Basic rules of exposure

[Robert Horwell]

Ok, i have read lots on the forums regarding exposing images on the RED, i have to say, lots of it is beyond me. Linear light etc...I come from a film background and as such am very used to relying on my trusty light meter and the knowing the inherent qualities of different stocks.

So far on my RED i have been finding my exposures using, my meter ( mainly for contrast ratios ) the traffic lights in my EVF and my eye and instinct. I am not used to exposing from a monitor so when i find myself doing it it just feels so wrong, but i am getting better at looking at my monitor more often.

It would be great to have straight forward bread and butter descriptions of your methods of exposure, and can i highlight the philosophy of KEEP IT SIMPLE, i can only speak for myself here, but i get very lost very quickly as soon as the speak gets to tech orientated.

So to clear up my humble request.

1. What in camera meters do you use, what is your understanding of what those meters do.

2. What other methods do you use, what equipment, and how do you use it.

3. Favourite in camera settings.


In my quest for exposure, in advance i thank you.

 

[David Rasberry]

Personally I think using film exposure techiques with good light meters is still the best way to go. Even for normal EFP style video shoots I do this. Monitors are poor exposure references unless they are accurately calibrated and in a controlled environment, like a studio control room or production truck. On set they are unreliable for accurate exposure.
I use Sekonic incident light meters mostly. If the camera (I don't own a red yet) has metering capability, I use it as a spot meter zoomed in to check highlights and shadows.
I would like to know more about using histograms as exposure references, this is something I have used in audio engineering, but not for photography.
On video cameras I calibrate to an 11 step log grayscale chart using a waveform monitor to get ASA/ISO rating for different setups. I don't know how or if this could apply to setting up a RED.

cheers

Razz

 

[Sidney L. Plaut]

I use the false color exposure meter in the camera. very simple in my opinion. Shadows are dark blue to lighter blue. 18% grey is green. "hot" areas are yellow to Red (burnt out).

So I expose using this and the stoplights to make sure im not clipping any channels.

I aim low - avoiding Red. kinda like super zebra´s.

Of course if you want contrast ratio a meter is prob. still the easiest. My gaffer uses the meter. I expose with false color.

 

[Rob Castiglione]

Hi Sidney,

This exposure business is in my view the trickiest thing about the camera and is driving me a little bit crazy with all the options. I have largely been using the histogram.

The false meter sounds great but there was a recent post suggesting that it was not an accurate representation of the raw capture. How have you found it in terms of an accurate representation of exposure once you have had an opportunity to actually view your footage in post?

Rob

 

[donatello b]

1) use luminance .. RAW meter ... keep eye on stop light .. view redspace or 709 (both affect stop light & luminance meters-so keep that in mind) ...

2) use meters to set lights (incident & spot) .. use meters to get approx. exposure ...
final F stop is "experience" = based on the meters and Red meters ... i do not light to monitors ..
for run n gun ... no hand meters .. go with Reds 3 meters in #1

3) don't use any of the custom settings ...basically set WB to 3200K or 4400K ( sometimes 3800k) or 5600K
... 1/48th shutter ( sometimes 1/40th ) .... ISO to 320 ... up to 640 night exteriors ..

 

[spitfire44]

I use a combination of the following three..

Eye, Meter, Zebra.

I trust my eye enough to not worry about being under...the Zebra keeps it inside and the meter, looks good on the hip and is used for a ballpark on overall exposure.

As far as the monitor for exposure, possible but don't count on it.

Eric

 

[kay madsen]

my basic rules of exposure

my basic rules of exposure

The RED is the equivalent to a 320 ASA, 5000K Film stock. That´s the setting where it performs best.

To expose it I mainly use the histograms and the false color mode. I use my light meter only to check contrast ratios when lighting interior sets.

Check this file for a simple description of histograms:
http://madsenfilm.de/downloads/histograms.zip

False color mode is like a more advanced zebra. Took me a while to get my head around it, but I have fallen in love with it eventually. Think of it as a colored version of Ansel Adams "zone system". It shows you exactly where in your image your shadows (shades of blue), mid tones (green equals 18% grey, shades of grey and pink) and your highlights are (yellow, orange and RED for clipped highlights).

Important note: the histograms and false color mode are tied to the color space you are working in. If you work in Red Space or REC709 your highlights will clip much earlier than in RAW. Set one of your user buttons to RAW view so you can quickly check how much more leeway you still have in the RAW file before you clip. This can be very useful in shots with extreme contrast. For example a slightly overcast hot sky in the background and a person with no extra fill light in the foreground. In situations like that you want to make sure you use the maximum latitude of your RAW file to retain as much detail in the sky as possible without underexposing the face too much. That´s when false color mode is a great tool to help you balance your exposure.

I always carry a couple of ND grad filters and a 2-3 stage mattebox with rotatable filter trays to selectively shade certain parts of the frame and control contrast.

And last but not least it´s important to get your head around the fact that out of the 4096 tonal values you have from black to white in your linear 12 bit file, half of them are used for the brightest exposure value in your image. Your brightest highlights use 50% of the information in your file, from 4096 down to 2048 tonal values. The rest of your image has only 2048 tonal values left to represent the rest of the information in your image, your midtones and your shadows. For each f-stop less your amount of tonal values is divided by two:

4096 - 2048 = your brightest highlight - 2048 tonal values
2048 - 1024 = 1024 tonal values
1024 - 512 = 512 tonal values
512 - 256 = 256 tonal values
256 - 128 = 128 tonal values
128 - 64 = 64 tonal values
64 - 32 = 32 tonal values
32 - 16 = 16 tonal values
16 - 8 = 8 tonal values
8 - 4 = 4 tonal values
4 - 2 = 2 tonal values
2 - 1 = 1 tonal value

The amount of tonal values you have to represent your midtones and shadows in the linear 12 bit file is substantially smaller compared to the amount you have for your highlights.

Check the uncorrected image I attached. I made this shot with a single 1200 W HMI chimera soft light. I tried to be really clever and save every hightlight detail, even in the brightest reflections on the edge of the piano. The problem with my exposure: 2/3 of the tonal values are used for just the highlights in the face and the few pings, the brightest two stops in the whole image (see 2nd version of image, colored in red). The rest of the image, all the midtones and shadows, have only around 900-1000 tonal values left out of 4096. As soon as I color correct that image I get a lot of noise in the midtones (especially in the wall around the microphone stand).

Looking back I should have used a smaller contrast ratio, a bit more fill light for the background or I should have opened up the lens at least one more stop. I would have lost detail in the hightlights on the piano, but who cares? I would have gotten more tonal values in the mids and shadows and less overall noise.

What does it mean? The more you underexpose (histogram leaning towards left side) the less tonal values you use. The more you overexpose (histogram is leaning toward the right side) the more tonal values you use. Use more tonal values and you will get less noise and less artifacts in general. If possible try to light with smaller contrast ratio, you can always dial in more contrast in post production.

That´s my two cents. Comments are welcome...

Cheers,

Kay

 

[Robert Horwell]

This is BRILLIANT please keep it coming....



MERRY CHRISTMAS!! HO HO HO..

 

[Graeme Nattress]

Kay, that's a great idea to mark on the histogram, but you've forgotten that the bit allocation is in linear light, but the histogram is representing the image of the gamma corrected data - unless you did some image manipulation to put in a linear light histogram?

Graeme

 

[kay madsen]

Kay, that's a great idea to mark on the histogram, but you've forgotten that the bit allocation is in linear light, but the histogram is representing the image of the gamma corrected data - unless you did some image manipulation to put in a linear light histogram?

Graeme

Hi Graeme,

thank you for your remark. I was completely ignorant of the fact that there is a difference between the two. But you are right. I guess the images still help to get across the general idea of exposing to the right of the histogram in order to use as much tonal values as possible.

I have made another screenshot of the same image using camera RGB as color space and linear light as gamma space... is that the correct way to represent the linear bit allocation? If it is, I used even less tonal values the way I exposed my image... looks like sever underexposure... no wonder I get a lot of noise when color correcting that image...

I checked a couple of different shots with those settings and was a bit shocked to see that in most shooting situations, even those that are very bright with flat light, rarely much more than half of the tonal values are used... looks like half of the 12bits get wasted most of the time... another reason to expose to the right of the histogram...

btw, one more question for you... if I scrub through the histogram in RedAlert, I get 0-65535 tonal values. Is that because RedAlert uses 16bit processing?

Cheers,

Kay

 

[Graeme Nattress]

Yes Kay, if you're looking at bit allocation, your image is now correct!
REDAlert uses float for a lot of the internal processing, resulting in a 16bit image. It's that 16bit image that is used for the numbers you see.

Graeme

 

[Robert Horwell]

Kay, that's a great idea to mark on the histogram, but you've forgotten that the bit allocation is in linear light, but the histogram is representing the image of the gamma corrected data - unless you did some image manipulation to put in a linear light histogram?

Graeme

Okay this is where i start to get confused...Graeme, Linear light? Please, how does this work, a laymans description. Thanks.

 

[Robert Horwell]

Hi Graeme,

thank you for your remark. I was completely ignorant of the fact that there is a difference between the two. But you are right. I guess the images still help to get across the general idea of exposing to the right of the histogram in order to use as much tonal values as possible.

I have made another screenshot of the same image using camera RGB as color space and linear light as gamma space... is that the correct way to represent the linear bit allocation? If it is, I used even less tonal values the way I exposed my image... looks like sever underexposure... no wonder I get a lot of noise when color correcting that image...

I checked a couple of different shots with those settings and was a bit shocked to see that in most shooting situations, even those that are very bright with flat light, rarely much more than half of the tonal values are used... looks like half of the 12bits get wasted most of the time... another reason to expose to the right of the histogram...

btw, one more question for you... if I scrub through the histogram in RedAlert, I get 0-65535 tonal values. Is that because RedAlert uses 16bit processing?

Cheers,

Kay

So. kay, lets say we are shooting a high contrast scene, how do we get more info info in the image to the right of the histogram, i.e using more of the 16 bits available...more light? then pull the image about in the grade?

 

[Graeme Nattress]

Linear light basically means that if you double the brightness, you double the code number that represents that brightness.

Graeme

 

[kay madsen]

Yes Kay, if you're looking at bit allocation, your image is now correct!
REDAlert uses float for a lot of the internal processing, resulting in a 16bit image. It's that 16bit image that is used for the numbers you see.

Graeme

If I look at the highlights in the linear light version of my image, I end up with apx. 7%-99% of the tonal values used only for the brightest highlights... more than half of the 4096 tonal values are wasted by the reflection on the edge of the piano... only around 130-150 tonal values are left for the midtones and the shadows...

If I had opened up the lens by one stop, clipped the hightlights on the piano and effectively moved the histogram to the right, I would have gotten apx. 500 more tonal values for the shadows and mid tones...

Comparing the linear version of the image with the gamma corrected version I saw on my monitor on set is quite a difference... with high contrast images like this it really makes sense to check the RAW image and make an informed decision about which highlights are worth protecting...

 

[chuck colburn]

Linear light basically means that if you double the brightness, you double the code number that represents that brightness.

Graeme

So basically the same as a f stop. The reciprocal of the square.

 

[kay madsen]

Okay this is where i start to get confused...Graeme, Linear light? Please, how does this work, a laymans description. Thanks.

Just remember the table I posted before. Double the brightness (open up one stop) and you double the values.

4096 - 2048 = your brightest highlights - 2048 tonal values
2048 - 1024 = 1024 tonal values
1024 - 512 = 512 tonal values
512 - 256 = 256 tonal values
256 - 128 = 128 tonal values
128 - 64 = 64 tonal values
64 - 32 = 32 tonal values
32 - 16 = 16 tonal values
16 - 8 = 8 tonal values
8 - 4 = 4 tonal values
4 - 2 = 2 tonal values
2 - 1 = 1 tonal value

 

[Thom Steinhoff]

Linear light basically means that if you double the brightness, you double the code number that represents that brightness.

Graeme

So, let me get this straight.

Based on the adjusted linear histogram, he is really looking at very little image information and would have had a far richer shot if he just brought the overall light level up and spread the histogram right a bit--even if, in post he brought out the blacks again?

I've noticed in my own shooting that the camera LOVES light and I really try to bring up the base levels and fill in the shadows a bit--even if I'm going to take it away later.

 

[Thom Steinhoff]

Kay, that's a great idea to mark on the histogram, but you've forgotten that the bit allocation is in linear light, but the histogram is representing the image of the gamma corrected data - unless you did some image manipulation to put in a linear light histogram?

Graeme

Graeme,

Since his marked histogram represents linear light, but the hisogram is displaying gamma corrected data which wouldn't be right...

Would it be possible to adjust the halving of detail lines to properly represent what is going on with the gamma correction? I assume they would slide right a bit and spread out a little.

I think lines like this would be very useful. Is it possible? Would it be essentially static--i.e. something you could post a picture of and we use it for reference, or would it be dynamic based on camera settings?

Best regards and happy holidays,

Thom

 

[kay madsen]

linear light

linear light

I tried to elaborate a little on my graphical representation of the linear light issue... here is the same underexposed linear image with the overlay grid representing 4096 tonal values of the 12bit image, overexposed in one stop increments you can see pretty well how the values double with each stop of overexposure...