http://www.youtube.com/watch?v=FXjrXdi6 ... e=youtu.be
So I'm wanting to expose 16mm film to a computer screen (using the above), and I'm asking myself what pixel value (what shade of grey), I can display on the computer screen, to use as the equivalent of a grey card for my light meter (because obviously I can't do a conventional incident light reading as there isn't any incident light - the screen is self luminous and the rest of the space is in darkness). But as it turns out I can use an incident light reading but first - the journey ...
So I'm more interested in the theory on this rather than the practice. In practice I can just bracket some shots and eyeball the results (and I'll be doing that) - but what would I use in theory?
So as a start I measured a number of pixel values with my light meter (using reflected light readings) for two different screen settings (mid contrast and low contrast). There's a bit more info in the diagram than this (relating to film sensitivity and remapping such back to pixel values) - the light meter measurements are in the yellow and red curves. The vertical axis on the left is pixel values (0-255) and the horizontal axis along the bottom is light meter values in EV.
So for example, a pixel value of PX:128 reads as
EV:5.3 on the low contrast screen, and
EV:7.4 on the medium contrast screen
Now while a PX value of 128 is the middle of the pixel values (0 to 255) that doesn't mean it corresponds to the same light reflected from a grey card.
What exactly is reflected from a grey card? A grey card is defined and designed to reflect 18% of the incident light otherwise falling on it. Light meters, perversely, are calibrated to treat reflected light readings as if they were reading 12.5% (rather than 18%) of the incident light. So when you take a reading off a grey card, the light meter will actually infer an incorrect incident light since it will be multiplying the reading by 800% (reciprical of 12.5%) rather than 555% (reciprical of 18%). The multiplication is typically done physically rather than electronically, ie. by having the reflected light head admit 800% of the light that would otherwise be admitted by the incident light meter head.
What is 12.5% of the incident light? In terms of stops it is 3 stops less than 100%. 18% is 2.5 stops less. So the light meter is designed to assume, when doing a reflected light reading, that the object is reflecting 3 stops less light than the incident light.
This is useful information. But what do I treat as the incident light level for a computer screen? In reality there is no incident light. It is zero. But in terms of the range of pixel values the incident light can be regarded as the light emitted by the brightest pixels ie. by white pixels.
So theoretically all I have to do is take a reflected light measurement of white, in terms of EV, subtract 3 stops, and then look up the corresponding pixel value in the graph. And use that pixel value as grey.
This turns out to be, PX:74 for both the high contrast screen, and the low contrast screen.
But what about the fact that grey cards reflect 18% and not the 12.5% that light meters are calibrated to assume? If a grey card reflects 18% of the incident light, it is therefore reflecting half a stop more light than the light meter assumes it is reflecting. The light meter, multiplying this 18%, by 800%, would arrive at an incident light reading that was half a stop brighter than the reading it would otherwise provide were it taking an actual incident light reading (all else being equal).
What are the reasons for this difference?
I trace it back to a difference between the way print theory works and the way light theory works. A value of 12.5 % is logarithmically neat and tidy. It is exactly 3 stops down from 100% (100/2 = 50 > 50/2 = 25 > 25/2 = 12.5). So where does 18% arrive? I suspect it's related to Ansel Adams zone system (and the history from which that is derived) where one subdivides a range of values in terms of zones (or what is called "bins" in computer image processing). In other words Adams is more interested in the region or zone between two values, rather than the values which otherwise frame the zone. So for example, in relation to incident light (treated as 100%), the first zone would be between 100% and 50%, the next between 50% and 25%, the next between 25% and 12.5%. Given a zone in which grey appears between 25% and 12.5% the midpoint of this would be 2 ^ [ log2(25%) - 0.5 ] = 17.678%, or rounded: 18%.
That's the origin of the 18% figure I reckon.
Ultimately it doesn't matter what figure you use as a reference. What is important is that the grey card and meter should match each other, either both 18%, or both 12.5 %, but perversely (historically) they do not. If you mix and match incident light readings off incident light, and reflected light readings off a grey card, you'll get 1/2 stop jumps in exposure between the two (all else being equal).
So back to the computer screen. Since we're not using a grey card, and only a light meter, all that matters is that we satisfy what the reflected light meter assumes, and that is that it's reading 12.5% of the incident light, or 3 stops less light than white.
But if all this is correct then it's also all unnecessary - if the theory is correct then one should be able take an incident light reading instead, with the meter facing the screen, (with the screen displaying 100% white rather than some notional grey level), and use that. A quick test between measuring PX:74, using reflected light reading (and a whole heap of measurements to find this pixel value), and measuring PX:255 (white), using a single incident light reading (facing the screen) gave exactly the same EV measurement. As the theory otherwise implies and predicts.
So if the theory is correct then one can forget about measuring a whole heap of grey levels to find the equivalent of a 12.5% grey card and measuring that, or measuring white with reflected light meter and subtracting 3 stops ... for one can just measure white using an incident light reading instead (facing the screen) and one will arrive directly at exactly the same EV (ie. same camera setting).
The next task is to see what happens with actual filmstock.
Carl