I've been thinking quite a lot more about this, hence the long post.

A: Accuracy

We expect and get surprisingly high accuracy we expect from a foot-ruler (maybe +/-1% at 1 inch or 25mm, as little as +/-0.15 per cent at 12 inches or 30cm for a finely graduated engineer's steel rule) and from a quartz clock: a mimute a week, entirely commonplace, is 1 part in about 10,000. These are WAY greater than the accuracy we expect or get from an exposure meter.

B: Standard source

The original standard candle was of spermaceti wax, weighed 6 to the pound, and burned at 6 grains per hour. By modern scientific standards it was hopelessly inconsistent but as far as I am aware a modern paraffin wax household candle of similar weight, with the wick trimmed for maximum brightness without smoke, is very close to a standard candle, and a standard candle exhibits remarkably little variation, probably less than the +/- 1/3 stop of an 'in-spec' light meter: the Ainger Hall spot meter used a candle flame for regular calibration. With a standard candle and a darkened room, it should be feasible to meter the flame directly; or to construct a transmission target without further assumptions (measuring the opacity of the diffuser directly); or to make a reflection target with the assumption that bright white paper reflects 90 per cent of the light falling on it.

C: The usefulness of such calibration

First, it is at a single (low) colour temperature. From memory the standard white light for photometry is filtered to 4850K, and variations in colour temperature could make for significant variations in meter reading, depending on the cell's sensitivity.

Second, it is a single fixed point. The intelligent way to use it would be for shadow speed point calibration for negative film and highlight speed point calibration for colour slide, but many would want to use it to calibrate to an 'average' reflectivity of 14 per cent and a good few would use 18 per cent because they know no better. The big problem is that the meter might have very different sensitivity at different illumination levels, and there is no calibration for linearity.

Third, the exposure meter equations make a LOT of assumptions. The first, concerning image brightness at the film plane, includes object-to-lens distance, lens transmittance, camera flare, lens barrel vignetting, and the average angle of off-axis rays. The film speed equation also incorporates a constant K and a safety factor. K in turn was based on psychophysical testing, i.e. popular acclaim.

Better still, many of these assumptions have changed over the years with new generations of standards: for example, as well as K (now represented by K0) we have K1 (sorry, can't do subscripts) and the 1971 ANSI standard makes specific reference to different values of K for projected film.

All assumed values are based on 'typical' values and may depart quite widely from the actual values encountered by a specific photographer using specific equipment: this is why manufacturers clearly state that ISO (and earlier ASA or DIN) values are a starting point only. Anyone who wants to see the equations in all their glory, with an explanation of the associated assumptions, can find them (among other places) on page 79 et seq. of Perfect Exposure, Hicks & Schultz, David & Charles/Amphoto 1999.

All this explains why such things as the ZS and BTZS exit, and why personalized EIs are a good idea. It also explains why I suggested early in this thread that the best idea was to go out and take pictures using the meter. I should have added, 'after the most casual of tests to reassure yourself that it is working at all, whether you use Sunny 16 or compare it with another meter.'

Cheers,

R.