This is the project everything else on the boat is waiting on. Refrigeration, the Raspberry Pi nav stack, the autopilot, and every future idea that starts with "what if we ran…" all come out of the same battery bank. Until there's a real charging source that doesn't involve running the diesel, the boat's ambitions are capped.
How we got here — four designs deep
Build logs that only show the finished thing aren't much use to anyone, so here's the actual path. We're on the fourth idea, and every one of the first three taught us something we needed.
Design one: flexible panels on the bimini
We started with flexible panels — the obvious choice for a curved bimini, in theory. We sent them back for two reasons.
First, we'd have had to frame them anyway. A flexible panel that isn't supported across its whole area is one that works loose, flogs in a blow, and eventually cracks cells you can't see. Once you're building a frame, the main advantage of "flexible" has evaporated.
Second, flat is kinder. Rigid panels mounted flat sit in clean air instead of sandwiched against canvas cooking in their own heat, and flat cells don't fight their own geometry. So: rigid panels, mounted flat.
Design two: build our own arch out of pipe and connectors
The next plan was a whole second structure — a custom bimini frame built from stainless pipe and fittings, no tube bender and no TIG welder required. We still like that approach on its merits, and if this boat didn't already have an arch we'd be building it.
But it does. There's a perfectly good stainless arch aft with the mainsheet traveller riding on a plate across the top of it, and every hour spent building a second structure next to the first one was an hour of duplicated work, duplicated weight, and a second set of deck penetrations. The honest read was that we were building an arch because we'd already decided to build an arch.
Design three: a rack that clamps to the arch we own
So the current design is a bolt-on stainless rack that picks up on the existing arch — two clamps out in the bends up forward, two diagonal struts down to the legs aft — and carries the two panels flat, just below the traveller plate. The traveller stays entirely out of it. No bolts swapped, no new holes, no re-bedding, and the whole thing comes off with hand tools.
The canvas idea survives, just decoupled from the panels. Once the rack is up we'll add canvas to fill in whatever shade gaps are left over the cockpit. Shade is not a side benefit of this project — on a hot afternoon it's arguably the main one.
Design four: the panels come off before we sail
Rev F was a good rack. It was also 116 lb sitting about seven feet above the cockpit sole, and it was engineered to stay up there through a 70 mph gust — which meant thick-wall tube, which meant more weight, which is a loop that only runs one direction.
Then we did the sum we should have done first. This boat is going to Cabo. She's a 7,400 lb coastal cruiser with a comfort ratio of 17.6 — lively is the polite word — and every pound bolted on top of the arch comes straight out of the righting moment that keeps her the right way up when it turns rude offshore.
So Rev G inverts the requirement. The panels only have to be up there when the boat is standing still. They come off before we sail and go back on when the anchor is down. The frame stays, because bare round tube has almost no wind area — and because a frame that never has to survive a gale with 20 sq ft of glass bolted to it can be built from much lighter tube. Less weight, less structure, less money, and the heavy part of the array spends every passage in the quarter berth where it belongs.
Rev K adds the missing middle. A panel in the quarter berth makes no power, and on a four-day run that's four days of alternator. So one panel now has a third home on the coach roof — flat, low, latched down, and still plugged in.
Design five: stop reaching around the plate
Rev L isn't really a fifth design so much as a photograph that made four revisions of cleverness unnecessary. Looking up at the underside of the arch, there are four empty bolt holes already drilled through the traveller plate, 18 in in from each end, left over from a fitting nobody remembers. Every version of this rack had been contorting itself to pick up outboard of that plate, because clamps can't grip where a plate is welded on. It turns out we didn't need to grip anything. The front of the frame now bolts straight into holes that were there all along.
That's the argument for publishing an unfinished design, incidentally. Nobody would have gone looking for those holes if the drawing hadn't said, in writing, exactly which dimension it was still guessing at.
Why publish the whole design?
Because that's the point of open source boating. Every dimension, every part number, and every load calculation is on this page, so you can copy it, argue with it, or adapt it to whatever arch you've got. If you build one, tell us how it went — and subscribe on YouTube to watch us actually cut the tube.
The weight-aloft math
Every pound on the arch works on a long lever. The rack sits roughly seven feet above the cockpit sole — call it 9.3 ft above the waterline, and about 8.9 ft above the boat's centre of gravity. Weight times that lever is the moment working against 2,550 lb of ballast.
| Condition | What's aloft | Moment above CG | Rise in CG | Initial stability lost |
|---|---|---|---|---|
| Rev F, sailing | 116 lb at 8.9 ft | 1,032 ft·lb | 1.65 in | ~3.0% |
| Rev G, sailing | 55 lb frame at 8.9 ft + 48 lb of panel at 1.4 ft | 557 ft·lb | 0.89 in | ~1.7% |
| Rev G, at anchor | 103 lb at 8.9 ft | 917 ft·lb | 1.47 in | ~2.7% |
| Rev H, sailing with the radar pole | 55 lb frame at 8.9 ft + 39 lb of pole and radar at ~8.6 ft + 48 lb of panel at 1.4 ft | 892 ft·lb | 1.42 in | ~2.6% |
| Rev K, sailing one panel on the coach roof | 55 lb frame at 8.9 ft + 39 lb of pole and radar at ~8.6 ft + 34 lb of panel and cradle at ~4.4 ft + 24 lb of panel at 1.4 ft | 1,005 ft·lb | 1.60 in | ~3.0% |
| Rev M, sailing bolted pickup, 5/16 flat bar | 53 lb frame at 8.9 ft + 39 lb of pole and radar at ~8.6 ft + 34 lb of panel and cradle at ~4.4 ft + 24 lb of panel at 1.4 ft | 990 ft·lb | 1.58 in | ~2.9% |
| Rev N, sailing radar on the mast | 53 lb frame at 8.9 ft + ~21 lb of radar, mount and cable on the mast, ~364 ft·lb + 34 lb of panel and cradle at ~4.4 ft + 24 lb of panel at 1.4 ft | ~1,019 ft·lb | 1.63 in | ~3.0% |
| Rev F plus the same pole what we avoided | 116 lb at 8.9 ft + 39 lb of pole and radar | 1,367 ft·lb | 2.17 in | ~4.0% |
The radar is why this mattered. A HALO20 on an eight-foot pole was about 335 ft·lb that goes up and never comes down again — radar is a safety system, and it stays fitted. Rev G freed up 475 ft·lb; the radar spends about seven-tenths of it. Sailing with both, we were still 14% below where Rev F had us without any radar at all. Had we kept Rev F and added the pole, we'd be 32% worse than Rev F. In that sense Rev G paid for the radar.
Rev N moves the radar off the pole and up the mast, and the bill barely changes: about 21 lb at around 18 ft above the centre of gravity comes to roughly 364 ft·lb, against the pole's 335. Much less weight, much higher. The radar page has the arithmetic, and the honest caveat that the mount's weight is still a placeholder until it's on a scale.
Sailing, that's 46% less weight-moment aloft than Rev F — a bit over three-quarters of an inch less rise in the centre of gravity, and roughly 1.7% of initial stability given up instead of 3%. Three-quarters of that comes from the panels going below; the last quarter is the lighter tube that the easier load case pays for.
Rev K spends what's left, and it's worth saying out loud
Put a panel and its cradle on the coach roof and the sailing figure goes from 892 to 1,005 ft·lb — 114 more, because 34 lb sitting four and a half feet up is not nothing. That lands us 2.6% below Rev F, which is to say: level with where we started, two revisions ago, before any of this.
We're making that trade deliberately. What Rev F had for its 1,032 ft·lb was 48 lb of glass doing nothing at all while we sailed. What Rev K has for its 1,005 is a radar and 200 watts still working on passage. That's a better boat for the same stability budget.
But the budget is now spent. The next thing that wants to go up there — a wind generator, a second radome, a dinghy on davits — has to come out of something else, and we'll say so on this page when it does.
Rev L hands a little of it back, and Rev M gives some of that up again. Two rail clamps, two riser straps, two U-bolts and 16 in of C1 come off the arch — about 3 lb at the worst possible height. Then the plate measured 5/16 in, which pushed every flat bar on the page from 1/4 to 5/16 and put roughly a pound back. Net: sailing goes 1,005 → 990 ft·lb, 4.0% below Rev F rather than 2.6%. It isn't much. It is, however, the first change in three revisions that made the boat lighter aloft instead of heavier, and it came from deleting parts rather than adding them.
Rev N gives a little of that back, on purpose. Moving the radar up the mast adds roughly 29 ft·lb, so sailing goes 990 → about 1,019 ft·lb — still 1.3% below Rev F. What it buys is the radome about six and a half feet higher, which is roughly an extra nautical mile of radar range on every target. That is the kind of thing this budget is for.
At anchor the numbers go back up, and that's fine. At anchor you are not being knocked down by a breaking sea; you're swinging to a rode with the rig pointed into the wind. The condition that matters is the one where the boat is being driven hard, and in that condition the heavy panel is below and the light one is flat on the roof.
A number that doesn't measure what you'd think
The Hunter 290's Capsize Screening Formula is 2.21, and anything over 2.0 is the conventional flag for thinking twice about going offshore. It's tempting to say the rack makes that worse. It doesn't — and it can't.
CSF is just beam ÷ (displacement ÷ 64.2)1/3. Beam and displaced volume, nothing else. Add 116 lb anywhere on the boat and displacement goes up, so the number goes down: 2.209 becomes 2.198. By CSF, bolting lead to the masthead is an improvement. That tells you exactly how much the formula is worth for this particular question.
The concern is real; the metric just isn't the one that sees it. What sees it is righting moment — weight times height above the centre of gravity — which is the table above.
Where these numbers come from
The moments are solid: they're weight times a measured height, and the comparison between rows is reliable even if the absolute lever is off by a few inches. The percentages are softer. They assume the boat's centre of gravity sits about half a foot above the waterline and a metacentric height of 4.5 ft, which is typical for this beam but is an estimate, not a measurement. Nobody has inclined this boat. If you want the real figures for your own hull, that's the test to ask for.
Rev K adds one more estimate to the pile: the coach roof is taken as 4.5 ft above the waterline. That one is easy to check with a tape and we will.
Why a wider slot fixes the structure
This is the part of the design that surprised us. It came out of Rev F, and it survived into Rev G completely untouched — the geometry was right for reasons that have nothing to do with how the panels are fastened.
The arch will only let the frame clamp at about X = ±42 in from centreline, out past the traveller plate where the tubes start bending down. That leaves a long lever between the clamps and the fore-and-aft arms, and it was pushing the front crossbar down to a safety factor of 1.9. There are only two ways to shorten that lever: move the clamps inboard, which the arch forbids, or move the arms outboard — and the arms can only go where the panels are.
So the gap between the two panels turns out to be a structural control, not a styling choice. Widening it slides both panels outboard, the arms follow, and the lever shrinks. The load is identical; the geometry just gets better. And the same change hands you a genuinely usable window to look up through at the masthead.
Rev L stands this argument on its head — and 18 in survives anyway
Everything above is true for a rack held at X = ±42, out in the bends, with the arms inboard of the supports. Rev L moves the front pickup to X = ±21.5, which puts the arms outboard of the supports instead. C1's ends stop being a span and become a cantilever, and the sign of the trade flips: from here on, a wider slot costs margin rather than buying it.
We're leaving this section standing rather than quietly rewriting it, because the reasoning is the interesting part and it was right for the geometry it was written for. The recommendation doesn't change — 18 in is still the answer — but the reason it stops at 18 now is that 22 would drop C1 below this design's 3.5× floor. The numbers are recomputed below.
| Slot | Array width | Arms at | Worst crossbar moment | Margin, 0.083 wall | Crossbar proud of glass | Standing room |
|---|---|---|---|---|---|---|
| 6 in | 66.2 | ±30.0 | 1,081 in·lb | 1.9× | 8.9 in | ±3.2 in |
| 10 in | 70.2 | ±31.0 | 991 | 2.1× | 6.9 | ±5.3 |
| 14 in | 74.2 | ±32.0 | 900 | 2.3× | 4.9 | ±7.4 |
| 18 in | 78.2 | ±33.0 | 810 | 2.5× | 2.9 | ±9.6 |
| 22 in | 82.2 | ±34.0 | 720 | 2.8× | 0.9 | ±11.7 |
18 inches is where it stops paying
At 18 in the array is 78.2 in wide — just inside the 79 in traveller plate. Nothing sticks out past the arch, so there's no new snag point for sheets, dock lines, or anyone coming aboard. Go wider and the panels start overhanging the arch itself, which buys a little more margin and a lot more nuisance. That natural stopping point is why it's 18 and not 22.
| Rev F | Rev G | |
|---|---|---|
| Panels held by | 12 edge clamps, 24 bolts | 2 hooks + 2 pins per panel |
| Time to strike both panels | ~40 min with a wrench | ~10 min, no tools |
| Design case | 70 mph gust, sailing, panels on | 50 kt gust, at anchor, panels on |
| Design uplift | 391 lb | 264 lb |
| C1 front crossbar | 0.120 wall | 0.083 wall |
| C2, C3 crossbars | 0.120 wall | 0.065 wall |
| A1 arms | 0.120 wall | 0.083 wall |
| Frame weight | ~68 lb | ~55 lb |
| Weight aloft, sailing | 116 lb | 55 lb |
| Worst margin | 3.2× | 3.6× |
| Geometry — slot, arms, stack, clamps | — | unchanged, every dimension |
Rev G is mostly Rev F on a diet
Nothing moved. The 18 in slot, the arms at X = ±33, the clamps out in the bends, the 5/8 in build rule, all five drawings below — all of it carries straight over. What changed is the wall thickness of the tube and the way the panels attach to it. If you already built Rev F, you don't have to start again; you retrofit the quick-release kit and you get most of the benefit.
The front pickup — bolted, not clamped
Every revision from F through K reached around the traveller plate. No clamp can wrap anything inside X = ±39.5 in — the plate is welded along the tubes for that whole span — so two articulating rail clamps went out on bare tube past the bends, with riser straps climbing 6 to 7 in to reach frame height. It worked on paper. It was also the only joint in the whole design that held by friction rather than by a bolt, and the only one that could creep without ever breaking.
There were already holes in the plate
Four of them, found on a photograph taken from underneath the arch: two on each side of the traveller, 18 in in from the end of the plate, drilled clean through and empty. You can see sky through all four. Somebody put them there for a fitting that is no longer aboard.
The plate is 79 in wide, so they land at X = ±21.5 — deep inside the span no clamp could ever reach. That single fact deletes the rail clamps, the riser straps, their U-bolts, and the whole "measure where straight bare tube begins" problem that three revisions carried as an open item.
And the standing rule survives untouched: the traveller stays entirely out of it. These holes are its neighbours, not its fasteners. No traveller bolt is borrowed, swapped, loosened or re-bedded, and nothing we do shares a load path with the mainsheet.
Why a hole you didn't drill is better than a clamp you bought
- It's a positive connection. A rail clamp on polished 1-1/4 in tube holds by squeeze. Bolts in reamed holes hold by bearing. At these loads neither is close to failing, but only one of them can quietly walk down a leg over a season.
- No new holes anywhere. The arch is not drilled, the plate is not drilled, the traveller is not disturbed. We are using what's there.
- The riser problem disappears. No cut-to-fit flat bar, no dependence on where the bend ends, no "what's the tube OD at that station", no risk of a clamp rocking on a curve.
- It's reversible. Four bolts out and the boat is exactly as it was, which is more than could be said for anything clamped hard to a polished rail for five years.
- C1 gets shorter. With nothing to reach out to, the front crossbar now stops flush with the plate at ±39.5 instead of running to ±42. It finishes 0.4 in proud of the glass instead of 2.9, which retires the last of the snag-point worry.
The plate measured 5/16 — and that decides the whole stack
The build rule has always been that C1's top surface sits 5/8 in below the top face of the traveller plate. Rev K achieved that by holding tube in the air and cutting risers to fit. Rev L got it from a tab plus a shim. Rev M gets it from two numbers that happen to be the same.
| Layer | Thickness | Running depth below P |
|---|---|---|
| Traveller plate measured | 0.3125 | 0.3125 |
| F1 tab, 5/16 bar | 0.3125 | 0.6250 |
| Top of C1 — the build rule | — | 0.625 |
No shim, no cutting, no judgement
A 5/16 in plate and a 5/16 in tab come to 5/8 exactly. The packer shim Rev L needed is deleted. Bolt the tab to the underside of the plate, lay C1 against the tab, and the build rule is satisfied by the thickness of two pieces of metal. Check it with a straightedge if you like, but there is nothing left to adjust.
Keep the arithmetic in the build sheet anyway, because it's the check: plate + tab = 0.625. If your plate isn't 5/16, that equation tells you what bar to buy, or what shim to add.
The knock-on: every flat bar on this page goes to 5/16
A 5/16 in tab hanging under a 5/16 in plate puts its top face at P − 0.3125. The panel's underside has to sit at or below that, or it fouls the tab — and the panel's underside is set by the thickness of the hook plates it rests on. So P1, P2 and P3 go from 1/4 in bar to 5/16 in bar.
That is the whole consequence, and it's a good one: one stock size for every flat part on the boat, and the panel now rests on the tab's top face in the same plane as its own hooks. The glass finishes at P + 0.87 instead of P + 0.81, which still leaves 1.13 in under the 2 in ceiling. It costs about a pound of stainless.
What the tab actually is
A flat bar, and not much of one. It lies against the underside of the plate, picks up the existing holes, runs aft past the plate's edge, and finishes by wrapping C1.
- F1 — 316 flat bar, 5/16 × 2 in, about 7 in long. Two holes drilled to match the plate's own pattern, 1-3/4 in apart fore-and-aft. The aft end is formed over C1's crown and down its aft face, ending in a drilled ear beneath the tube.
- F1b — a short keeper strap wrapping C1's forward-lower quadrant, with a matching ear.
- F3 — one 1/4-20 cross bolt below the tube, closing the two ears around C1. Nothing stands proud above the tab's top face, which is the constraint that decided this detail: the panel sits in that plane.
At 5/16 × 2 in bar and a 3 in lever from the aft bolt to C1's centreline, the tab runs at 5.4× against 316 yield. It's the same forming operation as the P1 hook plates — a 1.30 in throat around 1-1/4 in tube — so the fabricator is already set up for it.
The hole pattern, and the one thing that nearly didn't fit
Four holes a side, arranged as two pairs 1-3/4 in apart fore-and-aft, straddling the traveller rail with about 1/8 in of clear plate either side of it. The holes measure somewhere between 1/4 and 5/16 — which is almost certainly 9/32, the standard clearance drill for a 1/4 in fastener. Whatever used to live there was bolted down with quarter-inch hardware.
A measurement that checks itself
Two rows 1-3/4 in apart, each hole 9/32 wide sitting 1/8 in clear of the rail, works back to a traveller rail base of 1.75 − 0.281 − 0.250 = 1.22 in — which is exactly what a 1-1/4 in track base is. Three independently measured numbers landing on a fourth that nobody measured is the best evidence you get that a tape measure was read correctly.
The screws have to be flush, because the car nearly touches the plate
This is the constraint that sizes the fastener, not the load. A button-head 1/4-20 stands about 0.15 in proud and the traveller car will find it. The answer is a countersunk 82° flat head, seated dead flush — and then the question becomes how big a countersink will fit in the 1/8 in of plate between the hole and the rail.
| Flat head | Countersink dia | Depth cut | Plate left | Grows past the hole by | Clear to the rail |
|---|---|---|---|---|---|
| #10-24 | 0.385 | 0.060 | 0.253 | 0.052 | 0.125 — fits easily |
| M6 | 0.472 | 0.110 | 0.203 | 0.095 | 0.125 — 30 thou spare |
| 1/4-20 | 0.500 | 0.126 | 0.187 | 0.109 | 0.125 — 16 thou spare |
Go with #10-24 A4-70 flat heads. They carry 1,142 lb of proof load against the 104 lb this joint asks of them — eleven times over — while cutting a countersink only 0.060 in deep and growing barely fifty thou past the existing hole. The 1/4-20 would work too, on paper, with sixteen thousandths to spare against the rail. That is not a number to design a boat around.
Two useful side effects. A flat head self-centres in its cone, so the slop between a #10 screw and a 9/32 hole vanishes and the joint ends up with no play at all — the cone also carries any fore-and-aft shear. And a 0.060 in countersink takes almost nothing out of a plate that already has a full-depth hole through it.
First, though: look at the top of the plate. If whatever used to be mounted there was held by flat heads, the countersinks are already cut and this is a shopping trip, not a machining job.
Check the 5/16 within each pair — as measured, the two holes very nearly merge
Two 9/32 in holes on 5/16 in centres leave one thirty-second of an inch of metal between them. That isn't two holes, it's a ragged slot, and two #10 countersinks on those centres would certainly break into each other.
Most likely the 5/16 is the gap between the holes rather than the centres, which puts the centres at about 19/32 and makes the whole thing ordinary. Worth one minute with a caliper to settle.
Either way the design doesn't move. If the pairs really are that close, use one hole from each pair — two screws per tab, 1-3/4 in apart, which is the couple that actually resists the cantilever. Two #10s is still eleven times what this joint needs. The spare holes get a dab of butyl and a rest.
Which way the load runs now — and the one thing it costs
This is the part to be straight about, because it isn't all free. With the pickup at ±42 the arms at ±33 sat inboard of their supports, and C1's ends were a 9 in span. With the pickup at ±21.5 the arms sit outboard, and C1's ends become an 11.5 in cantilever.
Same load, longer lever: C1's worst bending moment goes from 547 to 699 in·lb, and its margin against 316 yield from 4.6× to 3.6×. That is still above this design's 3.5× floor, it is exactly level with the arms — which have governed since the first sketch — and it needs no change of wall thickness anywhere. The worst margin on the page is still 3.6×, and it's still an arm.
| Slot | Arms at | Cantilever | C1 moment | C1 stress | Margin, 0.083 wall | Was, at ±42 |
|---|---|---|---|---|---|---|
| 6 in | ±30.0 | 8.5 in | 517 in·lb | 6,200 psi | 4.8× | 1.9× |
| 10 in | ±31.0 | 9.5 | 577 | 6,930 | 4.3× | 2.1× |
| 14 in | ±32.0 | 10.5 | 638 | 7,660 | 3.9× | 2.3× |
| 18 in | ±33.0 | 11.5 | 699 | 8,390 | 3.6× | 2.5× |
| 22 in | ±34.0 | 12.5 | 760 | 9,120 | 3.3× | 2.8× |
Read the two slot tables side by side and the columns run in opposite directions, which is the honest signature of a design change that moved a support past a load. Eighteen inches is still the answer in both. In Rev G it was the widest slot that kept the array inside the plate; in Rev L it is also the widest slot that keeps C1 at or above 3.5×. Two independent reasons arriving at the same number is about as much comfort as this kind of thing ever offers.
Keep the hooks at ±33. Do not drift outboard.
The cantilever is measured from the tab to wherever the panel's load actually enters C1, so the hook plate positions are now structural, not cosmetic.
- Hooks at ±33 — 3.6×, as designed.
- Hooks at ±36 — 881 in·lb, 2.8×.
- Hooks at ±39 — 1,064 in·lb, 2.4×.
Set the hook plates at the arm stations and nowhere else. If a panel's own frame holes force them outboard of ±34, step C1 to 1-1/2 in OD × 0.083 rather than accepting the margin.
Does the arch mind?
Fair question, because the front reaction now enters the top rail 18 in inboard of the bends instead of right at them. Two reasons not to lose sleep, and one thing to check.
First, the plate is welded along both tubes for the whole ±39.5 in span. The load doesn't travel through the plate to the tubes; it goes into the composite of plate-plus-two-tubes, which is a far deeper section than either on its own.
Second, and more persuasively: this rail already carries the mainsheet. A hard-sheeted main on a 29-footer puts several hundred pounds into the traveller car, right in the middle of the span, on every beat. Our two tabs add about 61 lb each, further outboard, in the same direction. If the plate can take the mainsheet — and it has, for twenty-odd years — it is not going to notice the solar rack.
The thing to check is local, not global: get under there and look at the plate-to-tube welds either side of each hole group, and at the holes themselves for any sign that a previous fitting tore at them. Clean stainless, no crazing at the weld toe, sound edges. If a weld looks tired, that's a fabricator's job before anything bolts to it.
One number left
Everything that sizes this joint is now measured: the holes are 18 in in from each end of a 79 in plate, so X = ±21.5; the plate is 5/16 in; the holes are 9/32 in two rows 1-3/4 in apart straddling the rail. The last one is the distance from the aft row to the plate's aft edge, because that plus the 2 in to C1 is the tab's lever.
We have assumed 3 in, which puts the 5/16 × 2 in tab at 5.4×. If it measures longer than about 3.5 in, widen the tab to 2-1/2 in rather than thickening it — the thickness is spoken for by the build rule and the width is free, as long as it still passes between the traveller's nuts.
And while you're under there with a tape: edge distance and weld condition. Two hole diameters of sound plate to any edge, and a good look at the plate-to-tube welds either side of each group.
The vertical stack
Datum P is the top face of the traveller plate. Everything on the boat gets measured from there, because it's the one surface you can lay a straightedge on. The boom sets a hard ceiling 2 in above it; the top of the panel glass finishes at P + 0.87, so there's 1.13 in in hand.
| Level | Height | What sets it |
|---|---|---|
| — ceiling — | P + 2.00 | The limit. Nothing comes near it. |
| Top of panel glass | P + 0.87 | 1.13 in of margin in hand |
| Underside of panel | P − 0.31 | Panel is 1.18 in thick |
| Top of hook plate / clevis pad | P − 0.31 | 5/16 in 316 — sized to match the F1 tab's top face, so the panel rests in one plane across both |
| Top of the F1 tab | P − 0.31 | The plate's own underside. Measured: the plate is 5/16. |
| Top of crossbars C1–C3 | P − 0.63 | The build rule. It falls out of the stack: 5/16 plate + 5/16 tab. Nothing to shim and nothing to measure down. |
| Top of the traveller plate | P | Datum |
| Crossbar centreline | P − 1.25 | — |
| Top of fore-and-aft arms | P − 1.88 | Arms hang under the crossbars in cross clamps |
| Arm centreline | P − 2.50 | Clears beneath both arch tubes |
The frame
- F1 tabs bolted into the traveller plate's existing holes at X = ±21.5, wrapping C1 from above. New in Rev L — they replace the two rail clamps and their risers.
- C1 at Y = +2.0, running out to ±39.5 — flush with the ends of the plate.
- C2 and C3 ride on top of the arms at 26 in and 48 in aft.
- A1 arms at X = ±33, hanging under all three crossbars — under the panels' outer thirds, and 11.5 in outboard of the tabs.
- S1 struts from the arms at C3, running down, forward and outboard to the arch legs, well below the panel plane.
- R3 brace clamp — deleted in Rev N. Rev H tied the port arm's aft end to a radar pole on the aft deck. The radar has moved up the mast, so there's no pole and nothing to tie to.
The traveller control line — a hard clearance
The line down the arch corner is traveller control and must stay completely clear. Rev L halves this problem — the two H1 clamps that used to land in exactly that region of the leg are gone, so only the strut feet are left down there.
- Clock the S1 foot clamps so their bodies and bolt heads sit on the face of the leg the line does not run on.
- Check with the car walked fully to each end before cutting anything — the lead moves as the car travels.
- Run the panel cable down the opposite leg where you can, never in a shared clamp.
The sight slot, and what it does now
Rev F fought hard for this slot, because a solid ceiling over the cockpit means you can't see your own rig. Rev G quietly wins that fight outright: when we're sailing there are no panels up there at all, so the view is completely clear. Better than any slot.
The 18 in gap still earns its keep, though, and for four reasons that have nothing to do with helming:
- It's still a structural control — but it now pulls the other way. The slot sets where the arms sit, and since Rev L moved the pickup inboard the arms cantilever outboard of it. Eighteen inches is the widest slot that keeps C1 at 3.6×; go wider and the margin falls rather than rises.
- It's how you reach the inboard pins. Two of the four quick-release pins sit on the inboard corners. Through an 18 in slot you can reach them standing on the cockpit seat. Through a 6 in slot you couldn't.
- Air under the cells. A panel that can breathe on both faces runs cooler and makes more power. The slot plus the 1 in leading gap keeps air moving across the underside.
- At anchor you still look up. Wind vane, anchor light, halyards slatting at 0200 — the slot is how you check the rig without leaving the cockpit.
Everything below is the Rev F sight study, kept because it's still the at-anchor case and because it's the number that set the slot width in the first place.
Tube — sizes, walls and cut lengths
All 316/316L, polished ornamental or seamless. Rev G runs two wall thicknesses: 0.083 where bending governs — the front crossbar and the arms — and 0.065 everywhere else. Y is measured aft from the traveller plate's aft edge, X from centreline.
| Mark | Qty | Member | Tube | Cut length | Position |
|---|---|---|---|---|---|
| C1 | 1 | Front crossbar | 1-1/4 in OD × 0.083 wall | 80 in cut long | Y = +2.0. Trim to 79.0 — ±39.5, flush with the ends of the traveller plate. Carries the panel hooks. |
| C2 | 1 | Mid crossbar | 1-1/4 in OD × 0.065 wall | 81.0 in | Y = +26.0. Rest pads only — nothing fastens here. |
| C3 | 1 | Aft crossbar | 1-1/4 in OD × 0.065 wall | 81.0 in | Y = +48.0; strut nodes and the four pin clevises |
| A1 | 2 | Fore-and-aft arm | 1-1/4 in OD × 0.083 wall | 52.0 in | X = ±33, under C1–C3, Y = +1.2 to +53.2. Still the governing member. |
| S1 | 2 | Diagonal strut | 1-1/4 in OD × 0.065 wall | 60.0 in cut long | Arm at C3 to the arch leg, ~30 in below the plate |
| CR1 | 2 | Coach roof rail | 1-1/4 in OD × 0.065 wall | 34.0 in | Runs fore-and-aft on the coach roof, on its own through-bolted feet. The two set 46.0 in apart athwartships — exactly C1 to C3, so the panel's own hooks and pins fit. Detailed below. |
| Stock | Qty | Yields | Offcut |
|---|---|---|---|
| 1-1/4 × 0.083 × 12 ft | 2 | C1 (80) from one; both arms (52 + 52) from the second | 64 / 40 in |
| 1-1/4 × 0.065 × 12 ft | 3 | One crossbar plus one strut from each of two (81 + 60); both coach roof rails (34 + 34) from the third | 3 / 3 / 76 in |
| 316 flat bar, 5/16 × 2 × 48 in | 2 | Two F1 pickup tabs and their keepers, four hook plates, four pin tabs, six clevis brackets. One thickness for every flat part on the boat — set by the traveller plate measuring 5/16. | trim to suit |
| 1/4 in aluminium or G10 plate | 1 | Four 5 × 5 in backing plates for the coach roof rail feet | trim to suit |
Three substitutions
- If you want C2 and C3 stiffer: take them in 0.083 instead. Costs 3 lb, lifts them from 3.7× to 4.5×, and roughly triples what you can hang off the middle of C3 without thinking about it. Worth it if an antenna, a light bar or a dinghy-motor hanger is anywhere in your future.
- If the hook plates have to sit outboard of X = ±34: C1's cantilever grows enough to matter. Step C1 up to 1-1/2 in OD × 0.083 and use 1-1/2 × 1-1/4 cross clamps where the arms meet it. The numbers are in the front pickup section.
- If you decide to leave the panels up all season: then you're building Rev F, not Rev G. Go back to 0.120 wall throughout. Don't run Rev G's tube on a Rev F duty cycle.
Never substitute 0.049 wall — that's bimini tube, and it will crush under a clamp.
Mounting hardware
Everything wetted is 316 / A4-70. The traveller stays entirely outside this build — no bolts swapped, no new holes drilled, no re-bedding, and in Rev L nothing clamped to the arch legs at the front either. The frame fittings below are permanent; the quick-release kit in the next table is the part you actually handle.
| # | Qty | Item | Spec |
|---|---|---|---|
| F1 | 2 | Front pickup tab | 316 flat bar 5/16 × 2 in, ~7 in long. Two holes at 1-3/4 in centres to match the plate; aft end formed over C1's crown and down its aft face to a drilled ear below the tube. Replaces the H1 clamp and riser of every earlier revision. |
| F1b | 2 | Keeper strap | 316 flat bar 5/16 × 2 × ~3 in, formed to C1's forward-lower quadrant with a matching ear. Closes the wrap. |
| F2 | 0 | Packer shim — deleted | Not required. The plate measured 5/16 and the tab is 5/16, so plate + tab = 0.625 closes the build rule on its own. Keep the equation on the build sheet as the check. |
| F3 | 2 | Cross bolt | A4-70, 1/4-20 × 1-1/2 in with nyloc, through both ears under C1. Nothing stands proud above the tab. |
| F5 | 4 | Plate screws | #10-24 A4-70, 82° countersunk flat head, into the plate's existing 9/32 holes — one per hole pair, two per tab. Fender washer and nyloc below. Heads must finish flush: the traveller car passes within very little of the plate. Countersink 0.060 in deep, 0.385 dia. Butyl in every hole; Tef-Gel the threads. |
| H2 | 6 | 90° cross clamp | 316, 1-1/4 × 1-1/4 in — each arm to each of the three crossbars |
| H3 | 2 | Swivel tube fitting | 316, 1-1/4 in, strut head at the arm/C3 node. Two-axis swivel. |
| H4 | 2 | Swivel rail clamp | 316, 1-1/4 in × [leg OD], strut foot ~30 in below the plate. Clock it off the traveller control line. |
| H5 | 4 | U-bolt, 1-1/4 in | 316 with saddle — one per clevis bracket, onto C3 |
| H8 | 2 | Tube end caps, 1-1/4 in | 316 — C1's ends, now only 0.4 in proud of the glass |
| R3 | 0 | Radar pole brace clamp — deleted | Not required from Rev N. The radar is on the mast; there is no pole to brace. |
The quick-release kit
This is the whole idea of Rev G, and it is four moving parts. Each panel hooks at its forward edge and pins at its aft edge: two J-shaped hook plates wrap under the front crossbar, and two quick-release pins drop through clevises on the aft crossbar. The middle crossbar carries nothing but a pair of rest pads. Two pins per panel, four for the whole array, and not a tool among them.
The forward hooks take the uplift because they wrap under C1 — a panel trying to fly can't get past them. To release, you lift the aft edge and slide the panel aft about an inch, which is a motion the wind can't do for you by accident.
| # | Qty | Item | Spec / note |
|---|---|---|---|
| P1 | 4 | Hook plate | 316 flat bar 5/16 × 2 in, formed into a shallow J with a 1.30 in throat (slip fit over 1-1/4 in tube plus EPDM). Bolts to the panel's forward frame flange through the existing mounting holes. Two per panel, at the arm stations. |
| P2 | 4 | Pin tab | 316 flat bar 5/16 × 2 × 3 in, drilled 3/8 in. Bolts to the panel's side flange at the C3 station. |
| P3 | 4 | Clevis bracket | 316 flat bar 5/16 × 2 in folded into a U with 1-1/2 in ears, drilled 3/8 in. U-bolted to C3. The ears also locate the panel athwartships. |
| P4 | 4 | Quick-release pin | 3/8 in × 2 in 316 ball-detent pip pin, each on its own lanyard to the frame. A pin you can drop overboard is a pin you will drop overboard. |
| P5 | 4 | Rest pad | 1/8 in EPDM or white nylon, bonded to C2 where each panel crosses it. No fastening — it just stops the middle of the panel drumming. |
| P6 | 2 | Panel tether | 3 ft of 3/16 in Dyneema with a snap shackle, panel to frame. Clipped on before the pins come out, every single time. This is the part that stops a $240 panel becoming a reef ornament. |
| P7 | 2 | Padded stow sleeve | Sunbrella over 1/4 in closed-cell foam, carry handle, flap closure. The canvas idea from Design One, finally given a job it's actually good at. |
| P8 | 16 | Panel frame bolts | A4 316, M6 or 1/4-20 to suit the panel's existing frame holes, with fender washers both sides. Four per hook plate and pin tab pair. |
| # | Qty | Item | Spec / note |
|---|---|---|---|
| H9 | 20 | Hex bolt 1/4-20 × 1-3/4 in | A4-70 316 — frame fittings. The panel bolts are P8. |
| H10 | 4–8 | Plate bolts — see F5 | A4-70 316, diameter to suit the plate's existing holes. Button head up, fender washer and nyloc below, butyl in the hole. |
| H11 | 28 / 8 | Nyloc nuts 1/4-20 / 5/16-18 | A4 316 — nylocs only, no split washers |
| H12 | 56 / 16 | Fender washers 1/4 / 5/16 | A4 316 |
| H13 | 4 ft | EPDM strip, 1/16 × 1 in | Under every clamp touching the arch legs. Not under the F1 tabs — that joint is a metal-to-metal bolted stack and rubber in it would only relax. |
| H14 | 1 tube | Tef-Gel or Lanocote | Mandatory on every 316-on-316 thread. Stainless galls. |
| H15 | 1 | Loctite 243 | On any set screw a fitting still has |
| # | Qty | Item | Spec / note |
|---|---|---|---|
| E1 | 60 ft | 10 AWG tinned duplex, marine | Two runs now, one per panel. Under 2% drop at 6.4 A over a 25 ft run. |
| E2 | 2 set | MC4 connectors / short extension | One pair per panel. No series link in Rev K — each panel runs to its own controller. |
| E3 | 2 | MPPT controller | Victron SmartSolar MPPT 100/20 — one per panel. 100 V PV ceiling against a 36.5 V panel, 20 A output against a 14.5 A peak. Bluetooth and VE.Direct on both. |
| E4 | 2 | 2-pole DC breaker, 15 A | One per controller — each panel gets its own PV disconnect |
| E5 | 1 | Deck cable gland + butyl | The only bedding job in this build |
| E6 | 10 | 316 cushioned cable clamps | Down an arch leg, never sharing with the traveller line |
| E7 | 2 pr | MC4 service loop at each panel | 18 in of slack so you can unplug a panel without straining the run or the gland. This is what makes the panel a removable component instead of a wired-in one. |
| E8 | 4 | MC4 dust caps | On the parked leads whenever the panels are off. Salt in an open MC4 is a resistance problem you'll chase for a season. |
| E9 | 1 | Lead pocket on the arch leg | Sunbrella pouch, so the disconnected leads aren't swinging around the cockpit |
| E10 | 1 pr | MC4 extension, 20 ft | Lives rolled up with the coach roof cradle; reaches from the roof panel back to the deck gland leads. |
| E11 | 2 | VE.Direct to USB cable | One per controller, into the Raspberry Pi. This is what turns "we'll publish the measured numbers" into something we can actually do. |
What it has to survive
Rev F was sized for a 70 mph gust with the panels bolted on and the boat being sailed. Rev G's panels are only up there at anchor or alongside, so the design case becomes a different question: what's the worst squall that can arrive faster than we can get them off?
We set that at 50 knots — 57.5 mph, dynamic pressure 8.5 psf. A flat panel with airflow on both faces carries a net coefficient near ±1.5, so 12.7 psf across 20.8 sq ft: 264 lb of uplift instead of 391. That 32% reduction is exactly what pays for the thinner wall.
The removal plan is part of the structure. The rule is that panels come off at a 35 kt forecast, which leaves the 50 kt case for the squall nobody saw coming. If you won't keep that rule, don't build Rev G's tube — build Rev F's.
| Check | Wall | Demand | Stress | Margin vs yield | Rev F was |
|---|---|---|---|---|---|
| C1, bending over the 11.5 in tab-to-arm cantilever | 0.083 | 699 in·lb | 8,390 psi | 3.6× | 4.1× |
| C2 / C3 crossbar, bending at centre | 0.065 | 547 in·lb | 8,030 psi | 3.7× | 4.1× |
| A1 arm, bending under the C2 reaction | 0.083 | 697 in·lb | 8,370 psi | 3.6× | 3.2× |
| A1 arm, midspan deflection | 0.083 | 0.09 in | — | negligible | 0.10 in |
| F1 tab, bending at the aft screw (3 in lever) | 5/16 × 2 bar | 182 in·lb | 5,600 psi | 5.4× | 3.5× |
| S1 strut, axial | 0.065 | 141 lb | buckling at 3,300 lb | 23× | 16× |
| P1 hook plate, bending at the panel flange | 1/4 bar | 33 in·lb | 1,580 psi | 19× | — |
| P4 pip pin, double shear | 3/8 in | 33 lb | 150 psi | >100× | — |
| F5 screw, tension in the forward row | #10-24 | 104 lb | 1,142 lb proof | 11× | governed |
| Traveller plate under the countersink | 0.253 left of 0.3125 | — | 0.060 in cut | 81% of section kept | — |
Every member still sits at or above 3.5× despite losing a third of its wall, and the arm — the weakest link since the first sketch — still governs at 3.6×, with C1 now exactly level with it. The quick-release hardware isn't close to being a limit: each of the eight attachment points carries about 33 lb, which a 3/8 in pin finds funny.
Rev L retired the sentence this page carried for four revisions: the tube isn't the limit, the two front clamps are. There are no front clamps. The front pickup is now four countersunk screws working in tension at eleven times their demand, and the thing to watch is no longer a friction joint creeping down a rail — it is simply keeping the hook plates where the drawing puts them.
And the bare frame, at 70 mph?
Fine, which is the whole point. With the panels off, the frame shows about 4.1 sq ft of round tube to the wind — roughly 61 lb of drag at 70 mph, and essentially no uplift, because a cylinder doesn't generate lift the way a flat plate does. The bare frame isn't the governing case for anything on this page. It just goes along for the ride, which is exactly what we wanted it to do.
It is not a handhold, and it is not a step
Worth saying plainly, because it's the mistake everyone makes with anything that looks like a radar arch. A 250 lb person landing in the middle of C3 puts about 4,100 in·lb into a 66 in span. That is past yield for 0.065 wall — and past yield for Rev F's 0.120 as well. Neither version of this rack was ever something you could hang off.
Safe working point load at mid-span is about 50 lb with a 2× margin, and a good deal more if you put it near an arm at X = ±33. An antenna, a solar light, a horseshoe buoy: fine. A person coming aboard from a dinghy: no. Grab the arch itself — that's what the arch is for.
Almost all of it still lands on the arch
The front pickup puts its reaction into the traveller plate, which is welded along both tubes; the struts put the aft reaction into the legs below. In Rev F through Rev M, one hundred percent of that 264 lb ends up at the arch's two deck bases. Moving the pickup inboard changed where the load enters the arch, not where it leaves. Get under the deck and look at those bolts and backing plates before you build — that hasn't changed and it never will.
Rev H briefly moved a piece of it: a radar pole on the aft deck clamped to the rack's port aft corner and gave that corner's share — roughly 66 lb of the 264 — a direct path to the hull. Rev N takes that back. The radar is on the mast now, the pole is gone, and the arch's two deck bases carry all 264 lb again — which is the case the rack was always sized for.
The rack was always sized as if the pole weren't there — and now it isn't
Every margin in the table above was computed with the original four support points and no help from the radar pole. We said at the time that the pole was upside, not an assumption, because a rack that only stands up while a radar pole is bolted to it fails the first time the pole comes off. Two revisions later the pole came off for good. Nothing on this page had to change except deleting one clamp — which is exactly why it was worth refusing the help.
Three honest costs
- Shading. The traveller track and car stand above the panel plane, so with the sun low and forward you'll get a shadow band across the leading edge. Two things soften it: the ShadowFlux cells are the anti-shading type, and each panel now has its own MPPT, so a shaded panel no longer drags its neighbour down with it. A real trade for the boom clearance, and a cheaper one than it was.
- Half the array is asleep while you sail. This was the real price of Rev G, and Rev K only half-fixes it. One panel goes flat on the coach roof and keeps working; the other is in the quarter berth making nothing. Call it 200 W of nominal capacity offline on every passage, and more than that once the boom starts shading what's left.
- A ritual you have to actually perform. A removable panel you never remove is just a heavy panel in a worse mount — and Rev K adds a second ritual on top of the first. Ten minutes to strike the array, another ten to cradle one panel on the roof. If we find the discipline slipping, the honest fix is to go back to Rev F's wall thickness, not to pretend.
Wiring the two panels
Earlier revisions of this page argued about series versus parallel. Rev K throws that argument away: each panel gets its own charge controller. Two Victron SmartSolar MPPT 100/20, one per panel, instead of one bigger unit with the two wired into a string.
Both numbers sit comfortably inside the box. A single ShadowFlux panel is Voc 36.5 V — even on a freezing morning, with the cold-temperature rise, that's less than half the controller's 100 V ceiling. And 200 W delivered into a 12 V bank at charging voltage is about 14.5 A, inside the 20 A rating with headroom for a cold bright day when a panel briefly beats its nameplate.
Why two controllers beat one bigger one
- Independent MPPT on each panel, which is the whole argument. In a series string, the shaded panel drags the unshaded one down to its operating point. On this boat the panels get shaded differently all day — the traveller and car throw a band across the arch array, the boom and mast work over the coach roof panel, and at anchor the rig sweeps across as she swings. Two trackers means each panel finds its own maximum and the other one never knows.
- It makes the stowed panel a non-event. With one panel on the coach roof and the other in the quarter berth, there is no string to unbalance — you simply switch off the controller belonging to the panel that's below. No bypass diodes conducting, no half-voltage string, nothing to think about.
- Redundancy. A dead controller costs you half the array instead of all of it. On a boat whose whole electrical plan rests on solar, that's worth something for what it costs.
Both are Bluetooth, so VictronConnect gives you live figures on a phone, and both carry a VE.Direct port — which means the Raspberry Pi can log actual production rather than this page continuing to quote rules of thumb. That's how the measured numbers promised further down this page are going to get measured.
Wiring is now two separate runs rather than one: each panel to its own controller, each on its own two-pole breaker. Current per run is only 6.4 A, so 10 AWG is generous — it's sized for volt drop over the length, not for heat. Run both down the arch leg in cushioned clamps, the opposite leg from the traveller control line, through a bedded gland into the aft locker. The B&R rig helps: no backstay to lay a shadow down the middle of the array.
Never unplug MC4 connectors under load
This matters far more in Rev G than it did in Rev F, because now you're doing it twice a trip instead of once a season. A panel in sun is a current source that does not care that you've stopped wanting it: pull a live MC4 and you draw an arc that pits the contacts, and a pitted contact is a hot joint for the rest of its life.
Breaker off first, every time. Each panel has its own two-pole breaker at its own controller — that's what they're there for, and with one controller per panel you only ever have to kill the one you're working on. Off, then unplug, then cap. Reverse it going back on. If you want belt and braces, throw a towel over the panel first.
The ten-minute drill
A removable panel is only as good as the routine around it. Here's the whole thing, in the order it actually happens.
Measured: the panels do fit below
This was the one that could have sunk the whole idea, and it has nothing to do with the rack. The panels are 49.69 × 30.08 × 1.18 in and the companionway is nowhere near 30 in wide, so getting them to the quarter berth depends on a thin rigid sheet going through a narrower opening on the diagonal, the way plywood goes through a doorway.
It works — checked with a cardboard cut-out before anything was ordered. If you're copying this design, do the same test on your own boat first; it costs a sheet of cardboard and it settles the question in ten minutes.
Had it failed, the fallback was the coach roof cradle for both panels rather than one, keeping about three-quarters of the stability benefit and losing only the ability to get them right out of the weather.
Coming off — before you sail
- Breaker off at the controller. Not negotiable — see the warning above.
- Unplug the two MC4s at each panel, cap them, and park the leads in the pocket on the arch leg.
- Clip the tether from the panel to the frame. Before the pins. Every time, including the time you're in a hurry, especially the time you're in a hurry.
- Pull the two pip pins at C3. They stay on their lanyards; you never hold a loose pin over water.
- Lift the aft edge about two inches, slide the panel aft roughly an inch to free the forward hooks, and lift it clear. One person can do this from the cockpit seat; two is nicer.
- Straight into its padded sleeve — don't set a bare panel down on a winch.
- Below, into the quarter berth, glass to glass with the two sleeves facing each other so nothing rubs a cell.
- Repeat for the second panel. Ten minutes for the pair once you've done it twice.
Going back on — when the hook is down
- Panels up, sleeves off, tether clipped first again.
- Hook the forward edge over C1 and slide the panel forward until the J-hooks seat. It will hang there on its own while you walk aft — that's the point of hooking first.
- Drop the aft edge so the pin tabs land between the clevis ears, and drop both pins.
- Check the 18 in slot and the 1 in leading gap with a rule. If either is off, a hook has not seated.
- Plug in, uncap, breaker on. Watch the controller pick up before you go and open a beer.
Two things we'll learn the hard way, and report back
Whether ten minutes is honest, or whether it's ten minutes on a flat day at the dock and twenty-five at anchor in a chop. And whether the tether is long enough to be useful but short enough that a dropped panel doesn't swing into the pushpit. Both go in the video.
The third place a panel can live
Rev G gave the panels two homes: on the arch when the boat is standing still, in the quarter berth when it isn't. That left an honest hole in the middle. A panel in the quarter berth makes no power, and on a four-day run down the Baja coast that's four days of charging the bank off a 55 amp alternator that realistically delivers twenty-five.
Rev K fills it. One panel gets a third home — latched flat to the coach roof, about four and a half feet above the water instead of nine and a half, out from under the boom's swing, and still plugged in.
| State | Where the panels are | When | Weight aloft | Nominal output |
|---|---|---|---|---|
| 1 — Array up | Both on the arch rack | At anchor, alongside | 103 lb at 8.9 ft | 400 W |
| 2 — Passage | One flat on the coach roof, athwartships just forward of the companionway; one in the quarter berth | On passage, reasonable conditions | 55 lb at 8.9 ft + 34 lb at 4.4 ft | ~200 W |
| 3 — Stowed | Both in the quarter berth | Heavy weather, laid up | 55 lb at 8.9 ft | nothing |
The neat part is that the panel needs no new hardware at all to do this. The cradle is two rails of the same 1-1/4 in tube as the rack, set 46.0 in apart — exactly the C1-to-C3 spacing on the arch. So the panel's own P1 hook plates and P2 pin tabs land on the coach roof cradle the same way they land on the rack.
There's one difference, and it's a good one. The panel lies athwartships up there — long axis across the boat, on the molded non-skid just forward of the companionway — so the two rails run fore-and-aft. Which means that when the panel isn't on them, they are a pair of grab rails exactly where you want grab rails. One rail takes the hooks, the other takes the pins; you slide the panel outboard an inch to release it instead of aft.
We were wrong about the boom
Earlier drafts of this page called the boom the governing constraint and said the whole idea stood or fell on measuring it. It's been measured, and the boom is 3.5 feet above the coach roof. It is not a constraint at all. The in-mast furler helps too — there's no flaked mainsail hanging below the spar, because the sail lives inside the mast.
Publishing a wrong call and then correcting it in place is the point of doing this in the open. We'd still rather be right the first time.
The real constraint is one line, at six inches
The lowest thing over the coach roof isn't the spar — it's the line that secures the vang, which clears the coach roof by 6 in. The cradle puts the top of the glass at 3.18 in, so there are 2.8 in between the panel and that line. It fits. It is not generous.
And a line is not a spar. A spar at a fixed height either clears or it doesn't; a control line sags, swings and chafes. Two and a half inches above a sheet of glass is room enough for a slack line to lie down on the panel, and a loaded line dragging across a panel's edge is a chafe problem for the line and a snag hazard for you.
What to check next is whether that line crosses the panel's footprint at all. The vang is at the mast, which is 17 in forward of the panel's leading edge, and if the line leads aft outboard through the deck organisers then it never passes over the panel and this whole paragraph is moot. If it does cross, there are three fixes in increasing order of effort: re-lead it outboard, stand it off on a low fairlead so it can't sag, or build the low cradle below and buy another inch and a half.
Why we stopped trying to clamp the handrails
The first plan was to hang the cradle off the teak coach roof handrails with Sea-Dog 327199-1 removable rail mount clamps — nylon body, stainless thumb screws, shimmed for 3/4, 1 and 1-1/4 inch round tubing, and marketed by Sea-Dog for solar panels specifically. Good part, wrong boat. Our handrails are teak: rectangular in section, and no amount of shimming makes a round-tube clamp grip a rectangular timber.
We went looking for an off-the-shelf teak handrail mount and there isn't one. Every marine "rail mount" on the market assumes round tube. That left fabricating a saddle to wrap the teak — doable, and we drew one — but it puts a clamped fitting on a piece of twenty-five-year-old wood whose only real job is to stop somebody going over the side.
So we stopped clamping and started bolting. The rails get their own feet, through-bolted into the molded non-skid. It's four more holes than we wanted, but it's a stronger mount, it leaves the handrails alone to be handrails, and it hands the coach roof a second pair of grab rails into the bargain. The Sea-Dog clamps go in the spares box — they'll find a job on the arch one day.
| # | Qty | Item | Spec / note |
|---|---|---|---|
| CR1 | 2 | Coach roof rail | 1-1/4 in OD × 0.065 wall 316, 34 in cut long, running fore-and-aft. The two set 46.0 in apart athwartships, symmetric about centreline. Same section as C1 and C3 — that is the whole reason the panel's hooks fit. Doubles as a grab rail. |
| CR2 | 4 | Rail foot | 316 base, 3 in round or square, four 1/4-20 holes, as low as the pattern allows. Through-bolted into the molded non-skid. See the deck note below — this is the real work. |
| CR3 | 4 | Backing plate | 1/4 in aluminium or G10, minimum 5 × 5 in, under each foot. Nylocs, fender washers. Never wing nuts on a structural deck fitting. |
| CR4 | 2 | Clevis bracket | Identical to P3, U-bolted to whichever rail takes the pins. |
| CR5 | 2 | Quick-release pin | A second pair of P4 pip pins, on lanyards, so you aren't robbing the arch every passage. |
| CR6 | 2 | Cam strap | 1 in polyester webbing, over the panel and down to the rails. The belt to the latch's braces. |
| CR7 | 1 pr | MC4 extension, 20 ft | E10 in the electrical table. Rolled up and stowed with the cradle. |
| CR8 | 16 | Bolt, 1/4-20 | A4-70 316, length to suit the deck thickness plus backing plate. Fender washers both sides. |
| CR9 | 1 | Butyl tape + thickened epoxy | The consumables that decide whether this installation lasts twenty years or three. |
Sixteen holes in a cored deck is the whole job
The coach roof is a cored laminate. Sixteen bolts driven through it and sealed with a squirt of 5200 is sixteen future leaks and a soft deck in three seasons. Every hole gets the same treatment, and none of it is optional:
- Drill the hole oversize — 1/2 in for a 1/4 in bolt.
- Dig the core back about 1/2 in all round with a bent nail chucked in a drill, leaving both skins intact.
- Fill the void with thickened epoxy. Let it cure properly, not overnight-ish.
- Re-drill 1/4 in through the solid epoxy plug, and chamfer the top edge so the bedding has somewhere to sit.
- Bed the foot in butyl tape — not silicone, and not 5200, which you will regret the first time you need the fitting off.
- Backing plate, fender washer, nyloc below. You've confirmed the headliner comes down, so this part is straightforward.
Done that way the bolt never touches core and the deck stays dry for twenty years. Done the quick way, you'll be grinding out wet balsa before the Cabo trip.
The coach roof is crowned; the cradle must not be
The roof curves athwartships. Both rails sit the same distance either side of centreline, so in principle they land at the same height — but the feet still want checking with a straightedge before anything is drilled, and the two rails must end up parallel and coplanar.
A rigid glass panel pulled down onto a twisted cradle cracks cells. It cracks them quietly, somewhere under the laminate where you can't see, and you find out from the charge controller three months later. Set the rail heights with a straightedge across both and shim the feet individually. Don't cut four feet the same length and hope.
What it has to hold
The design case here is 45 knots of apparent wind over the deck — dynamic pressure 6.9 psf, net coefficient ±1.5, so 10.3 psf across one panel's 10.4 sq ft: 107 lb of uplift, or about 27 lb at each of the four attachment points. Above that the rule is the same as it has been since Rev G: the panel goes below.
| Check | Demand | Margin |
|---|---|---|
| CR1 rail, bending between its feet | 134 in·lb | 15× |
| P1 hook plate — same part as on the arch | 27 lb | 23× |
| P4 pip pin, double shear | 27 lb | >100× |
| CR2 foot, uplift each | 27 lb | 3.8 psi under the base |
| CR8 bolts, four per foot | ~7 lb each | absurd |
| Deck core and bedding | — | governs, and it isn't a number |
The stainless is comically over-strong here, which is what happens when you reuse tube sized for a different job. That's fine — the point of keeping the 1-1/4 in section was never strength, it was that the panel's hooks fit it. Nothing in this cradle will fail by breaking. If it fails it will be because water got into the core around a bolt, which is why the drill-fill-drill note above is the longest thing in this section.
If the boom is tighter than about 6 inches: the low cradle
Give up the shared hardware and the panel goes lower. Four low chocks bolted straight to the non-skid, the panel's aluminium frame resting on them, pinned through the frame's own mounting holes. Glass top about 1.8 in above the deck instead of 3.2 — an inch and a half bought at the price of a second set of fittings on the panel and a second procedure to remember.
We'd rather not. One interface in both places is most of what makes Rev G work at all. But an inch and a half is an inch and a half, and a boom doesn't negotiate.
Where it goes, and where it must not
- On the molded non-skid, just forward of the companionway, panel athwartships. Confirmed clear of the hatches — which matters, because a panel over a hatch is a ventilation and escape-route problem and no amount of watts is worth it.
- Reefing is not affected, which is worth saying because it looked like it would be. The main is an in-mast furler reefed from the cockpit, so there are no reefing lines flaked across the coach roof waiting to be buried under a panel. The panel does bridge the halyard and furling-line runs, and that stays an accepted cost — you can get a hand under it — but the job you'd actually be doing at three in the morning happens at the cockpit end.
- The mast is 17 in forward of the panel's leading edge. That's the standing room you get when you go forward to furl at the mast instead of from the cockpit. It's enough to work in, but it puts your feet right at the panel's edge — see the next point, and mean it.
- Not stepped on. A panel on the coach roof is exactly where your foot wants to go heading forward in a hurry. It's glass on an aluminium frame and it will not take it. Mark it, brief the crew, route around it.
- The boom is 3.5 ft up and irrelevant; the vang's securing line at 6 in is not. Check whether that line crosses the panel footprint, on both gybes. See the warning at the top of this section.
- The new rails are extra handholds, not a replacement. The teak handrails stay exactly as they are and stay usable, which is one of the better arguments for having stopped trying to clamp them.
Two controllers make this part trivial
In an earlier revision this was a trap worth warning about: with the panels wired in series into one controller, leaving the stowed panel connected meant a dark panel in a string with a sunlit one — bypass diodes conducting, half the voltage gone, and no obvious symptom.
Rev K's twin SmartSolar 100/20s delete the problem outright. Each panel has its own controller and its own breaker. In state 2 you switch off the breaker belonging to whichever panel went below, cap its leads, and the panel on the coach roof carries on into its own tracker as though the other never existed. Nothing to reconfigure, nothing to get wrong at two in the morning.
What it actually buys
One 200 W panel at the same three-times-rating rule of thumb is about 600 Wh a day — roughly 43 Ah delivered at 13.8 V. Then derate it, honestly and hard, because a panel on the coach roof of a sailing boat spends a good part of the day in the shadow of the main and the boom. Call it 25 to 30 Ah on a passage day.
Against the 79 Ah underway consumption in the energy budget below, that covers roughly a third. It does not make the boat self-sufficient on passage and this page is not going to pretend it does. What it does is turn the alternator from the only charging source into a top-up, which on a four-day run is the difference between running the engine daily and running it every second or third day.
We'll publish the measured number once we've sailed with it. The ShadowFlux cells should handle the boom's moving shadow better than a conventional panel would — that's the whole reason they're on the boat — but "should" is doing real work in that sentence until there's a shunt log behind it.
The energy budget
This section exists because someone on a forum took the design apart, and he was right. The rack drawings above are the easy half of a solar project. The half that decides whether it works is arithmetic, and until now this page didn't do it.
What the array actually harvests
The rule of thumb sailors use — Nigel Calder's, and it matches most people's logs — is that a panel returns about three times its nominal rating in watt-hours per day, averaged across seasons and weather. Better on a bright June day, far worse under December overcast.
So: 400 W × 3 = 1,200 Wh/day.
That gets quoted as "100 amp-hours" by dividing by 12 volts, and that's where a lot of solar plans quietly overstate themselves. Charging doesn't happen at 12 V — it happens at 13.8 to 14.4. Divide by the voltage the current is actually delivered at and 1,200 Wh is closer to:
An earlier version of this page said "100 or more amp-hours a day." That was optimistic by about 13%, and the correction matters because of what comes next.
What the boat actually draws
Honest numbers for what's aboard, not a wish list:
| Load | At anchor | Underway | Notes |
|---|---|---|---|
| Refrigeration | 40 | 40 | The whole argument in one line. A BD35-class compressor at 4–5 A on a 40–50% duty cycle, and it never stops. |
| Raspberry Pi nav stack | 6 | 12 | ~1 A continuous while it's on. OpenPlotter, Signal K, AIS, plotter. |
| Autopilot | 0 | 8 | Roughly 1 A average while steering. More in a seaway, much more with a badly balanced helm. |
| Lights — anchor, nav, cabin | 8 | 8 | All LED. Would be three times this with incandescent. |
| VHF, instruments | 3 | 4 | Standby draw plus occasional transmit. |
| Water pump, charging, sundries | 7 | 7 | Phones, tablets, the pressure pump cycling. |
| Total | ~64 Ah | ~79 Ah |
The fridge is more than half of it. If you want cheaper amp-hours than solar can sell you, another inch of insulation in the box beats another panel on the arch — and it costs a tenth as much.
The gap
Harvest ~87. Consumption 64 at anchor, 79 underway. So on an average day the array covers the boat and hands back 20-odd amp-hours at anchor, and almost nothing underway.
That's fine as far as it goes — the array does the job it was bought for, which is to stop the engine being a charging appliance. What it doesn't do is refill a deeply discharged bank in any useful time. And with lead-acid, refilling completely is not optional.
It's worth being blunt about the size of the bank underneath all this. There are two Group 27 AGMs aboard, not three — 180 Ah nominal, 90 Ah usable at a sane depth of discharge. Against 64 Ah a day at anchor, a day and a half without sun takes the bank to its limit. The array is what keeps that from happening; it is not a reserve.
Note what Rev K does and doesn't change here. The 87 Ah figure is the full array at anchor, which is still state 1. On passage you have one panel on the coach roof making perhaps 25 to 30 Ah against a 79 Ah draw — better than the nothing Rev G offered, but still a deficit the alternator has to close.
The absorption problem
This is the part that kills AGM banks on solar boats, and it isn't about daily energy balance at all.
A lead-acid battery takes charge greedily up to about 80% state of charge and then stops cooperating. Above 80–85% its charge acceptance collapses — the current it will take falls away steeply no matter how much you offer it. Finishing that last 15% means holding absorption voltage for another three or four hours at steadily declining current.
Solar gives you five or six useful hours a day. If the boat's own loads eat most of the harvest, there is never a long enough surplus to run that absorption phase to completion. You end up cycling between, say, 60% and 85% forever — chronic partial state of charge, which sulfates the plates, and a bank that should last six or seven years is finished in one or two, losing capacity the whole way down.
| Figure | Why | |
|---|---|---|
| Bank, nominal | 180 Ah | Two Group 27s at ~90 Ah each. The bank that's actually aboard, and it stays for now. |
| Usable | 90 Ah | 50% depth of discharge. Going deeper trades life for capacity. |
| To replace it | ~108 Ah | Lead-acid charging is about 85% efficient — 1.2 Ah in for every 1 Ah out. |
| At +20 Ah/day surplus | 5–6 days | And the last stretch is the slowest, because acceptance is falling. |
| Off the alternator | 6+ hours | The 55 A Hitachi realistically delivers ~25 A at cruise, and tapers as voltage rises. |
What this means for our actual use
We're coastal — weekends and short trips, back on shore power every week or two. That matters, because a proper shore charger takes the bank to 100% and holds it there, which is exactly the medicine AGMs need. In that pattern the array is offsetting daily draw rather than being the sole charge source, and the design works.
Change the mission and it stops working. Anchor out for a week or more with the fridge running and there is no path back to 100%. The bank would live in partial state of charge and degrade, and no amount of rack engineering fixes that.
Where this points: lithium
The honest conclusion is that the limit isn't the array. It's the chemistry underneath it.
LiFePO4 takes full charge current right up to about 95% state of charge, so there is no absorption tail to lose. It's about 99% charge-efficient instead of 85%. It gives 80–100% depth of discharge instead of 50%. And — the part that actually matters here — sitting at partial state of charge does it no harm at all.
| 2 × AGM | 200 Ah LiFePO4 | |
|---|---|---|
| Nominal | 180 Ah | 200 Ah |
| Usable | 90 Ah | ~180 Ah |
| Charge efficiency | ~85% | ~99% |
| Absorption tail | 3–4 hours | None to speak of |
| Harmed by partial charge | Yes — badly | No |
| Weight | ~120 lb | ~30 lb |
| Cycle life | 300–600 | 2,000–4,000+ |
| Cost | ~$500 | $500–1,400 |
With only two AGMs aboard the comparison stops being close: one lithium gives twice the usable capacity of the existing bank, takes 90 lb off a 7,400 lb boat, and every amp-hour the array makes actually goes in. Against AGMs you may be replacing every year or two under partial-charge cycling, the money is a wash on the first replacement and clearly favourable after that.
Even so, don't read this as 400 W solving everything. With lithium and this load list the array is comfortably positive at anchor and roughly break-even underway. It buys independence for long weekends, not indefinitely. If we ever want to sit out for a fortnight with the fridge running, the answer is a smaller fridge load before it's a bigger array.
Credit where it's due
The arithmetic in this section came out of a forum critique of the design — charge acceptance, the 85% efficiency figure, the alternator's real output, and the point that AGMs need to reach 100% regularly or die young. All of it checked out. Publishing a design and having someone competent pull it apart is worth more than another revision of the drawings.
Measure before you order
Eighteen measurements stood between this design and a purchase order; most are now taken. If you're copying this for your own boat, these are the numbers that will differ — the first five decide the front pickup, and the last six decide whether the panels have anywhere to go.
- The existing holes in the traveller plate. Done: 18 in in from each end, 9/32 dia, two rows 1-3/4 in apart straddling the rail with 1/8 in of clear plate each side. Re-check the spacing within each pair with a caliper — as measured it leaves a 1/32 in web, which would mean using one hole per pair.
- Traveller plate thickness. Measured: 5/16 in. It sets the tab thickness, deletes the shim and sets every other flat bar on the boat.
- Traveller car clearance over the plate. Confirmed tight — the screws must finish flush. Check whether the holes are already countersunk before reaching for a countersink bit.
- Aft hole row to the plate's aft edge. The last one. Sets the tab's cantilever; we've assumed 3 in from the aft screw to C1's centreline, and past 3.5 in the tab goes to 2-1/2 in wide.
- Clear gap between adjacent traveller nuts at each hole group, measured athwartships. The tab has to thread between them with 1/2 in to spare.
- Edge distance and weld condition around the holes — two bolt diameters of sound plate minimum, and a good look at the plate-to-tube welds either side.
- Which face of the leg the traveller control line runs down, at the S1 strut foot stations. Rev L put nothing else on the legs.
- Confirm the plate's aft edge sits on the aft tube's centreline.
- Plate aft edge to the transom edge, horizontally. The array reaches 50.7 in aft — this is the check most likely to force a change.
- Traveller car fully eased, both ways. Nothing within an inch of the panel's leading edge; control line clear of C1 and the struts; and the car body clear of both F1 tabs at the ends of its travel.
- Boom and vang clearance sheeted hard in, and with the boom right out on a run.
- Plate top above the cockpit sole, and standing eye height at the wheel. Confirms the sight slot — which in Rev G is an at-anchor rig check rather than a helming sightline, but the geometry is the same.
- The arch's deck bases — bolts, backing plates, gelcoat crazing. Everything the array does in a blow ends here.
- The companionway, with cardboard. Done — the panels go below. If you're copying this, cut a rectangle 49.69 × 30.08 in and carry it down before you order anything.
- The quarter berth. Clear width and height at the opening, and the flat length and width inside. You need to land two panels 50 × 30 × 1.2 in, side by side or stacked, without them sliding about on a beat.
- Boom and vang clearance over the coach roof. Measured: boom 3.5 ft, and the vang securing line does not cross the panel. The coach roof is clear — nothing overhangs the cradle at all.
- The molded non-skid field, forward of the companionway. You need 46 in athwartships between the rail feet, symmetric about centreline, landing on flat gray rather than out on the curve — plus about 34 in fore-and-aft for the rails. Lay the cardboard panel there and check it misses every hatch.
- The crown. Straightedge across the roof at the rail stations, and measure the gap at each end. This tells you how much to shim the feet so the two rails come out coplanar.
- What's behind the headliner at each of the four foot positions — wiring, a bulkhead top, anything you'd rather not drill into from above.
Build sequence
- Dry-fit on the dock, not on the water, and leave everything finger-tight until step 9.
- Confirm the plate is 5/16 and the tabs are 5/16 — that is the whole build rule, and it wants checking with a caliper rather than assuming. Then look at the top of the plate: if the holes are already countersunk, skip the next step.
- Countersink the four holes 82°, 0.385 dia, about 0.060 deep, from above, using a piloted bit that centres in the existing hole. Test-fit a #10 flat head and lay a straightedge over it — it must not stand proud. Then walk the traveller car across before you go any further.
- Offer each F1 tab up to its holes dry. Check it threads clear of the traveller nuts with the car walked to both ends of its travel, and that it lies flat on the packer rather than rocking on a weld fillet. Dress the tab, not the plate.
- Trim C1 to 79.0 in, wrap it with both tabs, and run the screws up finger-tight. Check the build rule with a straightedge across the plate: C1's top should read 5/8 in down with nothing adjusted. If it doesn't, the plate or the bar isn't the thickness you thought — measure again before shimming.
- Check C1 is level athwartships and square to centreline, then fit the F1b keepers and the F3 cross bolts.
- Hang the two arms under C1 at X = ±33 in cross clamps, running aft. Level them fore-and-aft with a torpedo level.
- Add C2 and C3 on top of the arms at 24.0 and 22.0 in. Square the frame by measuring the two diagonals — match within 1/8 in.
- Fit one strut. Before marking it, walk the traveller car fully to each end and confirm the control line stays clear of the strut and its foot clamp. Then mark, cut both to match, deburr and polish.
- Strip it back down. EPDM under every clamp on the arch legs, butyl in every one of the plate holes, and Tef-Gel on every 316-on-316 thread — every one.
- Reassemble and torque: 1/4-20 to ~75 in·lb, 5/16-18 to ~130 in·lb (lubricated values — never dry-thread figures on stainless). Cap C1's ends.
- Fit the quick-release kit by offering the panel up, not by working off the drawing. Hold a panel in position, mark where its forward flange crosses C1 and where its side flange crosses C3, and set the hook plates, pin tabs and clevises from those marks. A quarter inch of accumulated error here is the difference between a panel that drops on and one you have to fight.
- Hang both panels: hook forward, drop aft, pin. Check the 18 in slot and the 1 in leading gap with a rule. Bond the C2 rest pads once you know exactly where the panels land.
- Stand at the wheel and look up through the slot — at 18 in you should find the masthead from any normal stance. If you can't, something is off in the height assumptions and it's worth stopping before you torque anything.
- Wire, gland, breaker, controller. Check polarity with the panels covered before connecting anything.
- Run the removal drill twice, with a stopwatch, before you call the job done. If it isn't roughly ten minutes for the pair, something in the kit needs moving — and better to find that at the dock than at anchor with a front coming through.
- Measure the boom before you touch the coach roof. Sail flaked, vang normal, boom centred and then right out on each gybe. That number decides whether you build the rail cradle or the low cradle, and everything after this step depends on it.
- Now set the coach roof cradle out from the panel, not from the drawing. Lay the panel on the non-skid where it will live, athwartships, forward of the companionway. Mark where its hooks and pin tabs fall. Those marks set the 46 in rail spacing far more reliably than a tape does.
- Straightedge across both rail stations and shim the feet individually so the two rails come out parallel and coplanar. Dry-fit the panel on them and sight along it before a single hole is drilled.
- Drill-fill-drill every one of the sixteen holes per the deck note: oversize, core dug back, epoxy, cure, re-drill, chamfer, butyl, backing plate, nyloc. This is the step that decides whether the installation lasts. Do not shortcut it because the weather is good.
- Hang the panel, fit the CR6 cam straps and leave them rigged. Run the CR7 extension and watch the roof panel pick up on the controller.
- Re-torque everything after the first day sail and again after a month. Rev L made this a shorter list — bolts in holes don't migrate the way a clamp on a polished rail can — but check the F5 nuts, the strut feet and the coach roof feet for weeping all the same.
Still to sort out
Being honest about what isn't decided yet, because build logs that only show finished work aren't much use to anyone:
- Settled: the front pickup is #10-24 countersunk flat heads into 9/32 holes in a 5/16 plate. What's left is one tape measure from the aft hole row to the plate's aft edge
- Charge controller selection and exactly where it mounts in the aft locker
- Decided: the two AGMs stay for now. The energy budget still says lithium, and with only two batteries aboard the case is stronger than ever — one 200 Ah LiFePO4 is twice the usable capacity. It's a question of when, not whether, and it isn't this season.
- Real-world output measurement, panel temperature, and how much the rig shades the array at anchor as the boat swings
- Where the canvas goes back in to fill the shade gaps once the rack is up
- Settled: the panels fit below, the arch geometry is measured, and the front pickup is no longer a guess — it bolts into holes that already exist at X = ±21.5
- Settled: the vang securing line does not cross the panel. The coach roof is clear and the drill can come out
- How much the boom actually shades the coach roof panel on each point of sail — the number that decides whether 25 Ah a day was optimistic
- Whether bridging the halyard and reefing runs turns out to be as livable as it sounds at the dock
- How the drill holds up when it's blowing 20 and we want to leave — the honest test of the whole idea
All of it goes on YouTube as we work through it — the mistakes included, since those are the useful part — and the longer write-ups land on Substack.
Read this before you copy it
Rev N, 2 October 2026, supersedes Rev M. One change: the radar moves from an eight-foot stern pole to a mast mount, so the R3 brace clamp is deleted, the rack carries all of its load into the arch again, and the sailing weight-moment row picks up the radar at its new height — about 1,019 ft·lb, 1.3% below Rev F. No tube, fitting, dimension or margin on the rack itself changes.
Rev M, 30 September 2026, supersedes Rev L and Rev K of the same day, Rev J, Rev H of 28 September (the radar pole tie), Rev G of 27 September (the removable panels) and Rev F of 23 August. Rev L moved the front pickup; Rev M measures it. No wall thickness, slot, arm position or aft geometry has changed since Rev G:
- The two H1 rail clamps, their riser straps and their U-bolts are deleted. The frame bolts to four holes already drilled through the traveller plate, 18 in in from each end — X = ±21.5.
- The traveller plate measures 5/16 in, so a 5/16 in tab closes the build rule with no shim at all: plate + tab = 0.625. The packer is deleted.
- Every flat bar on the page goes from 1/4 to 5/16 — hook plates, pin tabs, clevis brackets — so the panel rests in one plane across its hooks and the tab. Glass finishes at P + 0.87, still 1.13 in under the ceiling.
- The holes are 9/32, in two rows 1-3/4 in apart straddling the traveller rail with 1/8 in of clear plate either side. Fasteners are #10-24 A4-70 countersunk flat heads, flush, because the traveller car clears the plate by very little.
- C1 shortens from ±42 to ±39.5, flush with the plate's ends, and finishes 0.4 in proud of the glass instead of 2.9.
- The arms now sit outboard of their supports. C1's worst moment rises from 547 to 699 in·lb and its margin falls from 4.6× to 3.6× — level with the arms, above the 3.5× floor, no change of tube.
- About 3 lb comes off the arch and roughly 1 lb goes back on in thicker flat bar. Sailing weight-moment drops from 1,005 to 990 ft·lb.
- The vang securing line does not cross the coach roof panel. That clearance question is closed.
Rev K, which this supersedes, folded in the first tape-measure session: bare tube starting 6 in below the top of the arch, the boom 3.5 ft above the coach roof, the panels fitting below, two SmartSolar MPPT 100/20 — one per panel, and a house bank of two Group 27 AGMs rather than three. All of that still stands.
An earlier draft of this revision hung the cradle off the teak coach roof handrails on proprietary rail clamps. Photographs of the boat killed that: the handrails are teak, no off-the-shelf rail mount grips a rectangular timber, and the panel wants to lie athwartships forward of the companionway rather than fore-and-aft. The cradle is now through-bolted to the molded non-skid. The handrails are left alone.
Dimensions in inches; datum P is the top face of the traveller plate, Y measured aft from the plate's aft edge, X from centreline. All structural members 316/316L.
Two different load cases live on this page. The rack is sized for a 50 kt design gust with the panels fitted and the boat at anchor; the coach roof cradle is sized for 45 kt apparent over the deck on passage. Both at a net pressure coefficient of ±1.5, with margins against annealed 316 yield rather than ultimate. The bare frame was separately checked at 70 mph and is not governed by it. Member checks assume the F1 tabs at X = ±21.5, a 3 in tab lever, and the hook plates at ±33. All of this assumes the panels actually come off before you sail. If they live up there full time, build Rev F.
One dimension is still open: the distance from the aft hole row to the plate's aft edge, which sets the tab's lever. We have assumed 3 in. Past about 3.5 in the tab widens to 2-1/2 in; it cannot change anything else. One spacing is worth re-checking with a caliper — the 5/16 in between the holes within each pair, which as measured would leave a 1/32 in web. Everything else has been measured. Strap the coach roof panel as well as latching it, and take it below when it's going to be rough.
One thing genuinely wants a second opinion. The front reaction now enters the traveller plate 18 in inboard of the bends rather than out at them. We think that's untroubling — the plate is welded along both tubes, and it already carries the mainsheet at greater load closer to midspan — but it is the one judgement in Rev L that a surveyor should confirm rather than a website.
Stability figures are estimates, not measurements — see the note in the weight-aloft section. Weight-moment comparisons between conditions are reliable; the absolute percentages are not.
This is not a stamped engineering drawing. Have a marine fabricator or surveyor review the arch attachment — the traveller plate, its welds to the tubes, and the arch's deck bases — before you trust any of it offshore. Bolting to holes somebody else drilled is convenient, not automatically safe. Look at them first.
Panel data: Renogy ShadowFlux 200 W · ENF datasheet. Cradle hardware: Sea-Dog 327199-1. Hull data: Sailboatdata Hunter 290 · Hunter 290.
The radar, up the mast
A HALO20 on a mast mount and a GO9 XSE at the helm — more range than the pole, for a quarter of the price.
Related buildFresh water aboard a Hunter 290
What $1,000 actually buys in a 12-volt watermaker — and why we spent $400 instead.
Related buildOff-the-grid air conditioning
Why 400 watts isn't the array you'd need for DC air conditioning — and what we did instead.
Back toSkipper Scott's home
Travels, how-tos, and the rest of the open source boat.