@Bania I have not looked at the footage but I would like to mention that the issue at ISO 800 that you mentioned would not be the result of the ISO bug, while the others might.

The camera has some issues handling severe underexposure compared to some other models, so you might want to look to that first. A lot of people use noise reduction programs (the 3 top ones would be Neat Video, Purifier and Magic Bullet Denoiser II all of which are slightly different) and in my experience GOP1 settings have an easier time in denoising than higher GOP settings.

If you are shooting at very high ISOs or in dark places, the current setting that achieves the best results is Canis Majoris Night (which conversely has some issues with blues if you are shooting in daytime light but is good for shade, night, artificial light and sunrise/sunset on skies without much blue).

Neat Video helps alot to reduce noise just in the color channel and exposure range you have trouble with. So you can tread the noise in shadows - especially in the blue channel - without hurting important detail.

This plugin is so amazing, it takes some time to dig in, but then its pure magic.

Thanks for the responses guys.

While I was aware of the camera bug it is possible that for a few of the shots I may have had the camera set to iso160. Is this likely to cause the amount of noise visible in the third shot from the clip I uploaded?

@thepalalias & Meierhans: Thanks for the tip re: neat video. Hopefully this will clean the footage up a bit.

@LPowell. You'll have to take my word for it that the noise in these ProRes clips is more or less the same as with the original MTs files.I'm presuming you're referring to the first part of the clip when you say "dark, featureless video". I'm not sure what you mean by the lens being blocked. I was shooting in very dark conditions with a f/2.8 lens at iso800. I probably should have increased the iso but was wary of too more noise. In any case, is the lesson here that even marginally underexposed footage is prone to noise on the GH2? If I were to "properly expose" that first shot, would it not be just as noisy due to the higher iso?

All wisdom much appreciated :)

@Bania: man we're like brothers from another dimension or something. First shoot on GH2 on a rather involved project (10min short): check Flowmotion 2.02 untested beforehand: check Underlit night INT/EXT (DoP or lack thereof issues): check ISO800 f/2.8 (no faster wide lens available): check Well lit INT (f/4 ISO640 shutter/125) still has some noise.

So I get craploads of noise when underexposed, AND (this seems to happen with Flowmotion only) a "snow" effect (very low frequency noise) on top of it. The "snow" happens even with substantial available light coming through a window at ISO400 f/1.8. Sorry lpowell but I will not touch Flowmotion again. I loaded stock just to have a baseline on what the cam actually does on stock firmware. There is still some "snow" but Flowmotion exacerbates it.

Bania, the CURE is indeed NeatVideo. I have a history with the plugin, if there is one plugin I wouldn't shy away from paying again for another NLE host or OS (which somewhat outrageously they do indeed charge for) this is it. It fixes ALL kind of noise (luma/chroma), high/mid/low frequency, including the bullshit snowing which drove me mad. I ran a quick check on the raw footage and it will save my project as it always seems to do; kinda NV-man to the rescue. It may even deliver blowjobs with an advanced combination of options. It even sharpens the image and removes the flicker (I shot 23.98 with lighting on a 50Hz powergrid and there is no safe shutterspeed it seems, not 1/50, not 1/100, certainly not 1/125 I shot with).

I'm currently testing Valkyrie 444 ZERO1 (GOP=3) for night capabilities, and I'm not impressed. I was actually looking for Canis Majoris Night, that's how I hit this thread.

@bania @thepalias: My lessons for a night shoot with mainly ambiental street lighting:

-f/1.8 or faster lens

-Canis Majoris Night

-Neat Video

In that order.

@bania I checked out the video "GH2 Flowmotion.mov" again, and found I needed to use Quicktime Player on my Win7 system to see it. It's possible that the noise may be exaggerated by the Quicktime gamma bug, so you may want to investigate that potential issue. The noise in Clip 1 is the result of underexposing the GH2. The exposure in Clip 2 is good and the noise in the shadows is typical. Clip 3 is a little overexposed, which makes the shadows somewhat milky. Of course, the contrast can easily be increased in post, which would reduce the noise visible in the shadows.

@woodybrando Not really. This is a demonstration of the BMC killing every other camera under 20k. Noise is incidental compared to the DR and recorded detail being demonstrated here. Would you rather have your noise in RAW or baked into macroblocks? I'd take RAW noise personally.

@thepalalias

middle third of the histogram whenever possible

Would you know what luma values this maps to? You might have seen my Avisynth script for looking at baked in noise at varying luma values. The lower the luma the coarser the noise blocking becomes until in worst cases (low datarates and encoder starvation) it posterises and macroblocks. I think this coarse artifacting is from the compression rather than the sensor but I would need low luma footage from an external recorder to confirm this once and for all.

I am entertaining the idea of an online webapp for posting out of camera footage. The online app would transcode footage into optimal visualisation templates for comparing encoder performance as it relates to gradability and other important metrics. Was just reading about Vapoursynth which is a functional port of avisynth into Python. VS could be the building block that makes it all possible.

@vicx the noise doesn't look to be a deal breaker but brought me back down to reality. no perfect cam. But I know the bmcc is young and i'm excited to see what the new firmware brings. but, no way i'm going to use a raw workflow. I've got 23 hard drives filled up and I shoot h.264 and avchd.