Cold panel voltage: why 2 in series can hit 109V

Why voltage goes up when it gets cold, not down

Everything else about a cold panel gets worse: current sags a little, efficiency drops, output power falls. Voltage is the one number that moves the wrong way. A solar cell's open-circuit voltage is a function of the semiconductor band gap, and that band gap widens as the cell cools — the physics is the same reason silicon diodes shift with temperature. The practical result: every crystalline silicon panel's datasheet lists a temperature coefficient of Voc, almost always between about −0.27%/°C and −0.35%/°C, meaning voltage rises roughly a third of a percent for every degree below the 25°C the panel was tested at.

That sounds small. A single panel rated 48.2 V at 25°C reads 54.4 V at −20°C — a 12.9% increase. Wired in series with a second panel, the string's voltage moves by the same percentage, but the absolute gap doubles: 96.4 V becomes 108.8 V, a 12.4 V overshoot on a controller rated for 100 V.

The formula

V(t) = V25 × (1 + coefficient × (t − 25))

V₂₅ and the coefficient come straight off the panel's datasheet — nothing here is estimated. t is the temperature you actually need to survive, and this is where most string calculations go wrong: people plug in a mild winter morning, not the coldest night the location actually sees. Two numbers matter, not one:

• Design low — the average of the single coldest night each year, over a multi-year record. This is what a string should be sized against day to day.
• Record low — the coldest night ever measured there. A string that fails only at the record low, not the design low, is a known and disclosed risk, not a design defect — but it should be a choice, not a surprise.

Multiply the per-panel result by the number of panels in series, and compare it to the controller's maximum PV input voltage — an absolute limit, not a soft one. Above it, a controller's input protection either shuts down (the survivable outcome) or, on cheaper hardware, doesn't (the outcome that ends with a burned MPPT stage and a warranty claim that gets denied because the string was wired outside spec).

A concrete example

Take a REC REC380TP2SM 72 (Voc 48.2 V, temperature coefficient −0.285%/°C) on a 100 V-limit MPPT controller — a common pairing, since 100 V is the most widely sold controller input class.

The crossover — the exact temperature where two panels stop being safe on this controller — works out to about 12°C. Below that, ordinary and unremarkable weather for most of the continental US in winter, the string exceeds the controller's rated input every single time the sun comes up cold. This is not a rare edge case reserved for the Arctic; it is the default state of a two-panel 48 V string on a 100 V controller anywhere that sees a normal winter.

The fix in this specific case is not a bigger controller — it's one panel per string instead of two (48.2 V label, 55.8 V at a brutal −30°C, comfortably under 100 V at any temperature this panel will plausibly see), wired in parallel instead of series to keep the same total wattage. Parallel strings don't add voltage; they add current, which has its own limit (the controller's rated charging current and the wire's ampacity) but never approaches the input-voltage failure mode series wiring does.

Why the coefficient itself varies by cell technology

Not every panel drifts by the same amount. Standard p-type monocrystalline cells typically sit around −0.30 to −0.35%/°C. N-type TOPCon cells — increasingly common on newer 12V portable panels, including several in this site's catalog at −0.25 to −0.28%/°C — run noticeably flatter, because the wider band gap and lower dark saturation current in an N-type cell make its voltage less sensitive to temperature in the first place. That difference sounds academic until it changes an answer: two otherwise-identical 48V-class panels, one at −0.35%/°C and one at −0.25%/°C, diverge by nearly 3 V per panel at −20°C — enough on its own to move a borderline string from safe to over the limit.

This is also why a coefficient can never be assumed or copied from a similar-looking panel. It is printed on the specific model's datasheet, changes between product generations of the same brand, and is the one number in this whole calculation that has to come from the manufacturer rather than a formula — everything else here follows mechanically once that one input is right.

A second example: healthy margin instead of a thin one

Not every string is this fragile. Take a 40 V-class panel with a flatter −0.32%/°C coefficient, two in series on the same 100 V controller: 80 V at label, and even at a brutal −30°C the string only reaches about 94.1 V — inside the limit, with a working margin instead of none. The difference between this case and the REC example above isn't the physics; it's the starting voltage relative to the controller's headroom. A 96.4 V label string on a 100 V controller has essentially no room to give before it gets cold; an 80 V label string on the same controller has 20 V of slack, which this specific coefficient can't burn through even at record-cold temperatures.

The lesson generalizes: the danger isn't the temperature coefficient in isolation, it's how little headroom the string leaves between its label voltage and the controller's limit. A string sized to leave at least 10–15% headroom at label conditions survives almost any realistic coefficient; a string sized to just clear the limit at label conditions is betting on a mild winter.

Design low vs. record low, and where those numbers come from

Every location page on this site derives design low the same way: the average of each year's single coldest night, over as many complete years of NOAA NCEI GHCN-Daily station data as are available (a minimum of 25 years, to smooth out any one unusually mild or harsh winter). Record low is the coldest night that station has ever measured, full stop — no averaging. Sizing a string against design low accepts that, in an unusually severe winter, the string might briefly exceed a controller's limit and trip its input protection for a night; sizing against record low accepts zero exceedances at the cost of either fewer panels per string or a higher-voltage controller. Neither choice is wrong, but it should be a choice made with both numbers in view, not just whichever one happens to be a colder-sounding tagline on a weather site.

Does this apply to PWM controllers too?

The same formula applies to a PWM controller's input just as much as an MPPT's, but PWM systems rarely get built with more than one panel in series in the first place: a PWM controller clamps the array's operating point down near battery voltage, so anything beyond a single panel in series (for a 12V battery) mostly wastes the panel's extra voltage as heat rather than converting it to charging current. Because of that, PWM stringing decisions tend to naturally stay far under any voltage limit that would matter here — the trap in this guide is specifically an MPPT problem, precisely because MPPT controllers are the ones designed to accept and use higher string voltages.

Try it with your own panel and controller

Enter a panel's Voc and temperature coefficient from its own datasheet, the controller's input limit, and a temperature to check.

What this means for wiring a string

Three practical consequences follow directly from the formula, not from a rule of thumb:

1. Use the design low, not the coldest day you've personally seen there. A string that has run fine for three winters can still fail on the fourth if that year's cold snap is worse than the previous three — this is exactly why design low is defined as a multi-year average of annual minimums, not a single observed number.
2. Check the record low too, and decide on purpose. If a string is safe at the design low but not the record low, that's a real, disclosed tradeoff — a rare hard freeze might trip the controller's protection for a day, which is inconvenient but not damaging on hardware with real overvoltage protection. Wiring past the limit even at the design low is a different, worse category of risk.
3. A higher-voltage controller (or fewer panels in series) is the only real fix. Deliberately undersizing the array to keep voltage down costs generation every single day the string operates normally, to guard against a handful of cold mornings a smarter wiring choice would have handled for free.

The panel catalog lists this exact calculation — series count at 25°C, −10°C and −30°C — for every panel against every controller in the database, so the check above isn't limited to the one example panel here.