String Sizing for Solar Inverters: The Temperature Mathematics of Voc, Vmp and MPPT Windows

A module datasheet gives voltage at a cell temperature of 25 °C. An inverter datasheet gives a maximum DC voltage, a window in which it tracks the maximum power point, and a narrower window in which it delivers full power. Your roof is almost never at 25 °C. String sizing fits one set of numbers inside the other across every temperature the plant will see.

The errors are not symmetric. Too many modules in series, and a cold morning lifts the open-circuit voltage above the inverter's absolute limit, which can destroy it. Too few, and a hot afternoon pulls the string below the range where the inverter delivers rated power. This guide derives both limits from the temperature behaviour of a silicon cell, then covers current limits, reverse current and fusing, and the DC/AC ratio with its clipping loss.

Every module and inverter figure below is an illustrative typical datasheet value, not a recommendation. Use your own datasheets and site temperatures. If the I-V curve is new to you, read how a solar module works first. The wiring side of the same design is in solar cable sizing and voltage drop.

Series and parallel rules for an array

The module is the smallest unit here. Symbols for this section:

• $N_s$: modules in series in one string; $N_p$: identical strings in parallel (both dimensionless).
• $V_{\mathrm{mod}}$, $I_{\mathrm{mod}}$: voltage (V) and current (A) of one module.
• $V_{\mathrm{str}}$, $I_{\mathrm{str}}$: voltage and current of one string. $V_{\mathrm{arr}}$, $I_{\mathrm{arr}}$, $P_{\mathrm{arr}}$: voltage (V), current (A) and power (W) of the whole array.

Assume identical modules, the same irradiance and cell temperature on all of them, and no wiring loss. Mismatch, shade and cable resistance break these assumptions slightly. Kirchhoff's current law gives every module in a series chain the same current, and his voltage law makes the voltages add:

Strings joined at the same two terminals share one voltage, and their currents add at the junction:

The array power is therefore the sum of all module powers:

Worked example. Take 19 modules per string and 2 strings of the 550 Wp module, whose maximum-power-point voltage $V_{mp}$ is 41.8 V and current $I_{mp}$ is 13.16 A. The string voltage is 19 × 41.8 V = 794.2 V, the array current is 2 × 13.16 A = 26.32 A, and the power is 794.2 V × 26.32 A = 20,903 W. Sanity check: 38 modules × 550 W = 20.9 kW.

How temperature and light move module voltage

Symbols for this section:

• $T_c$: cell temperature in °C. $T$: the same temperature in kelvin, $T = T_c + 273.15$.
• Voc, Vmp, Isc: open-circuit voltage, maximum-power voltage and short-circuit current. The suffix STC means 25 °C cell temperature and 1,000 W/m².
• $\beta$, $\gamma$, $\alpha$: temperature coefficients of Voc, of maximum power and of Isc, each in 1/°C. Datasheets quote them in %/°C.
• $G$: irradiance on the module plane (W/m²). $n$: diode ideality factor, about 1.0 to 1.2. $N_c$: cells in series in the module.
• $k = 1.380649\times10^{-23}\,\mathrm{J/K}$ and $q = 1.602176634\times10^{-19}\,\mathrm{C}$: Boltzmann constant and elementary charge. $E_g$: silicon bandgap energy (J), so $E_g/q$ is a voltage.

Where the temperature coefficient comes from

At open circuit, the light-generated current $I_{ph}$ flows entirely through the cell's own junction diode. Solving the diode equation for voltage gives, per cell:

$I_0$ is the diode's reverse saturation current, the leakage that grows as the cell warms. For silicon, with $B$ a constant:

Substituting, with $n = 1$ to keep the algebra short:

Differentiate with respect to $T$, treating $I_{ph}$ as constant because its own change is small, and replace the logarithm term using the equation itself:

The sign is negative: a warmer junction leaks more, so less voltage balances the photocurrent. Our module has 144 half-cut cells as two parallel halves of 72, so $N_c = 72$ and the cell voltage is 49.6 V / 72 = 0.689 V. With $E_g/q \approx$ 1.2 V (the value extrapolated to 0 K) and $T$ = 298.15 K:

Over 72 cells that is −0.142 V/K, or −0.286%/K of 49.6 V, within 6% of a typical datasheet value of −0.27%/K. In practice use the datasheet figure, which was measured on the real product.

The linear translation equations

Across the design range, voltage is close to a straight line in temperature:

Temperature differences in °C and kelvin are equal, so the brackets need no conversion. With $\beta$ = −0.27%/°C, the Voc factor at 10 °C is 1 + (−0.0027)(10 − 25) = 1.0405.

How good is a straight line? The full expression has slight curvature from the $T \ln T$ term. Against a line with the same slope at 25 °C, the full model differs by 0.01% at 10 °C, 0.04% at 0 °C and 0.14% (0.06 V) at 70 °C. That is far below the uncertainty in the design temperatures, so the linear model is sound.

The Vmp coefficient. Datasheets list $\beta$, $\alpha$ and $\gamma$, but rarely a Vmp coefficient $\beta_V$. Since $P_{mp} = V_{mp} I_{mp}$, taking logarithms and differentiating gives $\gamma \approx \beta_V + \beta_I$, where $\beta_I$ is the small positive Imp coefficient. So $\beta_V$ is slightly more negative than $\gamma$. This guide uses $\gamma$ = −0.34%/°C for Vmp as a stated approximation and tests its effect below. Use a listed Vmp coefficient if you have one.

How light changes voltage

Short-circuit current is proportional to irradiance, so $I_{ph}$ scales with $G$. Define the module thermal voltage $V_{t,mod} = n N_c kT/q$. Writing the open-circuit equation at $G$ and at 1,000 W/m² and subtracting, $I_0$ cancels:

For a cell at 10 °C with $n$ = 1.1 (assumed), $kT/q$ = 0.02440 V, so $V_{t,mod}$ = 1.1 × 72 × 0.02440 V = 1.932 V. Starting from Voc = 51.61 V at 1,000 W/m²:

Sanity check: a five-fold drop in light costs only 6% of the voltage. A module wakes up at nearly full voltage long before it makes much current, and that is the key to the cold-morning limit.

The cold limit: how many modules in series

The inverter's maximum DC voltage, $V_{dc,max}$, is a hard limit, and exceeding it even briefly can damage the input stage. An unloaded string at open circuit sits above its working voltage, so Voc is the figure that must stay inside the limit.

Why the coldest morning? Overnight the cells cool to ambient, sometimes below it. At sunrise the inverter has not yet started, so the strings sit at open circuit while irradiance climbs. The table above shows that at 200 W/m² the voltage is already within about 6% of its full-light value, while the cells have warmed by only a few kelvin. The coldest morning therefore gives the highest string voltage of the year.

The design minimum temperature $T_{min}$ is a site decision. Use the lowest ambient temperature in IMD climate data for your location, with a margin for unusual years. For a Tirupati-area example this guide uses a conservative, illustrative 10 °C. It is not a quoted record. The condition is that the cold string stays within the limit:

Dividing and rounding down to a whole module gives the longest permitted string:

Worked example. At 10 °C, Voc = 49.6 V × 1.0405 = 51.61 V. With $V_{dc,max}$ = 1,100 V, 1,100 / 51.61 = 21.31, so $N_{max}$ = 21. Twenty-one modules give 1,083.8 V, which is 98.5% of the limit. The module's own maximum system voltage, commonly 1,500 V, is a separate check and does not bind here. A colder site tightens the result:

Datasheet Voc also carries a tolerance. Adding 3% at 10 °C gives 1,100 / (51.61 × 1.03) = 20.69, so $N_{max}$ falls to 20.

The hot limit: minimum string length and the MPPT window

On a hot afternoon, cell temperature rises, voltage per module falls, and a short string can drop below the range where the inverter tracks or delivers rated power. Symbols:

• $T_a$: ambient air temperature (°C). NOCT: nominal operating cell temperature (°C), a datasheet value measured at 800 W/m², 20 °C ambient and light wind. About 45 °C is typical.
• $T_{c,max}$: design maximum cell temperature (°C). $N_{min}$: the shortest permitted string. $V_{lo}$, $V_{hi}$: the lower and upper voltage limit being tested (V).
• $V_{mppt,min}$, $V_{mppt,max}$: limits of the inverter's MPPT range (V). $V_{full,min}$, $V_{full,max}$: the narrower range where it delivers full rated power (V). $V_{start}$: the voltage needed before it starts (V).

An energy balance puts the cell above the air by an amount proportional to irradiance, and NOCT calibrates the constant:

Worked example. For $T_a$ = 40 °C, $G$ = 1,000 W/m² and NOCT = 45 °C, $T_c$ = 40 + (25/800) × 1,000 = 71.25 °C. This guide therefore uses an illustrative $T_{c,max}$ of 70 °C. Close-mounted modules can run hotter, and the NOCT formula is good to only a few kelvin.

The voltage that matters now is the working voltage at the maximum power point, which the tracker holds the array near. At 70 °C:

Requiring the hot string to stay at or above a lower limit $V_{lo}$, and rounding up, gives the shortest string:

For the bottom of the MPPT range, then the bottom of the full-power range:

Below 550 V the inverter may still track but cannot deliver full power, so the second figure protects yield. To test the Vmp approximation, use $\beta_V$ = −0.39%/°C ($\gamma$ minus an assumed Imp coefficient of +0.05%/°C). Then Vmp(70) = 34.46 V, and 550 / 34.46 = 15.96 still rounds up to 16. The answer holds, with almost no margin.

The upper limits use Vmp at the cold end, $\mathrm{Vmp}(10)$ = 41.8 × [1 + 0.0034 × 15] = 43.93 V:

Last is start-up. At 25 °C and 100 W/m², Voc is 49.6 V + 2.035 V × ln(0.1) = 44.92 V, where 2.035 V is $V_{t,mod}$ at 25 °C. For $V_{start}$ = 250 V, 250 / 44.92 = 5.57, so at least 6 modules. For lengths that fit the full-power window, start-up never binds. How the tracker finds the maximum inside this window is covered in the MPPT guide. Combining every condition:

That is max(6, 16, 6) = 16 up to min(21, 22, 19) = 19. The valid range is 16 to 19 modules, and the full-power window, not the 1,100 V limit, sets the upper end.

Current limits, bifacial gain and reverse current

Short-circuit current rises slightly with heat, by the coefficient $\alpha$. Define $I_{sc,max}$ (A) as the highest short-circuit current a module delivers in design conditions, and $b$ (dimensionless) as the fractional extra current from light on a bifacial module's rear face. An illustrative $b$ is 0.05 to 0.10, depending on roof reflectance and mounting height:

With $\alpha$ = +0.05%/°C, a hot cell gives 14.0 × (1 + 0.0005 × 45) = 14.315 A. A 10% rear gain brings $I_{sc,max}$ to 14.315 × 1.10 = 15.75 A. Irradiance briefly exceeds 1,000 W/m² at cloud edges, so cable and protection design adds a further margin, covered in the cable sizing post.

The per-input current check

Each MPPT input has a maximum input current $I_{mppt,max}$. If $N_{str}$ strings share an input, their currents add:

Here two strings carry 31.49 A (from the unrounded 15.7465 A) against 32 A. It passes with 1.6% to spare, and a rear gain above 11.8% would breach it. Many inverters list a separate, higher short-circuit rating; this example conservatively uses the 32 A figure. The operating current is lower: 2 × 13.16 A = 26.32 A.

Reverse current in parallel strings

Suppose one of $N_p$ parallel strings develops a fault, such as a cable shorted to earth. Its voltage collapses, so it no longer opposes the shared bus voltage. The healthy strings now drive current through it. A healthy string's I-V curve is nearly flat at $I_{sc}$ at low voltage, so each delivers about $I_{sc,max}$ into the fault. The reverse current $I_{rev}$ (A) is therefore:

This current flows backwards through the faulted string's modules. Each module datasheet gives a maximum series fuse rating $I_{fuse,max}$, often 20 to 30 A. Taking 25 A as an illustrative value, no extra protection is needed only if:

For $I_{sc,max}$ = 15.75 A, 25 / 15.75 = 1.59, so $N_p$ is at most 1 + 1 = 2.

This is the usual reasoning for string protection at three or more parallel strings. The exact rule depends on the module datasheet and the standard, so check the current edition of IEC 62548.

DC/AC ratio and clipping

Modules rarely deliver their rated power, because heat, soiling, wiring loss and sun angle cut output for most of the day. Designers therefore install more DC capacity than the inverter's AC rating. Symbols:

• $P_{dc}$: installed module capacity at STC (W). $P_{ac}$: inverter rated AC output (W).
• $r$: the DC/AC ratio. $s = 1/r$: its inverse (both dimensionless).
• $T_d$: day length from sunrise to sunset (h). $t$: time since sunrise (h). $x = \pi t/T_d$: the same instant as an angle (rad).
• $E_{day}$, $E_{clip}$: DC energy available in a day, and the part above $P_{ac}$ (Wh).

Model a clear day as a half-sine. Assume irradiance follows a sine of time, constant module temperature, no losses, and an inverter that caps output exactly at $P_{ac}$. These are strong simplifications, so the result is an upper bound:

The day's energy is the area under the half-sine. With $x = \pi t/T_d$, so $dt = (T_d/\pi)\,dx$:

Clipping starts where the sine reaches $P_{ac}$, at angle $x_1$, and by symmetry ends at $\pi - x_1$:

The clipped energy is the area between the sine and the flat line $P_{ac}$ over that interval. The integral of $\sin x$ gives $2\cos x_1$:

Dividing by $E_{day}$ gives the clipped fraction $f$. Using $\cos x_1 = \sqrt{1 - s^2}$ and $\pi/2 - x_1 = \arccos s$:

Worked example, $r$ = 1.2. Here $s$ = 0.8333, so $\sqrt{1 - 0.6944}$ = 0.5528 and arccos(0.8333) = 0.5857 rad. Then $f$ = 0.5528 − 0.8333 × 0.5857 = 0.0647, or 6.5% of the clear-day DC energy. Check: at $r$ = 1, $s$ = 1 and $f$ = 0, as it should. A numerical integration of the half-sine agrees with the closed form to three decimal places.

Why the annual loss is much lower

The ideal-sine column overstates real clipping for three reasons. First, DC peak never reaches the STC rating. At 70 °C the module power factor is 1 + (−0.0034)(45) = 0.847, and wiring, soiling and mismatch take a few percent more. For the 62.7 kWp array of the worked example, the peak is about 62.7 × 0.847 = 53.1 kW, or roughly 50.5 kW after 5% further losses. Second, most hours are below peak. Third, many days are hazy or cloudy.

The fifth column applies the first effect alone. The last column adds the others through a toy mix of days: 35% peaking at 85% of rating, 35% at 60% and 30% at 30%, each weighted by its energy. It is only an illustration, since an hourly yield simulation is the proper tool. The trend is clear: modest oversizing costs little energy and adds output in weak-light hours, which is why 1.2 to 1.3 is common. The system sizing guide and the inverter overview cover the trade-off in plainer terms.

Worked example: a 50 kW three-phase inverter

All values below are illustrative typical datasheet figures.

• Module, 550 Wp: Voc 49.6 V, Vmp 41.8 V, Isc 14.0 A, Imp 13.16 A; $\beta$ = −0.27%/°C, $\gamma$ = −0.34%/°C (used for Vmp), $\alpha$ = +0.05%/°C; assumed rear gain 10%.
• Inverter, 50 kW AC: maximum DC 1,100 V; MPPT range 200 to 1,000 V; full-power range 550 to 850 V; start-up 250 V; four MPPT inputs of 32 A each, two strings per input.
• Site: $T_{min}$ = 10 °C and $T_{c,max}$ = 70 °C, both illustrative.

Step 1: the string length window

The module values at the extremes are Voc(10 °C) = 51.61 V, Vmp(10 °C) = 43.93 V, Vmp(70 °C) = 35.40 V and $I_{sc,max}$ = 15.75 A, as derived above. Every condition then gives one bound:

Step 2: choose a layout

For a DC/AC ratio near 1.25, the array needs 1.25 × 50 kW / 0.55 kWp = 113.6 modules. Six strings of 19 give 114 modules and 62.7 kWp, so $r$ = 62.7 / 50 = 1.254. Put them on three MPPT inputs, two strings each, and keep the fourth free for expansion or a differently oriented roof face. Eight strings of 16 would give 70.4 kWp and $r$ = 1.408, with 14.9% ideal clear-day clipping and a thin hot margin.

Step 3: check the results

Common mistakes and what to do next

The derivations reduce to a short checklist. Fix $T_{min}$ from your site's recorded low and $T_{c,max}$ from your mounting. Translate Voc, Vmp and Isc with datasheet coefficients. Find the window from the cold ceiling and hot floor, with margin at both ends. Check per-input current with rear gain, and reverse current for every group of parallel strings. Choose the DC/AC ratio from a site yield simulation, not from the clear-day sine alone.

This calculation is routine, but it needs real datasheets and site data. Ray2Volt can run it for your plant: send the module and inverter datasheets with your roof layout, and our engineers will check the string design as part of a commercial solar design, or book a free power audit and we will start from your site. Figures here are illustrative; confirm every limit against your own datasheets and the current standards.

Frequently asked questions

Ray2Volt Solar Private Limited

We design, install, and service rooftop solar for homes and businesses across Tirupati district and Andhra Pradesh, from PM Surya Ghar residential systems to commercial and industrial plants. Every enquiry starts with a free power audit.