Off grid system earthing / grounding 1 is a somewhat complicated topic. Floating the DC side is currently permissble in NZ so long as the array voltage is under 120VDC (be aware that you need to fuse both current carrying conductors if you go this route). And, there are problems with DC side GFCI systems, which are better suited to grid tied transformerless inverters than to off grid setups. So, in general, the industry standard is to earth bond both the DC and AC sides.
1 I use the term grounding and earthing more or less interchangeably. The colonies use Protective Earth (PE), while the americas use GEC/EGC. However you need to get clear about the difference between gounding and bonding. Bonding is the tieing together of all metal parts to reduce electric shock hazards, whereas grounding is the physical conenction to earth, whose only purpose is to dissipate over voltage. Both terms have historically been lumped together, and this page relates to the bonding side of the equation.
Thus connecting all metal DC side system components such as the charge controller and disconnect to ground, and bonding the battery negative to ground is what we are generally trying to achieve. However when it comes to grounding the inverter, you cant just treat an inverter like any other metal box. Because it is exposed to both the AC and DC sides of the system, extra attention to its grounding is required.
One way around this is by using a double insulated inverter such as the Stecas. But for your average DIY installed inverter, the temptation is to equipment bond the thing to the nearest earth bus in the disconnect and call it good.
However to understand this, we really need to look at the full range of fault conditions that can occur inside an inverter, which is something like the following:
a) DC side short,
b) AC side short,
c) a DC positive to chasis short,
d) an AC hot to chassis short,
e) DC current carrying conductor to AC hot short.
Of course these are all pretty obscure/unlikely scenarios (manufacturing defects and/or foreign body ingress are possible causes) but should they occur, your grounding system needs to be able to cope with them.
For (a) and (b) they are straight forward, and handled by your circuit protection on the DC side in the case of (a), and by the inverter faulting and shutting down for (b). For (c) and (d), should an inverter develop an earth fault, providing an adequate current path for the resulting fault current is important, as well as ensuring that that current is disconnected correctly.
On the AC side a hot to chasis short will result in the following: In the first instance the chasis becomes live and a shock risk. Immediately after, current will flow through the inverter's EGC, navigate its way through your earth system to find the earth to neutral bond link, thus allowing the current to return to the inverter via the neutral conductor. A few tenths of a second later, the AC breaker in the hot conductor will pop, and/or the inverter will overload and shut down, which ever is the fastest.
The same thing occurs on the DC side, with the current returning either via the DC negative to earth bond jumper, or via a ground fault detection and interuption device. The nature of these earth fault currents and the exact path taken is where it gets interesting.
For the AC side, the current that a typical residential off grid inverter can produce is limited to the rated capacity of the inverter, maybe 10 or 20 amps (at 230V). So the EGC needed there is relatively small gauge. (However with the higher voltage, there is still a high arc energy and associated fire risk.)
On the DC side, the ability of the circuit breaker which protects the inverter to safely interupt potentially 10,000 or more amps is critically important. The short circuit potential of your battery bank is limited only by the internal resistance of the cells, and the resistance of the associated cabling. If you have a DC GFI, that depends on how it is connected. For example the GFI in the Midnite Classic will high resistance sever the earth, which stops the fault current, but leaves the fault in place.
The question that was on my mind one rainy sunday as i was bolting the VFX to the wall, was, do i equipment bond the inverter to the DC earth bus or the AC earth bus? To make matters more complex the VFX, unilke many/most, has both AC and DC earth terminals. On the other hand cheaper inverters are lucky to have one, usually on the AC side.
Article 690.47(C) (NEC 2011) specifies three alternative earth arrangements for systems involving DC and AC circuits.
"[Y]ou must bond the dc grounding system to the ac grounding system. Section 690.47(C) specifies three methods for providing that bond.
Separate and bonded. Per Section 690.47(C)(1), a separate dc grounding-electrode system can be installed, provided it is bonded to the ac grounding-electrode system, as shown in Option 1 in Figure 2. The ac and dc GECs must be sized based on their respective NEC sections, and the grounding electrodes must be bonded using a conductor no smaller than the larger of the ac or dc GECs. Note that the dc GEC and/or the bonding conductor between the grounding electrodes cannot replace the required EGCs.
Separate and bonded grounding-electrode systems are used on residential and commercial retrofits, as well as large utility-scale systems. A dc grounding electrode, such as a new ground rod or ring, is installed as part of the PV system and connected via a dc GEC to the marked point in the inverter. The dc grounding electrode is then bonded to the premise's existing ac grounding-electrode system, such as a ground rod or Ufer, or, in large-scale systems, to the newly installed ac grounding electrode on the secondary side of a medium-voltage transformer that is between the inverter and the utility grid.
Common grounding electrode. NEC Section 690.47(C)(2) allows for a common grounding electrode (or grounding-electrode system) to serve both the ac and dc systems, as shown in Option 2 in Figure 2. The dc GEC cannot replace the required ac EGC. In the event that the ac grounding electrode is not accessible, the dc GEC can connect directly to the ac GEC per Section 250.64(C)(1), which allows for irreversibly crimped or exothermically welded connections.
This method is widely employed for a variety of PV system types, from residential to utility-scale systems. For example, a ground ring around an inverter and transformer pad can serve as both the ac and the dc grounding electrode. The allowance for connecting to the ac GEC is helpful when the existing grounding electrode is a Ufer. Be aware, however, that if an existing ac grounding electrode is inaccessible, then there may be no way to verify that the resistance to ground is less than or equal to 25 O, as required in Section 250.53(A)(2). In such cases the AHJ may require a supplemental grounding electrode; if so, this must be installed at least 6 feet away from the existing electrode and bonded to it.
Combined grounding conductor. A combined grounding conductor can serve as both the dc GEC and the ac EGC, as shown in Option 3, Figure 2. While Section 250.121 specifies that an EGC cannot be used as a GEC, Section 690.47(C)(3) amends this general Code requirement and provides an allowance exclusively for PV systems. On the face of it, installing one conductor instead of two seems like the simplest and least expensive method. In practice, the requirements associated with installing a combined grounding conductor may complicate things.
Since the combined grounding conductor serves two functionsdc GEC and ac EGCit must be sized according to the function that requires the largest conductor: The size of the ac EGC is determined based on NEC Table 250.122, and the size of the dc GEC is determined based on Section 250.166. Further, the combined conductor must either run unspliced or irreversibly spliced from the inverter to the grounding busbar in the associated ac equipment, which refers to the ac equipment that the ac GEC connects to. As a result, you must crimp pigtails to the combined grounding conductor whenever you need to connect an EGC to any electrical equipmentincluding disconnect switches, production meters, switchgear and so forthlocated between the inverter and the final termination point at the ac GEC. Lastly, all other GEC installation requirements still apply, such as bonding ferrous raceways.
While the option to use a combined grounding conductor is applicable to all PV system types, it is most common on systems that use microinverters with an internal dc system bonding jumper. While the microinverter trunk cable may include an EGC, the grounding conductor connected to the microinverter chassis runs from the PV array to the associated ac equipment, serving as both the dc GEC and the ac EGC
Here it is in graphic form:
Another diagram, this one from Bluesea, perhaps easier to understand:
The block-diagrammy-ness of these clauses need some intepretation. NEC method C was more or less what i was i had in mind. However, reading the fine print, that method prohibits a screw terminal connector (in the form of our DC earth bus) between the AC earth bus and the inverter. Some of what they are trying to cater for is:
1. If you run independent DC and AC GECs to the same earth electrode, with only a single single inverter ECG then current passing between the AC and DC earth systems is forced to navigate the single most corosion prone connection, at the stake. Such current occurs in the event of several of our fault types above. You can get around this by crimping or welding the two GECs, but 3 terminal crimps and exothermic welding gear are not necessarily easily accessible.
2. However, you must also factor the differing magnitude of fault currents on the AC and DC sides. If the only earth terminal on the inverter is on the AC side and fits only small guage wire, any earth fault on the DC side will result in hundreds of amps attempting to navigate the AC EGC, fow which it is not rated. This is an ampacity scenario with unhappy consequences.
Thus, the VFX has it right, a large gauge DC earth terminal, and a small guage AC earth terminal. Bonding each to their respective buses, not only provides an appropriate path for the respective fault depending on which side the fault occurs, but when combined with a bond between the AC and DC ground buses, provides a level of redundancy, to doubly make sure that everything is at the same potential. The one remaining concern is that this creates a parallel path in the earth system, or ground loop if you will. Does a fault on the DC side, take the DC EGC, or the longer AC ECG, AC earth bus, AC/DC bond route. Well, clearly a fault current would take both routes, but with ampacity and resistance being generally proportional, the lower resistance DC would take the larger share.
Which brings us to the sizing of inverter EGCs. What is obvious is that the ampacity of the earth wire should meet the rating of the circuit breaker protecting the current carrying conductors on that side of the inverter. In practice because the EGC only carrys fault currents for brief periods some derating is allowed, often 25 or 30%. However, the code also makes a relationship between the size of the current carrying conductors and the EGC. Again some derating is allowed. I have not seen an explanation of the reason for this ruling, however the paragraph above goes some way to an answer. The bigger the DC EGC the less risk that the AC EGC will carry current beyond its ampacity. But again in practice if you have quality circit protection, the duration of fault currents in your earthing system will be very very brief, and its possible that smaller EGCs are manageable in practice.
Heres a diagram from the Magnum inverter manual, which was used to derive my solution.
And a diagram of my adopted solution.
Lastly, what happened to fault case (e), you ask? Well thats messy alright, and its a form of earth fault in that the fault current also traverses the earth system. The internal AC hot short, bypasses the AC circuit protection, and yet doesnt source sufficient current to trip the DC side protection. The inverter will detect the high output current, and likely high input voltage, and then shutdown, and the peak fault current is limited to the rated inverter output as with fault type (b). What a cheap inverter with poor output protection might do is another matter altogether. But whether the inverter survives or not, exposed voltages should be removed.