One equation, and the reason it exists
Voc cold = Voc at STC × (1 + β × (Tmin − 25))
β is the temperature coefficient of open-circuit voltage in percent per degree Celsius, expressed as a decimal in the equation, and it is negative. Tmin is the coldest cell temperature the array will see, in Celsius. Because both terms are negative, their product is positive, and the voltage goes up.
Take a 49.5 volt module with a coefficient of −0.28 percent per degree at a record low of −18 °C. The swing from the 25 °C rating point is −43 degrees. Multiply by −0.0028 and you get +0.1204, so Voc cold is 49.5 × 1.1204, which is 55.5 volts. Ten modules that measure 495 volts on a mild afternoon will present 555 volts on that morning, and against a 600 volt input limit the margin has gone from comfortable to 45 volts.
Silicon photovoltaics behave this way because the band gap widens slightly as the cell cools, which raises the voltage the junction can develop. Current barely moves. This is not a defect and it is not avoidable; it is what the material does.
Why this is the check that gets skipped
Everything about the commissioning environment hides it. The array is installed in fair weather. The meter reads a plausible number. The inverter starts, produces, and reports no faults. The condition that breaks it may not occur for months, and when it does the array is at its coldest and brightest simultaneously, because the cold clear morning that produces the record low is the same morning with no cloud between the array and the sun.
The other reason is that the failure is quiet and terminal. Input capacitors and switching devices in an inverter or a charge controller are rated to a hard ceiling. Above it there is no derating and no protective shutdown to rely on; there is a component that has been over-stressed, and the resulting failure is usually excluded from warranty because the design fault is visible in the string layout. Nobody gets a second attempt at this calculation.
Choosing the record low, and choosing the cell temperature
Two temperatures matter and they are not symmetrical.
| Which end | What to use | Why |
|---|---|---|
| Cold, for Voc | The extreme minimum for the site, not an average winter low | Voc peaks at sunrise before the module has warmed up, so cell temperature equals air temperature. Extreme-minimum datasets exist for this purpose; a nearby station record is the common stand-in. |
| Hot, for Vmp | Cell temperature, roughly ambient plus 25 to 35 °C | A module in still summer sun runs far above air temperature. Flush roof mounts with little airflow behind them run hottest; ground-mounted and standoff arrays run cooler. |
Notice the asymmetry: for the cold end you use air temperature directly, because at dawn the module has been radiating to the night sky and is at or slightly below ambient. For the hot end you must add the temperature rise, because a module in full sun is nowhere near air temperature.
The other end of the window
Strings can also be too short. A maximum power point tracker needs the string voltage to stay above its window floor to operate, and Vmp falls with heat for the same reason Voc rises with cold. A string that sits comfortably above the floor in March can drop below it in August, and the symptom is a string that stops producing in the middle of the best solar day of the year. That is why the calculator reports a minimum as well as a maximum, and why the useful answer is a range of string lengths rather than a single number.
When the range is empty — the minimum exceeds the maximum — the module and the equipment are genuinely incompatible at your temperature extremes. No arrangement of panels fixes it, and pretending otherwise by ignoring one end is how the first failure gets built.
What this settles and what it does not
This page settles voltage. It does not settle current, conductor sizing, overcurrent protection, disconnect location, grounding, rapid shutdown, module-level electronics, or whether the roof can carry the array. String-level optimisers and microinverters change the voltage question entirely, because each module or pair is managed separately and the series string voltage is either fixed by the electronics or absent. If your design uses them, the string limit is whatever the manufacturer specifies for their product and this calculation does not apply.
Once the string length is settled, the current side belongs on the inverter and charge controller sizing calculator, the array total on the solar panel array sizing calculator, and the conductor run on the wire size calculator or the voltage drop calculator. A PV array is energized whenever light falls on it. There is no switch on the panel that turns it off, and a module lying face up on the grass is a live source. DC arc faults do not self-extinguish the way an AC arc does, because DC current never passes through zero. A battery bank can deliver thousands of amps into a short circuit without any warning noise, and a dropped wrench across two terminals is a serious burn and fire hazard rather than a spark. Lithium cells that have been damaged, punctured, or charged below freezing can fail violently, and the failure is not always immediate. Nothing on this page is a wiring, commissioning or battery-assembly procedure, and none of it should be used as one.
Anything that ties a solar system to a building supply or to the grid is decided by your Authority Having Jurisdiction and your utility, not by a calculator. Interconnection, transfer switching, backfeed protection, disconnect placement, grounding, rapid shutdown and conductor sizing all fall to them, and in most places the work is licensed. Net metering terms, export credit rates and incentive programmes differ by state, by utility and by year, so treat nothing here as a statement of what yours will do.
Questions people ask
Where do I find the temperature coefficient of Voc?
On the module datasheet, in the electrical characteristics block, usually written as a percentage per degree Celsius alongside coefficients for Isc and Pmax. Crystalline silicon modules are commonly in the range of -0.25 to -0.35 percent per degree, and the coefficient has improved slowly over the years as cell technology has changed, so an old rule of thumb applied to a new module will be pessimistic. Some datasheets give it in millivolts per degree instead: divide that by the module Voc in millivolts and multiply by 100 to convert. If a datasheet does not state it at all, that is a reason to be suspicious of the datasheet rather than a reason to guess.
Can I just use the number of modules the inverter manual suggests?
Only if the manual states the module it assumed and the minimum temperature it assumed, and both match yours. Manufacturers publish string tables that are correct for their assumptions, and those assumptions are frequently a moderate climate and a mid-range module. A colder site or a higher-voltage module moves the answer by one or two modules, and one module is the entire margin in many designs. Run the arithmetic with your own module and your own record low, then compare against the table. If they disagree, find out why before building.
What happens if the string exceeds the maximum input voltage?
Nothing, until it does. The equipment operates normally at every temperature above the threshold, so the array can run for a full summer looking healthy. When the cold morning arrives the input stage sees a voltage its components are not rated for, and the result ranges from a fault code and a dead MPPT channel to a failed inverter. It is a hard limit rather than a derating point, which is why the specification calls it absolute maximum. It is also a design error visible from the string layout, so warranty coverage is rarely forthcoming.
Does the maximum apply to Voc or to operating voltage?
To open-circuit voltage, which is the higher of the two and is what the string presents whenever it is illuminated and not delivering current — before the inverter starts in the morning, during a fault, after a shutdown, while the array is being worked on. That is precisely the moment the cold-morning maximum occurs, which is what makes the combination dangerous. The MPPT window, by contrast, describes operating voltage while the tracker is running, and it has both a floor and a ceiling that behave quite differently: below the floor you lose production, above the ceiling you usually lose a little production, and above the absolute maximum you lose equipment.
Do microinverters and optimisers have the same problem?
Not in the same form. With microinverters each module has its own conversion stage, so there is no long series string and no accumulating Voc; the constraint becomes how many units can share an AC branch, which the manufacturer specifies. Power optimisers keep a DC string but manage its voltage electronically, and the maximum string length comes from the optimiser specification rather than from module Voc arithmetic. In both cases the manufacturer figures govern and this calculation does not. What does not change is that the modules themselves are still live in daylight, and the wiring between module and electronics is still DC.