Three generations of photoelectric cell have powered hand-held and in-camera exposure meters, and each one fails differently. A meter is only as honest as its cell, and the gap between selenium, cadmium sulphide, and silicon is not antiquarian trivia: it decides whether the needle can be trusted in a dim interior, under tungsten, or just after the cell has spent a minute pointed at the sky. Get it wrong and the error lands in the negative as misplaced shadows. Point an uncorrected silicon meter at a tungsten-lit room on Tri-X and it reads the infrared the film cannot see, recommends a faster exposure, and drops your Zone III shadow detail down toward Zone II. The cell, not the photographer, made that decision.

Selenium: Self-Powered, Eye-Matched, Blind in the Dark

The selenium cell was the first practical exposure-meter cell, introduced by Rhamstine and Weston in the early 1930s through the Weston Electrical Instrument Corporation; the Weston Master series became its canonical form. It is a photovoltaic sandwich: an iron base plate, a layer of selenium, and an ultra-fine semi-transparent gold electrode on top. Light passing through the gold drives a current between the gold and the iron, read directly by a galvanometer with no battery in the circuit. Output current is proportional to incident light, which is why a fifty-year-old Weston still works in a fully mechanical body.

Spectrally selenium is well behaved. Its response peaks in the green around 560 to 570 nm, leaning slightly to the red of the photopic luminosity function, which peaks at 555 nm. That close match is why selenium ignores infrared and gave the “forgiving” daylight readings these meters were known for; the 1949 J. Sci. Instrum. paper on selenium spectral-correction filters confirmed the response was near enough to photopic that only modest glass filtering was needed.

The failure is sensitivity. On a Weston Master III the bright range runs from 25 to 1600, dropping to a low range of 0.2 to 50, but below a needle reading of about 10 the scale divisions bunch together and become unreadable. That is the practical floor: a selenium meter handles domestic interior light at a push and cannot register candlelight or moonlight at all. Ageing compounds it. The thin gold electrode and the seal degrade over decades, and the cell suffers light-induced fatigue, so old selenium cells drift downward, reading low. A meter that recommends half a stop more exposure than its neighbours is usually a tired selenium cell, not a calibration to be trusted.

Cadmium Sulphide: Sensitive Enough for TTL, but with a Memory

A cadmium-sulphide (CdS) cell is a photoresistor: its resistance falls as light rises, so it needs a battery to drive the bridge circuit. The payoff is sensitivity far beyond selenium, and small enough to sit inside a camera. That is why the Asahi Pentax Spotmatic of 1964 — among the first production cameras with through-the-lens metering, following the Topcon RE Super of 1963 — used twin CdS cells behind the prism, and why CdS displaced selenium for TTL and available-light work through the decade.

CdS has a bandgap near 2.42 eV, giving a spectral peak around 515 nm and useful response from roughly 515 to 730 nm, closely matched to the eye. But it carries two faults. The first is speed: maximum spectral response time is about 100 ms, and on top of that sits a memory effect. Resistance depends on the cell’s recent light history; after a bright-to-dim transition it can take 30 seconds to a couple of minutes to read darkness correctly, and high-sensitivity types drift for hours. Meter a sunlit street, step into a doorway and read immediately, and the cell still half-remembers the sun, reporting too much light and underexposing the shadows by a stop or more until it settles. Storing the meter in light before use, rather than capped in a dark bag, reduces the lag.

Some meters used cadmium selenide (CdSe) instead, peaking further into the red at 690 to 730 nm with a faster 10 ms response for extra low-light reach, but CdSe resistance is far more temperature-sensitive than CdS, so the gain came at the cost of stability in cold or heat.

The second CdS fault is its power supply. The bridge circuits were calibrated against the flat, constant 1.35 V of a PX625 or PX13 mercury cell, and many had no voltage regulation. Drop a 1.5 V alkaline into that socket and the reading shifts by roughly half a stop to a full stop. Mercury battery sales were banned in 1996 on toxicity grounds, so any inherited CdS meter taking a PX625 is now living on a substitute: use a 1.35 V zinc-air cell or an adapter, not a bare alkaline, or the cell’s accuracy is gone before the memory effect even gets a vote.

Silicon: Fast, Linear, and Hungry for Infrared

The silicon photodiode is photovoltaic but generates far less voltage than selenium, so it depends on an amplifier and a battery. In return it answers in microseconds, has no measurable memory effect, and stays linear across a very wide range. The speed is decisive for flash: a flash burst lasts a fraction of a millisecond, and a CdS cell with its 100 ms lag physically cannot integrate it, whereas a silicon cell registers the whole burst. Gossen built the silicon-blue cell into the Profisix and Luna-Pro SBC in 1977, and the Lunasix F / Luna-Pro F of 1981 added flash metering on the strength of exactly that response time. Silicon had displaced CdS in most meters by the end of the 1980s.

Its weakness is spectral. Bare silicon responds from about 200 nm in the ultraviolet out to roughly 1100 nm, with peak responsivity deep in the near-infrared, commonly around 850 to 980 nm and around 0.4 to 0.7 A/W, far outside what panchromatic film records as luminance. Uncorrected, it over-reads any infrared-rich source, tungsten worst of all. The fix is an integral colour-correction filter that cuts the infrared and reshapes the response toward photopic, and the result is sold as a silicon-blue cell (SBC) or SPD. It solves the same photopic-matching problem the selenium-filter literature addressed, but by the opposite means: where selenium needed a touch of filtering to add correction, silicon needs aggressive filtering to subtract its infrared appetite. A silicon meter is only as accurate as that filter.

Choosing by Failure Mode

Match the cell to the job and the failure it cannot help. For a daylight landscape, selenium or a silicon-blue cell both read honestly; selenium needs no battery and ignores infrared by nature. For a dim available-light interior, reach for CdS or silicon for the sensitivity, but give CdS its 30 seconds to a couple of minutes to forget the brighter scene you just left. For flash, only silicon will do; selenium and CdS are both too slow to catch the burst. For an inherited or aged meter, distrust selenium first for downward drift, and check what battery a CdS unit expects before believing a single reading. Under tungsten, trust the SBC’s filter over a bare silicon cell or you will underexpose the shadows by the stop that infrared invented. Knowing which cell sits behind the needle explains most of the disagreement between two meters pointed at the same scene.