GH2 Film Modes and ISO

Posted on 14 November 2012

0


Before discussing film modes it is important to get your head around Dynamic Range or latitude. DR (dynamic range) is about the number of stops between detail in the darkest shadow and the brightest highlight that can be captured by the sensor or displayed by your monitor: each additional stop represents a doubling of lattitude.

Another thing to understand is curves, which means: how much detail will be recorded. Sensors work lineair: 16x the amount of light that hits the sensor light is 16x the electrical charge that is built up in the photo site. But our perceived brightness works more logarithmic, like film: 16x the amount of light is only 4x brighter. This means that the human eye is more receptive to details in shadows, so you might want as much detail in shadow as you can get.

This is why the standard rec.709 gamma is giving more bits to shadows, and less bits to highlights. Rec.709 also limits the dynamic range to about 6 stops, to make the image look good on every HD TV. As taking 12 stops and simply squashing it into 6 stops will create a disappointing, flat and dull looking image, almost all camera’s will give you a pretty high contrast image with crushed blacks and clipped whites. In other words:  there is no headroom, your highlights will ‘blow out’, ‘clip’ or you will get ugly crushed highlights. The highlight problem is partly solved by the knee function in most camera’s, which gives it a nice highlight roll off and might expand the DR with one or two stops. This might get you  to the 8 stops DR of most modern rec.709 camera’s, and this reduced contrast will be enough for normal viewing but it will become a big problem if you want to grade your film.

Custom Curves

So for grading you would want  a flat (washed out) image with lots of detail in highlights (not to be confused with a nice highlight roll off). This approach is the ‘film’ technique: with film you will need to develope and print the negative and it that proces you can make choices. That’s why log curves are developed: it stretches blacks and compresses highlights even more then rec.709, at the expense of less detail. See full explanation here and here are a few good links by Steve Castle about S-log. You will find that the S in S-log does not refer to the shape of the curve, but only to the name Sony! The log part stands for logarithmic curve, which works nice with 10 bit recording. The new thing is lineair high DR recording that needs 12 bits or more. The compromise for 8 bit recording is cinegamma or hypergamma: it’s the same as the rec.709 gamma for shadows and midtones, but the curve bends gracefully to squeeze as much highlight latitude into the top bits of your recording, by compressing it. Alister Chapman has a lot of nice explanations on xdcam-user.com, like this one.

If you want these images to look ‘normal’ on a rec.709 TV you will have to apply a LUT or color grading to get it from the 12+ stops of latitude back to the 5 stops of your TV. There are two important things to note however:

  1. Recording in a curve like S-log is giving your material a higher DR (dynamic range), but it doesn’t change the dynamic range your sensor can capture. If the sensor of the camera is not capable of capturing the DR of the scene, you will still have to choose between clipped highlights or no detail in the shadows. The limitation with the GH2 is the dynamic range of the sensor: it can only take 8 stops of latitude (nostalgic film mode).
  2. Recording a flat image for color grading is great with uncompressed 10 bit RGB data, but in 8 bit studio swing 4:2:0 YCC color space with AVCHD compression it might not be the best idea. My first believe is that it’s best to get the image in the camera as close to the final ‘look’ you are going for and only use grading for final corrections. But on the other hand there is this great tutorial video that tells us to avoid the highlight clipping and shadow noise by exposing for 20-80% (in 8 bit this is 4 stops, while skins 30-65% = 2 stops) and then correcting that image to the look the way you want. This comes down to only 235*60% =  140  gradients = about 7 bit.

If you don’t want to grade after editing, you might want to have vivid colors and high contrast. High contrast is the opposite of high dynamic range: with high contrast there is less detail in highlights and shadows. This is where a nice highlight roll off  is so important: it gives a nice transition into the clipped highlight area (100% white). Try it in photoshop: if you set the contrast higher you will loose all the details in the light and dark area’s: it will just be black and white.

ISO

This is to set the amplification of the signal, like the ‘gain’ setting on videocamera’s. Gaining the signal will produce noise. So it is good to know the native ISO value: the value without analogue or especially without digital amplification. There are two things to note here: the native ISO value for video might be different then the photography value, due to pixel binning or line skipping to get the 1920×1080 video resolution. The other thing to note is that ISO values differ with white balance settings (see below) and that it will be better to gain in the camera before reducing the signal to a 8 bit studio swing YCC signal and compressing it. More about this later.

Also be aware of the ISO bug. If you want to be safe, just stay at the 200-400-800 etc range and don’t go up to 320, 640 or 1250 ISO. A good workaround is to set your camera (using C1…C3 memory) to 1600 ISO default and then dial down to the ISO number that’s best. Never move up in ISO.

One other thing to keep in mind is that noise is not necessarily a bad thing. Shooting at 160 ISO might be the theoretical best thing to do (assuming the native ISO is 160) but this might introduce more gradient banding then say 400 ISO.

Film modes

With the GH2 you get a few film modes, like the picture profiles on the Canon camera’s. There is no S-log (Sony) or Technicolor CineStyle profile (Canon DSLR) but there are a few film modes, each with it’s own (dis)advantages. One of the big users on PersonalView is Driftwood. He uses nostalgic, cinema, smooth (in that order).

  • Nostalgic is said to be the best choice, but others report a yellow cast on the midtones;
  • Cinema is said to have bad highlight roll off (oversaturation that looks like color banding) and crushed blacks;
  • Smooth is said to be the best ‘flat’ profile but comes at the expense of a bit of noise;
  • Nature should give the best skintones;
  • Vibrant has a high contrast and saturation, and could be the least noise of the pack.

Here is a comparison of the curves done by DP review (so this is in photography mode). You can see clearly that nostalgic gives you the highest latitude (8 stops) and that cinema has the worst highlight roll off. There is one note to make: this is recorded in daylight while this particular sensor is at its best in tungsten light.

To line up three modes next to each other (to avoid skipping through modes you don’t use you can save two as ‘my film 1’ and ‘my film 2’ and select the third as ‘multi film 1’. When recording motion pictures the multifilm 2 and 3 will not be used.

Contrast, Sharpness, Saturation and Noise Reduction

For each film mode you can set details by setting the Contrast, Sharpness, Saturation and Noise Reduction.

  • Contrast and saturation are important subsettings that can be set from -2 up to +2. I’ve read in reviews that the saturation changes are too subtle, but maybe this is only with contrast set to -2. I’ll dig into this later.
  • Set the sharpness to -2 because otherwise you could get choppy movements (see my test of Nostalgic below). I’ve found out the hard way when using the cake 2.0 profile. Even then the sharpening could make it choppy, so it’s best to keep the shutter at 1/25 or 1/30.
  • Most users say it is best to keep noise reduction set this to -2 as they say it is better to do the noise reduction in post with a plugin like NeatAE. I am not sure yet about this for the same reason as setting the ISO value right. But have not tested it yet.

Test: Nostalgic

This profile has the highest DR. Personal-view user davidhjlindberg was using Nostalgic -2+2+2-1 with Orion V4b CBR. I’ve tested this profile on a job with cake 2.0, my leica, canon and panasonic lenses and found that it chopped up my pans and other fast movements, even with the shutter set to 1/50. I’ve read a lot about other users having the same problems with the sharpness settings. So i would always keep the sharpness at -2 with this film mode! Also because I read in the manual on page 117 that ‘In Film Mode, the camera may do something equivalent to pull processing or push processing to create a picture with particular characteristics. The shutter speed may become very slow at this time.’ Could it be that the shutter speed gets faster to give us the higher DR for highlights of this Nostalgic film mode? I don’t know but this seems to be a clue.

The other thing I found is a lot of noise in the shadows (even at ISO 160 and 200) and the yellow cast on the midtones. I could not really get rid of it without the image turning magenta. I’ve read you could correct for this with setting the white balance to +2 Blue. But for me, when I don’t get a lot of postproduction time (see this guide to go through neat-video, sharpening etc) it’s a no go on Nostalgic because of the noise.

Test: Smooth

Personal-view user terry2 uses smooth -2,-2,-2,-2 at 320 ISO. I’ve tested it outdoors with my Leica lens and later with cake 2.0 at ISO 160 ETC mode with my fujinon ENG lens. I knew the fujinon was giving a glow at f/1.7 so I used that as my ‘look’: a little over-exposed. So far I am really happy with this look.

Test: Vibrant

Vibrant might be the best option for a vivid image with the least noise. It might come at the expense of a reduced DR, but it might not be a bad thing if it doesn’t crush the blacks and has a nice highlight roll off.

Advertisements
Posted in: GH2