How Light Meters work

Trying to explain why the hot shoe TTL and Commander/Remote CLS Nikon SB-800 flash pictures sometimes come out underexposed.

D300, all images used Center metering. All were RAW, and the only processing was to click Flash White Balance in Adobe ACR, and the smaller resample to JPG.

You can click on any image to see a larger picture.

A scene with much brighness (a white towel fills most of area). Manual flash mode, with exposure manually metered with hand held flash meter, at 1/8 power at f/5.6. Note that an incident meter or a flash meter meters the LIGHT, and not the subject. An incident meter does not see the subject. dsc_1443.jpg 170.36 KB	Same scene with much brightness This is hot shoe TTL metered flash (on SC-17 hot shoe extension cord) It is about two stops underexposed, which gives me displeasure, but the reflective camera meter is influenced by the white towel, so I think it is doing what it is designed to do in this situation (see card section at bottom of this page). This is an expected result. dsc_1448.jpg 160.46 KB
The only one change is the metering method. Incident vs reflected To show how much is one stop on the histogram:
Scene with much brightness CLS Remote flash with +1 EV flash compensation in D300. dsc_1449.jpg 174.41 KB	Scene with much brightness CLS Remote flash with +2 EV compensation in commander (a little too much, detail in white towel is blown out) dsc_1450.jpg 164.35 KB
Shifting from hot shoe iTTL to Remote CLS operation (shows same effect on same subject)
Scene with much brighness CLS Remote TTL flash, with no flash compensation dsc_1451.jpg 160.10 KB	NOT an untypically bright scene CLS Remote TTL flash, with no flash compensation Hard to fault it. dsc_1452.jpg 284.44 KB
The only one change is that the white towel is removed. (see card section at bottom of this page)
NOT an untypically bright scene Flash in Manual mode - exposure metered with flash meter at 1/8 power at f/5.6/ Good, the pink paper remains pink. Same exposure as first shot of white towel at top of page. dsc_1455.jpg 287.83 KB	Overall scene FWIW Averages to near normal middle gray expectation. TTL flash via hot shoe SC-17. dsc_1456.jpg 231.08 KB

What Light meters do

Camera TTL light meters do not have human brains to know what we are pointing them at. The meter cannot recognize a snow covered mountain from a black Lab dog. Human brains know expected result instantly, but the meter cannot recognize the subject, and does not know if the subject should be black or white or pink, or if it should be light or dark. They only see some degree of reflected light intensity coming back through the lens. But it turns out that most average scenes contain a range of various degrees of brightness distributed over the full width of the histogram (a shiny face and dark hair for example, or sky and mountain for example), and specifically, all the tones in most scenes typically do average out to be near a middle gray overall.

Reflective light meters simply always assume every scene averages out overall to middle gray, and the light meter's goal in life is to make the picture exposure average out to be middle gray. Normally this is correct for average scenes, but there can still be exceptional non-typical scenes when this fails somewhat.

Techie stuff, but note that middle gray is NOT at the midpoint of the histogram range (not 128 on a 0..255 scale). An 18% gray card should be at 18% in liner data (at the RAW camera sensor, before gamma encoding), and which comes out at 117 after gamma encoding, 0.18^(1/2.2)x255 = 117 (46%). This is near enough the midpoint that it confuses us, but the coincidental reasons are quite different. The ANSI standard for light meters is 12% reflectance, not 18%. Kodak says metering from their 18% gray card should be compensated with 1/2 stop more exposure. One stop down from histogram maximum right end is at 128 in linear data, but what we see is at 0.5^(1/2.2)*255 = 187 (73%) after gamma encoding. Two stops down is 0.25^(1/2.2)x255 = 136 (53%). All RGB images are gamma encoded, we see nothing else. See Adobe PDF Link.

NOTE middle gray does not necessarily mean gray, except when discussing color balance. When discussing exposure, it could be red or blue or green (a mid-tone tonal concept, between darkest and brightest), and it refers to the equivalent tone value present in the (gray) histogram.

The design of any reflective light meter is that when metering something all gray, all black, or all white - whatever it might meter - all are supposed to come out all middle gray. This is how simply reflective light meters work. It may not be the way humans think however. :) Reflective meters are designed to measure an overall composite scene which OUGHT to average out to middle gray, but which may not always be the case.

Incident light meters (like flash meters) instead are pointed the other way, at the light instead of the subject. They instead directly measure the brightness of the light (at the subjects location), and give a corresponding exposure so that middle gray things should come out middle gray, and white things come out as white, and black things come out as black. This is a much more consistent reading but cannot be implemented using the camera lens.

Card Section

Photographing three Gray, Black, and White cards with camera light meter

Gray card with TTL flash on SC-17 cord The small pink paper with text disrupts the metering, guarding against over exposure of the peaks. Middle gray is made to be middle gray. dsc_1457.jpg 129.04 KB	Black card (a fun foamy) with TTL flash on SC-17 cord. The metering used much more flash power (exposure) to make the black card be middle gray. It makes couch and pink paper be white. dsc_1458.jpg 202.44 KB
White card (a fun foamy) with TTL flash on SC-17 cord. The metering used much less flash power (exposure) to make the white card be middle gray. Makes couch black and pink paper dark. dsc_1459.jpg 53.64 KB	The three cards used (TTL metered). Really is black and white and gray. dsc_1460.jpg 127.98 KB

The point is, this is simply what light meters do... tries to average every scene to middle gray. This will usually come out near correct for average scenes. But it can guarantee underexposure for bright scenes, and overexposure for dark scenes (speaking of the "color" and "reflectivity" of the scene, not the light that is on it).

This example uses camera TTL flash metering, but you would get similar results with tungsten or sunlight... any reflective meter strives to give middle gray result for any scene it sees. Most typical scenes do in fact average out about at middle gray, so it normally works out fine.

We need to understand this, regarding the amount we compensate the flash. The flash has no idea what the scene is, or how it should look, it only sees the overall average, which is assumes should be middle gray. But we know what the scene is, so it becomes our job. If metering white things (say a wedding dress), you proably need +1EV compensation to keep it white. If metering black things, then probably -1EV compensation to keep it black. This depends, check your results.

Seeing this here is one thing, but to really grasp it and make it yours, you should of course repeat this simple experiment yourself.