C100: Noise or Compression artifacts?

Tim Nitsch

Well-known member
So tonight is a full moon, so I went out and filmed a little bit of it just for fun.

I brought it into 5D to RGB and converted it to ProRes HQ (I normally do LT) 23.98.

And while playing it I noticed a lot of noise (which I guess is understandable being that I'm filming a black sky) but when I went frame by frame I noticed this blocky-ness:

(NOTE: I cropped the image quite a bit)

Moon.jpg

The strange thing is, out of three frames, one frame doesn't have the blocky-ness while the other two do.

What's causing this? The AVCHD Codec, transcoding, or just the lack of light?

If I play the unconverted .MTS file in VLC:
1. I can bearily see any noise at all
2. I don't really see the blocking

I also noticed that if I transcode using Toast, the edge of the moon will get really bad aliasing, Even if I use the exact same settings I do in 5D to RGB.

I know 5D to RBG is supposed to help bring out the "Full Range" of the image, but would that introduce all this noise and blocky-ness?
 
Last edited:
This blockiness is the thing that I dislike about the C100 the most. My guess it is macro blocking from the AVCHD encoder due to low nitrate.. If someone can verify that this is not present on the HDMI output I will buy an external recorder.
 
This blockiness is the thing that I dislike about the C100 the most. My guess it is macro blocking from the AVCHD encoder due to low nitrate.. If someone can verify that this is not present on the HDMI output I will buy an external recorder.
its not
 
This is the first time I've really noticed it. If what ever I'm filming is even half way lit you really don't see it.
 
looks like macroblocking. what iso were you shooting? underexposed areas can indeed show more nastiness in higher isos, or even normal isos. that plus low bitrate avc encoder aren't exactly a match made in low noise heaven.
 
I also noticed that if I transcode using Toast, the edge of the moon will get really bad aliasing, Even if I use the exact same settings I do in 5D to RGB.

Sounds like Toast is doing what FCPX does - it misinterprets the PSF setting and assumes it should do a savage 'field doubling' deinterlace, thus reducing your resolution by half. To be avoided.

I know 5D to RBG is supposed to help bring out the "Full Range" of the image, but would that introduce all this noise and blocky-ness?

I think you're missing some detail in the VLC rendition. I've stepped through footage from the C100 and can see the 'Good, bad, bad, Good, bad, bad' cycle and that's part of Canon's implementation of AVCHD compression. It's absolutely fine when the compressor isn't stressed, but begins to fall apart when things get tough. I was seeing what I thought were interlace artefacts on frame by frame inspection, and this is on PSF.

I do believe Canon's AVCHD compressor does nasty things when stressed. So... I try not to stress it.

If I light well, if I don't have to dial in huge amounts of gain in low contrast situations (which seems to be the main issue), life is sweet. I've yet to play fully with the noise reduction settings in the C100, but from what I've seen so far, I'd prefer to do my noise reduction in post. I'm also working on a hypothesis that, when dealing with noisy situations, it may be better to record externally and do noise reduction in post rather than rely on the built-in codec - and that THIS is why we need a Ninja:

Low contrast, high gain = problem for internal codec. Dial out noise reduction, record clean feed to ProRes and use Neat Video to sort things out.

Dialling in NR to the C100 makes the picture look more and more like the FS100, so I can see where we're going there. Just like abdicating Sharpening or Detail to the post process, perhaps we now need to understand and deal with noise in a similar fashion.

I'll take the C100 over the FS100, though. :happy:
 
I saw this on my C100 clips, one frame looks fine (like I-frame) and other two frames look as they were compressed with lower bitrate.
Yes, I think that it is AVCHD with GOP, so for better result you need an external recorder. But I found that this noise is not very visible.
 
Back
Top