Calibrating the darkroom timer

Purpose

To leave the bench with a signed certificate saying what this timer and this lamp actually deliver, and below what exposure they stop. Why the commanded, delivered and received exposures differ is on the calibration experiment.

When to use it

At commissioning, and again after a firmware change, a repair, or a change of head. The offset belongs to a lamp, so a second head means a second calibration and a second line on the certificate. It is not a session procedure: a printing evening reads the certificate rather than remaking it.

Before you start

You should already have built and flashed the instrument, read the switched output, and copied the head’s warm-up rule and its exposure repeatability in stops off its own certificate onto tonight’s first notebook page.

Declare your acceptance criterion in writing, before the first reading. No source in this course establishes an acceptance limit for a darkroom timer, so the number is yours; writing it afterwards is fitting the criterion to the result.

On the bench: the probe board, one ground wire to the timer and its signal on GP15; the photodiode head and its converter; a hand-held clock; ten sheets from one packet, pencilled on the back before the lights go out, two of them controls; developer, stop and fixer at working dilution, one pair of tongs each; a thermometer; the labelled waste containers.

Level B, for a high-brightness source switched repeatedly for two and a half hours and for low-voltage electronics beside three trays. Nothing on the mains side is opened. Arm 5 needs the supply voltage: measure it at the low-voltage side, or with a plug-in meter that is itself a sealed appliance, and where neither is available the term goes on the certificate as unmeasured.

Procedure

Setting up

1. Write the firmware version and all eight constants as set into the notebook.
2. Write down the present value of OFFSET_MS, set it to zero, and reload. A run made with a correction already applied measures the residual and then reports it as the offset.
3. Switch the head on and start the warm-up clock. No measurement precedes the warm-up rule.
4. Prove the probe: run the loopback command with the probe watching, and confirm the pulses arrive.

The electrical and optical arms

5. Arm 1. Command a fixed interval twenty times at each of four lengths across the instrument’s range; record the mean overshoot, the worst, and the spread.
6. Arm 2. Note a clock reading as one exposure ends and again an hour or more later as another does, and subtract the two log tick counts. Record both readings and the difference in seconds and in parts per million. Once, at commissioning.
7. Arm 3. Log the photodiode head through single exposures at five lengths, capturing a second of darkness before and after each. Write down the converter’s actual sampling interval before you look at any curve.
8. Compute the equivalent-rectangle time for each interval, recording the two constants it is made of separately as well as their difference.
9. Plot equivalent against commanded time and mark where the line departs from slope 1: the linearity floor.
10. Arm 4. Soak ten exposures at each of six commanded intervals with the probe watching, and record the spread in milliseconds and again converted into stops.
11. Sort each column and look at the extremes rather than the standard deviation: one outlier in ten is one ruined sheet in ten.
12. Arm 5. Log supply voltage and easel illuminance together, paired, through an evening’s switching. On a constant-current driver, confirm that the second does not follow the first rather than assume it.

The paper arm

13. Make a five-band third-stop strip at a base exposure inside the region arm 3 found constant, with a control sheet in the batch.
14. Make a second strip from that negative with every band’s time multiplied by four and the aperture closed two stops.
15. Process both in one batch, dry them, and read every band against base plus fog.

Closing

16. Take the largest of the linearity, repeatability and reciprocity floors as the shortest honest exposure, and name on the certificate which one bound it.
17. Write the certificate to the field list on the calibration page, and check that it claims neither accuracy from a repeatability figure, nor a resolution the paper never saw, nor traceability from a chain with one link, nor an offset good for a head it does not name.
18. Set OFFSET_MS to the measured value, write it in the notebook and on the certificate, reload.
19. Empty the trays into their labelled containers, disconnect the probe, let the head cool. Local regulation governs the routes — check yours, under the disposal caveat.

How you know it worked

• OFFSET_MS was zero for every reading and carries the measured value when you leave.
• The mean overshoot at the shortest and the longest interval is similar in milliseconds.
• Every offset is written as two constants and their difference, per light source, with the head named.
• The certificate quotes a spread, not a standard deviation, and a bound for the timebase.
• The shortest honest exposure names which floor set it, beside the criterion you declared first.

If a step fails

Step 4, the probe reports nothing. Meter the signal pin against the probe’s ground while the timer exposes. One ground wire between the boards is not optional, and the active sense may be inverted.

Step 5, the mean overshoot is milliseconds rather than microseconds. Something inside the exposure loop is blocking. Disable the metronome, re-run, and record the difference.

Step 6, the hour check disagrees by minutes. A clock-frequency or divisor error, which is the only fault this arm exists to catch. Read the board’s actual system clock. A disagreement of a second or two is the resolution of a pair of hand readings: record it as a bound. A negative subtraction means the tick counter wrapped, so take a shorter run.

Step 7, the light curve looks like a staircase. That is the converter’s sample rate, not the lamp. A rise you cannot resolve is written as zero within plus or minus the resolution, never as no offset.

Step 9, the plot bends everywhere. A single constant is then the wrong model across your whole range. Record the range over which it is straight and set the floor at its lower end.

Step 10, repeatability is far worse at one interval only. Something periodic is interacting with that length. Re-run it with the metronome disabled and compare.

Step 15, the two strips disagree everywhere rather than only at the short end. That is not reciprocity: the aperture change was not exactly two stops, or the two were processed differently. Re-make both in one batch and verify the aperture by measuring illuminance rather than by counting clicks.

Step 18, the constant was measured and never fitted. The notebook line from step 2 is what catches it.

What to record

Everything the certificate carries, and the raw material behind it: the three boards’ serial logs in one dated folder; the strips in a dated envelope with it; the paper packet and batch; the developer, dilution, temperature, time and agitation; the ambient and head temperatures at the start and the end. A certificate that does not point back at a notebook page is a claim rather than a record.