Solar Panels Facing Different Directions: How to Size Strings

Panels that face different directions can share one inverter safely — the electrical checks are the same ones you run for a south roof. What changes is how you wire them. Four rules cover almost every roof:

1. One string, one direction. Never put east- and west-facing panels in the same series string. In our simulation it costs 13 % of the year's energy.
2. One roof face per MPPT tracker. An east–west roof on a two-tracker inverter is the textbook layout: each tracker follows its own face of the roof.
3. If two faces must share a tracker, wire them in parallel with the same number of panels. Equal parallel strings lose under 0.1 % a year; a string one panel shorter than its neighbour loses about 4.4 %.
4. Check every string at the worst case. Maximum voltage on the coldest morning and current in full sun do not depend on direction. The one figure that does — the combined current of two faces on one input — never relaxes the safety limit.

The worked example below uses the panel and inverter our visitors chose most often this month — the LONGi LR7-72HVH-650M and the Deye SUN-6K-SG05LP1-EU-AM2-P. Every limit check comes from the same engine as our string calculator; the energy figures come from an hour-by-hour simulation on five years of PVGIS weather data.

Before any wiring decision, separate the numbers that depend on direction from the ones that do not. A string's safety limits are set by the extremes — the coldest dawn and full sun on the hottest day — and every roof face meets both at some point in the year.

The rest of this article works through the table row by row: how to wire the faces, how long each string may be, what one shared tracker costs, and how far you can oversize the inverter.

In a series string the same current flows through every panel. With the sun in the east, the east panels could deliver their full current while the west panels, facing away from the sun, deliver only a fraction of it. The string cannot run at two currents at once, so the weakest panels set the pace:

Bypass diodes stop the weak panels from blocking the string completely — once the current rises above what a panel can make, its diodes conduct and the panel is skipped — but a skipped panel contributes nothing, and the inverter now has to find the best of several humps on the power curve. Either way, part of the roof is always working below its potential.

We simulated it hour by hour over five years of London weather (PVGIS, 2019–2023) with a single-diode model of the LONGi 650 W panel. A string of 4 east + 3 west panels makes 13.0 % less energy than the same seven panels split by direction onto two trackers; 4 south + 3 west loses 10.7 %. It is the single most expensive wiring mistake on a multi-direction roof — and it hides from any voltage check, because the string's voltage is perfectly fine.

An MPPT tracker is an independent input that finds the best operating point for whatever is connected to it. Give each face its own tracker and each one follows its own sun: the east string works at its best in the morning, the west string in the afternoon, and neither drags the other down. That is why a two-tracker inverter is the natural match for an east–west roof.

Our example is the pair our visitors picked most in the calculator this month: a LONGi Hi-MO X10 LR7-72HVH-650M panel on a Deye SUN-6K-SG05LP1-EU-AM2-P hybrid inverter, on a London gable roof with 7 panels on each 30° slope. Design temperatures are −10 °C in winter and 35 °C in summer.

Each tracker is checked on its own, exactly as if it were a separate inverter. Both strings have 7 panels, so the numbers are the same for east and west:

All nine checks pass. The array is 14 × 650 W = 9.1 kWp on a 6 kW inverter — a DC/AC ratio of 1.52, well inside what a hybrid inverter is designed for and below the 12 kW of panels Deye allows on this model. In the calculator's Advanced mode, the layout reads like this:

The same arithmetic gives the range each tracker accepts: at least 4 panels (4 × 40.88 V = 163.5 V clears the 150 V minimum) and at most 8 (8 × 57.67 V = 461.4 V stays under the 500 V limit). With 8 panels the calculator warns, because on a −10 °C morning 461 V sits above the 425 V tracking ceiling and the inverter waits for the panels to warm up before it starts — safe, but not free. Seven panels per face is the sweet spot for this pair.

It is tempting to reason that a west-facing string never sees the cold morning sun and can therefore take a panel or two more. It cannot. Open-circuit voltage depends on light only logarithmically: in our single-diode model of the LONGi panel, Voc at 100 W/m² — a tenth of full sun, the diffuse light of a clear dawn — is still about 91 % of its full-sun value. A west string on a frosty morning sits in exactly that light.

Temperature does the rest. The calculator takes the cell temperature on the coldest morning to be the air temperature itself, because at first light the panels have not warmed up yet. That moment comes to every face on every clear winter morning, so the maximum-voltage check is identical for east, south, west and even north strings.

The hot end is the mirror image. A west face collects full sun in the warmest hours of the day, so on a summer afternoon it runs hotter than the east face did at 10 am, and its operating voltage is the lowest on the roof. The calculator assumes full sun at your highest temperature for every string — exactly the west string's situation — and checks the MPPT minimum against that:

This is where small roof faces bite. A west-facing dormer with room for 3 panels gives 3 × 40.88 V = 122.7 V on a hot afternoon — below the 150 V minimum, and the calculator fails it. The fixes are to find room for a fourth panel, to put the small face on a microinverter, or to leave it empty; adding it to another face's string is the one option rule 1 forbids.

Sometimes there are more faces than trackers: an L-shaped roof with east, south and west slopes on a two-tracker inverter, or an east–west roof on a single-tracker off-grid inverter. Then two faces have to share an input, and the question is what that costs in energy. We ran the same five-year, hour-by-hour simulation for the usual cases, always against the ideal of one tracker per face:

Parallel strings of equal length lose almost nothing. A panel's maximum-power voltage hardly moves with light — it moves with temperature: at 200 W/m² the LONGi panel's MPP voltage is still 43.8 V against 44.56 V in full sun. Two strings in parallel share one voltage and simply add their currents, so as long as their best voltages are close, the tracker loses nothing by choosing one. The east string runs a little warmer in the morning than the west string, which shifts its best voltage by a few percent — and near its peak the power curve is so flat that a few percent of voltage costs a fraction of a percent of power.

Length is a different matter. A 6-panel string's best voltage is a seventh lower than a 7-panel string's, far too wide a gap for the flat top of the curve: 4.4 % of the year's energy is gone. It is just as gone for two strings on the same face (4.7 %), because direction was never the problem. And mixing directions in series, at 11–13 %, is worse again by an order of magnitude.

Sharing a tracker is cheap in energy but expensive in current. Two strings in parallel feed their currents into the same input, and the input has two ratings: the current it can use, and the short-circuit current it can survive. For our 650 W panel:

So on the SUN-6K, the three-face layout — east and west in parallel on tracker 1, south on tracker 2 — is incompatible. Its trackers also have one connector each, so the second string would need a Y-connector, and 21 panels (13.65 kWp) exceed the 12 kW the datasheet allows. With a location set, the calculator shows one more thing: under each current check, a 'Realistic max' of 25.16 A.

That realistic figure is real. Given directions and a location, the calculator sweeps the sun across your sky on the longest day and finds the moment when the strings on a tracker deliver the most current together. East and west slopes are never both in full sun, so their combined peak — around noon — is 81 % of the naive sum. 25.16 A is below the 27 A short-circuit rating, and the check still fails. That is deliberate: the short-circuit rating protects the inverter's input hardware, and IEC 62548 sizes protection for the worst case — a cloud edge that lifts irradiance above 1,000 W/m², snow reflecting light onto both slopes, a fault. The realistic figure is for the other rating — the input current, which only costs energy when exceeded, as the next example shows. The safety check never reads it.

The fix is an input built for two strings. The second most chosen inverter in our calculator this month, the Deye SUN-8K-SG05LP1-EU-AM2-P, has two connectors per tracker rated 32 A / 48 A. With two 6-panel strings — one east, one west — on tracker 1 and a 7-panel south string on tracker 2, every check passes: 31.21 A against 32 A and 48 A, and 12.35 kWp against a 16 kW connection limit.

When may the realistic figure decide? Only for the input current. Past that rating the inverter simply holds the extra current back — it costs energy, never hardware — so if strings facing different directions stay inside it at their shared peak, nothing is lost. Take the third most chosen panel this month, the LONGi LR8-66HGD-620M, whose 16.05 A is a little higher than our example's: seven east and seven west in parallel on tracker 1 of the SUN-8K add up to 2 × 16.35 A = 32.7 A in the worst case, just over the 32 A input. Their realistic shared peak is 26.35 A, so the calculator passes the check and says why — and with 32.7 A far below the 48 A short-circuit rating, the whole layout is compatible.

This is not wishful thinking. Over five years of hourly London weather the same pair lost 0.000 % of its energy to the current limit on a 32 A input — and would lose nothing on a 28 A one either. Strings facing the same way get no such allowance: two south strings on a 28 A input lose about 0.24 % a year, and the calculator keeps its warning.

How much the combined current drops depends on the roof pitch and the latitude. The steeper the slopes, the less the sun can light both at once; the lower the latitude, the higher the noon sun and the more both faces see. Here is the peak current of an east ‖ west pair as a share of the two strings' sum, from the same clear-sky sweep the calculator runs:

On a 10–15° flat-roof rack, east and west strings together reach 88–98 % of the sum — practically the same as two south strings. Only on steep roofs at high latitudes does the reduction become large, and even there it is about a quarter, not the half many people expect.

Where orientation really helps is the inverter's power rating. The two faces peak hours apart, so the array never delivers its nameplate power at once. The chart shows a clear June day in London for our 9.1 kWp example, next to the same 14 panels facing south:

On an ideal day the east–west array peaks at 6.9 kW — only 76 % of its nameplate — and has 3.0 kWh above the 6 kW AC limit, against 11.3 kWh for the south array. Real days are hazier, and real systems lose power in cables, dust and heat before it reaches the inverter, so over a whole year the difference is even clearer. Hourly PVGIS output for London, 2019–2023, with 14 % system losses:

In northern Europe an east–west array can run at a DC/AC ratio of about 1.8 and lose under 2 % to clipping — less than the south array loses at 1.5. The south array still makes more energy per panel; the east–west one makes better use of every inverter kilowatt.

The calculator stays conservative here. Its DC/AC check compares the nameplate array with the inverter's AC rating against a fixed threshold per inverter type — 1.5 for string inverters, 2.0 for hybrids — with no credit for orientation, and it warns when the array exceeds the PV input the datasheet allows. Treat that datasheet limit as firm whatever your roof looks like: it is the manufacturer's, and the warranty depends on it. Orientation lets you use the room below it with less clipping than the ratio suggests.

Direction changes the yield far more than the wiring does. Annual output per kWp from PVGIS for three European cities (SARAH3 satellite data, 14 % system losses), with the share of a south-facing 35° roof in brackets:

The pattern holds from London to Madrid: south-east and south-west keep about 93–95 % of the south yield, a single east or west face about 77–80 %, and an east–west pair on a low 15° rack 82–83 %. East usually edges out west slightly, because panels run cooler in the morning and afternoons are often hazier.

Per panel, an east–west roof is the weaker choice. Per roof it can be the better one: on a flat roof, low east–west racks stand back to back without the row gaps a south-facing rack needs to avoid shading its neighbours, so the same area holds more panels — and production spreads over the day, which suits households that use power in the morning and evening.

The calculator handles every rule above in its Advanced mode, where you describe each MPPT tracker separately. Here is the east–west example, step by step.

1. Running one string across two roof faces

   It passes every limit and quietly costs 11–13 % of the year's energy. One string, one direction.

2. Paralleling strings of different length

   Two strings on one tracker must match in panel count and model; one panel short costs about 4.4 %. Strings on different trackers may differ freely.

3. Trusting the realistic current for the short-circuit rating

   Two faces together peak below their sum, but the short-circuit rating protects hardware and is sized for the sum. The calculator keeps that check at the worst case on purpose.

4. Lengthening a west or north string

   Voc is nearly full in dawn light, and every face meets the coldest morning. The maximum-voltage check is the same for every direction.

5. Wiring a tiny roof face as its own string

   Three panels give 122.7 V on a hot afternoon in our example — below the 150 V MPPT minimum. Add a panel, use a microinverter, or leave that face empty.

6. Sizing the inverter to an east–west array's nameplate

   An east–west array never reaches its nameplate at once, so a smaller inverter clips very little. Stay within the datasheet's PV-input limit and let the orientation do the rest.