I recorded on the camera's on board SD cards on AVCHD PH 1080/25p
and also recorded on the Nano Flash at 100Mbps & 180Mbps
both had a noisy black...Nano01.jpg01.jpg

might it be that the specific camera i have in the rental house has a problem in the electronic circuits ? or is it the normal image of this camera.

Check out the black part in the pictures i posted.
 
I am going to pull a Barry... Have you tried viewing it on another monitor? And or is your monitor calibrated? Also, post some Tiff grabs if you can, full res, the jpeg is too overly compressed to be able to draw a conclusion. Because as is, all I see is compression artifacts, no noise at all really.
 
Barry Green should write a book about THE CYCLE... it would only be one page long, but, it would be a great read for everyone.

You probably aren't seeing things: there's noise in the image. And, it probably gets worse when you add Tungsten Light.

Do a black balance and proper white balance for the scene you're shooting in, see how it changes.
 
Pedestal was on 0, detail coring i tried with various settings which didn't really make any difference.
Black balance was done, and white balance was done

I have the same picture grabs in TIFF format, but i didn't manage to upload it here.
 
dcloud said:
Settings? Master pedestal, detail sharpness and coring? Could be those
Click to expand...

As per Kohli, you're seeing noise. This being a Panasonic camera - from those shots Id adjust some settings here. The AF100 is very very sharp at the outset.

Try this - got me noise free when I used it

Detail -7
Detail Coring +2 / 3

Should fix you up. As I said when I picked up the AF100 it was extremely sharp at standard settings - turn that baby down. Camera will still be sharp, just clean.

Also how did you output these stills?
 
Tried also putting the detail at very low levels, and also the detail coring at + levels, but it didn't help.
It just looks like very very noisy on the blacks.

viewed it on various monitors like : Astro, Panasonic 8.4", Panasonic 17", Cinedeck, and various editing monitors that are used for post production.
the blacks look like there is snow movement in them.

Did anyone encounter the same problem at all ?
or is it just us here with that specific DEMO camera...
 
If you have a demo unit, could be that. I just went through all my shots from the day I used the AF100, most of which was just me shooting anything - the only similar example I had was an unbalanced shot with virtually no light at all and a high iso.

Hmm what ISO was your camera set at?

Heres just one sample image has a lot of gradient, all tungsten light. THis has some compression artifacts from being converted to a jpg, but I hope you see what I mean.

AF100.jpg
 
This is disappointing and exactly the reason I wasn't in any pre-order line. I had hoped Panasonic could fix this noise issue with the larger sensor. But, if they haven't, the following will either eliminate or dramatically reduce the noise issue.

DRS = OFF. I think DRS was available in most or all modes on the AF100. I think I read that somewhere. Bottom line -- TURN IT OFF -- ALWAYS. It was designed by the best technicians in the world to add noise to your images, including stuff shot in broad daylight.

Use gamma curves B.PRESS or CINE-V. I'm guessing you were using Cine-D, and that is the absolute worst in terms of noise. Try switching to CINE-V (second to best relative to noise) and/or B.Press (best, it crushes blacks). Again, this is disappointing, I had assumed the new large sensor would make Cine-D useable.

SHARPNESS. To quote Philip Bloom -- "digital edge enhancement is the devil." (or something close to that). Turn BOTH the "Detail" and "V.Detail" down to -7 (all the way down). Do all your sharpening in post. As most of you already know, "sharpness" is about in-camera edge enhancement, it has nothing to do with resolution. -7 has the exact same image detail as +7 -- the only difference is in-camera processing. Again, do your sharpening in post. In-camera edge enhancement enhances the edges of the noise.

CORING. Keep it at 0 or -2. Anything above 2 and coring really does destroy image detail.

In theory -- do the above and the difference will be night and day. Although, with the large sensor of the AF-100, it's almost heart braking that you need all the old tricks from the 200/170/150 to reduce the appearance of noise.
 
AGMedia Im going to agree with you on some things and disagree on others. See my image above your post? Cine-D.

DRS on low setting is very useful and I didnt notice much effect to noise - maybe because my camera settings were optimum. On high, yeah if I was really desperate I might consider it but outside of that desperation - never.

I agree on the sharpness, insofar as you dont really need it. With detail down to -7 as I did, image was sharp.

Coring - youre back to front. A positive coring setting reduces it, So should be around +2 as per my image above.

Old tricks of the 200/170/150? I cant recall a single camera Ive ever used ( wow its a lot now I think about it), where I didnt tweak the crap out of it to get what i wanted. And I didnt mind applying a few old tricks to make a $5k camera do what I want. Remember these points are all relative. Is it Alexa - nope. But so what?
 
Not really IMO Matthew. They should know what they are renting anyway - all cameras are different and work best set at A, B or C for shooting X, Y or Z.
 
I opened both samples in Photoshop and pedestal isn't at zero and one sample is even clipping in the white. See same photos with histogram:
http://www.squarepixel.com/pics/Nano01_histogram.jpg
http://www.squarepixel.com/pics/01_histogram.jpg
If pedestal would be correct you wouldn't be seeing this.

Now to make the noise more visible I elevated the shadows of both photos:
http://www.squarepixel.com/pics/Nano01_elevated_shadows.jpg
http://www.squarepixel.com/pics/01_elevated_shadows.jpg
This clearly shows that it isn't noise, but compression artifacts which you normally see in heavily compressed video like satellite TV. If you do this with a framegrab from a DV or S-VHS camcorder you won't see this. I don't think it should be here in the first place but I'm not an expert on this. I do know that jpeg compression can't cause this. So I'm curious if these articats are also visible in video recorded from the AF100 SDI/HDMI output directly. Jourman are you able to duplicate your problem on HDMI output and show some grabs from that?
 
Here's what I'd like to see, before anyone says anything about noise -- play the footage back from the camera on a properly calibrated monitor.

Why?

Because I'm starting to think that there's some shortcuts being taken in the h.264 playback on some computers, or something like that.

Why?

Because this is sort of familiar, and I'm flashing back to Raylight/HVX/P2 Viewer days. P2 Viewer made footage look like complete @$$. Hella noisy blacks. And Marcus van Bavel of Raylight actually discovered why; it was because (for speed) the DVCPRO-HD decoder in P2 Viewer was "cheating" -- it was using hardware MPEG-2 decoding for the DCT coefficients, which gave vaguely similar but generally inaccurate results. But, it was "close enough" for a simple viewer program. Raylight made a big deal about the quality difference of their codec (which did proper decompression) vs. P2 Viewer's decompression, which was optimized entirely for speed. Look near the bottom of this page: http://dvfilm.com/raylight/raylightRes.htm

This became a problem for people who were using CineForm, because CineForm used the P2 Viewer codec -- and so everything that they transcoded, had crud in the blacks -- crud that is *not* in the original footage!

I think it's entirely possible that some h.264 decoders are cheating in a similar way, to speed up their playback or whatever. And it's affecting the footage in the same way -- artificially creating noise in the blacks, that isn't there in the first place.

So the only way to absolutely know whether you have noise in your footage, is to play back the footage in-camera, from the original source files, onto a properly-calibrated monitor. If you see noisy blacks there, then blame the footage. If you don't see noisy blacks there, but you see them on your computer, then -- blame the computer.

How do we solve it? (Because it needs to be solved). I don't know yet. But step one is identifying where the problem is. Once we identify what post processes are creating the cruddy noise, and what post processes DON'T create it, then we'll be on track to isolating and eliminating the problem and getting the various software manufacturers to make appropriate solutions. Premiere Pro (up to CS5) had a nasty bug that ruined the chroma from 1080/24P AVCHD footage, and they fixed it. If any NLE has an issue where they're introducing crud into the footage, then they'll have to fix it (or at least we can make the end users well aware of what's causing the problems).
 
Last edited:
But the OP said he also recorded to nanoflash at 180Mbps and saw the same issues. That would obviously bypass the internal H.264 codec and cheapy viewer applications. Wouldn't another test be to put an external recorder on ProRes 422 and view the much higher bandwidth data right in FCP? That would be taking as high of a bandwidth signal this camera outputs right into a DI that would be used on the NLE.
 
Barry is saying to play back from the camera first -- removing any transcoding or computer play back from the chain to see if there is a difference in the way the footage is decoded for viewing.
 
Zak Forsman said:
Barry is saying to play back from the camera first -- removing any transcoding or computer play back from the chain to see if there is a difference in the way the footage is decoded for viewing.
Click to expand...

Oh I agree, we definitely need to do this. But I'm proposing after that test, then let's see how it looks right into ProRes (bypassing H.264) because if it appears perfect out of the cam right to a monitor, our next step will be to determine if this quality can be replicated. I.e., it's an issue if we can only get the best result by only plugging the cam into a monitor but never editing it or getting it out of the camera :)
 
You must log in or register to reply here.
Back
Top