• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

Norwood's dome revisted

Doug Kerr

Well-known member
In another thread in this forum I showed a picture of a Sekonic L-398A exposure meter, the latest (and possibly last) model in a series that is described as the "Norwood director" series (after the name of several early members of the family). Here we see it:

ND_L-398A_Front_F42515-S400.jpg


Douglas A. Kerr: Sekonic L-398A exposure meter

The photo clearly shows what is the hallmark of the Norwood Director series, a translucent essentially-hemispherical dome "light receptor". In fact today many exposure meters, in their "incident light" configuration, feature a similar dome. (The Norwood Director meters are "incident light" meters in their basic configuration.) It is interesting to reflect on what this ingredient does, and why we want that.

It is widely (but incorrectly) believed that this "dome" receptor is used to give the meter the ability to accurately determine the illuminance on a certain surface, or a certain plane, regardless of how the illumination may be composed of components arriving from different angles. Technically, to do that requires that the instrument have a "cosine" directivity.

But the dome does not give the instrument a cosine response. In fact, the Norwood Director meters have an alternate "receptor", almost flat, that in fact gives the meter a cosine directivity, so it can truly measure the illuminance upon a certain surface (or a certain plane). Some Sekonic meter models have a dual-mode receptor, a dome in its basic configuration, that can be physically changed to yield a cosine directivity, for actual measurement of illuminance.

So what directivity pattern does a hemispherical receptor give, and why do we want that?

Well theoretically, it will give an close approximation to a cardiod pattern (a pattern found, in the electro-acoustic domain, in many microphones).

And it has been determined, empirically, that this directivity pattern, in the context of the key-fill lighting technique, for a human subject (with special attention to the face), over a range of keylight locations (in terms of angle from the camera axis), seems to make the meter give exposure recommendations that produce "good" exposure results.

Now of course we have to deal with the fact that, especially in the case of substantial keylight angles, just what constitutes a "good" exposure result is very subjective. Do we want the side of the face nearest the keylight to be brighter in the image than we would want the whole face to be with "head-on" lighting? Maybe yes, maybe no.

Now, is there some theoretical relationship that wouild show us why this is so? No. And in fact the seminal paper on this matter, published by Don Norwood, who introduced the hemispherical receptor meter, if it were free from error, would show that the "optional" result, in the sense I mentioned above, would come from a slightly different directivity pattern. However Norwood, in that paper, made two errors such that is seems as if the pattern to be expected from a hemispherical receptor would be ideal - exactly ideal (fancy that).

Now, the discrepancy caused by Norwood's errors is not substantial, and we are talking here about a highly subjective assessment of the result anyway, so we might reasonably conclude that a hemispherical receptor is very suitable to pursue Norwood's concept.

Now, I have said that the objective of Norwood's concept is to make the meter provide good exposure recommendations for a human face (head, actually) over a range of keylight positions in the overall context of key-fill lighting. What about in other lighting situations?

Well, to take an extreme alternative, consider the situation of multiple-source lighting on a flat subject (perhaps a painting). Then, we are more likely to get a "desirable" exposure recommendation from a meter with a "cosine" directivity (which measures the actual, uniform illuminance on the subject. So with a Norwood Director type meter, we put on the alternate receptor. For certain other Sekonic meters, we switch the multi-mode receptor to the "cosine pattern" mode.

What about, for example, using incident light metering on a human subject in a typical outdoor setting? Will the "cardiod" directivity pattern (with the "hemispherical" receptor) or the cosine pattern (with the alternate receptor) give the "best" exposure recommendation? It is hard to tell. One's mileage may vary.

Those interested in further details of this matter may wish to red the technical article, "Norwood’s dome: a revolution in incident‑light photographic exposure metering", available here:

http://dougkerr.net/Pumpkin/index.htm#NorwoodsDome

Best regards,

Doug
 

Jerome Marot

Well-known member
Interesting.

I think that the dome works because human skin is not a perfect diffusor. The camera does not record incident light, it records light reflected from the model's face. If human skin was reflective like a mirror, exposure readings would not be done with an incident meter. If human skin were a perfect diffusing surface, probably the cosine response would be appropriate. But human skin, as all natural surfaces, is intermediate between a mirror and a perfect diffusing surface: it diffuses but is still a bit shiny. It may be that the pattern of the dome is close to that.
 

Doug Kerr

Well-known member
One of Don Norwood's early (but I think not initial) visions was that the hemispherical receptor was an approximate proxy for the portion of a human head that could be seen from the camera position.

The, for any lighting situation, the amount of light collected by the receptor (and the meter reading would be proportional to that) would be proportional to the total amount of light striking the visible-to-the camera portion of the subject head.

That has a very nice sound to it.

Norwood then jumps to, "Thus, the meter reading will tell us the photographic exposure needed for a proper exposure result for the shot."

But there is no clear reason why that should follow.

One of the complications of metering for a key-fill lighting situation (where the key light is aimed from some angle to the side of the subject and the fill light, typically less "potent", is typically aimed from the camera location or thereabouts) is just what do we want to be the exposure result on, for example, the part of the head facing the key light and on the part facing the fill light and the camera.

There is no direct answer to that.

Norwood, in a set of tests done quite a while after he introduced his hemispherical collector concept, attempted to get an empirical answer to that question. He reported the results in a seminal paper before the Society of Motion Picture and Television Engineers in 1950.

He used a test setup in which a human head (and he used several subjects over the entire series of tests) was illuminated by a fill light from the camera position and a key light (always 8 times as potent) aimed from a variable angle to the side.

Then, for each subject, a reference shot was taken with the key light "head on", with the exposure set by metering to follow the classical norms of "standard exposure". Then the key light was moved to various positions to the side. For each key light position, several shots were taken with various exposures above and below that used for the "head on" shot, at 1/2 stop intervals .

Then, as Norwood tells the story, a panel of viewers were asked, for each series of exposures, for each keylight position (angle), to judge which exposure gave a "matching result" with that from the "key light head on" reference shot.

At this point we need to consider in just what way the viewers were asked to judge a "matching effect" It couldn't be that the two prints looked "the same", as the pattern of brightness across the image of the subject's face and head would differ substantially over the various key light positions. So perhaps the viewers were asked to choose which shot seemed to have an overall exposure result that was "equally as nice" as for the head-on reference shot.

And of course that might vary considerably with different individual preferences for "dramatic" lighting. We just don't know.

In any case, for each key light position, Norwood determined the average photographic exposure, compared to that used for the "head-on" reference shot, that yielded what the viewers judged to be "a matching result" compared to the head on shot.

Norwood then took the inverse of that, and considered that be a factor of the "relative effectiveness" of the overall illumination for that key light position (angle). He then reasoned if the relative response of the exposure meter to light from that angle (as dictated by the "directivity pattern" of the meter) followed that "relative effectiveness" factor as a function of the keylight angle, then the meter reasoning would, for any key light position (angle) lead to the optimism recommendation of photographic exposure.

(There was in fact a clinker in the details of this, but I will not bore you with it now.)

Norwood then went on to state that the theoretical directivity pattern of a hemispherical-receptor meter would be exactly the needed pattern (fancy that). (In fact, here he misreckoned the theoretical reactivity pattern of a hemispherical-receptor meter.)

But fortunately, these two discrepancies were not large, numerically, and in any case this chain of reasoning was predicated on a greatly-subjective assessment of the results of different exposures in the different lightning situations., so even if all steps were done "correctly" this would not have led to an "exact" answer.

But we have to wonder whether, in Norwood's paper, he was not a little too anxious to demonstrate an exactly ideal relationship.

But in any case, it seems that Norwood, having had (for who knows what initial reason) the vision of the benefits of aa hemispherical receptor, was lucky in that it seems to empirically produce good exposure results over a range of key light positions when using key-fill lighting of a human head.

Best regards,

Doug
 

Doug Kerr

Well-known member
This little parable will help to illuminate (!) the difficulty of trying to devise an "ideal" scheme for incident-light metering.

Conceptually, incident-light metering seeks to provide an appropriate exposure result for each scene element based on its reflectance. As we often say, it seems to make a white cat on a snowbank look like a white cat on a snowbank, and a black cat on a coal pile look like a black car on a coal pile.

Suppose we begin by saying, as we often do, that our objective is that a spherical neutral object (a "ball") with a reflectivity everywhere of 18 percent, when exposed with the photographic exposure recommended by our meter, will end up with an exposure result we consider "reference gray".

Now suppose that, for some kind of product shot, our main subject is a spherical ball with a "diffuse" gray surface, interestingly-enough with a reflectivity of 18% (and by "diffuse" I mean specifically Lambertian). Recall that if it is uniformly illuminated (in the three-dimensional sense), it will exhibit the same illuminance across its entire "disk", regardless of the angle from which it is observed.

In our first test, we in fact illuminate the ball with absolutely-uniform illumination (imagine from the inside of a white integrating sphere). We measure the illuminance on the ball with our meter, and set the camera per its exposure recommendation. We take the shot, and sun of a gun - right on the money. In the image, the disk of the object has everywhere the "reference gray" exposure result. A theoretical triumph.

But a practical failure. Instead of seeing a "ball", we see a "uniform gray disk". Oh, crap.

So we give up the uniform lighting and instead light the ball with some combination of head-on and side lighting. And sure enough, now the image looks like a ball.

But what would we now consider "correct" exposure? Should the nearest portion of the ball (we will assume for a moment that is the brightest) exhibit "reference gray" in the image? Or maybe something else? Our classical ideal of proper exposure by way of incident light measurement doesn't really seem to be applicable to such a real-life object. It would probably be good for a "flat" object.

Now, if we, like Don Norwood, seek to contrive an exposure meter that, in a single measurement, will give us an exposure recommendation that will consistently lead to the "correct" exposure result, for multiple-source lighting of three-dimensional objects, just what might that correct result be?

And therein is a dilemma that infests and frustrates this whole matter.

Finally, so it shouldn't be a total loss, I close with an image of a subject that is definitely Lambertian, but certainly not even close to diffuse, gray, or flat:

7489c734a5d54d462ea46f69700da37d.jpg


Miranda Lambert​

I would judge the exposure here to be good.

Best regards,

Doug
 
Top