DIY Section: build it, part by part

Ten parts of the Zeeman-EIT bench are made in the lab and not bought as finished instruments: the solenoid, the cell stack, the ITC502 cables, the degaussing coil, a backup current source, the lock-box divider, the fluxgate supply, the RF control box, the AOM platform, and the beam dumps and covers. Each one below has a purpose, a parts list (no prices), tools, dimensioned drawings and wiring schematics, numbered build steps with a check after every step, tests, the usual mistakes, and the items you must still measure or confirm before you cut or solder.

How to read the drawings

Safety for every part

The twelve parts and the order to build them

The order puts the cables and the tests first, so that everything after them can be checked, and the optics last, when the beam path is known. Times are estimates for one person who already solders and are counted without waiting for machining.

The solenoid makes the uniform magnetic field along the beams that splits the Zeeman levels: 50 mG gives a two-photon splitting of 70 kHz (1.4 kHz per mG). The EIT line is only about 7 kHz wide, so the field must be the same everywhere in the cell to a fraction of 0.2 % (0.1 mG, a 140 Hz shift). The design below was computed ring by ring with the exact Biot–Savart formula (checked against a brute-force sum to 10⁻¹²) and gives 0.20 % over the cell.

What the design does

• One helical groove in a plastic tube holds the wire exactly: 36 turns at 3.6 mm pitch in the middle and 3 turns at 1.8 mm pitch at each end. The closer end turns make up for the field that falls off at the ends of a finite coil: the error over the cell drops from 0.42 % to 0.20 %.
• Four end turns would give more margin. The design just meets the 0.2 % target at the very ends of the 25 mm path (about half of it on average). Four turns at 1.8 mm at each end (a groove 144 mm long, 44 turns) give 0.15 %: use them if you cannot centre the cell to better than 1 mm.
• Two return wires 180° apart run back along the axis, each carrying half the current. A helix carries a net axial current; a single return wire would add a transverse field of 0.10 mG per mA at the axis, and two wires at 180° cancel it.
• Plastic only (PEEK, Delrin or printed PETG): any steel or nickel near the shield magnetises it and shifts the field by amounts far larger than 0.1 mG.

Parts and materials

Tools

• Lathe with a thread-cutting tool, or a CNC lathe or mill, or a 3D printer (the groove is a non-standard 3.6 mm pitch with a 1.8 mm pitch at both ends, so ask the workshop first).
• Digital calliper (0.01 mm), steel rule kept away from the shield, a winding rig: the former on a rod in two supports, turned by hand, with the wire spool on a friction brake (50 to 100 g pull).
• Soldering iron 25 to 40 W, multimeter with a 4-wire resistance range if you have one, brass or plastic tweezers, a small compass to test for magnetic parts.

3D-printed former (PETG, Bambu Lab A1, 0.2 mm nozzle)

The former can be printed instead of machined. Print it upright, cable end on the plate: the helix then comes from the Z axis of the printer, which is its most accurate axis, and the upper flank of the groove is made at 45° so it needs no support. At the surface the groove is 1.00 mm wide; the wire seat at the bottom is 0.60 wide and 0.40 deep, as in the drawing. A small slot at the cable end marks 0°, where the helix starts and the leads leave; the notch of the cable-end spacer ring goes over it, and both point up, towards the cable holes of the shield caps. A shallow ring marks the mid-plane. Inside, a channel 10 mm wide and 0.6 mm deep runs along the top of the bore (on the 0° line, over the lead mark) from the cable end to 86 mm, just past the mid-plane: the stem of the cell (D2) slides along it, so the stem has 1.1 mm to spare instead of 0.5 mm, and the bore may print a little undersize. Under the wire seats the wall is still 1.0 mm thick. Both ends of the bore have a 1 mm chamfer.

Print time is roughly 8 to 10 hours for the former at these settings (2000 layers of about 10 loops), about 3 hours for the coupon. Coupon check: ten pitches 36.0 ± 0.2 mm (correct the Z scale if needed), the Delrin tube slides into the bore, and the enamelled wire sits flush in the seat along a whole turn.

Shield cradle and the 100 mm beam height

Every optic on the bench sits at one beam height, 100 mm above the breadboard. It is set by the shield: its outer layer is about Ø68 mm, the outside groups of the degaussing coil (D4) reach about 41 mm from the axis, and a cradle needs room under it. The shield rests on a printed cradle with its axis at 100 mm; the posts of the other optics (BOM M03) are set to the same height.

Drawings

Build steps

1. Get the shield data. Ask the maker, or read the drawing, for the innermost bore, the inner length and the end-cap holes: a Ø14 mm beam hole on the axis and a Ø12 mm cable hole 16 mm off the axis, lined up in all three caps at both ends. The beam goes through the beam hole; the coil leads, the cell cables, the fluxgate cable and the degaussing bundle go through the cable holes, so nothing crosses the beam. The former needs a bore of at least Ø46 mm (the former plus 3.45 mm of radial clearance), an inner length of at least 250 mm, and the winding then stays about 55 mm from each end cap. A smaller bore needs a thinner former and a new field calculation, because the winding radius enters directly.Check You have the bore, the length and the hole positions in writing.
2. Make the former. Follow the former drawing. Cut the groove as one right-hand helix, 0.60 wide and 0.40 deep, rounded at the edges (a sharp edge cuts the enamel). Scribe a light ring at the mid-plane and mark the cable end (the end where the leads leave).Check wall 2.0 ± 0.1 mm; ten pitches measure 36.0 ± 0.2 mm; the groove starts and ends 70.2 mm from the mid-plane (140.4 mm long); no burrs.
3. Prepare the wire and the rig. Cut off 5.9 m for the helix with its start lead (0.8 m of lead + 4.98 m of helix + a little spare) and keep the rest of the reel for the two return wires (0.2 m each) and the finish lead (0.8 m). Put the spool on a friction brake so that the wire runs under 50 to 100 g of pull. Handle it only with plastic or brass tools.Check the wire runs without kinks and the enamel is not scraped.
4. Wind the helix. Start at the cable end (z = −70.2 mm). Lay 3 turns at 1.8 mm pitch, then 36 turns at 3.6 mm, then 3 turns at 1.8 mm; the finish is at z = +70.2 mm. Count aloud and mark every tenth turn with a dot of marker on the former.Check exactly 42 turns. The resistance of the helix alone is about 0.44 Ω (4.98 m at 0.0876 Ω/m).
5. Fix the end turns. Put a drop of cyanoacrylate or epoxy on the first and last 3 turns and on every tenth turn. Do not glue the other turns: glue stress on the wire shifts it in the groove.Check no turn can move when you push it with a plastic tool.
6. Lay the two return wires. At the far end, scrape the enamel from the helix end and from one end of each return wire, twist the three together and solder (one joint). Lay the returns along the axis at 90° and 270° (see the winding plan), tape them every 40 mm with polyimide tape, and at the cable end solder them together into the finish lead.Check resistance from start lead to finish lead: 0.58 Ω ± 0.03 (the two returns in parallel add only 0.007 Ω).
7. Twist the leads. Twist the start and finish leads together, about one twist per 20 mm, and solder the plug of D3 (LD OUT). Mark the start lead red.Check no loop of area in the leads: they run as one twisted pair.
8. Check the field outside the shield. Put the former on a plastic stand a metre away from steel, with the fluxgate (D7) at the centre. Apply +10.00 mA, record the reading, apply −10.00 mA and record again. The difference removes the Earth field (about 0.4 G) and the sensor offset.Check B(+10 mA) − B(−10 mA) = 67.8 mG ± 5 % (3.39 mG/mA × 20 mA).
9. Mark the field direction. With the same measurement, note which current polarity gives a field along the beam direction (z, away from the laser) and engrave or label “B+” on the former.Check the label agrees with the fluxgate sign.
10. Install and calibrate in the shield. Push the former into the innermost layer with the cell stack (D2) inside, centre it with the two spacer rings (former slot and ring notch up, towards the cable holes; the ring recesses then face sideways, where the flat degaussing bundle lies) and close the end caps with all cables through the cable hole. The shield rests on the printed cradle with its axis at the 100 mm beam height. After the degauss (D4), put the fluxgate plate (D7) in place of the cell and measure B against I at ±5, ±10 and ±15 mA.Check a straight line B = k·I; use the measured k (a few per cent above 3.39 mG/mA, because the shield pulls the flux in) in every later experiment, not the formula. The fluxgate scale is good to ± 2 %: confirm k with the atoms (the EIT line moves 1.4 kHz per mG).

Tests before use

Mistakes to avoid

• Steel tools, steel screws or nickel-plated connector shells near the former or the shield: they magnetise and the field no longer follows the formula.
• A helix without the two return wires, or with only one: the transverse field is 0.10 mG per mA, 1.5 mG at 14.7 mA.
• Winding under too much pull: the groove edge cuts the enamel and two turns short together.
• Different pitch at the ends: the error over the cell rises from 0.20 % to 0.42 %.
• Loose, untwisted leads that move: the loop area changes and the field at the cell follows.

The cell sits on the axis in the middle of the solenoid, warm enough for an optical depth near 10 (55 to 65 °C), with a heater that makes no magnetic field, a temperature sensor the ITC502 can read, and enough insulation to hold the temperature with about 2 W. There is no metal near the cell. The heater is built as a bifilar pair (go and return side by side) so that its field cancels; a purchased foil is allowed only if its layout is paired and the test below passes.

What the design does

• NTC and not Pt100. The ITC502 resolves 1 Ω on its thermistor range: a Pt100 gives 2.6 K per ohm, a 10 kΩ NTC gives 14 mK per ohm at 65 °C.
• Foam sleeve and two Delrin tubes hold the temperature with 0.9 to 2.1 W and keep the whole stack on the axis: tube 63 + sleeve 34 + tube 63 = 160 mm, the length of the former.
• Diode in series with the heater. A resistor cannot cool: when the controller asks for cooling it would reverse the current, which heats again and runs away. The diode lets only one polarity through; its orientation is found with the dummy-load test of D3.
• Unpaired heater wire is a coil. One unpaired turn of 0.54 A at the end of the heater gives 97 mG at the cell centre, 2000 times what we can accept. That is why the pattern is paired.

Parts and materials

Tools

• Scalpel and a plastic ruler, a flat plastic board to build the heater on, brass or plastic tweezers.
• Soldering iron with a fine tip, multimeter, a thin brush for the epoxy, polyimide tape.
• Cure the epoxy as its data sheet says (room temperature, or a warm plate below 80 °C); no heat gun near the cell.

Drawings

Build steps

1. Measure the cell. Callipers on the real cell: diameter, length, window thickness, and the position and length of any stem. Keep the cell in its box until the heater is ready; never touch the windows.Check diameter 25.0 ± 0.3 mm, length 28 ± 0.5 mm; stem at mid-length, at most 4.0 mm above the glass and Ø8 at its base. A taller stem does not fit the former: stop and read “If the stem is taller” below.
2. Draw the heater pattern. On a Kapton film 68 × 24 mm mark 15 runs of 22 mm at a pitch of 4.4 mm, starting 3 mm from the edge (drawing “Heater”): the barrel between the windows is only about 24.8 mm long, so the runs stay 1.4 mm clear of each window joint. The film is shorter than the circumference so that it leaves a 10.5 mm strip free under the stem. Use a pencil, not a marker with metal ink.Check a sheet of 68 × 24 mm with 15 marked runs of 22 mm.
3. Lay the bifilar pair. Cut 0.80 m of wire and fold it in the middle. Lay the folded wire so that every run is a pair (go and return side by side, 0.4 mm apart), with the fold at the far end. Tape every run with polyimide tape. Leave the two tails at the near end only 10 mm long and solder them to the twisted copper pair of the heater cable (constantan needs an active flux); sleeve each joint with polyimide tape. Long constantan tails would add resistance and heat the leads.Check resistance of the pair: 12 ± 1 Ω.
4. Close the heater. Put the second Kapton film over the wire and press with a warm plate (below 80 °C) or tape it down. Add the NTC bead: glue it with thermally conductive epoxy to the cell glass 5 mm from one window, with its leads flat under the foil edge, and cure it.Check the NTC reads about 10 kΩ at 25 °C (about 8 kΩ at 30 °C); no short to the heater wire (> 10 MΩ).
5. Glue the stem NTC. Glue the second NTC with the same epoxy to the glass at the base of the stem, leads flat along the barrel towards the cable end.Check it reads the same as the first NTC at room temperature (± 1 %); no short to anything.
6. Wrap the cell. Wrap the heater round the barrel (circumference 78.5 mm, film 68 mm) so that the free strip is centred under the stem, tails towards the cable end. Fix with two turns of polyimide tape. No glue on the windows.Check the foil sits flat; no wire crosses a window.
7. Add the foam sleeve. Wrap the foam sleeve (ID 25.6, OD 31.6, length 34) over the heater with the stem through its Ø10 hole; the stem must not touch the foam edge. Run the heater tails and the NTC pair along the sleeve in the 1.2 mm clearance towards the cable end.Check sleeve ends 3 mm beyond each window; leads not pinched.
8. Add the two tubes. Push one tube on each side against the sleeve ends: they centre the stack in the 160 mm former. Put the groove of the cable-side tube over the leads.Check the stack is centred; the tubes do not touch the glass.
9. Test the heater on the bench. Before the stack goes in the shield: connect the heater and the NTC to the ITC502 as in D3, set I TEC LIM to 0.50 A (on the open bench; 0.33 A once the stack is inside the closed former and shield) and T SET to 2.100 kΩ (about 65 °C for this NTC), switch the output on, and watch the NTC reading fall.Check the NTC falls steadily to about 2.1 kΩ and the controller settles in about 15 minutes; at 65 °C the heater takes about 0.28 to 0.42 A (3.3 to 5.0 V across the 12 Ω pair).
10. Magnetic test of the heater. Before the real cell is used, wrap the finished heater round a plastic dummy tube of the cell’s size (Ø25, bore at least 20 mm) and push the fluxgate board of D7 through the tube with the centre of its detection coil in the middle of the heater. Put this in the closed shield, switch the heater current on and off (use the maximum current, not the working current) and read the field.Check the field changes by less than 0.1 mG.
11. Test-fit in the coupon. Push the finished stack, stem up and in the stem channel, through the 50 mm test coupon of D1 (same bore and channel as the former). If the stem or the foam catches, find out why before the wound former is used.Check the stack passes the coupon without touching the stem.
12. Install. Slide the stack into the former from the cable end, stem up in the channel over the lead mark, and the former into the shield. Mark the cable end and keep the NTC and the heater cables twisted as two separate pairs.Check cell centre at the mid-plane of the former (mark it with a pencil ring on the foam).

Tests before use

If the stem is taller

The cell must slide into the former after the winding, so the bore sets the tallest stem: bore radius = 12.5 + stem + 0.5 mm. A larger bore means a larger winding radius (the field gets less uniform, a few more end turns fix it) and a larger shield, because the winding must keep its clearance to the innermost layer. Ring model of D1, same 3.6 mm main pitch and 36 main turns:

The first row is the design on this page. A stem pushed through a hole in the winding was also calculated (Biot–Savart on the true helix, wire bent round the hole): in the beam (within 3 mm of the axis) the field spread stays about 0.2 %, but near the glass under the stem it reaches 1 to 4 %; the real problems are that the cell can no longer slide in and the stem tip leaves the heater. So the stem is limited by the bore, not passed through the winding.

Estimates on this page: the loss at 65 °C (2.1 W in free air, 0.9 W inside the former and the shield) and the time constants (166 and 379 s) come from a simple model (foam k = 0.05 W/m/K, h = 8 W/m²K, heat capacity 8 J/K). Design for the larger loss; measure your own stack and replace them.

Mistakes to avoid

• A heater wound as one coil, or a foil with a return bus along one edge: that is an unpaired turn (97 mG at the cell from 0.54 A). Always run the on-and-off test.
• A Pt100 on the ITC502: 1 Ω is 2.6 K, so the temperature would step by 2.6 K.
• No diode: a resistor heater runs away when the controller tries to cool.
• A cold window: Rb condenses on the coldest glass, and the Rb vapour density follows the temperature of that coldest point. It should be the stem (in its heater gap, a little colder than the barrel), never a window: keep the windows inside the foam sleeve ends, and if Rb droplets appear on a window after a day at temperature, add foam over the window rims.
• Cutting a slot in the former or the winding for a stem: the wrapped cell could then never slide into the wound former, and a stem that sticks out of the heater becomes the coldest point and lowers the Rb density. Order a short stem instead.
• Thermistor leads running through the heater field region as a loop: twist them and keep them flat.

The lab owns two ITC502 controllers. The first drives the DFB laser. The second can drive the cell heater on its TEC OUT jack (15-pin) and the solenoid on its LD OUT jack (9-pin): both are low-noise and already in the lab. Neither jack comes with a cable, so each plug is wired by hand. The drawings follow the pin numbering of the ITC500 manual (chapters 1.9.1 and 1.9.2) and show the solder-cup side of the plug, the way you hold it while soldering.

What the design does

• Heater: TEC pins 5, 6, 7 are tied together as “+”, pins 13, 14, 15 as “−” (the manual says always use all three pins), the thermistor on pins 4 and 3. The diode is in the heater + wire.
• Solenoid: anode pin 8 (polarity CG) and ground pin 3 carry the current. Pins 9 and 6 can sense the voltage at the coil (optional). A link between pins 1 and 5 is the interlock: without it the output does not switch on.
• Never plug a laser diode into the coil cable. It has no monitor-diode wiring. Label the plug.

Parts and materials

Tools

• Soldering iron 25 to 40 W with a fine tip, flux-cored solder, helping hands.
• Wire strippers, heat gun or lighter for the sleeving, multimeter.
• A permanent marker and labels: every plug gets a label.

Drawings

The same plugs are also drawn on the real rear panel of the ITC502 in the wiring figures of the Questions and answers (section 09).

Build steps

1. Read your own rear panel. Look at the TEC OUT and LD OUT jacks of the second ITC502: pins or sockets, and the number engraved beside the first pin. The figures show the numbering of the manual; the numbers moulded into your plug win.Check you know which gender to buy and which row is “wide”.
2. Make the two links. For the 15-pin plug cut two bare tinned wires and fix them with sleeving: one joins cups 7, 6 and 5, the other cups 15, 14 and 13.Check the links do not touch the neighbouring cups.
3. Solder the heater cable. Red conductor to the middle cup of the 7–6–5 link (cup 6) through the diode; black to cup 14. Put the diode anywhere along the red wire with sleeving, anode towards the controller for the first try.Check continuity cup 6 to the red wire end through the diode in one direction only.
4. Solder the thermistor cable. Blue to cup 4, green to cup 3, drain wire to the 15–14–13 link. Do not connect the shield at the cell end.Check cup 4 and cup 3 are not shorted to anything; at the far end the NTC reads 8 to 10 kΩ at room temperature (25 to 30 °C).
5. Close the 15-pin plug. Pull each wire gently, close the plastic hood with the clamp on the outer jacket (not on a single wire).Check no strand stands out.
6. Solder the 9-pin plug. Tin the six cups. Red (coil +) to pin 8, black (coil −) to pin 3 with the drain wire, white to pin 9, green to pin 6. Solder a yellow link between pin 5 and pin 1 and sleeve it. Pins 2, 4 and 7 stay open.Check pin 5 to pin 1 reads about 0 Ω; pin 8 to pin 3 is open.
7. Label the plugs. Label “TEC OUT, cell heater, max 0.5 A” and “LD OUT, solenoid, max 200 mA”.Check the labels are on the plugs.
8. Run the dummy-load tests. Follow the test figure. For the TEC cable use the 10 Ω load and the 10 kΩ and set the I TEC limit to 0.30 A; for the LD cable use the 100 Ω with the limit at 20 mA. Decide the diode orientation here.Check the readings of the figure are reached (0.30 A at 3.7 V; 0.50 V at 5.00 mA).

Tests before use

Mistakes to avoid

• Pin numbers read from the wrong side of the plug: the figures show the solder-cup side, mirrored from the front. Trust the numbers moulded into the plug.
• Pin 12 of the TEC OUT jack supplies Thorlabs mounts: do not connect it to anything.
• A shield connected at both ends: a ground loop. Connect it at the ITC502 end only.
• Using the heater with no current limit: set I TEC LIM first (0.50 A on the bench, 0.33 A in the shield), and I LD LIM to 20 mA first for the coil.

Opening the shield lets the Earth’s field (about 50 µT) magnetise the inner layers a little; after reassembly a remanent field of 10 to 100 nT remains (Twinleaf). Our line needs less than 20 nT (0.2 mG). Degaussing removes it: drive the layers into saturation with an oscillating field, then bring the amplitude down to zero slowly and smoothly. The coil has to pass through the shield so that the flux circulates round the wall of every layer, and it should be spread round the circumference: a toroidal winding through the bore and back outside in at least three groups.

What the design does

• Toroidal winding. The field in the wall of a layer at radius r is H = N·√2·I / (2π r): 193 A/m peak at 1 A rms in the outer layer. Mu-metal saturates at a few tens of A/m, so 1 A rms is already well above saturation in every layer, and the procedure uses up to 2 A rms.
• Find the knee, then ramp. Twinleaf: start low and raise the amplitude step by step; saturation shows as the coil current rising sharply above an applied voltage. Then use an amplitude above the knee and ramp down linearly in 100 to 300 s.
• 50 Hz has a limit. The skin depth in mu-metal at 50 Hz is only about 0.4 mm (relative permeability near 2 × 10⁴), less than the wall of a 1 mm layer. The mains drive works best on thin layers; for thick layers use the low-frequency alternative (1 to 10 Hz).

Parts and materials

Tools

• Wire strippers, soldering iron, crimp tool for the mains terminals, insulation tester or the highest resistance range of a DMM.
• A helper to hold the shield while you thread the wire, a spool holder, a tape measure.

Drawings

Build steps

1. Measure the shield. Length of the layers, the outer radius, and the cable holes of the end caps (Ø12 mm, 16 mm off the axis, lined up in all three caps). The inner bundle (30 wires of Ø1.3 mm, flat: 39 mm wide, 1.3 mm thick) has to lie in the 3.45 mm gap between the former and the bore, and the round bundle (about Ø8 mm) has to pass the cable holes at both ends; it must never cross the beam hole on the axis. If the holes are smaller, use a thinner wire or fewer turns and raise the current (the ampere-turns matter: 30 turns × 2.8 A peak = 85 A).Check the cable holes are at least Ø12 mm and line up through all three caps.
2. Wind the coil. Cut 24 m of wire and leave 0.8 m for each lead. Pass the wire through the bore, back along the outside in group 1, through the bore again, back in group 2, then group 3, and repeat until there are 30 passes through the bore (10 turns per group).Check 30 wires in the bore, 10 on the outside in each group.
3. Lay and tie the bundles. Lay the 30 inner wires flat against the inner wall, tie them every 50 mm with polyimide tape, and tie the three outside groups to the outer layer at 120° with Velcro, cross-wires 10 mm apart.Check resistance of the coil 0.83 Ω ± 0.05; no wire touches a steel part.
4. Build the drive box. With the cord unplugged: mains live through the 2 A fuse and the switch to the variac IN L; the neutral to IN N; the transformer primary across the variac OUT and COM; the secondary through the 3.15 A fuse, the 4.7 Ω resistor and the ammeter to the two output sockets. Join mains earth to the box, the variac frame and the transformer core.Check earth to box, variac frame and core: below 0.5 Ω; mains to box: above 10 MΩ.
5. Dry test on the bench. Connect the coil in the air, away from the shield, switch on with the variac at zero and raise it slowly.Check the current rises linearly with the setting; 2 A rms at a setting of about 106 V; the resistor is only warm after a minute.
6. Find the knee. With the shield closed and everything inside non-magnetic, raise the variac slowly from zero while you read the ammeter. The knee is where the current starts to rise much faster than the voltage.Check you can write down the variac setting of the knee.
7. Degauss. Go to about 1.5 times the knee (our rule of thumb; Twinleaf only says “high enough to ensure saturation”), at most 2 A rms. Hold 5 seconds, then turn the knob down steadily to zero in 100 to 300 seconds. Do not stop and do not jerk.Check the ammeter falls smoothly to zero.
8. Disconnect. Variac at zero, mains switch off, then unplug the coil from the drive. Twinleaf’s commercial unit disconnects the coil with relays when it is finished, so that no noise couples into the shield.Check the coil is open-circuit.
9. Repeat once and measure. Repeat the degauss, then read the field with the fluxgate plate (D7) or by looking at the atoms (the EIT line against the coil current, reversed once).Check two runs in a row give the same field to ± 0.1 mG.

Tests before use

Mistakes to avoid

• Stopping or jerking the knob during the ramp: the last small oscillation leaves the magnetisation behind. Turn it down smoothly to zero.
• A coil that does not go through the shield: the flux then does not circulate in the wall and the layers are not degaussed.
• Steel clips, steel screws or a steel table leg close to the shield: they re-magnetise it.
• Leaving the coil connected to the transformer: it picks up noise. Unplug it after each run.
• Opening the shield after the degauss: every opening magnetises the layers again. Degauss last, with the shield closed.

Alternative drive without mains wiring

An arbitrary waveform generator makes a sine of 1 to 10 Hz with a slowly decreasing amplitude (for example by amplitude modulation with a ramp), and a DC-coupled power amplifier drives the coil with up to 3 A. The series resistor and the same ammeter are used. An audio amplifier is AC-coupled and can leave a small direct current in the coil, which would magnetise the shield: do not use one. This route also suits thick layers better than 50 Hz.

The first choice for the solenoid current is the LD OUT of the second ITC502 (D3): noise below 1.5 µA rms and drift below 10 µA in 24 h. If that unit is not available or not quiet enough, this source is the backup. It is a current sink: an op-amp compares the voltage across a 100 Ω sense resistor with a set voltage from a 2.5 V reference and a 10-turn potentiometer, and a MOSFET carries the coil current, so I = V_set / 100 Ω. A 9 V battery gives the quietest supply.

What the design does

• The sense resistor sets the stability. A 100 Ω metal-film resistor of 0.1 % and 25 ppm/K drifts only 0.38 µA per kelvin at 15 mA.
• The reference sets the rest. The LM4040-2.5 runs from 9 V through 6.8 kΩ (0.96 mA) and the potentiometer loads it by 0.25 mA only.
• A simple test source (battery, 560 Ω and a 100 Ω trimmer, 15 mA) is drawn on the same sheet for a first look; it is not for measurements.

Parts and materials

Tools

• Soldering iron, multimeter (6½ digits if possible), oscilloscope with a 1× probe.
• A dummy load of 100 Ω for the first test (the coil is only 0.58 Ω).

Drawing

Build steps

1. Check the parts. Measure R_s, R1, R2, R3 and the potentiometer with the DMM. Look up the pin order of the MOSFET and of the op-amp board.Check R_s within 0.1 % of 100 Ω.
2. Build the reference. V+ through R1 to the LM4040 and to GND. Add the potentiometer between the reference node and GND.Check the reference node reads 2.50 V ± 0.01 V; the wiper runs from 0 to 2.5 V.
3. Add the filter and the op-amp. Wiper to R2 (1 kΩ), C1 (1 µF to GND) and to the + input. Put 100 nF at the supply pins of the op-amp, wires short.Check the + input follows the potentiometer within 1 mV.
4. Add the MOSFET and the sense resistor. Op-amp output through R3 (1 kΩ) to the gate; source to the sense resistor and to the − input (TP1); the sense resistor to GND; the drain to the coil; the coil to V+.Check with a 100 Ω dummy in place of the coil and V_set = 1.000 V: TP1 reads 1.000 V and the dummy carries 10.00 mA.
5. Test with the coil. Replace the dummy by the solenoid. Set V_set = 1.475 V.Check TP1 = 1.475 V; the current from the DMM in series reads 14.75 mA.
6. Measure the noise. AC voltage range of the DMM across R_s with the coil connected, then the oscilloscope with a 1× probe.Check below 50 µV rms across R_s (0.5 µA).
7. Close the box. Fit the battery, the switch and the coil socket; label the dial setting for 50 mG after the calibration in the shield.Check the box is closed, the shell is grounded to the circuit ground only at one point.

Tests before use

Mistakes to avoid

• The coil in the source leg instead of the drain: the feedback then sees the coil voltage and the current is wrong.
• BS170 and 2N7000 have opposite pin orders: a reversed MOSFET gives a current that does not follow V_set.
• A switching supply or a long cable on the battery side: the noise adds to the field.
• Calibrating B against the current with the fluxgate only: its accuracy is ± 2 % ± 0.3 µT. Calibrate with the atoms (1.4 kHz per mG).

The Stefan Mayer FLC 100 is a bare sensor board: it needs a 5 V supply, a way to read its differential output, and a holder. It measures the field along its long axis only. The data sheet gives ± 100 µT, noise below 3 nT peak-to-peak (0.1 to 10 Hz), zero drift below 2 nT per kelvin and an accuracy of ± 2 % ± 0.3 µT, so it is excellent for differences (coil current reversed, heater on and off, one degauss against the next) and not for the absolute zero: for that, the atoms are better.

What the design does

• Quiet supply. An LM7805 from a linear ±12 V supply (or a 9 V battery and a low-dropout regulator) followed by a 10 Ω and 470 µF filter (corner 34 Hz). The supply ripple at 50 and 100 Hz would otherwise sit in the reading.
• Differential reading. OUT− is a 2.5 V reference and the signal is OUT+ − OUT−: read it with a floating DMM across the two pads, never OUT+ against ground.
• A plate in place of the cell. The PEEK plate carries the board into the former with the centre of the detection coil at z = 0, on the axis, where the atoms will be.

Parts and materials

Tools

• Soldering iron with a fine tip, magnifier, multimeter with a differential input (floating), oscilloscope.
• Machine or hand tools for the plate (non-magnetic bits and screws).

Drawing

Build steps

1. Build the supply. On perfboard as in the figure. Power it from the +12 V of a linear supply or a battery.Check the output reads 4.98 V with a 2.5 kΩ dummy load (2 mA); no more than 100 µV ripple on the scope.
2. Wire the sensor. Solder four AWG 30 wires to the pads 0V, +5V, OUT− and OUT+; leave SYNC open. Twist 0V with +5V and OUT− with OUT+. Fix the wires with a drop of epoxy for strain relief.Check continuity pad to wire end; no short between neighbouring pads.
3. First power-up, outside the shield. Away from steel, switch on and read OUT+ − OUT− with the DMM. Turn the sensor end for end.Check the reading changes sign and by about twice the component of the Earth’s field along its axis (20 mV per µT, up to ± 2.5 V).
4. Mount the plate. Screw the sensor into the pocket with nylon screws so that the centre of the detection coil is at the middle of the plate. Mark the direction of the arrow B (positive output) and the cable end.Check the arrow is marked.
5. Use in the shield. Take the cell stack out, slide the plate in with the coil centre at z = 0, close the shield, and read the field with the coil current at +I and −I (D1 calibration) or after a degauss (D4).Check B(+I) − B(−I) = 2 k I with the same sign convention each time.
6. Find the offset. Turn the plate end for end (the coil centre stays at z = 0) and average the two readings, or use the atoms.Check the two readings differ by twice the offset.

Tests before use

Mistakes to avoid

• Reading OUT+ against ground: the 2.5 V of OUT− is then in the reading and the sign of the field is lost.
• Steel or nickel-plated parts in the plate: screws, a connector shell, header pins. They magnetise or distort the field.
• Long untwisted leads and a noisy supply: 50 and 100 Hz ripple on the signal.
• Taking the absolute reading for the residual field: the sensor offset can be ± 3 mG. Use differences.

Each AOM arm needs 80 MHz at about +30 dBm (1 W). The chain is the synthesizer (R01), an RF switch (R04), a step attenuator (R03), an amplifier (R02), a coupler for monitoring (R06) and the AOM. The Mini-Circuits parts need things the BOM does not list: the switch needs +5 V and −5 V and a TTL gate, the attenuator needs +3 V and −3 V and six logic lines, the amplifier +24 V at 0.9 A. This part builds a small bias and control box that makes the four rails from ±12 V, holds the DIP switches of the attenuators and the gate inputs, and gives the wiring of the whole chain, with a commissioning order that protects the amplifier and the AOM.

What the design does

• Levels. With the synthesizer at +1 dBm the switch (insertion loss 1.3 dB typ at 80 MHz) and the attenuator (1.5 dB typ at 0 dB) leave -1.8 dBm, and the amplifier (gain 29 dB minimum, 32 dB typical; the data sheet gives no maximum) brings it to about +30 dBm, 1.05 W. The amplifier takes at most +10 dBm at its input, and the attenuator sets the working level from 0 to 31.5 dB in 0.5 dB steps.
• Safe state first. The DIP switches start with all six bits on (31.5 dB). The attenuator is wired in direct mode (latch enable tied high), so its state follows the switches and, at power-up, is whatever they say.
• Gate polarity. With the TTL input of the switch high, OUT 1 is off and OUT 2 is on (datasheet). The AOM sits on OUT 2, so gate high means RF on, and OUT 1 gets a 50 Ω load.
• Rails from ±12 V. The ±5 V come from an LM7805 and an LM7905; the +3.0 V and −3.1 V from an LM317 and an LM337 fed from the raw ±12 V, because 5 V − 3 V leaves only 2 V of head-room, which the adjustable regulators do not have.

Parts and materials

Tools

• Soldering iron (fast, for the DC pins of the attenuator: they overheat, Mini-Circuits AN-40-010), multimeter, oscilloscope with 50 Ω inputs or feed-through terminators.
• Drill, step drill and a nibbler for the DIP cut-outs, SMA wrench.
• A spectrum analyser or an RF power meter is useful but not needed: the coupler and the scope are enough.

Drawings

Build steps

1. Check the parts against their data sheets. Read the pin tables of the four regulators and of the three Mini-Circuits boards; write the pin numbers on the drawings.Check ZASWA-2-50DR+: RF IN pin 1, OUT 1 pin 3, OUT 2 pin 6, TTL pin 4, +5 V pin 2, −5 V pin 5. ZX76: LE J1-1, C1 J1-2, C0.5 J1-3, C16 J1-5, GND J1-6 and J1-7, C4 J1-8, C8 J1-9, C2 J1-10; P2 Vdd, P4 Vss.
2. Build the four rails. On perfboard as in the supply drawing; the 7805 and 7905 on a small aluminium angle. Power from ±12 V, nothing connected yet.Check +5.00 V (4.9 to 5.2), −5.00 V (−4.9 to −5.2), +2.99 V and −3.14 V (limits 2.7 to 3.3 V and −2.7 to −3.3 V).
3. Load the rails. Put 220 Ω (switch) and 1 kΩ (attenuator) dummies on the rails.Check the voltages move by less than 20 mV.
4. Build the control part. The DIP banks and the pull-downs as in the control drawing; LE of each attenuator tied to the +3.0 V rail; the 100 Ω and 10 kΩ at each gate input.Check with all six switches ON the six control lines read +3.0 V; OFF reads 0 V.
5. Drill and assemble the box. Follow the panel drawing: rectangular cut-outs for the DIP banks, Ø9.5 for the BNCs, Ø5 for the LEDs, Ø9.5 for the binding posts. Mount the boards on stand-offs.Check no wire touches a regulator tab; the box lid closes.
6. Wire the chain. Follow the wiring list: synthesizer to switch IN, switch OUT 2 to attenuator, attenuator to amplifier, amplifier to coupler IN, coupler OUT to the AOM, coupler coupled port to a scope input through a 50 Ω feed-through. A 50 Ω load on switch OUT 1.Check every SMA connector is tight; the cables are 0.3 to 0.5 m.
7. Commission in this order. (1) bias box alone, (2) gate and DIP with the multimeter, (3) one switch and one attenuator with the synthesizer at 0 dBm into a 50 Ω scope input, (4) the amplifier with a 50 Ω load, (5) the AOM last. Attenuator at 31.5 dB at every power-up.Check the switch gates the signal; the attenuator gives 0.5 dB steps; the coupler shows the level of the table.
8. Label and log. Label each arm A or B on every box and cable, and write down the attenuator setting that gave the best diffraction efficiency.Check a log sheet next to the bench.

Tests before use

Mistakes to avoid

• An open or shorted amplifier output: the data sheet says an open load can damage it and derates the maximum input by 20 dB. Always have the AOM or a 50 Ω load connected.
• Powering the amplifier before the bias box, or the synthesizer before the amplifier: use the order bias box, +24 V, synthesizer; and the reverse to switch off.
• The synthesizer above 0 dBm: 0 dBm already gives about +30 dBm out, and the amplifier takes at most +10 dBm in.
• DC on an RF line: the switch and the attenuator both require the RF ports to be held at 0 V or DC-blocked.
• Overheating the DC pins of the attenuator when soldering; or soldering the wrong rail to the wrong pin (P2 Vdd +3 V, P4 Vss −3 V).
• Mixing the two arms: label them, and keep the 35 kHz difference between 80.000 and 79.965 MHz in the synthesizer, not in the amplifiers.

An AOM diffracts well only when the beam meets the sound fronts at the Bragg angle θ_B = λf / 2v: 7.57 mrad (0.43°) at 795 nm and 80 MHz, with v = 4.2 mm/µs (Gooch & Housego data for the 3080 family). The acceptance of a 1.5 mm beam is only about 0.5 mrad, so the AOM needs an angular adjustment of about 0.1 mrad about the vertical axis. A micrometer pushing a tail 80 mm from the pivot gives 0.125 mrad per 10 µm division and ±81 mrad of range. After the platform, the double pass is aligned: a quarter-wave plate, the f = 100 mm lens as a cat’s eye, a retro mirror, a razor blade that stops the 0th order, and the polarizing beamsplitter that picks the returned beam.

What the design does

• Lever and micrometer. A carrier plate turns on a brass pivot; the micrometer spindle pushes its tail and a spring plunger pushes back, so there is no play. One division (10 µm) at 80 mm is 0.125 mrad, about a fifth of the Bragg angle, so the maximum can be found by hand.
• Cat’s eye. The lens puts the retro mirror at its focus, so a change of the RF frequency (which changes the deflection angle) does not walk the returning beam. With f = 100 mm the two orders are 1.5 mm apart at the mirror; a razor-blade edge placed between them on a small stage stops the zeroth order.
• Polarization. The quarter-wave plate is crossed twice, so the returned beam is turned by 90° and the polarizing beamsplitter sends it out of the arm. The two passes through the AOM therefore have crossed polarizations. The G&H drawing of the 3080-125 asks for the input polarization at 90° to the mounting plane, so one pass is always the other way: ask G&H for the efficiency in both polarizations (the double-pass efficiency is the product of the two).
• Which order. Use the −1 order in both passes (−160 MHz), as decided for the lock offset. Whether a given side of the zeroth order is +1 or −1 depends on the direction of the sound (from the transducer, marked on the AOM drawing) relative to the beam: ask the supplier, or check the sign once on the bench.

Parts and materials

Tools

• Drill press or mill, M6 and M4 taps, calliper, a reamer for the pivot.
• IR viewer card, optical power meter (E05), two irises, the oscilloscope with the photodiode.

Drawings

Build steps

1. Read the AOM drawing. The G&H outline drawing 97-01598-01 rev D gives the housing as 2.00 × 1.00 × 0.53 in (50.8 × 25.4 × 13.5 mm), drawn on the platform; from the drawing or the unit take the hole pattern, the height of the aperture above the base, which faces carry the windows, the position of the SMB jack and the maximum RF power. If the supplier builds a custom AOM, use its drawing instead.Check you have the real numbers on paper.
2. Make the plates. Cut the base and the carrier from aluminium, drill the pivot hole (Ø4.1) in the middle of the carrier, the M4 holes for the AOM, and the M6 holes of the base on the 25 mm grid. Deburr.Check the carrier turns without play on the pivot.
3. Fit the micrometer. Clamp the micrometer in its bracket on the base at the top edge so that its spindle touches the tail 80 mm from the pivot; fit the spring plunger on the other side of the tail.Check the spindle and the plunger hold the tail with light force.
4. Mount the AOM. Screw the AOM to the carrier with nylon or brass screws and a thin shim set so that its aperture is at the beam height (100 mm). Plug in the SMB-to-SMA cable with some slack.Check the aperture centre is at 100 mm ± 0.5 mm above the breadboard.
5. Find the first order. Beam Ø1 to 1.5 mm, collimated, through the middle of the aperture. RF at 80 MHz with the attenuator at 16 dB. Turn the micrometer slowly: the first order appears 1.5 mm per 100 mm off the zeroth.Check a bright first-order spot on the IR card.
6. Maximise, then raise the RF. Maximise the first-order power with the micrometer, then lower the attenuator in steps and stop at the plateau.Check single-pass efficiency of about 80 % or more (measure with the power meter).
7. Build the cat’s eye. Put the lens one focal length (100 mm) from the retro mirror, the quarter-wave plate between the AOM and the lens, and the razor blade at the mirror plane between the two foci; set the plate at 45°.Check the first order hits the mirror centre; the blade stops the zeroth order only.
8. Close the double pass. Walk the retro mirror until the returned beam retraces the incoming one through the irises; set the polarizing beamsplitter so that the returned beam leaves through the side port.Check the returned beam is on the output axis; its polarization is turned by 90°.
9. Measure and tune. Measure the double-pass efficiency and trim the plate angle and the RF level. Change the RF by ± 5 MHz.Check the output spot moves by less than 0.1 mm at 1 m (the cat’s-eye property); efficiency of 40 to 60 % overall is typical.
10. Lock and log. Clamp the micrometer and write the setting on the platform.Check the setting is on the label.

Tests before use

The G&H listing of the 3080-125 gives a 2 mm aperture, 415 to 900 nm, 4.2 mm/µs, 80 MHz, 1 W saturation RF power at 633 nm and 0.65 W at 515 nm; with the λ² scaling used here, about 1.6 W at 795 nm is an estimate. The 3080-122 has a 1.0 × 2.5 mm aperture and is for 780 to 850 nm; a 1.5 mm beam would be clipped by it. Rise time is about 159 ns per mm of beam diameter (238 ns for 1.5 mm).

Mistakes to avoid

• Adjusting the tilt (pitch) instead of the yaw: the first order walks up and down and the efficiency stays low. The micrometer must turn the AOM about the vertical axis.
• A beam bigger than the aperture: with the 3080-125 keep it under 1.5 mm; with the 3080-122 under 0.8 mm. The planned fibre collimator (f = 11 mm) gives about 2.1 mm: decide the beam reduction first (shorter collimator or a 2:1 telescope, see the distance calculator).
• Tightening the clamp of the micrometer with a steel set screw near the cell: use the clamp of the head.
• Hurrying the warm-up: 1 W of RF warms the crystal, and the efficiency drifts for the first 10 to 15 minutes.
• The retro mirror not exactly at the focus: the returned beam then depends on the frequency.

The bench has 12 beam dumps: 4 for AOM zeroth orders, 1 for the Glan–Taylor reject, 1 for the SAS pump and 6 for PBS leakage ports. The strongest is the zeroth order of the control arm when its AOM is switched off: about 10 mW with the laser at its full 15 mW.about 10 mW once the 100 mW of the Yanban is turned down at the source; the dump right after the isolation takes the rest, about 90 mW, and needs a bore that spreads it (a Thorlabs-type beam trap or a deep anodised block, never a thin card). Stray 795 nm light is invisible and it scatters onto the detector. A dump is a black block with a deep bore and a tilted entry, so the beam bounces many times in the matte black wall and dies. Over each AOM arm a hardboard cover keeps room light and draughts out and stray beams in.

Parts and materials

Drawings

Build steps

1. Drill the blocks. Drill the Ø10 bore 36 mm deep; the standard 118° point gives the end. Do not polish or ream the bore.Check depth 36 mm; the wall is rough.
2. Blacken. Anodise black or paint matte black, inside and out; let it dry fully.Check no gloss at any angle.
3. Make the mount. A cross hole for two M4 screws, or a 30 mm clamp, so that the bore axis is tilted 10° to the beam.Check the front face reflection does not return along the beam.
4. Test each dump. Beam in, IR viewer card around the block, power meter near the entrance.Check nothing visible comes back and the block is only lukewarm.
5. Cut the covers. From the cut list: top 300 × 110, two sides 300 × 160, two ends 110 × 160; holes Ø14 (in-beam, end panel; out-beam, one side panel; both 100 mm above the breadboard), Ø9 for the SMA cable.Check the pieces fit around the arm with a 5 mm overlap on the table.
6. Assemble and label. Join with black tape or plastic angles; mark every cover “795 nm invisible laser”. A beam leaves a cover only through a hole closed by an SM1 tube or a black card.Check no light leaks when you look with the viewer card.
7. Detector tube. A black SM1 × 50 mm tube in front of the photodiode keeps room light off it.Check the dark signal does not change when the room light is switched.

Tests before use

Mistakes to avoid

• A polished or glossy bore: it reflects.
• A dump with its axis along the beam: the front face sends the reflection straight back; tilt it by 10°.
• Steel screws in the covers near the shield.
• Leaving a hole open: a beam leaving the cover is a laser-safety hazard.

The SAS reference cell works at room temperature and sits outside the magnetic shield, so a printed PETG holder is fine: there is no heat to soften it, and the lock does not need a milligauss-quiet field. The holder carries the cell on its axis at a fixed height above its base, clamps it gently through felt pads, and covers it against room light and draughts, which move the line through the cell temperature. Two Ø10 mm holes let the pump and the probe through. Real cells have a torch-sealed fill stem on the side, at mid-length: it points up between the saddles, under the closed top of the cover.

Files

Parts and materials

Print settings

• Any PETG profile works: these parts are not precise to 0.01 mm. With the 0.2 mm nozzle use 0.12 to 0.16 mm layers; with a 0.4 mm nozzle the standard 0.20 mm profile is fine.
• 4 wall loops, 30 % infill, no supports in the orientations of the files, 5 mm brim for the cover.
• Check the seat with a Ø19 mm rod or the cell box before you fit the pads: the cell must drop in with the pads in place and not rock.

Build steps

1. Read the cell drawing. Tube and window diameters, length over the windows, window tilt, and the position and height of the fill stem. The holder takes a stem at mid-length up to 13 mm high and Ø8 at its base (the middle 40 mm between the saddles are free and the cover is 24 mm high above the axis), windows up to Ø23 (they sit outside the saddles), and up to 100 mm over the windows inside the cover. For comparison, Thorlabs’ drawing of its Ø19 × 75 mm cell: stem Ø6 at most 10 mm high, windows Ø21 × 3 mm tilted 11°.Check the stem lies between the saddles and clears the cover.
2. Print and clean. Cradle, two straps, cover. Remove the brim and any strings from the seats and the beam holes.Check the cover sits on the base with its pegs in the holes.
3. Fit nuts and pads. Press the four M3 nuts into the traps under the base. Stick a felt strip into each saddle and each strap.Check the nuts do not turn; the pads lie flat.
4. Mount the cradle. On a ½-inch post with the M4 screw (head in the counterbore), or with two M6 screws on the breadboard. Post height = beam height − 20 mm = 80 mm.Check the base is level and the cell axis is at 100 mm.
5. Lay in the cell. With gloves, stem up, centred between the saddles; touch only the barrel.Check the cell rests on both pads and does not rock.
6. Fit the straps. One over each saddle; tighten the nylon screws evenly until the strap just meets the saddle top. Not more: the pads, not the screws, hold the glass.Check the cell does not turn by hand; no screw is forced.
7. Align the beams. Pump and probe through the cell along its axis. The 2° wedged windows deviate the beam slightly: steer with the mirrors, not by tilting the holder.Check both beams pass the full cell without clipping the windows.
8. Fit the cover. Over the cradle, pegs in the holes, beams through the Ø10 holes.Check no beam touches the cover; the SAS signal does not change when the room light is switched.

Mistakes to avoid

• Clamping the glass hard: tighten only until the strap meets the saddle.
• A stem pointing into a saddle or the base: turn the cell so it points straight up.
• Glossy filament or no cover: room light and reflections reach the SAS photodiode.
• Using this holder for the EIT cell: that cell is heated to 65 °C inside the shield (D2), where PETG would be too close to its softening point and the cradle does not fit.

Cage-mounted cubes are expensive, and most of the price is the housing. An unmounted ½-inch cube in a printed holder does the same job (the Spain-style bench also uses uncaged cubes). A cube needs little precision: it must sit square to the beam within a fraction of a degree and must not creep, and every face it uses must stay open. This holder carries the cube on a flat seat, locates it with four corner pillars that touch only the outer 1.5 mm of each face, and holds it down with a light clamp plate on top. It screws onto the M4 stud of a ½-inch post through a brass heat-set insert (or, without an insert, with an M4 screw from the top as in D11), and the post sets the height and the rotation.

Files

Parts and materials

Print settings (Bambu Lab A1, 0.2 mm nozzle, PETG)

• Your usual PETG profile, 0.10 to 0.12 mm layers, 4 wall loops, 40 % infill, no supports: the insert pocket is a short bridge, the pocket and the ribs are vertical walls.
• Print the holder base down and the clamp flat, as exported. The base bottom (on the post) and the seat (under the cube) are then flat surfaces of the print.
• If the cube is tight or loose in the first holder, change the X-Y hole compensation in the slicer by the difference (or ask for a file with another pocket) rather than sanding the ribs.
• Matte black or dark filament: a clear or glossy holder guides stray light along the plastic.

Build steps

1. Measure the cubes. All three edges of two cubes with a calliper. The holder is drawn for 12.7 mm with 0.2 mm clearance per side.Check the cubes are 12.7 ± 0.1 mm; if not, ask for a file with another pocket.
2. Print and test the insert coupon. Press an insert into each of the three holes with a soldering iron at about the PETG printing temperature (230 to 250 °C), straight down, light pressure, until it is flush.Check one hole gives a straight, flush insert that does not turn when an M4 screw is tightened by hand; use that size (the holder file has Ø5.5).
3. Print one holder and one clamp. Remove any strings from the pocket and the ribs.Check the cube drops into the pocket without force and does not rattle; the ribs are straight.
4. Fit the insert. From below, flush with the bottom face of the base.Check the insert is not proud: a proud insert lifts the holder off the post and lets it rock.
5. Screw the holder onto the post. Hand-tight on the M4 stud of a ½-inch post, then set the post in its post holder with the post top 84.65 mm above the breadboard.Check the beam height mark (15.35 mm above the post top) is at 100 mm.
6. Put the cube in. With gloves; touch only the top and bottom faces. Turn it so the coating diagonal reflects towards the side you need (if the cube has a mark for the input face, follow it). Push it lightly into one corner.Check all four faces you use are free and clean.
7. Fit the clamp. Nut into the trap on the underside, nylon screw through it. Clamp on the pillars, two M3 × 12 screws into diagonal pillars, then the nylon screw down until it just touches the cube and a quarter turn more.Check the cube does not move when you push it with a finger; nothing is forced.
8. Align. Rotate the post for the direction of the reflected beam; steer with the mirrors before the cube, not by tilting the holder.Check the transmitted and reflected beams leave through the middle of the open faces.
9. Drift test (before printing the rest). Mark the reflected spot on a wall about 2 m away. Check it again after a day and after gently tapping the post.Check the spot moves less than 0.5 mm (0.25 mrad); then print the other holders.

Mistakes to avoid

• Tightening the nylon screw hard: stress in the glass makes it birefringent and lowers the extinction. Finger-tight is enough.
• A tilted or proud insert: the holder rocks on the post.
• Glue, tape or fingerprints on an optical face.
• Using the screw version without removing the stud from the post, or the insert version with the stud removed.
• Clear or glossy filament next to the beams.

These are the places where the drawings use an assumption. Each one is also in the “Confirm before you start” box of its part. Tick them off in your build log.

Calculations

• Solenoid field and uniformity: exact ring-by-ring Biot–Savart sum (elliptic integrals), checked against a brute-force sum to 10⁻¹²; the volume is the cell, ρ ≤ 12.5 mm and |z| ≤ 12.5 mm; the shield is not modelled (it raises the field by a few per cent): calibrate in place.
• Heater and thermal numbers: a lumped estimate with foam k = 0.05 W/m/K, h = 8 W/m²K and a heat capacity of 8 J/K, once for the stack in free air (upper bound of the loss) and once as a series chain through the former, the shield gaps and the outside. They are estimates: measure your own stack.
• RF levels: the arithmetic of the data-sheet values (typical gain 33 dB, insertion loss 1.3 dB and 1.5 dB); the coupler read-out is Vpp = √(8 · 50 Ω · 10^{(P − 19.5)/10} / 1000).
• AOM numbers: θ_B = λf / 2v with v = 4.2 mm/µs; the saturation power at 795 nm scales with λ² from the 633 nm value (an estimate).
• All numbers in the text and in the drawings come from the same set of parameters, so they agree with each other.

No prices are on this page: costs are kept in the group’s private records.