BEOS outtakes: Photographic Film

From the cutting room floor at Building Electro-Optical Systems, Third Edition:

Photographic Film
 Okay, okay.  Photographic film isn't a detector of the sort we've been discussing.  Film is so out of fashion, so inconvenient.  It needs messy chemicals.  Getting it to be highly sensitive requires all sorts of 1960s alchemy such as pre-flashing and hypersensitizing in a forming gas or hot hydrogen atmosphere.  Why do we care about it at all, in these days of 4k x 4k CMOS imagers?

 There are two reasons.  Firstly, a telescope has a lot more than 4k x 4k resolvable spots.  The Palomar Schmidt has a 6.6\degrees\ square field.  At 1 arcsecond resolution, its plates were digitized at 23040 pixels square for the Digital Palomar Observatory Sky Survey (DPOSS).  That's 530 Mpel, which is a lot of imager chips, but just one photographic plate.  The plate can be digitized later on a scanning microdensitometer that also has many more than 4k x 4k resolvable spots.  Even an ordinary 35-mm camera produces images equivalent to 30 Mpel--and that's real pixels, not Marketing Megapixels (TM)  (see Section 3.9.14). The defect density in photographic film is lower than in IC imagers, too, and it makes a nice archival record that is guaranteed to represent the measurement data well. 

Photographic film has a power-law response over a huge range of signals.  The contrast exponent gamma can be anywhere from 4 down to 0.5, which compresses the dynamic range and makes bright and dim objects visible simultaneously.  Using low-contrast developers such as POTA, photographic film can record images whose dynamic range approaches 10^6:1, optical, e.g. a bomb flash and its surroundings, which is a task beyond any silicon imaging sensor whatever.

Film has two sorts of noise: grain, which is analogous to the digital nibblies from CCD pixels, and fog which is analogous to dark current.  Fog is due to a few grains being rendered developable by a few loose electron/hole pairs in the emulsion, and contributes random noise in the same way.


The second reason to talk about film is that some modern alchemy has got photographic film up to a QE of 1.0, and a multiplication gain of 2, so that a single photon can expose a grain of silver halide.  This is the quantum efficiency of the best CCDs, so there's no waste of photons any more. {J. Belloni et al., Nature V. 402, p. 865 (Dec.  1999).}  The trick is to add formate ions to the emulsion to scavenge all the excess holes without increasing the fog.  Unlike other hypersensitizing tricks, this one works at room temperature and is stable indefinitely.

Next: The Hurter-Driffield Curve

Temperature Measurement is Hard

Measuring temperature is surprisingly subtle.  There are lots of sensors out there; Digikey sells thermistor sensors interchangeable to +- 0.1 C from several vendors for about $3 in onesies.  IC sensors tout good accuracy and linearity, and come in both analogue and digital versions for way under a buck.  So what's the issue?

The issue is: temperature sensors measure the temperature of the sensor, whereas what we want is the temperature of something else: air, fluid, or some solid object we're trying to control.  So the problem is to get the sensor temperature to track the temperature we actually care about.   IC sensors are especially bad, because they have stout leads made of copper (400 W/m/K thermal conductivity) and small packages made of plastic (0.1 W/m/K).  Thus they basically measure the temperature of their leads, and are horrible at measuring air temperature, for instance.

National Semiconductor used to put out a Temperature Measurements Handbook, but since TI bought them, it seems to have disappeared from the web, but here's the  2007 edition. Not much has changed about the properties of plastic and metal since then, so it's still very current.

Decap Photo of Terabeam APD Photoreceiver

Decap picture of a Terabeam CD3109 APD/TIA module, taken with a lens glued to a cell
phone camera

Thermal Runaway Found Useful

This odd circuit  is an on-chip temperature balancer that uses thermal runaway to force all the transistors on an array to run at constant temperature. BJT dissipation goes up at low temperature, with very high gain.  Here's its step response.

Sine Wave Generation with TANH Wave Shaper

Sine wave generation is a perennial problem.


Direct-digital synthesis (DDS) uses a bunch of counters, lookup tables, and DACs,
but that's a relatively heavyweight solution that doesn't fit all problems.
BJT differential pairs naturally have a hyperbolic tangent (tanh) characteristic, which can be used to round off a triangle wave into a very passable sine.  I'm not old enough to have invented this technique, but here are a couple of illustrations of how it works: TANH Sine Wave Shaper (PDF) and TANH Sine Wave Shaper (Mathcad).