my 5D night footage looks bad
 

One of the things that impressed me the most about the early 5D videos was the night footage. Mine however, looks like crap. It is muddy, noisy, and filled with the horizontal banding to the point where it is unusable.

Attaching a screen grab so you can get an idea of the noise issue. I'm shooting with settings straight out of the box and the 24-105mm F4 lens. The most visibile flaws (in the still) is the noise in the bottom right section on the road. The moving picture is terrible. Thoughts?

I know that there are automatic settings in movie mode, but what were yours at?

Dylan:

What lens ? What was ISO at, at shooting. Did you lock exposure ?

The first time I shot night, I walked outside of my home and tried to shot toward a street light. I shot full auto. On attached shot, I set camera in locked AE by pushing * button. Then I adjusted exposure wheel to get ISO down to around 1200 or 1000. No more than that, in my opinion.

See pic below for result

f4 isn't helping your cause any. You're losing multiple stops of light because of that, which in turn is being made up by the 5D2 by cranking ISO, and thus noise.

I'll second the f.4 thing - my initial reaction was the same as yours, but with an f.2 nikon lens the results were much better. 2 stops doesn't seem like that much difference, but considering what you're really doing is dropping the ISO from 6400 down to 1600 it has a big impact on noise.

I'll try again tonight with my F2.8 and will be borrowing a f1.8 tomorrow, and see what a difference it makes.

Thanks gents!

that will make a huge difference. keep in mind that the vincent laforet video used about $25k worth of lenses. thats something thats rarely mentioned.

I think I read somewhere that movie mode expands its useable ISO range if ISO Expansion is on. But don't quote me on that, don't think it was substantiated.

I demonstrated in some thread here that the ISO can go up to at least 51200. This happened when it said it was shooting 1/8 second or something. Once it gets down to 1/40 second it keeps it at 1/40 and raises the ISO as much as needed, without telling you.

And boy was that noise ugly!

Is there any way of telling exactly what settings movie mode is using at any time? Before/after you lock/set exposure?

Push the shutter button part way. Press the Info button multiple times, until you get the desired mode.

Roger that, but that doesn't display the actual shutter speed that the movie mode uses, just the still photo shutter... right? I should really RTFM again. :)


Oh, night shoot update: my results with the f2.8 were not really any better than the f4. Watching the info while dialing from F4 back to F2.8, the ISO doesn't change. Still not enough light. I'll be trying it with an F1.8 tomorrow, hopefully it will make a bigger difference.

Dialing the exposure down in combination with the faster lens gives a cleaner picture, I'm just disappointed in how noisy it actually is, compared to what I've seen.

I ran some shutter tests back in December, and came to these conclusions:

1/40 and 1/50 displayed is actually about 1/45.

1/60, 1/80, and 1/100 are all actually 1/80, though the gain is lowered in 1/100 to simulate the slower shutter.

1/125, 1/250 and 1/500 are accurate.

Here's the evidence: ~ The Murder of Dirk Snowglobe - Article: 5D Mark II Shutter Exposed! ~

I think there are at least two problems: expectations and line skipping.

Expectations can be a problem if you base them on 3-chip cameras. Mine, for example, were rooted in the use of 3-chip 1/3" cameras. The 5D2 seems like it has a much larger sensor area, right? For thin DOF, yes, the difference is as big as everyone thinks. But for low light performance, the difference is smaller because of two factors:

3-chip cameras have three times the sensor area for a given size. This increases low light performance (sensitivity) relative to a single sensor of the same size. It doesn't affect DOF.

The 5D2 skips two out of three lines. This doesn't affect DOF, but it reduces the effective sensor area for low light performance to one third of what it would be if it read every line.

If we compare the full-wide focal length on the XH-A1, which is 4.5mm. At f/1.6 the aperture is 2.8mm. The equivalent focal length on the 5D2 is 32.5mm. The f-number that would give it the same depth of field is f/11.6 (2.8mm aperture). To account for the three-fold increase in area due to 3-chip, divide f-number by sqrt(3): f/6.67. The 5D2 skips two out of three lines on the sensor, reducing the effective light gathering area by a factor of three again: f/3.9.

So when you account just for sensor area, you have to shoot the 5D2 at f/3.9 to match f/1.6 on the XH-A1. Of course, that doesn't account for any difference between the two sensors such as spectral sensitivity, read noise, in-camera processing, etc. If one or the other were superior/different in one of those areas, the necessary f-number would also be different.

If they had the same technology, f/3.9 would only be equal, not superior in low light. A wider f-number would be necessary for that.

I have seen other people say the same thing (night footage is not what they expected) From what i have heard try and keep the ISO below 1600, any thing over that starts getting noisy.

Setting up the shot before hand is key. I have used my cell phone or cheap white LED flash light for forcing the camera to the settings i want then locking them in and start shooting. It is a pain at first but you get used to it.

Laforet's movie was mostly in Time Square which is, even at night, bright as day. so the results are tuff to match.

Faster lens, will help but keep an eye on the ISO and keep it low.

There is no niose reduction in movie mode so you will never get the super hight ISO.

Just my 2cent
Mike