Revealed - How to get clean ISO 6400 and 12,800 low light video on the GH2 and 5D Mark II

[Andrew Reid]

[html]

[url="http://www.eoshd.com/wp-content/uploads/2012/01/gh2-iso12800-tips-small.jpg"][img]http://www.eoshd.com/wp-content/uploads/2012/01/gh2-iso12800-tips-small-660x412.jpg[/img][/url]

1080p frame at ISO 12,800 from the GH2 – click to enlarge

The GH2 is the best low light DSLR. It is a myth it cannot compete with the FS100 or Canon C300. I am currently experimenting a lot with my GH2 and Nikon J1. I’ve seen some new opportunities with these… With the GH2 I am refunding my techniques for getting cinematic extreme low light at ISO 12,800 and F0.95. Due to the way CMOS sensors behave you are actually able to get a very film like ISO 800 grain over well exposed areas of an image and such is the F0.95 aperture on the Voigtlander Nokton and the high ISO, it is very easy to get an exposure in the dimmest of light. The noise is hideous in the lows and red / blue chroma channels, but you crush that noise out with a high contrast black and white picture profile and it solves the issue.

[url="http://www.eoshd.com/content/6858/tips-for-clean-iso-6400-and-12800-on-the-gh2-and-5d-mark-ii/"]Read full article[/url]

[/html]

[miseducation]

I guess this is as good a time as any to ask this question that's always bugged me about sensor size and aperture.

I know the full frame sensor produces shallow DOF at higher apertures but I always assumed this meant that it also let in more light at these apertures as well. I have no idea if this is right, but if it is, wouldn't the 5D match the same light-sensitivity of an 0.95 lens on a M4/3ds sensor with a 1.4?

Btw, I too love the look the 12,800 iso hacked GH2. I intend to shoot a nightclub scene with it this weekend and I'm just thrilled at how unique it's going to look.

[joe]

My fastest lens is my nikkor 50mm f1.2, curious how that would affect the grain quality for those of us who dont have f0.95? on my GH2

[P4INKiller]

I was thinking, and Fuji has the right idea of how the RGB pattern on a sensor should be distributed.
Just imagine, every pixel being a 3x3 random pattern made of 3 reds, 3 greens and 3 blues.
That would be magical in low light.

[quote author=joe link=topic=188.msg1349#msg1349 date=1326822332]
My fastest lens is my nikkor 50mm f1.2, curious how that would affect the grain quality for those of us who dont have f0.95? on my GH2
[/quote]
I'm thinking of getting that lens.
1.2 is still a very fast aperture, I don't think it would be that much of a difference.

[joe]

[quote author=joe link=topic=188.msg1349#msg1349 date=1326822332]
My fastest lens is my nikkor 50mm f1.2, curious how that would affect the grain quality for those of us who dont have f0.95? on my GH2
[/quote]
I'm thinking of getting that lens.
1.2 is still a very fast aperture, I don't think it would be that much of a difference.
[/quote]

I love my nikkor 50mm 1.2 its really beautiful I'd recommend it, my second favorite lens is my lumix 20mm 1.7 but OMG that nokton 0.95 makes me green with jeolousy

[Andrew Reid]

F0.95 isn't actually A LOT brighter than F1.2 or F1.4 but it definitely helps. It is more for a shallower DOF and a slightly dreamy look that I shoot wide open so much on the Nokton. Great looking lens.

F1.4 or even F2 is fine for most low light shooting. ISO makes a bigger difference.

[Samuel H]

one thing to take into account is that "ISO" doesn't mean the same on all cameras

* GH2 at ISO 12800 can be brighter than ISO 20000 on the C300
* FS100 at 24db (ISO 6400) can be a lot brighter than any other camera at ISO 6400 (really, a lot), and even significantly brighter than the C300 at ISO 12800
* NEX-5N can be a lot brigter than any other camera at ISO 3200 (including FS100 and GH2)

My ordering in terms of brightness would be:
5N > FS100 > GH2(a) > F3 > 5D2 >= 7D >= AF100 = GH2(b) > C300

where GH2(a) is above ISO 3200, and GH2(b) is below ISO 3200 (this one seems to be very non-linear, starts out dark at low-ish ISO and gets quite bright at high ISO, probably because it captures less latitude)

[url=http://www.similaar.com/foto/iso/pbtest.html]http://www.similaar.com/foto/iso/pbtest.html[/url]

of course, the main thing is how much noise there is in the images, not how bright they are, but still: it's something to consider when doing comparisons

[Samuel H]

re: sensors for low light

first: with a lens at f/1.4 on both cameras, the 5D2 gets roughly 4 times as many photons hitting the sensor than the GH2, but that doesn't mean the final image will be any brighter (check my previous post); what it should mean is that the camera should be able to go to a higher ISO without displaying a lot of noise (that's why the 5D2 can get to a higher ISO than its APS-C brothers without more noise); but the 5D2 is quite old technology by now, and new cameras (GH2, 5N) can give you images with more detail and less noise even starting out with just a fraction of the photons

also, on the Bayer pattern thing: even better than a random array would be to have clear photosites (white, instead of R, G or B)
Kodak already has a few patents regarding this
and I have this post proposing black (ND) photosites too (in order to protect highlights and increase DR) and a ratio of color-to-white photosites and an ordering that delivers exactly the amount of color information that 4:2:2 color subsampling will keep
[url=http://www.similaar.com/foto/bayer/rgbwk.html]http://www.similaar.com/foto/bayer/rgbwk.html[/url]

[bg2b]

The 5D2 gets 4x more photons at the same f-stop as the GH2 in stills mode, but in video it's not so clear.  Because of line-skipping or other hacks for downsampling, the 5D2 probably has an effectively smaller sensor in video mode (still maybe not as small as the GH2's, but certainly not 4x the area).

[Samuel H]

^ correct

[ChrisJ]

The 5D does get more photons, but over a larger area. The number of photons per square mm is exactly the same on both cameras.

Noise is added to the image signal from the sensor, 60% of noise is caused by heat within the sensor, the more heat the more noise, it's electrical noise. Larger sensors need more energy to heat them up than smaller sensors do, therefore larger sensors have less noise. Used to be true, but not necessarily so these days, if you reduce the current the sensor requires it will run cooler and you'll get less noise. There are more efficient sensors than those made by Canon which are relatively old technology, the sensor in the GH2 is one of these more recent ones.

Most of the other 40% of noise comes from the A/D converter, or more especially the high impedance connection from the sensor to the A/D converters, by simply integrating the A/D converters within the sensor the noise from this source is virtually eliminated. The GH2 sensor has this feature too.

This noise threshold is ever present in all digital sensors, lower on modern sensors than older ones, the only 100% way to get zero noise is to use more light, then the exposure is well above the noise which is why movie companies spend loads on lighting the scene, if you want it to look dark, record with bright lighting and then turn the exposure down in post processing, result, dark moody image without noise.

Chris

[Andrew Reid]

Interesting explanation of the sensor tech Chris, enjoyed that.

[Samuel H]

very clear explanation indeed

my point was that, given a level of background noise in the sensor (as you explained, created both by heat and by inefficiencies in the ADC), a bigger sensor catches -overall- more photons, so the signal-to-noise ratio is boosted

but also that, given the GH2 uses a much more modern sensor (the background noise is lower), the end result can go both ways

(of course, the line-skipping thing noted above is a very big deal, as it nearly equalizes the number of photons hitting the used sensor area in a 5D2 and a GH2)

[miseducation]

Thanks a ton, Samuel and Chris - extremely informative.