Bed platform

1What the lift gives us, and what it forbids

Ours is already installed and was repaired: broken set screw replaced, trolley re-seated, limit switch bypass confirmed functional. So the track positions are not a design choice, they are a measurement. What follows is from the Happijac users manual.

Four rules from the manual that constrain the design, not just the operation. Never operate the bed with persons on the platform. Never hang more than 20 lb from the cross-connecting shaft, which rules out running the lighting loom or anything else across it, so cabling goes on the rails and drops from one articulated point. Never travel with anything but bedding on the bed. And the motor carries a manual brake release lever, which is the only way the bed comes down on a dead battery, so it has to stay reachable after the ceiling and trim go in.

Maintenance is one line: keep the trolley channels clean and give the inner faces a light silicone spray. The manual says it reduces noise and amperage draw, which makes it the cheapest thing on this page.

2The whole thing, in one picture

Two rails carry 4 struts. The oak and the struts both stand on the rails' bottom flanges, the oak outside against the leg and the struts butting into it. The 0.5 in ply deck lies on the struts and stops against the oak, which carries on up past it to form the edge.

That is the build-up. Everything below this section is the reasoning behind it.

The headboard hinges off the head oak at 5.5 above the deck and stands to 20, which shows 14 above the mattress. It folds inboard, flat onto the mattress, because outboard is the Happijac track and the platform edge sweeps that band every time the bed rises. Folded it adds 1.25 to the stack, which is the state it is in whenever the bed is travelling.

X is the only unknown, and it is the width across. It is H1 minus 2, where H1 is the clear width across the Happijac trolley tabs, which has not been taped. Everything running the other way is 60 and is real. The working figure for X is 74.6 and it comes from an assumed box, not a measurement.

3The width chain, which decides the mattress

Tope is a wide box with a flat floor and no wheel well intrusion, and the convertible zone is 60 in fore and aft. A platform 60 in fore and aft means the sleeper lies crossways, so the platform width is not the bed width. It is the sleeping length, and every inch lost to the walls and the trolleys comes off someone's feet.

So the mattress decision and the bracket decision are one decision, and both of them resolve the moment somebody takes measurement C8, the Happijac column faces. Until then the working assumption on this page is the short queen, because it is the one that fits without fighting anything.

4The edge, and the bracket flip

The factory method faces the trolley tab inboard and drops loose vertical pins through it. That costs width, it lets the frame float when the walls are out of square, and it rattles. Flipping the tab 180 degrees and bolting the rail to the trolley face buys the width back. The manual explicitly permits the flip: the tabs are designed to be adjusted up, down, or turned over.

What the flip actually costs. Three things, and only the first one is usually mentioned. The tracks have to be plumb and parallel or the trolleys bind, and the couple whose build this follows put that at under 1/4 in across opposing measurements. Bolting replaces the pins, so the platform stops being liftable off in one move, which matters the first time something behind it needs service. And the float goes away, which is exactly the float that absorbs a box that is not square. We have not confirmed the box is square.

The wall rub is a bigger risk here than in their build. On their first cycle with the crossbars fitted, the trailing frame edge carved into finished panelling. Our walls are 1/8 in ply under microcement, the one finish in this build that is known to crack. Cycle the lift with the frame bolted and the walls still bare, and find the clearance before the panels go on rather than after.

5The load budget

Built the way the reference build was built, at Tope's dimensions:

It works, and the lifting case is never close. The static case has 405 lb for two people, which is enough but is not a lot of margin for sitting on one edge with a third person visiting. The deck is the heaviest single item and it is the easiest to cut: at 25 in between beams, 1/2 in Baltic birch is far more than the span needs, and going to 3/8 returns about 15 lb straight to the occupant line.

6Section selection, and one claim to drop

One adult, 200 lb, concentrated at the midspan of a single beam over 74.6 in, which is deliberately pessimistic: the deck shares a real load across two or three beams. 6061-T6, yield 35 ksi, allowable stress taken at 0.55 of yield or 19.2 ksi. Deflection limit L/360, which on this span is 0.21 in. A bed is judged on stiffness, not strength: not breaking and not feeling like a trampoline are different tests, and the second one is harder.

Only 1.5 x 3 x 1/8 and 1.5 x 3 x 3/16 meet both criteria, and the lighter of those is 1.5 x 3 x 1/8.

That reverses an earlier reading on this page. The reference build chose 1.5 x 3 so the underside would read as exposed timber beams from below in daytime mode, and on stress alone 1 x 2 looks like it would do. It does not, and the reason is stiffness rather than strength. 1 x 2 sags 0.52 in against a 0.21 in limit, and even sharing that adult across two beams it still reaches 0.26 in, which misses L/360 and only just scrapes L/240. So 1.5 x 3 is not only a styling choice. It is the smallest section here that is comfortably inside an ordinary deflection limit at our span, and the timber-beam look is a bonus rather than the reason. Their aesthetic call and the engineering happen to agree.

Their claim that 1 in square tube is also structurally adequate stays wrong, and at our span it is not close: it lands essentially at yield with over 3 in of sag. Do not carry that sentence forward.

One detail is genuinely open. A 2.5 in angle and a 3 in beam cannot be flush at both top and bottom. Sit the beams on the bottom flange, as the reference build did, and the deck plane is set by the beams while the rail hides below it. Hang them from a top flange instead and the deck lands on rail and beam together. That choice decides the finished deck height, so it has to be made before the cut list, not during assembly.

7The frame, in plan

Two rails running fore and aft, one against each side wall, picking up two trolleys each. 4 beams spanning across between them, carrying the deck. The beams are spaced along the bed, not along themselves, which reads as obvious written down and was wrong in this model until the plan got drawn.

The one number missing from this drawing is H2, where the trolleys sit fore and aft. It decides how much of each rail spans and how much cantilevers past the last trolley, and it is not measured. Everything else follows from H1 and the 60 in depth.

8The rail, which nothing had checked

The cross beams had numbers. The rails did not, and they are the members that actually carry the bed into the trolleys. Each rail takes half the static limit, 300 lb, spanning between two trolleys. H2 is unmeasured, so this is run at the worst case, trolleys at the very ends of the platform.

An angle is not symmetric, so its two section moduli differ and the smaller one governs: 0.394 in³. That gives 5705 psi against 19250 allowable, and 0.120 in of sag against an L/360 limit of 0.167. It passes, but note that it is at 72 percent of the deflection limit and only 30 percent of the stress limit. Stiffness governs here too. Do not thin the rail below 1/4 in.

One thing an angle does badly is twist. Load lands on the horizontal leg, away from the shear centre, so a bare rail wants to roll. What stops it is the 4 cross beams bolted along it, which is a reason to bolt every beam to the rail web rather than just sitting them on the flange and trusting the deck.

9The deck, and the seams that stopped being a problem

What made that possible is the sheet. In 5 x 5 Baltic birch, nothing could span the crossways dimension, so a joint running fore and aft was forced, no strut could ever sit under it, and it needed a backing member of its own: four spine segments fitted between the struts. That member existed only to serve a seam the sheet size had forced on us. In 48 x 96, the 96 dimension clears the crossways span outright, so every seam can run the other way and land on structure that is already there.

2 sheets, and the second one is shared with the shower door. Two deck pieces sit side by side on the first sheet, using 38.5 of its 48 width. The third goes on the second sheet alongside the D1 door leaf, which is 19.5 wide, so 39.5 of 48. That is exactly the two sheets bought on Mr Plywood 556076.

Cut them oversize and trim on the frame. The slip says 20 across the board where the drawn sizes are 19.25, 20, 19.25, and that is right: the platform width is still a formula in H1, so the only honest final dimension is the one you measure off the built frame.

It has to stay a true all-birch core. The thickness below is calculated on Baltic birch properties. Birch-faced plywood on a softwood core is much weaker and will not hold that number, so 48 x 96 Baltic birch or an equivalent all-birch panel, not a birch-faced fir sheet.

Why 4 struts, and why not three

Asked twice, on 2026-08-28 and again on 08-29. The count came down from five to 4 on 08-29 because Jeff prefers how 4 looks, and the numbers already allowed it. Three does not work.

The deck is what decides this, not the struts. Sag goes with the cube of the pitch, so the three numbers above are eight, two and one times each other for pitches of 30, 20 and 15. At three struts the deck lands on 0.60 in, which is the exact figure this page rejected for 1/4 ply: not a floor, a trampoline, and visible from below because the soffit is the ceiling.

A correction worth recording, because the first version of this argument was rigged. It compared builds of EQUAL stiffness, which forces the ply thicker every time you drop a strut and makes fewer struts look heavier. Against the real standard, which is acceptable rather than identical, the ply does not change and the weight goes the honest way: 4 struts and 0.5 ply is about 87 lb against 95 for five. Three still fails, but on its own merits, not on that comparison.

The struts themselves were never the constraint. At the 75 in span a 1.5 x 3 x 1/8 section under 200 lb at midspan runs 4,630 psi against 19,250 allowed, and sags 0.164 against an L/360 limit of 0.207. Comfortable at four. It would have been marginal on a 1.5 x 1 section, which is why the ordered size matters.

Beams go at 0, 20, 40, 60 in from the front edge, a 20 in pitch. That is the spacing along the bed, not along the beams: each beam is 74.6 long across the vehicle.

Thickness, settled. One knee, 200 lb, at the midspan of a 12 in wide strip over the 20 in span. Baltic birch at 9 ksi bending and 1.5M modulus.

0.5 in is the honest minimum and 0.5 in is what to build. 1/4 comes out at a safety factor of 1.5 with 0.6 in of sag, which is not a floor, it is a trampoline, and it would be visible from below because this soffit is the ceiling. 3/8 works and returns about 14 lb. The reason to stay at 0.5 anyway is the fastener: a #10 countersink is 0.10 deep, which leaves 0.28 in under the head at 3/8 and 0.40 at 0.5, and those bolts are carrying the deck against a lift that swings it to the roof twice a day.

Fix the deck down with countersunk bolts, not screws. Flush heads so nothing telegraphs through the mattress, and bolts because 0.125 in wall is almost nothing for a screw to bite into.

Ventilate it, but not by drilling the ceiling. Foam on a sealed sheet traps body moisture, and we have Baja humidity, a shower in the same box and a bed that spends its days folded against a roof where no air moves. The obvious fix is to perforate the deck, and the obvious fix is wrong here: every hole is visible from the office below, because this soffit is the ceiling.

So ventilation goes on top, where nobody sees it. A hypervent or 3D mesh underlayment between foam and ply lifts the mattress off the deck and lets the air layer breathe sideways to the perimeter, which is how boat builders solve the identical problem in a berth. If more is wanted, slot the deck hard against the rails, where the eye does not go, rather than in the open field between beams. The reference build cut slits and accepted seeing them. We should not, because their bed underside was not lit and ours is.

10Slats, and why they are rejected

Slats over the same beams look excellent on paper. They delete the seam, the battens and the whole ventilation problem, and they take about 30 lb off the bed, which is more than any other single change available.

That last row decides it. With the bed up, this platform's underside is the office ceiling, and it has four pucks in it. Looking up at a slatted bed you see beams, then slats crossing above them, then mattress through the gaps. It is a bed seen from underneath, not a ceiling.

So the deck stays solid, and the 30 lb stays on the bed. The five beams standing proud below a flat soffit are the entire point of choosing a 1.5 x 3 section in the first place, and slats throw that away to solve a problem a mesh underlayment solves for twenty dollars.

Recorded because it was nearly built the other way. Ventilation and weight are real, and optimising for them in isolation produced an answer that failed a requirement neither of them could see.

11The underside, which is the room you look at

With the bed up this is a ceiling, lit by four pucks, seen from a sofa. It is the most looked-at surface on the bed and the only one with an aesthetic brief: heavy exposed timber beams over a flat soffit. Everything in the frame already serves that. Five 1.5 x 3 beams hang 3 in proud below a continuous plane at 20 in centres, with the rails closing the perimeter. That is a coffered ceiling, and it is why the section is 1.5 x 3 rather than something lighter.

The beams arrive as bare mill-finish aluminium, which reads as scaffolding, so the finish is a decision rather than a default.

The veneer wraps three faces of each member: both cheeks at 3 and the soffit at 1.5. The two end beams are the fore and aft edges and the perimeter oak is already their outboard face, so they show a cheek and a soffit rather than two cheeks.

15 sq ft, call it 17 with 15% for trimming and overlap. One 4 x 8 sheet of paper-backed white oak veneer is 32 sq ft and covers it with room to spare.

This is a deliberate exception to the black anodised trim family set by purchase, order 189513, and it is worth naming rather than drifting into. Black anodised governs trim: edges, reveals, the panel-to-Happijac joint. These beams are not trim, they are a surface, and oak beams over a birch soffit inside an oak perimeter is one material doing the room rather than two competing for it.

The soffit between the beams is Baltic birch, seen from below. Left sealed and clear coated it reads light against darker beams, which is the contrast doing the work. Do not microcement it: it is a moving panel on a lift and microcement crazes.

The 4 under-bed pucks let into this face, so their positions have to be set before the deck is drilled and not after. They go in a 40 x 40 grid, two by two. An earlier version of this page put all four in a line down the centreline, which lights the same floor and reads as a corridor. This is a room.

The grid is not free either. Fore and aft each puck has to sit in a bay between beams, and the outer two bay centres are the only symmetric pair far enough apart to read as a grid rather than a cluster. Crossways they are symmetric about the centreline at 20 either side, which makes it a true square. It was a rectangle until 2026-08-28, clamped by a deck spine that no longer exists.

Every puck sits 10.0 in clear of a beam fore and aft and clear of the spine crossways, and the whole grid is centred on the platform both ways. It is 40 by 40 rather than square because of the spine, which is close enough that nobody will read it as a rectangle. If the office below turns out not to be centred, sliding the grid costs nothing until the deck is drilled.

Wiring runs above the deck, not below it, because below it is the finished ceiling. Rout a 0.25 in groove in the deck's top face from each puck to the head edge and let the wire out through the head member. A wire lying loose on the deck under a mattress is a ridge you feel; a wire in a groove is not.

12Edges, and the mattress problem they expose

The perimeter is already aluminium. Head and foot are the two rails, the fore and aft edges are the first and last cross beams. So the frame closes itself, and none of it is any use for holding a mattress in: the beams land on the rail's bottom flange, which puts the rail top at 2.5 and the beam top at 3.25 against a deck surface at 3.75. The whole frame finishes 0.50 in BELOW the deck. Retention has to be added on top, it cannot be borrowed from the structure.

The heights are not arbitrary, and neither board gets ripped. The rail bottom sits 3.75 in below the deck surface, so a fascia of 3.75 in covers the aluminium completely and lands flush underneath. That fascia is the same on every edge, so the upstand is just whatever the board has left over: a 1 x 10 at 9.25 actual gives 5.5 above the deck, and a 1 x 6 at 5.5 gives 1.75. Both are heights we wanted anyway, so the boards are used whole. 1x stock already surfaces to 0.75, which is the member thickness. Crosscut to length and fit.

A caution on reading board sizes here, because it is an easy mistake to make in the other direction: a 1 x 6 does not give 5.5 in of upstand. It gives 1.75, because 3.75 of the board is fascia below the deck line. The board width and the height above the mattress are two different numbers and they are 3.75 apart.

Left, the head edge on its rail. Right, a fore and aft edge on an end beam, whose bottom sits 0.25 in higher, so the oak hangs that much below it and the bottom line stays continuous all the way round. The foot edge is the right hand member's height on the left hand rail.

Where the oak goes, and the view that decided it

With the bed UP this perimeter is overhead and prominent from the kitchen. It is the beam that frames the office below, so what matters is not what the section looks like, it is how many materials that face is made of and how many lines run along it.

Two earlier proposals both failed that test. A, oak sitting on the deck, is four bands and three lines: between the rail top at 2.5 and the deck underside at 3.25 there is an open 0.75 in gap you look straight into, then the ply edge, then the oak. B, oak on the flange, is tidier at two bands and one line, but the bottom 2.5 in is still veneered aluminium set proud of the oak above it. A third, hanging the oak outboard of the rail, wanted space that is not there.

Strut is 74.62, unchanged from the flange scheme. The deck goes to 75.12 across, 0.5 wider than the strut span, because it runs over the rail and butts the oak.

Where the oak sits

Where the oak sits, if B

It is a board screwed to the outside face of the rail. It is not sitting on anything. Its bottom edge finishes level with the bottom of the rail, which is why zero on the drawing above is both at once.

Fixed with stainless machine screws into tapped holes in the aluminium. Chamfer or radius the top inside arris so bedding does not chafe on a square edge every time the bed is made.

The head member takes the only real load, because someone leans on the headboard. A person putting 50 lb at the panel's mid height works out at about 124 psi in 0.75 in hardwood, against roughly 10,000 psi available. It is not close. The fixings matter more than the section: screw it every 8 in, not every 16.

So the mattress goes custom, cut to the platform. Inside the 0.75 in upstands with 0.25 in of clearance that is about 58 x 75. Custom foam is normal in a van, it is not expensive, and it buys building tolerance at the exact moment we need it, because H1 is still untaped and the crossways figure moves with it.

Order the foam last, off the finished inside dimension of the built frame, not off this page. Everything else here can be cut to a drawing. A mattress cannot, and it is the one part where being 1/2 in out is something you feel every night.

Sliding off the front, which is the only edge this applies to

The bed sits high. At its lowest it is resting on top of the small fridge, so nobody is sitting on this edge and standing up off it: you leave it by sliding down, and your weight crosses the front edge on the way. That is the motion the front upstand has to stay out of.

Which is why the front is a 1 x 6 and the walls are not. At 1.75 above the deck the front upstand stays under the foam even when the foam has given up most of its height: clear by 0.7 to 1.8 in. Nothing to cross. Go one board up to a 1 x 8 and it is somewhere either side of level depending on how the foam behaves, which is exactly the case where a mock-up decides it and a page cannot.

This wants a mock-up, and the mock-up should measure rather than judge. Put the real 6 in foam on something at the bed's real height, sit on the edge as if leaving, and have someone measure the foam height at the edge under load. That single number replaces the 2.4 to 3.6 range above and settles the front board on the spot. Then clamp a scrap at the candidate upstand and slide off it twenty times, in shorts, and once more at night half asleep, which is the actual use case.

Oak on all four edges settles the mattress question by making it. The alternative was bare fore and aft edges and thin aluminium angle at head and foot only, which is the only way a stock short queen fits. That is now dropped: it gave up retention on the two edges you actually roll towards, and it gave up the continuous timber line. So the foam is custom, and that is a decision rather than an option. Order it last, off the built frame, per the note above.

13The fold-down headboard

One fact sets this entire design: outboard of the head edge is the Happijac track, about 2.5 in away. There is nowhere for a panel to swing to on that side, and nothing can be fixed to the wall there either, because the platform edge sweeps that whole band of wall every time the bed goes up. So the headboard has to ride on the platform, and it has to fold inboard, over the mattress.

The hinge does not go at deck level. A panel pivoting from the deck cannot fold flat because the mattress is in its way. It goes at mattress top, on a 6 in upstand that earns its place twice: it carries the hinge, and it is the head-end mattress retainer.

The cost is 1.25 in of up-travel over a 14 in strip at the head end, because folded, the panel is the highest thing on the bed. Whether that is affordable depends on D5, the platform underside height in the UP position, which is unmeasured. If it turns out to be tight, the panel thins before anything else does.

No wiring in it. A loom across a hinge that cycles daily is a fatigue failure waiting to happen, and the lighting plan already solved reading lights a better way: leaning magnetic dock lanterns, zero rough-in. Leave the headboard dumb.

14Sourcing, Portland

Both aluminium sections are standard stock, not specials, and both are available cut to length locally rather than mail order.

Buy structural 6061 from a metal supplier, not hardware store extrusion. The 1/4 in wall on the rails is what stops them bowing under a moving sleeper, and the thin stuff sold in racks is not the same material. Salt air is the threat model for this build, so 6061 and stainless fasteners, never zinc plated.

15Cut list

Split by what the tape blocks. The fore and aft dimension is 60 and every part that runs that way is a real length you can cut today. Everything crossways is a function of the platform width, which is a function of C8, and those stay formulas so nobody cuts metal off an assumed 84. KNOWN means cut it; H1 means wait.

Buy structural 6061 from a metal supplier, not hardware store extrusion. The 1/4 in wall on the rails is what stops them bowing under a moving sleeper, and the thin stuff sold in racks is not the same material.

Veneer. 15 sq ft of surface, bought as 17 to allow 15% for trimming and overlap. Paper-backed white oak in a 4 x 8 sheet is 32 sq ft and covers the lot. Buy it whenever: the area moves a little with H1 but a whole sheet absorbs that.

Oak to buy. Two sizes, neither ripped. The head and foot are 1 x 10 at 5.0 ft each and are KNOWN, so that is 10.0 ft of 1 x 10 you can buy today. The back is another 1 x 10 and the front a 1 x 6, both at beam length, about 6.2 ft each at the current assumed width, and both wait on H1. Buy those two long: a member that is short is scrap, and their length is the one nobody knows.

Hold the front board specifically until the foam mock-up in section 11. If the loaded foam edge turns out taller than expected, the front can go up to a 1 x 8 and gain 1.75 in of upstand. That is a different board, not a recut, which is the good reason to leave it last.

16Open

• C8, the Happijac column faces. Decides the platform width, the mattress, and whether the bracket flip is even worth it. Everything else here waits on it.
• H1 to H4, the lift interface, now section 5 of the measure sheet: tab face to face across, trolley spacing fore and aft, tab height above the trolley body, and how far the track stands proud of the raw shell. H1 is what the cut list actually consumes.
• Are the limit switches in circuit? The repair note says the bypass was confirmed functional, which is the manual's diagnostic procedure. If a bypass was left in, the bed can drive into its own stops.
• Is the box square? The flip removes the float that covers it.
• Beam on the bottom flange or hung from a top flange. Sets the deck height.
• The platform height above the floor with the bed DOWN. Never measured, and it decides whether the bed edge is a comfortable perch or a clamber. It governs the exit question far more than the upstand does. Measure it before testing upstand heights, or the upstand gets blamed for a seat height problem. See section 11.
• How the head oak is held against the headboard load. The rail leg gives only 2.5 in of fixing and the member is 9.25 in tall, so everything above that is cantilevered off the screws. Blocking or a bracket, and neither is drawn. See section 2.
• The loaded foam height at the front edge. One measurement off a mock-up settles the front board, and nothing else on this page can settle it. The working range is 2.4 to 3.6 in and it is an estimate, not a spec. See section 11.
• Puck housing depth. The deck gives 0.5 in and a common recessed puck wants more. This decides whether the 2.5 in cutout is right and whether the mattress needs a local relief. Nothing gets drilled until a housing is in hand. See section 10.
• Where the office actually is. The puck grid is centred on the platform both ways, which is right until somebody puts a desk somewhere else. Sliding it is free until the deck is drilled.
• D5, the platform underside height with the bed UP. Decides whether the headboard's 1.25 in folded thickness is affordable.
• Which side is the head end, driver or passenger. Sets which wall the headboard lives against and which side you climb in from.
• Mattress thickness. 6 in is assumed and it sets the upstand height, the hinge line and therefore the headboard's total height.

17Sources

• Happijac chain drive bed lift users manual, ratings, precautions and the motor and brake test
• Happijac bed lift users manual, later revision, carries the rating scope footnote
• Ananda Living Experiences installation guide, the written version of the reference build, with the member sizes
• Happijac bed lift full installation tutorial, Nicole and Edden, the source of the outboard flip claim
• Lippert Happijac product page