Low Light Footage Settings Test

[Jay Rodriguez]

Wow, I think I'm going nuts here but I just sorted through all of the setting combinations from the top three gamma curves and all four matrices and they all look the same in noise. Some just looked darker then others but ALL looked the same in noise production....:crybaby:

B.PRESS + NORMAL 1, NORMAL 2, FLUORESCENT and CINELIKE
CINELIKE V + NORMAL 1, NORMAL 2, FLUORESCENT and CINELIKE
LOW + NORMAL 1, NORMAL 2, FLUORESCENT and CINELIKE

720pN
1/60 shutter
Iris open fully

From Detail Level:

-4
0
2
4
0
0
0
-5
0
DRS off
KNEE auto

Exported from FCP using QT conversion with the following settings:
compression: Sorenson video 3
quality: best
key frame rate: 24
bitrate: 6400
dimensions: 1280x720

File size is 117.4mb
Let me know what you think...... CLICK HERE (http://itchyhousefilms.com/films/gamma_test3.mov)


I just don't get it.... How do we reduce the noise on this camera?

[Jay Rodriguez]

Ok, I finished watching it again and did a frame by frame on most of them and it seems like out of the bunch that I did the ones that look "less" noisy are the low & cinelike and b press & cinelike combinations but damn, they still have a bunch of noise in the mids....

Looped comparison clip HERE (http://itchyhousefilms.com/films/gamma_test4.mov)

[streameru]

Is there anyway you can post a version that uses a intermediate codec... say ProRes or ProResHQ?

As you may know, Sorenson/mp4/H264 and other web delivery codecs can make noise look *worse* then what it is in the original footage...

Thanks for taking the time to test out the camera...

me

[Jay Rodriguez]

streameru, ok well here's a learning situation for me. What's an intermediate codecs compared to something like sorenson?

I'll go back in and export as ProRes or ProResHQ but what the difference between them?

[streameru]

Jay,

I'm going to skip over some techy stuff for sake of conversation. Think of this in "general terms". If you want the techy stuff, feel free to ask.

Intermediate codecs are used for things like capturing and editing. As a general rule, every frame is a "key frame", meaning each frame is a complete picture of the your scene. Much like a film camera. Another way to think of it, intermediate codecs maintain more of the visual integrity of the scene. For the must part, intermediate codecs don't throw much visual data. Yes, there are other issues, like color space, compression ratios, etc... but I'm talking in generalities here. DVCProHD, ProRes, DNxHD, are examples of intermediate codecs.

Codecs for delivery, (be it BluRay, mpeg, Sorenson, H264, web video, etc) follow the gerneral rule of recording a "key frame" once every x frames and then have different ways to record the "in between" frames. Because key frames take up so much data, these codecs have to "throw away" some visual data to be able not to take up a lot of storage space.

That make sense?

When you are taking your HPX high def footage and using Sorenson video 3 as a means to transport this footage, you are losing visual data. That is why it's hard to determine noise issues with the HPX when viewing Sorenson compressed video. Are we seeing noise that was actually there, amplified by Sorenson, or is Sorenson generating noise itself (which tends to happen with darkly lit scenes)???

ProRes offers better visual integrity then Sorenson. By taking your "raw" HPX footage and "wrapping it" in ProRes, we can reduce the impact the codec has on your footage and allows for better scrutiny. In other words, it's doubtful that ProRes will "dirty up" your video the way Sorenson can.

The difference between ProRes and ProResHQ is the bit rate. If your scene is complex (visually speaking), or your doing a lot of compositing work in your NLE, it might be wise to switch to ProResHQ. The increase in bitrate = larger files on your hard drive.

If you can handle the bandwidth and file size, for this test, use ProResHQ. Plus, apple offers the ProRes decoder for free on windows and mac...thus, less cross platform viewing issues


Hope that helps..


me

[arrestthisman]

Actually, I look at it differently.

Delivery codecs are where most of all of our footage will end up. I also find H.264 to have a smoothing effect on harsh detail, including high frequency noise.

Yes it would be great if we coud see exactly what he is seeing. But much the same as 35mm labs try to get the original as high quality as possible, knowing the theater prints will be substantially worse off in quality, that is how 99% of the footage will be delivered.

If the native footage is not good enough to output delivery clips, there's a problem.

[arrestthisman]

DAMN that is noisy! And it looks like classic HVX200 low light noise too. That's not the delivery. It's better than the HVX200, but not by enough.

Man.... I can't believe that. Why would panny not fix this by now?