Roof array planOcho Rios · rev 25 Oct 2026

Roof array — modules, strings, bonding

33 modules · 23.43 kW · 4 strings · 5 rail groups · plan view, dimensions in mm

Three things on this drawing can still change what gets bought

1 · The module tails are 300 mm. Both datasheets say 4 mm², 300 mm in length, length can be customized. Two tails give a joint 600 mm to play with, and most of the joints on this roof are 1,323 mm or 2,405 mm centre to centre. Every string sheet below counts its own shortfall; the total across the four strings is in the take-off. The cheap fix is to ask Sunpal for longer tails before the order ships — the datasheet says they can be customised, and it costs a line on an email instead of 30-odd made-up leads and 60-odd crimped MC4 ends on a roof.

2 · Which way do the rails run? Nobody knows, because the mounting system was quoted as one unnamed line item. If they run up-slope the array is 5 rail groups and the bonding is about 15 m of 16 mm². If they run along the rows it is 19 rail groups and about 45 m. Switch it below and watch the drawing change. That one unknown is worth roughly 30 m of wire and 13 lay-in lugs, which is why CART.md says do not order the 16 mm² yet.

3 · The terrace width. Measured off the drawing it is 8′-2″ and all of this fits. The drawing's own labels say 14′-0″, and at 14′-0″ zone B loses a whole column — 27 modules, not 33. One tape measure from the west wall to the primary bedroom's west wall settles it, and it should be the first measurement taken on site.

Which panel is it? Both are sold as “SP710M-66H”. The HJT runs 1.84 V per module hotter open-circuit, which is 16 V across a nine-module string.
Coldest cell temp 15 °C Voc rises as it gets colder. 15 °C is a fair worst case for Ocho Rios at 285 m; drag it to 10 °C to see the margin a cold morning eats.
Rails run Changes the bonding layer, the rail-group count and the lug count. Open question for Sunpal.
Layers
Click a module, a string, a rail group or a bonding run · drag to pan · pinch or scroll to zoom

Plan · looking down · north at the top, the array faces south · the two red hatched edges are real level changes · coloured chains are the four strings in series order, olive is the 16 mm² bonding network

Working from this drawing

  • Set out the rails from the walkway, not from a wall. The walkway and the 1′-0″ access gap are the only two places on this roof you can stand, and every dimension on this sheet is referenced off them. Get those two right and the rest follows; chase a wall line instead and the access gap closes up.
  • Nine modules is the ceiling, and string 3 is the one at the limit. Not by one rule but by three: 9 × 710 W = 6,390 W against a 6,600 W MPPT, and 9 modules of cold Voc is 448 V (TOPCon) or 465 V (HJT) against a 500 V inverter limit. A tenth module breaks both on either panel. There is no restring question left — 9 is simply the top.
  • Every string doubles back on itself. That looks like wasted cable until you remember that the + and the − both have to reach the junction box. A string run end to end leaves one home run at the far corner of the roof; a string that doubles back leaves both ends in the same place. String 4 is the clearest case: ordered the other way round it put a 45 ft home run across the whole array.
  • Rows 5 and 6 are one string on purpose. The 2 ft south parapet throws 1′-10″ at December noon and 2′-5″ at nine in the morning, and a landscape module shaded along its bottom edge drops a third of itself through a bypass diode. Putting both rows on one MPPT means a shaded edge costs that string and nothing else. It is free — rows 5 and 6 made a natural eight-module string anyway.
  • A string crossing a rail group does not bond it. This is the one thing on the sheet that is easy to get wrong by looking. String 2 runs from RG2 through RG1 into RG4, and none of that earths RG4 to anything. Only the 16 mm² does.
  • The chase cannot take four earths, which is why the junction box exists. Eight PV cores plus one 16 mm² is 29 % fill. Four 16 mm² is about 46 % against a 40 % limit — it will not go in. So the rail groups are bonded together on the roof and exactly one downlead goes down the chase. The junction box is load-bearing for the earthing design, not a convenience.
  • Conduit only where a conductor leaves the shelter of the array. PV cable is UV-rated and clips to the rails under the modules, which is cheaper and better than conduit — no water trap, no UV on the plastic, no new fixings through the membrane. That leaves exactly one conduit run on this roof: CR3, over the carport step. The earlier estimate of conduit along the whole walkway was wrong.
  • Stainless ties on CR1 and the perimeter, nylon everywhere else. The home-run bundle under row 1 is the one length of cable that is both permanent and reachable, so it is where a UV-cooked nylon tie actually drops a 1000 V conductor into a walkway. The 50 stainless ties are for that run and the array edges. Nylon under the panels, where sun never reaches.
  • Each module is 37.5–38.7 kg of bifacial double glass. About 40 % heavier than the panel most people have carried, with 2 mm of glass on the back instead of a plastic backsheet. Two people minimum, three over the carport step. There is no backsheet to take a knock — a dropped one is scrap.
  • Anti-oxidant compound on every copper-to-aluminium joint. Already in the cart. Dissimilar metals, 3.8 km from the coast, C4 corrosivity — an unprotected Cu-to-Al bond is the joint that fails first and the one nobody goes back to look at.
  • Label the eight cores at the junction box before they go in the chase. They arrive as eight identical black 4 mm² cores and they come out of the chase 40 m later in the plant room, where four of them have to land on the right MPPT. The flexible-ID wrap tape is bought for exactly this.
  • Nothing on this roof isolates the PV. Sun on glass means these conductors are live. The four C25 breakers in the SPD box protect what is downstream of them and cannot make the chase dead. Cover the array or work after sunset, then test with a meter.

The four strings

Every figure in this table moves when you change the panel family or the temperature above. Made-up lead is cable you have to cut and crimp because the module's own 300 mm tails do not reach.

Inverter 1 takes strings 1 and 2, inverter 2 takes strings 3 and 4 — matching the wall plan, where strings 1 and 2 land on inverter 1's PV1 and PV2. §4.9.2 forbids joining MPPT inputs, so each of these four is independent from its first module to its inverter terminal: no combiner, no paralleling, and no string fuses.

The module, whichever one it turns out to be

“SP710M-66H” exists in two families and the order does not say which. Datasheets are in 05-datasheets/. The HJT is the better panel for this climate — at a 70 °C cell it loses 11.7 % against the TOPCon's 13.5 % — but it also runs 16 V hotter across a nine-module string, which is the only place it matters electrically.

Bonding — and the one unknown that sets its price

A rail group is one continuous piece of aluminium. It is not a roof plane and not a string. What breaks a rail group is what breaks a rail: the access gap, the rotated portrait rails, the terrace, and the carport step. Each separate piece of metal needs the 16 mm² to reach it, and which way the rails run decides how many pieces there are.

Anodised aluminium is an insulator on its own surface, 15–25 µm thick. So either a listed clamp cuts through it — UL 2703, route A, and you buy lay-in lugs only — or a 4 mm² jumper goes round it, which is 32 jumpers and 64 bolted joints on a roof. The lay-in lugs are needed either way, because the 16 mm² has to land on copper whichever route the clamps take.

What goes down the chase

Nine conductors, one 1¼″ conduit, and the reason the junction box is not optional.

This is the run the wall plan labels 4 STRINGS FROM ROOF and ARRAY BOND DOWN THE CHASE. ⚠ The chase position is drawn at the junction box because that is where it should be — the wall plan puts the DC SPD box directly under it so the surge path to the arresters is as short as possible. If the plant room is not under this point, the chase moves and every home-run length on this sheet changes.

Roof take-off

Quantities follow the toggles above, so this is the take-off for the case currently drawn, not a fixed order. Cable is a Jamaica purchase — the Amazon basket is for small hardware only.

What this drawing cannot answer

  1. How long are the module tails, and where is the junction box on the back? The datasheet says 300 mm and that it can be customised. This is the one question on the list that is cheaper to ask than to solve — longer tails ordered before the modules ship replace about 30 made-up leads and 60 crimped MC4 ends. It belongs in the Sunpal email, which is still unsent.
  2. Which way do the rails run, and are they listed to UL 2703 for bonding? Decides 5 rail groups against 19, about 30 m of 16 mm², 13 lay-in lugs, and whether 32 bonding jumpers get bought at all. A WebSearch on 25 Oct found no Sunpal UL 2703 listing anywhere, while Unirac, SunModo and IronRidge surface immediately with public manuals. “Engineered in line with UL 2703” is not a listing.
  3. Are the lay-in lugs already in the mounting kit? Needed under either bonding route, so this is a quantity question, not a design one — 0 or 6, or 0 or 19.
  4. How wide is the terrace? 8′-2″ as drawn, 14′-0″ as labelled. The difference is six modules. First measurement on site.
  5. Where exactly does the chase come down? Drawn at the junction box. If the plant room sits elsewhere, the home runs lengthen and the walkway bundle grows.
  6. Is the roof membrane something you can fix to? Nothing on this sheet assumes a penetration except the chase — the array is drawn as a ballasted 3° system sheltered by the 2 ft parapet, which is the whole reason the 1′-6″ setback was defended. If the mounting system turns out to need through-fixings, the parapet-shelter argument and the ballast calculation both have to be redone.
  7. Are the three high-block modules really given up? Taking them back makes the array 36 and every string exactly 9 — the cleanest electrical layout available — at the cost of a sixth rail group, a fourth level change and a bonding run up the high-block wall. It is reversible right up until the frames go down.

Questions 1, 2 and 3 are one email. Questions 4 and 5 are one visit with a tape measure. Question 6 is for whoever supplies the mounting system, and 7 is a decision rather than a question.