Grounding and Bonding in Solar Racking: How the Path Is Formed, and What UL 2703, NEC 690 and IEC 62548 Require

A mounting system is a structure first. On the day the array is energized it is also a conductor, whether anyone designed it that way or not. Every module frame, rail, clamp, bolt and pile is metal, and metal in an array either carries fault current somewhere useful or sits at a voltage nobody planned. Which one happens comes down to a few square millimetres of contact inside each joint.

Metal to metal contact is not a bonding design. A compliant mounting system needs a deliberate, durable and verifiable electrical path from every exposed conductive part back to the project protective conductor. Two paths, in short form:

• Roof. Module frame, bonding clamp or earthing washer, rail, splice or jumper, foot or roof hook or seam clamp, grounding lug, then the EGC.
• Ground mount. Module frame, clamp, purlin or rail, a structural joint that was specified as a bond or a jumper across it, then an EGC run along the array to the earthing arrangement. On a tracker the same chain also has to survive bearings and pivots that move every day. We do not supply trackers. The bonding logic is the same, the fatigue problem is not.

What follows is how those paths are formed, where continuity is usually lost, and what UL 2703, NEC 690 Part V, IEC 62548-1, IEC 60364-7-712 and AS/NZS 5033 require. No universal torque value, conductor size, bonding resistance or earth electrode resistance appears anywhere in it, because no such number exists.

Grounding and bonding are two different jobs

Bonding connects metal parts to each other so they sit at the same potential. Grounding connects that bonded metal to earth. The equipment grounding conductor, the EGC, is the deliberate low impedance route back to the source so a fault gets cleared instead of waiting for someone to complete the circuit by hand.

PV makes this less forgiving than ordinary electrical work for one reason. A PV source circuit is a current limited source. A ground fault on the DC side often does not produce enough current to open anything. It produces a voltage on the array structure instead, and that voltage stays there as long as the sun is up. That is why the codes treat racking as safety equipment and not as a shelf.

There is a second reason. A ground mount is a large metal object standing in an open field. A rooftop array sits on the highest surface of the building. Even with no direct strike, an induced surge needs a bonded, low impedance frame to go somewhere.

The bond happens inside the joint, not on the drawing

Resistance across a metal to metal joint is not one number. Matrix Engineering breaks it into bulk resistance from the material, constriction resistance where current squeezes through microscopic contact points known as a-spots, spread resistance in the transition, and film resistance from oxides and corrosion products. Clamp load governs the whole thing, because it decides how many a-spots there are and how large they get.

Hence the sentence worth remembering from their work: a structurally sound joint is very likely a reliable electrical bond, and a loose one is not. They put roughly 44% of joint failures down to loosening, either relaxation, meaning pretension lost without rotation, or self loosening from dynamic load. A joint that loosens reduces contact area and breaks the seal keeping moisture out, which grows the film that raises resistance further.

Anodizing is the specific problem in aluminium racking. Matrix Engineering put the anodic layer at about 10 to 15 µm, with resistivity in the region of 10¹⁴ Ω·m, which makes it an insulator. It is applied precisely because it does not conduct and does not corrode. So a clamp that merely sits on an anodized frame is not a bond. The Solar ABCs module grounding study states the requirement plainly: the bonding connection shall penetrate nonconductive coatings, such as oxidized metal, clear coatings, paint, or vitreous enamel. Something has to cut through, and it has to stay cut through for the life of the plant.

This is where hardware choice stops being a detail. Generic external tooth star washers were widely used for years. The same Matrix Engineering review reports that they caused chronic loosening and module detachments by concentrating load and compressively deforming aluminium frames, and that in laboratory testing they showed erratic resistance readings and in some cases complete loss of continuity, against purpose built bonding washers.

Metallurgy at the bond matters just as much. The Solar ABCs work found that in salt mist most sample combinations corroded severely and failed within weeks, through galvanic, pitting and crevice mechanisms. Their compatibility screen kept combined electrochemical potentials below 0.6 V, and listed stainless containing at least 16% chromium as acceptable against aluminium alloys. We have written up the SS304 against aluminium pairing separately. Corrosion testing in this area is commonly run as 500 hour salt fog to ASTM B117.

Roof systems: more links than the drawing shows

On a pitched or flat roof the chain usually runs module frame, then mid or end clamp, then rail, then rail splice, then L foot or roof hook or seam clamp, then a lug, then the EGC back to the inverter and on to the building grounding electrode system.

Five places it breaks, in rough order of how often we see them raised on site:

• Clamp to frame. A clamp with no bonding pin resting on an anodized frame. It holds the module fine. It conducts nothing reliable. The part that does the electrical work is a purpose built earthing washer or a clamp with an integrated bonding pin. We supply earthing washers as mechanical components. We hold no bonding or grounding listing, UL 2703 or UL 467, so on a job under the NEC the listed bonding device has to come from whoever holds the listing.
• Rail splices. Two rails joined by an internal or external splice, often with a small designed gap for thermal movement. A sliding expansion joint is the opposite of a gas tight contact. Most systems need a bonding jumper across it, and a braided jumper rather than a solid one where the joint moves. A rail-less system does not remove this question. On SR-R2 the modules fasten to brackets screwed into the roof sheet, and an EPDM pad is glued in to seal that penetration, so the only metallic route to the sheet is the screw thread itself. That interface gets evaluated or it gets a jumper, and either way somebody has to decide which.
• Isolation pads. The EPDM or rubber pad under a foot that stops galvanic contact with the roof also stops current. That is by design, and it means the path has to be made elsewhere.
• Coated roofing. A clamp gripping a painted or coated standing seam is gripping paint. If the roof sheet is meant to be part of the path, the design has to say how the coating is penetrated.
• The last lug. A grounding lug landed on anodized rail without breaking the coating, or a lug not rated for aluminium. Grounding and bonding equipment is its own listed product category, UL 467 in North America.

Systems marketed as having integrated bonding have had their clamps and splices evaluated to carry the bonding current, so no separate jumper is needed between modules. Real, and it saves a lot of labour. The boundary that gets forgotten is that a UL 2703 listing is issued for a named mounting system tested with named module models. Change the module to a frame with different anodizing, flange thickness or clamping zone, and the listing does not travel with it.

Ground systems: fewer clamps, longer distances, harder joints

A ground mount reverses the problem. There are fewer module level clamps per kW, but the structure itself is longer, heavier and made of coated steel, and the array can run for hundreds of metres.

The structural joints are the first difference. Pile to beam and beam to purlin connections are bolted joints designed to transfer load, usually galvanized or ZAM coated. Both coatings are zinc based and conductive, but the joint still has slotted holes, shims, coating build up and a passivation layer. Such joints can make a good bond, and they are not automatically one, because nobody specified them as one. That is the distinction NEC 690.43(B) and (C) draw: structures listed and identified for bonding may bond PV equipment to the support, and using the structure itself as an EGC requires each separate section to be bonded with jumpers unless the structure is identified for equipment bonding.

The second difference is distance. Rows can sit tens of metres apart with no metallic connection between them at all. Most competent ground mount designs stop trusting a hundred structural joints in series and run a dedicated EGC along the array, bonding each row and each table to it. NEC 690.43(C) permits that conductor to run separately from the circuit conductors while it stays within the array.

The third difference is that the ground mount is already in the earth. Under NEC 250.52(A)(2) one or more metal in-ground support structures in direct contact with the earth vertically for 3.0 m, which is 10 ft, or more, with or without concrete encasement, qualify as a grounding electrode. Driven piles frequently meet that on their own. That does not remove the need for a designed earthing arrangement, and it does change what has to be added to get one. Our ground screw is the other case. SR-G5 runs 1200 to 2000 mm, so it never reaches the 3.0 m, and the electrode has to be designed in separately.

The fourth is movement. Single axis trackers rotate, so any bond crossing a bearing or a damper is a fatigue item. The 2026 NEC now calls for flexible, fine stranded conductors or braided straps for bonding on moving arrays. A solid jumper across a torque tube bearing will work on commissioning day and fatigue later.

The standards map, region by region

Scope notice before the standards map. What follows explains standards. It is not a claim that any Solaracks product is certified, listed or approved to any of them. Compliance attaches to a named model, never to a supplier. The seven questions at the end of this article are the list to send a supplier, ourselves included.

North America

Three documents sit under a mounting system here. UL 2703 is the racking level one, titled Mounting Systems, Mounting Devices, Clamping/Retention Devices, and Ground Lugs for Use with Flat-Plate Photovoltaic Modules and Panels. It is Edition 1, first issued 2015, and it is still being revised, with the latest revision published 28 April 2026. It covers systems up to 1500 V and evaluates the bonding and grounding path together with mechanical strength, fire classification and material suitability. UL 61730, harmonized with IEC 61730 in December 2017, covers the module, including its grounding provisions. UL 467 covers grounding and bonding equipment as components.

The number people ask about is the bonding path resistance limit. Under UL 1703, UL 61730 and UL 2703 the bonding path must not exceed 0.1 Ω, tested by passing twice the module series fuse rating through the joint and measuring the drop. UL 467 uses far larger currents for grounding equipment, sized to the conductor, in the order of 750 A for a #10 AWG conductor, against the same threshold. Two things follow. It is a laboratory value on a new, correctly torqued assembly, and the test current in the mounting system standard is modest, so passing it says nothing about what the joint does once the assembly has aged.

In the field the NEC governs, and the four subsections of 690.43 carry most of the weight. Devices that secure and bond module frames must be listed, labeled and identified for bonding. Metal support structures may bond equipment where identified for it. The EGC may run separately within the array, and must comply with 250.134 once it leaves the array vicinity. 690.43(D) disapplies the bonding bushing rules of 250.97 for metal raceways carrying PV DC conductors. Sizing comes from 690.45 via 250.122 and the overcurrent device rating, and 690.47 requires the building or structure supporting the PV system to have a grounding electrode system. In the 2023 edition 690.43(A) was shortened and gained an informational note, and 690.43(C) was retitled. Which edition applies is a jurisdiction question, not a national one.

The IEC world

The IEC set is structured differently, and the difference trips up suppliers shipping into both markets.

IEC 62548-1:2023, Photovoltaic (PV) arrays, Part 1: Design requirements, Edition 1.0 dated December 2023, consolidated as Edition 1.1 with Amendment 1 dated December 2025, holds the array level rules. Clause 7.3.2 covers earthing and bonding arrangements, with a decision tree for exposed conductive parts in Figure 13 and worked earthing configurations in Annex B. It hands off the protective conductor detail to IEC 60364-5-54.

IEC 60364-7-712:2017, Edition 2.0, is the installation rule. Clause 712.412.101 requires equipment used on the DC side to be Class II or equivalent insulation to IEC 61140. Clause 712.4.102 permits functional earthing of a live part on the DC side only where there is at least simple separation between AC and DC by a transformer with electrically separate windings, at a single point near the converter DC input, and the functional earthing cable must not be green and yellow. Clause 712.444.5.5.101 requires that where the array is earthed, the connection is made at a single point and taken to the main earthing terminal of the installation.

Read that against the NEC and the contrast is sharp. In the IEC world, Class II construction on the DC side is a legitimate route, and plenty of European rooftop arrays run with frames that are not earthed at all by design. Under the NEC, exposed metal is bonded to an EGC, full stop. A mounting system sold into both markets has to be able to do both, and a bill of materials written for one market is not automatically compliant in the other.

Commissioning is where the IEC set gets specific. IEC 62446-1:2016 puts continuity of protective earthing and equipotential bonding conductors in clause 6.1, as a Category 1 test, meaning it is part of the mandatory initial verification of every system, not an optional extra.

Lightning is handled separately under IEC 62305-3. The cross sections quoted consistently by European racking suppliers are at least 6 mm² copper or 16 mm² aluminium for equipotential bonding, rising to at least 16 mm² copper or 25 mm² aluminium where the connection carries lightning current. Separation distance s has to be calculated, not assumed: K2 Systems put typical values between 30 and 70 cm and warn that the empirical 50 cm rule cannot be used safely. S:FLEX publish that their mounting system may serve as a natural component of a down conductor where the cross bracing is continuously connected, verified for that product. A system specific verification, not a general property of aluminium rails.

Australia and New Zealand

AS/NZS 5033:2021 puts earthing in clause 4.6, where the 2021 revision replaced a simple rule with a decision tree in 4.6.5 and a sizing table in Table 4.6. Whether and how the array frame is earthed now depends on the lightning protection assessment under AS 1768, whether the system is earth referenced, whether the inverter is separated or non separated, whether the neutral is powered, and whether string level DC overcurrent protection is fitted. The practical consequence: on a large commercial system with no DC string protection the array earthing conductor has to match the AC earth conductor running to the inverter, which can mean 25 mm². Fitting a combiner with string level DC protection changes which row of the table applies, and with it the copper.

Working on a specific layout under any of these? Send the racking layout, the module model and the frame drawing. We will mark which joints in the delivered structure we specify as bonds and which we do not, and list the component documents we hold for them. The compliance call stays with your engineer of record and the authority having jurisdiction. Ask which joints we specify as bonds.

Testing it, and testing it again later

The commissioning test is a continuity measurement, and the useful version of it is not a beeping multimeter. Measure with a low resistance ohmmeter from the furthest module frame in the array back to the main earth bar, record the value, and record the torque used on the joints in that path. On a large ground mount, measure at the end of each row rather than at one convenient point.

One boundary matters more than the number. A field continuity measurement verifies the installed electrical path at the moment of testing. It does not replace product or system evaluation for fault current capability, environmental durability, mechanical retention, corrosion performance or roof fire classification. UL 2703 covers all of those alongside the bonding path. A low reading on a meter is evidence that a path exists today, and nothing else.

No standard covers the second half of the plant’s life. The listed bonding tests are run on new, correctly tightened assemblies, and nothing in them anticipates decades of thermal cycling, wind induced vibration and salt. That gap is the argument for re-torquing and re-measuring a sample of joints at maintenance, and for choosing bonding hardware designed to hold contact rather than hardware that merely passed once.

Seven questions worth asking a mounting supplier

1. Which specific mounting system does your bonding evaluation cover, and which module models was it tested with? Ask for the model list, not the certificate number.
2. Is bonding integrated at the clamps, or do you supply jumpers? At which joints exactly?
3. What happens at rail splices, and at any joint designed to move?
4. What is the assembly torque at each bonded joint, and what happens to the bond if it is installed under torque?
5. What metals meet at the bond, and what is the galvanic pairing in a coastal site?
6. For an IEC market, is the DC side designed as Class II, or are the frames bonded? If bonded, who supplies the 6 mm² path and the connection points?
7. What continuity value should the commissioning team expect, and from where to where?

One discipline underpins all seven. A listing or a test report attaches to a product and a configuration. It does not attach to a company or to a catalogue. So the answer to every one of those questions is a document with a scope written on it, and the scope is the part worth reading. That applies to us as much as to anyone, which is why the next section says plainly what we hold and what we do not.

Where our own hardware sits in this

Our aluminium rails and clamps are 6005-T5. Anodized surfaces on them are an insulator, so where a bond is required it is made by hardware that breaks the film, not by contact pressure alone. Fasteners are SS304 to A2-70, compliant with ISO 3506-1, with SS316 offered for high corrosion sites, which keeps the stainless side of any aluminium to stainless pairing above the 16% chromium threshold. On steel structures the standard coating is ZAM at 91% Zn, 6% Al and 3% Mg, used on systems such as our ZAM steel ground mount. Where ISO 12944-2 puts the site in C5, there is a duplex option, ZAM under 240 µm of paint, at the bottom of the ISO 12944-5 band for that category. We have not run the ISO 12944-6 laboratory qualification on that exact stack.

What we publish as third party evidence is mechanical test data with report numbers on it, for example the SR-C700Y clamp at 10.21 kN in an SGS pull test, report XMIN2403000515ML03, 2024, plus an engineering certificate covering the SRA-02-3 rail from GAMCORP, 5312/HS, 2018, and one covering roof mounting to AS 1170.2 from SPAD, V008_FL, 2019. We do not hold a bonding listing to UL 2703, and nothing above should be read as one. Where a project needs a listed bonding path, that belongs in the scope from the start, so the clamps, jumpers and lugs are chosen against it rather than assumed. The line between a joint that is a bond and a joint that is only structural has to be drawn before the commissioning test, not after it. That is what an EPC actually needs.

The bonding path is built by the same crew that builds the structure, with the same torque wrench, on the same day. It is worth specifying with the same care.

Standards referenced

• UL 2703, Mounting Systems, Mounting Devices, Clamping/Retention Devices, and Ground Lugs for Use with Flat-Plate Photovoltaic Modules and Panels, Ed. 1 (2015), latest revision published 28 April 2026
• UL 61730-1 and UL 61730-2, PV module safety qualification, harmonized with IEC 61730 in December 2017; UL 1703 remains the legacy module standard
• UL 467, Grounding and Bonding Equipment
• NFPA 70 National Electrical Code, Article 690 Part V (690.41, 690.43, 690.45, 690.47) and Article 250 (250.52(A)(2), 250.122, 250.134)
• IEC 62548-1:2023 Edition 1.0, consolidated as Edition 1.1 with Amendment 1 (December 2025), Photovoltaic (PV) arrays, Part 1: Design requirements, clause 7.3.2
• IEC 60364-7-712:2017 Edition 2.0, Low voltage electrical installations, Part 7-712: Solar photovoltaic (PV) power supply systems
• IEC 60364-5-54, Earthing arrangements and protective conductors
• IEC 62446-1:2016, clause 6.1, continuity of protective earthing and equipotential bonding conductors
• IEC 62305-3, Protection against lightning, physical damage to structures
• AS/NZS 5033:2021, clause 4.6 and Table 4.6; AS 1768 for lightning protection
• ASTM B117, salt fog exposure

Field data on joint behaviour is drawn from the Matrix Engineering review of electrical bonding in PV structural joints and from the Solar ABCs study report on grounding photovoltaic modules. Equipotential bonding cross sections are as published by K2 Systems, Mounting Systems GmbH and S:FLEX in their own lightning protection guidance.