Build: An LED Safelight

Purpose

To build a lamp you can defend. Not a red bulb: a deep-red emitter of known band, driven at a known current from a certified low-voltage supply, dimmed in a way that does not move its spectrum, inside a housing that emits light only where you pointed it — and carrying a label with a blank on it for the working time the next page will measure.

The label is the point of the build. Kodak’s account of a safelight has three parts: a housing that stops the source’s own light escaping anywhere but through the aperture, the filter, and the source. An LED lamp collapses the middle term into the first — the emitter is the colour, so there is no filter to fade — which leaves the housing and the source to get right, and one number to measure.

Until that number exists, the lamp is a light leak with a switch on it. Nothing you build today goes near a sheet of paper until the fog test has passed.

Learning objectives

By the end of this session you should be able to:

• read an emitter’s datasheet for the four figures that decide whether it is a safelight emitter, and say why “red” on a listing is not one of them;
• explain why the drive current is chosen well below the maximum, in terms of both temperature and spectrum;
• state what pulse-width dimming holds constant that current dimming does not, and name the second-order effect that spoils the claim slightly;
• build a housing that gives indirect and direct output from one emitter without the emitter ever being in direct view of the paper or the eye;
• find and close the four light leaks every home-made enclosure has;
• site a lamp — height, distance, aim — from published placement rules rather than from where the bracket happens to fit;
• argue the case for two dim lamps against one bright one;
• measure illuminance at the working plane at each dimmer setting and record it so that a setting can be returned to;
• say what a warm lamp does that a cold one does not, and why that makes warm-up part of the procedure.

Prerequisites

Safelight: spectra, distance, duration and the fog limit, in full. This page implements the decisions that page argued — deep red rather than amber, a single-die emitter rather than a phosphor white or an RGB strip, distance as the strongest lever — and re-argues none of them.

Low-voltage electronics for the darkroom, also in full. Constant-current driving, the LED as a diode, the heatsink as a photometric component, the five datasheet numbers, the soldering rules and the current budget are all established there. They are used here and not taught again.

Making a room dark, because there is no point measuring a lamp in a room that leaks.

Safety classification

Level A, and the letter had to be argued, because Part XIV places low-voltage electronics builds and high-brightness LED sources at Level B and this is a low-voltage electronics build with an LED in it.

The distinction is the one the rubric exists to make: a level is assigned to a procedure, not to a category. A high-brightness source means the hundreds of milliamps that Part XIV’s sensitometer and Part XVI’s enlarger head run at, where the emitter is an eye hazard at close range and the heatsink is a burn. A safelight is the opposite instrument. It is designed to be dim — the whole engineering problem is getting the illuminance at the paper down — and it runs at tens of milliamps into an emitter that spends its life behind a diffuser. Nothing is heated by intention, nothing is concentrated, no chemical is present, and the supply is certified and enclosed.

Two steps are declared higher, and both are avoidable.

Soldering is a declared Level B step under the mixed-level rule, for the iron, the burn and the flux fume, with Part XIV’s controls stated there and applying here unchanged: iron in its stand, ventilation or extraction at the joint, lead-free solder, and nothing touched for thirty seconds after it is made. It is avoidable: the whole lamp can be built with screw terminals and crimped ferrules, and the parts list gives that route.

First switch-on with the diffuser off is a declared Level B step for the eye. A deep-red emitter at 30 mA is not the hazard a white module at 700 mA is, but the correct habit is the same one and it is cheap: the emitter is never looked into at close range, and after stage 3 it is never uncovered again.

Hazards

Extra-low voltage, and nothing else. The lamp runs from a certified, enclosed 5 to 12 V supply with a moulded lead, or from a USB power bank. No mains enters this build, no mains enclosure is opened, and nothing is wired into a lighting circuit. HSE’s reasoning is the course’s: limit the supply voltage to the lowest that will do the job, because 230 V AC can kill and the risk is greatest in wet surroundings.

The iron and the joint. Declared above. Burns are the injury; the flux fume is the exposure.

The emitter at close range. Declared above. Diffuser on before the current goes up.

Warmth, not heat. At the currents here the emitter and its small heatsink become warm rather than hot, and the design reason for keeping them cool is photometric rather than thermal — see the last section of the build.

A lamp that has not been tested. The largest hazard on this page is not to a person. It is the temptation to use the lamp before the fog test, on the grounds that it looks about right. It is treated as a light leak until it has a number.

Required PPE

Eye protection while soldering and while trimming component leads, because a snipped lead leaves at speed. Nothing chemical, and that is a statement about this page: no solution is handled, so the controls here are the eye protection, the iron’s stand and the ventilation at the joint rather than gloves. A heat-resistant mat under the iron is equipment rather than protective equipment and is listed with the tools; it protects the bench, which is not the same job.

Ventilation is not among this page’s controls in the chemical sense, because nothing produces a vapour — except at the soldering step, where flux fume is extracted or the window is open, exactly as Part XIV requires.

Parts

Part	Quantity	What it does	Substitutes and notes
Deep-red LED emitter, on a star board	1	The source	The course names no part. Choose it on four datasheet figures: dominant wavelength, spectral half-width, radiant flux against forward current, and how the wavelength shifts with current and junction temperature. A star-mounted part is easier to heatsink and easier to replace than a bare 5 mm LED, and gives a flat surface to measure the temperature of
Constant-current LED driver, low voltage in, adjustable or fixed at 20 to 60 mA, with a PWM dimming input	1	Sets and holds the current	A resistor and a MOSFET switched by a microcontroller is the Part XIV route and works. A bought driver with a dimming input is less to go wrong
Certified, enclosed low-voltage supply, 5 to 12 V, with a moulded lead — or a USB power bank	1	Power	Check the plug carries BS 1363 and the fuse BS 1362, both marked; Electrical Safety First’s other test is that an appliance arrives with a proper UK plug or a conversion adaptor rather than a travel adaptor, and one that does not may not meet UK requirements at all. A power bank is a certified enclosure round a lithium cell, which is exactly why the course allows it and not a bare cell
Small aluminium heatsink, 20 to 40 mm, or a piece of 3 mm aluminium plate	1	Keeps the junction cool, which keeps the wavelength put	At these currents the plate is enough. Thermal adhesive or a thermal pad between star and plate
Opal acrylic sheet, 2 to 3 mm	2 pieces, one per aperture	The diffuser	Opal, not clear and not frosted-on-one-side. It has two jobs: it spreads the source so the lamp is not a point, and it stops the emitter being in direct view
Enclosure: a small ABS box, or panels cut from 5 mm black board	1	The housing	The housing’s job is Kodak’s first component: keep the source’s light from escaping anywhere but the two apertures you cut
Matt black paint or flocked card	one A5 piece	Everything inside the housing that is not the reflector	Matt. A gloss black interior is a mirror and will put the emitter’s image on the ceiling
Matt white card	one A6 piece	The small reflector behind the emitter in the indirect head	White below the diffuser, black above it — the same rule the densitometer head uses
Potentiometer or rotary encoder with a detented action, plus a knob with a pointer	1	The dimmer	Detented or marked. A continuous knob you cannot return to a setting makes every previous test void
Toggle or rocker switch with a distinctive shape	1	On and off, findable by touch	Not a push button: a switch whose position you can feel tells you the state of the lamp in the dark. Fit it where a hand naturally lands, and not next to the white-light switch — Kodak warn against placing white-light switches where they might be mistaken for safelight switches
Opaque tape, black, and heat-shrink sleeving	as needed	Blacking the cable exit, covering any indicator LED on the driver	Drivers often carry a power indicator; it is a second light source inside your safelight
Cable gland or grommet	1	Cable exit that does not leak	A tight grommet plus a loop of black tape is as good
Bracket, clamp or chain	1	Mounting	It has to hold aim, not just weight: a lamp that creeps is a lamp whose test has expired
Self-adhesive label, white	2	The certificate on the lamp, and a spare	Written in stage 7, filled in after the fog test

🛠Three things this page will not buy for you

A part number. No specific deep-red emitter available in the United Kingdom was verified against its own datasheet while this part was written, so none is named. What is named is the four figures to check, which will still be the right four when any part number has gone.

A commercial comparison. ILFORD publish an absorption curve for every one of their filters — charts of density against wavelength, which is real published data about the filter. The course found no commercially sold darkroom safelight publishing the emission spectrum of the finished lamp, so it cannot honestly compare a built lamp with a bought one on spectrum. It can compare them on the only measurement that decides anything, which is the working time, and that is the next page.

A flicker frequency from a standard. IEEE 1789 is the reference for flicker and stroboscopic effects and this course did not consult it. The frequency recommended in stage 4 is therefore given as the course’s own choice with its reasoning, and not as conformity to anything.

Tools

Tool	For	Notes
Soldering iron with a stand, and a heat-resistant mat	The emitter’s leads and the driver’s terminals	Only on the soldered route. The crimp-and-screw route needs a crimp tool instead
Side cutters, wire strippers, small screwdrivers	Everything else
Craft knife, steel rule, cutting mat	Board, opal sheet, flocked card	Score opal acrylic and snap it; do not try to cut through
Drill and step drill, or a hole saw	The two apertures and the cable exit	A step drill in acrylic and thin ABS beats a twist drill, which grabs
Multimeter	Setting and confirming the drive current	In series with the emitter, once, then written on the label
Lux meter or Part XV’s photodiode head	Illuminance at the working plane at each dimmer setting	See the caution in Testing and calibration: a photopic meter reads deep red poorly
Tape measure	Lamp-to-plane distances, which are half of every result	Measured to the paper’s position, not to the bench
A small torch	Leak-hunting the finished housing from the inside	The path works both ways

Estimated cost

Cost band £, and it is the cheapest build in this part by a wide margin. An emitter, a driver, a square of opal acrylic, a box and a supply you may already own.

This page quotes no component price, because the planner carries none. It prices no LED emitter, no driver, no low-voltage supply, no opal acrylic, no enclosure and no potentiometer; the planner’s own gap register names the electronics and the enclosure hardware among the things it could not price, and this page adds nothing new to that list because it needs nothing new.

The planner does price a bought mains safelight at £37.99 to £68.40, and that figure is the honest benchmark. A built lamp will come in below it. What the built lamp buys for the difference is not money but a spectrum you chose, a dimmer you can return to a setting, and a lamp that has no filter to fade.

Estimated consumables cost

Almost nothing is consumed here. The lamp is capital: it is meant to be re-tested rather than replaced, and every part of it survives a rebuild.

Consumed	This session	Sourced price	Cost this session
Solder and flux	under a metre	None. The planner carries no line for solder	—
Matt black paint, or flocked card	about 5 mL, or one A5 piece	None. The planner carries no line for either	—
Black opaque tape	about 1 m	None. The planner carries no line for it	—
Opal acrylic offcut	two small pieces	None. The planner carries no line for it	—
Thermal adhesive or pad	one pad	None. The planner carries no line for it	—
Self-adhesive labels	2	None	—

No row in this table carries a dated price, so the subtotal is zero and the subtotal is wrong. That is the honest form of the answer: one run of this page consumes a small amount of six things the price file cannot price, none of which is free. The emitter, driver, supply, heatsink, enclosure, switch and potentiometer are capital and are excluded on purpose, as is the lux meter — which the planner also could not price, and which is listed in its gap register.

The one consumable that is priced sits on the next page, not this one: the fog test’s paper and chemistry.

Alternative route

Buy a safelight and read this page anyway. A bought lamp is a lamp whose working time in your room is unmeasured, and everything in Testing and calibration applies to it unchanged: measure the illuminance at each setting and each distance, record them, and run the fog test. The parts of this page a bought lamp does not free you from are the placement rules in stage 6 and the label in stage 7.

Build only the housing. If the electronics are the barrier, a commercial deep-red LED module with its own driver and supply can go inside a housing you build, and the housing is where most of the light discipline lives. The baffle, the black interior, the sealed cable exit and the second aperture are all yours whichever way the light is made.

Use a USB power bank and no fixed supply at all. This is the route for a reader with no accessible socket near the working position, and it makes the lamp portable between a bathroom and a bedroom. Its own consequence is honest and belongs on the label: a power bank’s output voltage sags as it discharges, so the constant-current driver becomes load-bearing rather than a nicety, and a lamp run from a bank should be re-checked on the meter at the beginning of a long session.

No darkroom. Nothing in this build needs one; the lamp is made on a bench in daylight. What needs a dark room is the test, and the room page’s alternatives apply.

Build

Stage 1 — Choose the emitter, on paper, before you buy (20 minutes)

Four figures, in this order.

Dominant wavelength. The single number that says where the band is. Deep-red parts are sold under a range of dominant wavelengths and the difference between the short and the long end of that range is the difference between a lamp with margin and a lamp with a little less. Take the number from the datasheet’s own table, not from the listing’s title.

Spectral half-width. The width of the emission band at half its peak height. This is the figure that says how much of the emitter’s output is not at the dominant wavelength, and it is the whole reason a single-die LED is a better safelight than a filtered lamp. A narrow band with a short tail on the blue side is what you are buying.

Radiant flux against forward current. You need far less than you think — a safelight’s problem is excess, not shortage — so what this chart is really for is choosing a current low on its curve, where a small error in current is a small error in output.

How the wavelength moves. Two charts, and Cree’s XP-E2 family sheet is the worked example of where they live: relative chromaticity against forward current, and against junction temperature. Both exist because both effects are real. That is the physical basis of stage 4’s dimming decision and of stage 8’s warm-up.

Three parts to refuse, and the reasons are on the previous page: an RGB strip, whose blue and green dies are off only as a matter of software; a warm-white module behind red acetate, which carries a blue pump peak inside the paper’s strongest sensitivity; and any lamp described as “red” with no datasheet at all, which Kodak have been warning about since improvised safelights were bulbs dipped in lacquer.

Stage 2 — Drive it at a current you chose (25 minutes)

An LED is a diode: a small change in forward voltage is a large change in current, so a constant-voltage supply and a series resistor is a current that drifts with temperature. Part XIV settles this and the answer is a constant-current driver.

Choose the current well below the part’s maximum. Three reasons, and only the first is the obvious one. It runs cool, so the junction stays near ambient and the wavelength stays put. It sits low on the flux curve, where output is close to linear in current. And a safelight wants tens of milliamps, not hundreds — you will spend the whole of Testing and calibration trying to get the illuminance at the paper down, and starting with a lamp that is already dim saves you from building a lamp you then have to hide.

Wire it in this order and no other: supply, driver, emitter, and the dimming input last, with the supply unplugged throughout. Confirm the current with the multimeter in series once, at full setting, and write the reading on the label.

The safelight, wired

1. 1Certified, enclosed low-voltage supply, 5 to 12 V, moulded lead — plug marked BS 1363, fuse marked BS 1362; never opened, never modified
2. 2Switch with a distinctive shape, on the low-voltage side — findable by touch, and sited away from the white-light switch
3. 3Constant-current driver — holds the current the LED sees; a resistor cannot, because a diode's current is exponential in its forward voltage
4. 4Detented potentiometer into the PWM dimming input — detented so a tested setting can be returned to; a continuous knob voids every earlier test
5. 5Deep-red emitter on a star board — chosen on four datasheet figures in stage 1
6. 6Heatsink, thermally bonded to the star — a photometric component: junction temperature moves both flux and wavelength
7. 7Cable exit with grommet and a tape loop — one of the four leaks every home-made enclosure has
8. 8Black tape over the driver's own indicator LED — an untested light source living inside your safelight

Everything to the right of part 1 is extra-low voltage. Nothing in this drawing is opened, rewired or re-terminated at mains potential, and the supply is a sealed bought item.

Stage 3 — Dim it without moving the spectrum (20 minutes)

There are two ways to make an LED dimmer and they are not equivalent.

Reduce the current. The junction runs at a different current density and a different temperature, and the emitter’s dominant wavelength moves. The datasheet says so: that is what the relative chromaticity charts against current and against junction temperature are for.

Pulse-width modulation. The driver keeps sending the same peak current and varies only the fraction of time it flows. During each pulse the die is at the current you set and characterised. Average illuminance falls in proportion to the duty cycle, and the spectrum you measured at the bench is the spectrum you get at 10 per cent.

E(average) = E(peak) × D

Average illuminance under pulse-width dimming

E(average) is the illuminance at the paper averaged over many cycles, E(peak) is the illuminance while the pulse is on, and D is the duty cycle — the fraction of each cycle for which current flows. Because the paper integrates — its response depends on the total exposure and not on how it arrived — the average is what fogs it, which is why this is a legitimate way to dim a lamp and not a trick.

🔬The honest limit on the constant-spectrum claim

Pulse-width dimming holds the current constant, and that is most of the effect. It does not hold the junction temperature constant: a lamp at 10 per cent duty dissipates about a tenth of the power and settles cooler than the same lamp at full output, and the datasheet’s second chromaticity chart says temperature moves the wavelength too.

So the correct claim is not that PWM leaves the spectrum untouched. It is that PWM removes the larger of the two shifts and leaves a smaller one that runs in a known direction with duty cycle. The practical consequence is the one in stage 8: test at the setting you will use, and let the lamp warm up first.

Frequency. The course sets the switching frequency at 1 kHz or above, and gives its reasoning rather than a citation, because IEEE 1789 was not consulted. The reasoning: at a few hundred hertz a duty-cycled lamp can produce visible stroboscopic artefacts on moving objects — your own hand over the tray, a swinging print — which is fatiguing in a room you are already straining to see in; a kilohertz is comfortably above where that occurs and is trivially available on any driver’s dimming input. If you see banding or a shimmer on a moving hand, raise it.

And make the knob returnable. A detent, or a pointer against a scratched scale, or five marked positions. The fog test measures a lamp at a setting; a setting you cannot find again is a test you cannot use.

Stage 4 — The housing, and why it has two apertures (30 minutes)

One emitter, two outputs. ILFORD’s own DL20 does exactly this with two filters, a lower one for direct light and an upper one for reflected, and the arrangement is worth copying because the two kinds of light do different jobs.

Section through the head: indirect up, direct down, emitter never in view

1. 1Deep-red emitter facing UP, on its heatsink — facing up because the indirect output is the larger one and the direct output should be the harder one to get
2. 2Small matt white reflector above the emitter — spreads the die across the upper diffuser instead of imaging it as a bright spot
3. 3Upper opal diffuser, about 80 × 80 mm — the indirect output — aimed at a white ceiling; this is the light you actually work by
4. 4Black baffle between emitter and lower aperture — the specification is that NO straight line runs from die to lower aperture
5. 5Lower opal diffuser, about 30 × 30 mm — the direct pool — small on purpose: it lands on the tray and nowhere else
6. 6Every other interior surface matt black — gloss black is a mirror; white belongs only behind the emitter
7. 7Sealed cable entry, and every seam taped — two of the four leaks; the other two are the switch and the driver's indicator
8. 8Nothing leaves the back — the paper store is usually behind the lamp, and a lamp that leaks backwards fogs the box rather than the print

Drawn readable rather than to scale. The one dimension that must be right is the baffle's: check it by eye with the diffusers out and the emitter on at low current, from the position of the tray and from the position of the paper store.

Indirect is the light you work by. Bounced off a white ceiling it fills the room, has no hot spot, and puts the peak illuminance nowhere in particular. Kodak’s table asks for a higher wattage indirect than direct for exactly this reason: you lose most of it to the ceiling, which is the point.

Direct is a small pool over the developer tray, and it is small on purpose. Both manufacturers’ test procedures tell you to test at the tray precisely because that is where safelight illumination is brightest, and both tell you not to let a direct lamp fall on the enlarging easel.

The baffle is the part that is easy to get wrong. The rule is geometric and admits no judgement: no straight line from the die to either aperture. Check it with the diffusers removed, the current low, and your eye at the tray position and then at the paper store position. If you can see the die from either, add plate until you cannot.

Stage 5 — Close the four leaks (15 minutes)

Every home-made enclosure has the same four, and Kodak’s own list of unsafe-illumination sources names two of them.

1. The cable exit. Grommet, then a loop of black tape inside so that the cable’s own path bends.
2. The seams. Tape every joint on the inside before assembly; a 0.3 mm gap is a slot source.
3. The switch. Many illuminate. Black tape over the window from the inside, or buy an unilluminated switch.
4. The driver’s own indicator LED. Kodak name lighted dials on equipment controls as a fogging source. Cover it. It is inside your safelight and it is not deep red.

Then test the housing the same way you tested the vent trap: a torch inside, in a dark room, and your eye outside. Everything that shows is a leak, and it will show far more brightly from the inside than a leak ever shows from the outside.

Stage 6 — Mounting, and where the lamp actually goes (15 minutes)

Kodak’s placement rules are numbers and they are the starting point rather than the answer.

• Direct illumination no closer than 1.2 m to the working surface.
• No more than one general lamp per 6 m² of ceiling in a large room, lamps at least 2.5 m apart — which for most domestic darkrooms means one lamp.
• Not aimed at the enlarging easel.
• No bright pools against a dark background, which is about your eyes rather than your paper.

Height is set by the first of those and by your own reach: the lamp must clear your head at the tray and still be 1.2 m or more from the paper’s plane, which in a room with a 2.4 m ceiling means it hangs or brackets high and aims down and sideways rather than straight down.

Aim is part of the calibration. A bracket that lets the lamp creep is a bracket that voids the test. Tighten it, then mark the position with a pencil line across the joint so that a shift is visible.

Stage 7 — The label (5 minutes)

Fix a white label to the lamp and write, now: the emitter’s part and its dominant wavelength; the drive current you measured; the date of the build. Leave four blanks, to be filled in from the next page: the paper, the distance, the dimmer setting, and the safe working time.

The label is the reason the build exists. ILFORD ask for the installation date of a filter to be recorded for a lamp that has one; an LED lamp has no filter to age, and what ages instead is the configuration — the setting, the distance, the paper. The label is where that lives.

Testing and calibration

Four measurements, none of which is the fog test. The fog test needs paper, takes two hours and lives on the next page. These four are what make that test repeatable.

T1 — Drive current, once

Multimeter in series with the emitter at the maximum dimmer setting. Record the reading, compare it with the driver’s nominal figure, and write it on the label. If it is more than a few per cent above what you intended, correct it now: everything downstream is measured at this current.

T2 — Illuminance at the working plane, at every setting

Put the meter’s sensor where the paper will be — on the easel, and again on the surface beside the developer tray — and measure at each detent of the dimmer, with the distance recorded to the centimetre.

!Why a lux reading from a deep-red lamp is not what it looks like

Lux is photopically weighted: it counts photons by how much the human eye responds to them, and the eye’s response is collapsing exactly where a deep-red emitter is emitting. So a deep-red lamp that reads low in lux may be putting more light on the paper than an amber lamp reading the same number. Part XIV’s calibration page sets out the mismatch in full.

Two consequences. Use the readings as ratios rather than as absolutes — setting 3 against setting 5, position A against position B — because a ratio taken with one detector on one unchanged spectrum is faithful whatever the weighting. And where you have Part XV’s photodiode head, prefer it: a bare silicon photodiode is far more sensitive at 660 nm than the eye is, so it gives you a number with a usable signal in it.

The course could not source lux-meter accuracy classes, so it makes no claim about how good any particular meter is; what it claims is only that a photopic instrument is the wrong shape for this measurement and a relative one is the right shape.

T3 — The uniformity of the pool, and the fall-off across the room

Measure the direct pool at its centre and at 300 mm to each side, and measure the indirect illuminance at the easel, at the tray and at the paper store. You are looking for one thing: the brightest place in the room where paper sits, because that is where the fog test must be run and it is very often not where you expected.

T4 — Warm-up, and the repeatability it buys

Switch the lamp on cold and read the illuminance every minute for ten minutes at a fixed distance. Expect it to fall a little and settle. That is the junction warming: flux falls with junction temperature, and the wavelength moves with it too — the datasheet publishes charts for both.

Two conclusions, and they are the physics this build rests on. A warmed-up lamp at a fixed current is a repeatable lamp, so the fog test starts after the warm-up plateau and every later measurement does too. And the size of the drift is your heatsink’s report card: a lamp that keeps falling for ten minutes wants more metal.

Write the warm-up time on the label with everything else.

Troubleshooting

The lamp is visibly dimmer after ten minutes. Two candidates and one measurement separates them. If the current has held and the illuminance has fallen, it is thermal — the junction is hot, flux is down, and the answer is more heatsink or less current. If the current has fallen too, it is the supply or the driver: a power bank sagging as it discharges, or a driver in thermal fold-back of its own. Put the meter in series and you know within a minute.

The dimmer does nothing over the first third of its travel. Normal for many PWM inputs, which expect a specific voltage or duty range. Map the useful travel and mark the knob accordingly; do not try to fix it with a different potentiometer until you have read the driver’s own dimming specification.

A visible flicker, or banding on a moving hand. The switching frequency is too low. Raise it.

The room looks bright and prints are fine; the fog test still fails. Believe the test. Brightness is an eye judgement made by an eye that has been adapting for twenty minutes, and it says nothing about the illuminance at the paper. Move the lamp back first — the payoff is quadratic in distance — then reduce the setting, then reduce the time paper is out.

Light on the ceiling directly above the lamp is a bright patch rather than a wash. The white reflector in stage 4 is missing, misaligned, or gloss. An imaged die on the ceiling is a small bright source at the top of the room, which is not what indirect lighting is for.

A faint glow around the back of the housing when the room is fully dark. A seam you did not tape or a screw hole you did not black. Find it with the torch-inside test, not by eye — and until you have found it, move the paper store.

Questions

1. Your emitter’s datasheet gives a maximum forward current of 700 mA and you intend to run it at 30 mA. Give the three separate reasons this is the right decision for a safelight, and name the one circumstance in which running near maximum would be defensible.
2. Explain, to somebody who thinks the two are the same, why turning the driver’s current down and reducing the duty cycle are not equivalent ways of dimming this lamp. Then state the honest limit on the claim that PWM leaves the spectrum unchanged.
3. Your lux meter reads 0.4 lx at the tray from the deep-red lamp and 0.4 lx from an amber lamp. Can you conclude that the two present the same fogging risk to a variable-contrast paper? Give two independent reasons.
4. From an emitter’s radiant flux and the geometry of your diffuser, estimate the illuminance at 1.2 m, then compare it with what you measured. Say which of the estimate’s assumptions is most likely to be responsible for the discrepancy.
5. You have one emitter and a room 3 m by 2.5 m. Argue the case for splitting your output between an indirect head and a second dim direct head over the wet side, rather than putting all of it into one brighter lamp. Then give the argument against.
6. The driver you bought has a green power indicator on its board and you have taped over it. Name the other three light sources inside or on a typical home-made safelight housing, and give the test that finds all four at once.
7. Your lamp’s label reads “45 s at 1.0 m, setting 3, MULTIGRADE RC, 5 September 2026”. You move the lamp to a new bracket at 1.6 m and turn the dimmer to setting 5. Which parts of that label are now void, and what is the minimum work needed to make it valid again?

Further experiments

Measure the wavelength shift you have been told about. If you can borrow a spectrometer — many schools and makerspaces have a cheap grating instrument — record the emitter’s spectrum cold at 30 mA, hot at 30 mA, and at 300 mA. You will have measured, on your own part, the two effects the datasheet charts assert.

Build the second head and measure what it buys. Two lamps at half the output each, one indirect and one direct, against one lamp at full output. Measure the illuminance at the easel, at the tray and at the paper store in both arrangements, and then run the fog test in both. The prediction from the previous page is that the pair gives more usable light for the same fog because it puts less of the total where paper sits longest — and that prediction has not been tested by this course.

Try the amber emitter as well. Kodak are open that they recommend amber for paper because most workers judge print density better under it, at a small cost in protection. Build a second head with an amber part, measure both, fog-test both, and decide for yourself what the trade is worth on your paper. That is exactly the decision the manufacturers made on your behalf, made again with your own numbers.

Log the warm-up properly. With Part XV’s photodiode head and the Pico from Part XIV, record illuminance every ten seconds for half an hour from cold. The shape of that curve is the thermal time constant of your heatsink, and it tells you how long the warm-up in T4 really has to be.

Recap

An LED safelight replaces Kodak’s three-part lamp — housing, filter, bulb — with two parts, because the emitter is the colour. That removes the failure mode of a fading filter and leaves two things to get right: the source and the housing.

The source is chosen on four datasheet figures and driven at a current well below maximum, from a certified enclosed supply at extra-low voltage, with nothing mains opened anywhere. It is dimmed by duty cycle rather than by current, because duty cycle leaves the die at the current you characterised; the residual temperature shift is the honest limit on that claim, and warm-up is how you live with it.

The housing gives indirect light to work by and a small direct pool over the tray, with a baffle whose specification is geometric: no straight line from the die to either aperture. Four leaks — cable, seams, switch, driver indicator — are closed and then found again with a torch from the inside.

The lamp is placed by published numbers, aimed away from the easel, and labelled. And it is a light leak with a switch on it until the next page gives the label its last four entries.

Check your understanding

Question 1. Why does this build specify pulse-width dimming rather than a variable-current driver?

• Because PWM is more efficient and the lamp runs cooler
• Because PWM keeps the die at the peak current you characterised and varies only the fraction of time it flows, whereas reducing the current moves the emitter's dominant wavelength — a shift the datasheet publishes a chart for
• Because a variable-current driver cannot go low enough for a safelight
• Because the paper responds to peak illuminance rather than to average illuminance

Show the answer and why

Answer: Because PWM keeps the die at the peak current you characterised and varies only the fraction of time it flows, whereas reducing the current moves the emitter's dominant wavelength — a shift the datasheet publishes a chart for

An LED's dominant wavelength depends on current density and on junction temperature, and manufacturers publish relative chromaticity charts against both — Cree do so for the XP-E2 family. Reducing the current therefore moves the spectrum away from the one you chose the part for. PWM holds the current at its characterised value during each pulse and changes only the duty cycle, so average illuminance falls while the spectrum stays put. Option four inverts the physics that makes this work: the paper integrates, so it is the average that fogs it, which is precisely why duty-cycle dimming is legitimate rather than a trick.

Question 2. Your safelight is visibly dimmer after ten minutes of use. Which measurement distinguishes the two likely causes?

• Measure the illuminance again the next day
• Put the multimeter in series with the emitter: if the current has held and the illuminance has fallen, it is thermal and the answer is heatsinking; if the current has fallen too, it is the supply or the driver
• Look at the colour: a hot LED looks redder
• Swap the emitter for a new one and see whether it happens again

Show the answer and why

Answer: Put the multimeter in series with the emitter: if the current has held and the illuminance has fallen, it is thermal and the answer is heatsinking; if the current has fallen too, it is the supply or the driver

The two candidates are a hot junction, where flux falls with temperature at constant current, and a sagging supply or a driver in fold-back, where the current itself falls. One measurement separates them because they differ in exactly that variable. Option three would be a lovely diagnostic if the eye were an instrument, but the shift is small, the room is dark and you have been adapting for twenty minutes. Option four is the expensive way to find out what a meter tells you in a minute — and if the cause is thermal, the new emitter will do the same thing.

Question 3. What is the geometric specification for the baffle in the safelight housing, and how do you check it?

• It should block about half the light; check by eye that the lamp looks dim enough
• No straight line may run from the emitter's die to either aperture; check with the diffusers removed and the current low, from the position of the tray and again from the position of the paper store
• It should be at least 20 mm from the emitter; check with a rule
• It should be white, so that it reflects light forward efficiently

Show the answer and why

Answer: No straight line may run from the emitter's die to either aperture; check with the diffusers removed and the current low, from the position of the tray and again from the position of the paper store

The baffle exists so that every photon leaving an aperture has bounced off at least one matt black surface, which means the emitter is never in direct view of the paper or of the eye. That is a statement about lines of sight, so it is checked from the places that matter — where paper sits — rather than measured. Option four inverts the rule: white belongs only immediately behind the emitter, where it spreads the die across the diffuser; everywhere else matt black, because a gloss or white baffle would put the die's image straight back out of the aperture you were trying to protect.

Question 4. A lux meter reads 0.4 lx at the tray from your deep-red lamp and 0.4 lx from an amber one. Do they present the same risk to variable-contrast paper?

• Yes; the same illuminance means the same exposure
• No. Lux is weighted by the eye's response, which is collapsing in the deep red, so equal readings mean the red lamp is emitting far more radiant power; and the paper's own sensitivity differs between the two bands, with the amber lamp sitting nearer the green tail that also lowers contrast
• No; the red lamp is safer because red is always safer
• Yes, provided both were measured at the same distance

Show the answer and why

Answer: No. Lux is weighted by the eye's response, which is collapsing in the deep red, so equal readings mean the red lamp is emitting far more radiant power; and the paper's own sensitivity differs between the two bands, with the amber lamp sitting nearer the green tail that also lowers contrast

Two independent errors hide inside a photometric comparison of two different spectra. The first is the detector: lux weights photons by the photopic response, and that response is small at 660 nm, so a deep-red lamp reading 0.4 lx is radiating much more power than an amber lamp reading the same. The second is the material: paper responds in the blue with a slight green tail, so the amber lamp's photons are worth more per photon to the emulsion, and on variable-contrast paper the green end of that tail lowers contrast as well as raising fog. The two errors run in opposite directions, which is exactly why the answer is a fog test and not a meter reading.

Question 5. You have built the lamp, measured its illuminance at every setting, and hung it 1.4 m above the tray. May you now print with it?

• Yes; the placement meets Kodak's 1.2 m minimum
• No. Until the fog test has given a safe working time for your paper at that distance and setting, the lamp is an untested light source, and the manufacturers' distance figures are recommendations for their own lamps rather than a certificate for yours
• Yes, provided the dimmer is below half
• No, because a home-made lamp can never be tested

Show the answer and why

Answer: No. Until the fog test has given a safe working time for your paper at that distance and setting, the lamp is an untested light source, and the manufacturers' distance figures are recommendations for their own lamps rather than a certificate for yours

The 1.2 m figure is a minimum distance attached to a specific filter and bulb combination in a manufacturer's own product range. It constrains where you may put a lamp; it does not say anything about how long your paper may sit under yours. What the illuminance measurements bought you is repeatability — a setting you can return to and a distance you recorded — and the fog test converts that into the one number that governs printing. Option four is the opposite error: a home-made lamp is exactly as testable as a bought one, and the bought one needs the same test.

Question 6. What does a warm-up curve — illuminance measured every minute for ten minutes from cold — actually tell you?

• How long the supply takes to reach its rated voltage
• How much the junction temperature is moving flux and wavelength, and therefore how long the lamp must be on before any measurement or fog test is valid; the size of the drift also reports on the heatsink
• Whether the emitter is faulty
• The duty cycle of the dimmer

Show the answer and why

Answer: How much the junction temperature is moving flux and wavelength, and therefore how long the lamp must be on before any measurement or fog test is valid; the size of the drift also reports on the heatsink

At constant current, the flux of an LED falls as its junction warms, and the dominant wavelength moves with temperature as well — the datasheet publishes a chart for each. So the settling curve is the thermal behaviour of your particular emitter on your particular heatsink. It gives you two things: a warm-up time, before which no measurement is comparable with any other, and a diagnosis, since a lamp still drifting after ten minutes is telling you it wants more metal. That is why warm-up is written on the label with the current and the distance.

Check my answersClear