Panel build

The organising idea

Three voltage classes, kept apart

The panel is laid out in bands by voltage, not by function. Line voltage stays at the bottom where the feeds enter, 24 VAC sits above it, and the logic sits at the top furthest from the transformer. This colour code runs through every diagram and table below — and in the wiring figures it is doubled by stroke weight, line voltage drawn heaviest and signal lightest, so the distinction that matters most never rests on hue alone.

The parts

Every panel component, photographed

Nine of the eleven panel components have a chosen part number, and all nine are shown here as the manufacturer or distributor photographs them. The two without a photo are the two that are still undecided — the gap in this gallery is not a sourcing problem, it is the outstanding decisions, and it is worth reading that way.

no part chosen

no part chosen

Pi accessories are not shown: the microSD card, the M2.5 standoffs and the heatsink kit are chosen but belong to the Pi rather than to the plate.

Figure 1

Backplate, plan view

Why the parts sit where they do

Figure 2

Where the power goes

The HAT

Relay channels

Eight channels, fixed by the equipment rather than by preference. Channels 1–3 need changeover contacts because a three-wire actuator keeps one of two lines energised at all times; channels 4–5 must be isolated because the Raypak supplies its own low voltage on that terminal block.

                      top
   +-----------------------------------+
 8 |                                   | 1
 7 |          header, DIP              | 2
 6 |                                   | 3
 5 |                                   | 4
   +-----------------------------------+

Figure 3

What each channel actually switches

The same eight channels, drawn terminal to terminal. Read it left to right: what feeds the contact, what the contact is, and what it lands on. Five channels are fed from this panel's own buses, two are dry contacts that must touch nothing here, and one pulls a coil and nothing else.

Figure 4

One valve, traced end to end

Figure 3 draws all eight channels at once, which is the wrong scale for the question that actually comes up while holding a screwdriver: where does each individual wire go? So here is one valve — channel 1, the intake actuator — drawn as the complete circuit it is, out of the transformer, through the relay, out to the actuator and back.

It is a loop, and that is the part the plan views hide. Every other figure here draws a cable as a line going somewhere. Current does not go somewhere; it goes around. Twenty-four volts leaves one end of the transformer secondary, reaches the actuator through whichever conductor the relay has selected, passes through the motor, and comes back on black to the other end of the secondary. Two of the three conductors are in a complete circuit at any moment. The third is simply open.

The relay does not switch the valve on and off. This is the thing worth reading twice, and it is why channels 1–3 need changeover contacts rather than the ordinary on/off kind. COM is permanently fed with 24 VAC. The contact chooses whether that feed continues to N.C. or to N.O. — so one of white and red is live at all times, in both valve positions, including when the Pi is off. Energising the coil moves the feed from one to the other, and that is the entire command.

Which is why a permanently live wire does not cook the motor. The actuator's cam opens that direction's limit switch as it arrives at the end of travel. The line stays energised; the motor stops. This is also what lets ADR‑9 hold relay 3 energised through all of pool mode without a duty cycle problem — the 1 min ON / 8 min OFF limit governs movement, not sitting still with a line hot.

Which conductor gives which position is decided at the valve, not here. Red and white are just "the two directions"; the cam plate and the rear AUTO 1 / OFF / AUTO 2 toggle set what those directions mean physically. If the valve ends up backwards on first power-up, flip the toggle at the actuator — do not rewire the panel, and do not swap red and white at the terminals. Swapping them inverts the fail-safe, which is the one property here chosen deliberately: with the coil unpowered, the bypass must rest in flow.

Only two blocks on rail B belong to this valve. TB2a and TB2b carry white and red, each landing the field wire in one clamp and the wire up to the HAT in the other. Black shares the common bus with the other two actuators and the contactor coil, and the hot side is shared the same way. That is the whole reason for the jumpers: one arriving conductor, several departing ones, at one potential.

One thing this exposes. The terminal schedule on rail B reads TB1 COM · TB2–4 valves · TB5 coil, which named a common bus but no hot bus — though the hot leg needs exactly the same treatment, since it feeds four COM screws. Figure 6 settles it: three blocks for each bus, bridged by a comb, and a schedule that adds up to something you can order against.

Figure 5

The same panel, wired

Figure 1 says where the parts go and Figure 3 says what connects to what. This is both at once: every cable drawn on the real plate, entering at its own gland and routed through the gaps the parts leave. Colour is voltage class, as everywhere else.

Figure 6

Every 24 VAC conductor, block by block

Figure 4 followed one valve. This is the whole 24 VAC network at once — three actuators, the blower coil and the transformer — drawn at the level the confusion actually lives at: which conductor lands in which clamp, and what the jumpers tie together.

Positions here are chosen for legibility, not measured. Figure 5 is where the cables really run; this is what happens once they arrive.

The entanglement is two nodes. Everything else is point to point. The hot leg of the secondary arrives once and has to reach four COM screws; the common leg leaves once and has to collect three actuator blacks and the far side of the contactor coil. Those are the only two places where more than two conductors meet, and both are made by a comb rather than by twisting wires together in a clamp.

Two clamps per block, and the jumper is separate. A feed-through block has one clamp on each side of a single internal bar, so one block already takes two conductors — that is what the switched pairs use, HAT wire in one side, field wire out the other. The comb plugs into a third opening on the front and bridges neighbouring bars without consuming either clamp. So a node does not need a block per conductor.

But it does need a block per conductor leaving in the same direction, and that is what sets the count. The two clamps of a block face opposite ways. All four COM conductors go to the same place — up to the HAT — so all four want the upper clamp, and four upper clamps means four blocks. Dividing 5 conductors by 2 clamps gives 3, which is the number that fits the copper and the wrong number to build: one COM would have to leave on the face pointing away from the HAT and route back around the strip. The hot bus is four.

The common bus is five, and there is no choice about it. Every one of its conductors leaves downward — the transformer common, the contactor coil return, and the three actuator blacks routed under the strip to their glands. Five conductors on one face is five blocks. Its upper clamps are all spare, which is a good place to put a meter probe.

Read the table by column and the whole rule is visible. Eleven entries in the upper column, thirteen in the lower, and no cell doing double duty. Four COMs stacked in the upper column is why the hot bus is four blocks; five returns stacked in the lower column is why the common bus is five.

Sixteen blocks, one four-way comb and one five-way — 32 clamps for 24 conductors, so eight spare. That is the count to order against, and it is the first time this plan has had one: the BOM line still reads terminal blocks — no part chosen.

Size it before you buy it. Sixteen blocks at a typical 6.2 mm pitch is about 3.9 in, comfortably inside rail B’s 8 in even after two end brackets. But pitch runs from roughly 5.2 to 8 mm across the usual 2.5–6 mm² range, so this is a number that depends on the part and there is no part chosen yet. Add end plates: a feed-through block is open on the face the comb enters.

The comb is the one part that can short the transformer. It must be a four-way piece on blocks 1–4 and a five-way on 12–16, and neither may reach a neighbour. The gap drawn before block 12 is an end plate, and it is there for this reason. A comb that over-runs block 4 into block 5 ties the hot bus to CH1’s N.C. output, so white and red go live together and the intake actuator is driven both ways at once. One that reaches back from block 12 into block 11 ties CH6’s N.O. to common, and the first blower command is a dead short across the secondary. Cut them to length, count the ways twice, and leave a gap block if the part you buy cannot be cut cleanly.

Rail C adds four more. This figure is rail B only — the 24 VAC side. The panel’s own 120 V feed arrives once and has to reach two places, the transformer primary and the 5 V supply, so line and neutral are each a three-conductor node: two blocks apiece, bridged two-way. Four blocks, two combs. Nothing else lands there — the blower goes straight to the contactor and the light straight to CH7, both looped from the Intermatic subpanel, and every equipment grounding conductor goes to the bolted PE bar rather than to a rail. Twenty blocks on the panel altogether.

Two faces

Cable entries

Split across two faces: low voltage down the left, line voltage along the bottom. Ten entries need about 13.25 in of hole and clearance and the bottom wall offers roughly 11.5 in usable, so one row was never going to work — and splitting them puts the two classes through different walls, which is where they should have been anyway. Every grip seals only inside its cord range — a cable too small for the grip leaves an open hole just as surely as one too large.

Order of work

Assembly sequence

The order matters in two places: the plate is laid out before anything is drilled, and the panel is proven at 24 VAC before line voltage is ever connected.

1. Lay the parts on the plate, unmountedTransformer, contactor, PSU, Pi and rail, physically positioned. Confirm the diagonal separation and that DIN rail runs clear the transformer. Adjust on paper, not with a drill.
2. Mount DIN rail and the fixed partsRail first, then transformer on its feet, contactor, PSU. Pi and HAT on 11 mm standoffs last so nothing is dropped on it.
3. Drill entries and fit glandsOnly once the actual cables are in hand and measured. Bottom face, low voltage left, line right. Breather-drain at the lowest point.
4. Wire 24 VAC and signal onlyTransformer secondary to TB1–5, relay channels 1–6, RS-485, heater dry contacts. Nothing line-voltage yet.
5. Power the logic from a bench supply5 V to the HAT, Pi boots, supervisor runs. Exercise every relay channel and watch the actuators move before the panel is anywhere near the pad.
6. Electrician connects line voltageFeed in, blower out, light out. Every EGC to the ground bar, and the bar jumpered to the incoming ground — nothing in an all-plastic box is bonded on its own. Permit and inspection as your jurisdiction requires.
7. Commission njsPC against real equipmentPoint njsPC at /dev/serial0, set enable_uart=1, remove console=serial0, and set the HAT's TX/RX DIP switches on. The supervisor will tell you on screen if the port is still missing.
8. Seal, then measure the temperatureClose it, run it through a hot afternoon, and read the internal temperature. That is the one number the whole thermal design is still missing.

Not settled

What this plan does not decide