Microinverter vs String Inverter: What Your Roof Decides

The number does a quiet trick on the way from NREL's report into a sales pitch. The report says module-level electronics recovered 25% to 35% — of the shading loss. By the time it reaches a proposal it reads like 25% more electricity. If shade costs your roof 15% of its annual sun, the recoverable slice is about 4 to 5 percentage points of production. Real money, worth computing. Not a quarter of your array.

This post does three things with primary documents: pins down what each of the three architectures physically is, reads the NREL shading testbed the way it was written, and then shows why on many roofs the electrical code — not the shade — already made the choice. Every datasheet and report cited was downloaded and read on 2 September 2026 unless a different date is marked.

One series circuit, three places to break it

All three architectures start from the same problem: panels wired in series carry one current, and the weakest panel sets it.

A string inverter leaves the series circuit intact. Ten or twelve modules feed one maximum power point tracker (MPPT) in a box on the wall. Modern residential units carry more than one tracker — the SMA Sunny Boy 7.7-US datasheet (document SBxx-US-DS-en-41) lists three MPPT trackers with one string per tracker, plus a shade-management algorithm the sheet describes as "ShadeFix technology for string level optimization." String level. Within any one string, the mitigation hardware is the module's own bypass diodes, and NREL's testbed procedure spells out what those protect: "In a typical 60-cell module, each one-third of the module is protected by a single bypass diode." Shade one third and the diode routes current around it — you lose the third, not the string. Diode counts vary by module design, so the number for your panel is on its datasheet, not in a rule of thumb.

A DC optimizer system keeps the wall inverter but bolts a DC-to-DC converter under each module. Each converter runs its own MPPT, so a shaded module delivers what it can without dragging its neighbours' current down. SolarEdge's S-series datasheet for North America (DS-000018-NA, dated 1 March 2023) rates the S440 at 440 W input, 99.5% maximum and 98.6% weighted efficiency — a second conversion stage that is nearly, not perfectly, free, and one that multiplies with the wall inverter's own efficiency. A note on that link: solaredge.com would not serve its current file to the tools used here on 2 September 2026 — the download came back as a browser-check page — so the figures above are from the archived March 2023 revision and are labelled with it. If an S-series optimizer is on your quote, pull the current datasheet in an ordinary browser and check the revision date against this one.

A microinverter deletes the DC circuit entirely. Each module gets its own full DC-to-AC inverter; the roof wiring is ordinary AC. Enphase's IQ8 and IQ8+ datasheet (DSH-00207-5.0-EN-2025-08-14) puts the IQ8+ at 300 VA peak, 290 VA continuous, 97% CEC weighted efficiency, with "commonly used module pairings" of 235-440 W.

Granularity is the real spec in all of that. A microinverter designs down to one module. The S440 needs company: the same datasheet's design table requires at least 8 optimizers per string on single-phase inverters, at most 25. A string inverter needs enough modules in series to reach its MPPT window — 100-550 V on the Sunny Boy — per orientation. If your roof is four small planes and a dormer, that single column decides more than any efficiency figure.

Read the 25-35% the way the testbed wrote it

The document behind most shade-mitigation claims is NREL's Photovoltaic Shading Testbed for Module-Level Power Electronics: 2016 Performance Data Update (NREL/TP-5J00-62471, September 2016 — hosted at docs.nlr.gov since the lab was renamed in December 2025). Two identical arrays sit side by side, one with the device under test, one with a plain string inverter, and calibrated mesh shades both through 20-30 configurations. Three findings matter for a homeowner:

The shade bands are defined in irradiance, not vibes. "Light" shading is a 7% annual irradiance reduction, "moderate" is 15%-19%, "heavy" is 25%. When a shade report or a proposal tool gives you a percentage, you can place it on this scale.

The headline result is a recovery fraction. "The annual recovery of shading losses from the module-level electronics evaluated is 25%–35%, with the major difference between different trials being related to the number of parallel strings in the test installation rather than differences between the equipment tested." The equipment tested came from Enphase, SolarEdge, Maxim, and SMA — and the layout of the array moved the result more than the logo did.

The output plugs straight into a production model. The report says its shade-mitigation score "can be used directly in annual performance estimation software such as PVWatts or PVSyst." So use it that way. The PVWatts walk-through on this site built a reference roof — 7.2 kW in Denver, 11,473 kWh a year at the documented defaults — and the arithmetic below reuses it so you can swap in your own numbers.

PVWatts' default 14.08% system loss already contains 3% shading. Losses multiply, so first strip the shading term out to isolate the others:

non-shade loss factor = (1 - 0.1408) / (1 - 0.03) = 0.8592 / 0.97 = 0.8858

production at shade loss s:
  kWh(s) = 11,473 x (0.8858 x (1 - s)) / 0.8592

string system, s = 15% (moderate band):        10,054 kWh
MLPE recovers 25% of it, s = 11.25%:           10,497 kWh   (+443)
MLPE recovers 35% of it, s = 9.75%:            10,675 kWh   (+621)

So on this reference roof, with genuinely moderate shade, module-level electronics are worth 443 to 621 kWh a year. At an illustrative $0.20 per kWh that is $89 to $124 a year — replace the rate with the marginal rate off your own bill, and replace 11,473 with your own PVWatts run, before treating the dollars as yours. Then set the result against the actual price difference between the two quotes in your hand, not a published average: the delta is on the equipment lines, and a proposal that itemizes properly lets you read it off directly.

Run the same formula at the light band and watch it deflate: s = 7% gives 11,000 kWh for the string system, and 35% recovery lifts it to 11,290 — a gap of 290 kWh. At 3%, roughly a clean roof with the default soiling of real life, the recoverable slice rounds toward nothing.

Which is the honest summary: the shading benefit is real, proportional to your shade, and computable in four lines. Nobody's brochure gets to skip the multiplication.

On an unshaded roof, the code chose before you did

If the shade arithmetic above comes out small, you might expect the architecture choice to reopen. Mostly it does not, because a fire-safety rule already made it.

NEC 690.12 — rapid shutdown — exists so that a firefighter on your roof is not standing over live high-voltage DC. The requirement tightened in steps, laid out with section numbers in IAEI Magazine's history of 690.12 and UL 3741 (read 2 September 2026): the 2014 NEC controlled conductors beyond 10 feet from the array; the 2017 NEC drew an "array boundary" one foot around the array; and as of 1 January 2019, conductors inside that boundary must be controlled "to not more than 80 volts within 30 seconds of rapid shutdown initiation" — section 690.12(B)(2)(2), carried forward through the 2020 and 2023 editions.

A bare series string of ten modules cannot get under 80 V by itself. In the IAEI article's words, this is when the industry saw "a proliferation of module-level power electronics." The compliance is printed on the datasheets quoted above. Enphase: "UL Listed as PV rapid shutdown equipment and conforms with NEC 2014, NEC 2017, NEC 2020, and NEC 2023 section 690.12." SolarEdge: safety output voltage of "1 ± 0.1" volts per optimizer in standby — with a footnote that a string of more than 30 optimizers no longer meets the 30 V requirement outside the boundary. Even the string inverter's page concedes the point: SMA's accessories list includes "SunSpec Certified Rapid Shutdown Receivers," one per module, and Tigo's TS4-A-F sells as a "dedicated rapid shutdown device for 1 solar module" — a box that shuts down and does nothing else. A string system on a 2017-or-later code roof usually still carries per-module electronics; they are just dumber ones.

There is one genuine exit: a PV hazard control system listed to UL 3741, where a specific inverter, specific racking, and specific wire management are certified together and no module-level device is required. The standard was published in December 2020, and the listings so far cluster around commercial hardware — SMA's UL 3741 page (read 2 September 2026) is built around its Sunny Tripower CORE1 and Tripower X, both commercial three-phase units. For a house, in practice, the menu today is module-level something.

Two checks, then, before shade even enters your decision. First, which NEC edition your jurisdiction has adopted — 2014-era rules are looser, and adoption varies by state and sometimes by city, so ask the installer which edition the permit will be reviewed under and double-check with the local building department. Second, ground-mount systems sit outside 690.12's building-focused scope, which is why ground-mount quotes still feature plain strings.

The 440-watt module and the 290 VA ceiling

One more line deserves your calculator before you sign: the pairing ratio. Enphase's sheet pairs the IQ8+ (290 VA continuous) with modules up to 440 W and states, in footnote 5, "No enforced DC/AC ratio." That is a per-module DC-to-AC ratio of 440 / 290 = 1.52, where the PVWatts documentation's default is 1.2. Real modules rarely hit lab-rated watts on a roof, so headroom above 1.0 is normal and usually efficient — but output above the ceiling in cold, bright conditions is clipped, and how much that costs depends on your climate and tilt, not on a universal percentage. The optimizer side has its own version of the line: the S440's footnote 1 requires that module STC power not exceed the optimizer's 440 W rating, with +5% tolerance modules allowed. The ask for your installer is one sentence: did the production model use the actual pairing ratio on this quote, or a default?

And since the electronics multiply, so do the failure points and the paperwork. A string inverter is one device at ground level with, on the SMA sheet, a 10-year standard warranty; a microinverter roof is twenty-odd devices behind racking with a 25-year term; SolarEdge splits the difference at 12 years on the wall unit and 25 on the optimizers. Longer terms on harder-to-reach hardware, shorter on the easy box — and none of those terms include the labour to climb up. That trade runs on document numbers and exclusions, and it already has its own post.

The one thing to do with the quote in front of you: find the exact inverter or optimizer model numbers on the equipment page, pull each datasheet, and write three numbers in the margin — continuous AC rating against your module's wattage, the rapid shutdown compliance line with its NEC editions, and the warranty term. If any of the three is missing from the datasheet the installer can name, the architecture discussion is happening one document too early.