What if?
Pick one input and see how touch, step, GPR and grid resistance change across its range, with everything else held at the current design. The dashed line marks this design; shaded bands are where the voltage exceeds the allowed value.
What happens in the ground during a fault
When a phase conductor faults to grounded metal, the fault current has to get back to the neutral of the transformer that supplied it. Part of it goes home on metal: shield wires, neutrals, cable shields. The rest flows into the earth through the grounding system and spreads out through the soil.
Soil resists that current, so the grounding system rises in voltage relative to distant ("remote") earth. That rise is the ground potential rise (GPR): the grid current times the grid resistance. The ground surface above and around the grid rises too, but by less, and the further from the conductors, the less. A person standing on that surface and touching grounded metal bridges the difference. That difference, not the GPR itself, is what drives current through the body.
Use the Explore tab to change the soil, surface layer, fault, return paths and electrode and watch the touch and step voltages respond. Calculation shows the IEEE 80 working, What-if sweeps one input across its range.
The shock situations
IEEE 80 checks a person in each of these positions. In each case the voltage across the body is a difference between two surface or metal potentials, and the soil under the feet adds resistance in series with the body.
| Situation | Voltage across the person | What limits it |
|---|---|---|
| Touch | Grounded structure (at the GPR) minus the surface under the feet, standing within about 1 m of the structure. | Close grid spacing, conductors where people stand, surface layer, fast clearing. |
| Mesh | The worst touch voltage inside the grid, usually at the centre of a corner mesh where the surface sags most below the GPR. | Smaller corner meshes, perimeter rods, more conductors near the edge. |
| Step | Surface potential difference between the feet, 1 m apart, with no contact with metal. Largest just outside the grid edge and corners. | Burial depth, edge conductors and rods, surface layer. The step limit is much higher than the touch limit: IEEE 80 uses the same body current, but the two feet are in series (6Csρs) rather than in parallel (1.5Csρs). |
| Fence touch | A bonded fence is at the GPR; the person touching it from outside stands on ground that has already fallen away. | Run the grid (or a perimeter conductor) about 1 m outside the fence, extend the rock, or isolate the fence. |
| Transferred | A metallic path (communication shield, neutral, rail, pipe, fence) carries the GPR off site, or brings remote earth in. The difference can be nearly the full GPR. | Isolation (IEEE 487 for wire-line communications, fibre), bonding and insulating sections, keeping remote-earth conductors away from the station. |
| Metal-to-metal | Hand to hand between two metal objects at different potentials. No foot resistance helps. | Bond all metal that can be reached together. |
How much voltage a body can take
IEEE 80 uses Dalziel's fibrillation threshold: the current a person can survive is IB = k/√ts, with k = 0.116 for a 50 kg person (the usual choice where the public may be present) and 0.157 for 70 kg, valid from 0.03 to 3 s. The body is taken as 1,000 Ω and each foot as a 0.08 m radius plate on the surface, about 3ρs Ω. For touch the feet are in parallel, for step in series:
Faster clearing raises the allowed voltage as 1/√t; a surface layer raises the foot resistance. Two examples over 100 Ω·m soil, allowed touch / step in volts:
| Clearing time | Native soil, 50 kg | Native soil, 70 kg | 4 in (0.1 m) crushed rock 3,000 Ω·m, 50 kg | Rock, 70 kg |
|---|---|---|---|---|
| 0.1 s (6 cycles) | 422 / 587 | 571 / 794 | 1,528 / 5,010 | 2,068 / 6,781 |
| 0.25 s | 267 / 371 | 361 / 502 | 966 / 3,169 | 1,308 / 4,289 |
| 0.5 s | 189 / 262 | 255 / 355 | 683 / 2,241 | 925 / 3,033 |
| 1.0 s | 133 / 186 | 181 / 251 | 483 / 1,584 | 654 / 2,144 |
Cs is the surface layer derating factor: a thin, resistive layer over more conductive soil is less effective than an infinitely deep one. IEEE 80 gives Cs = 1 − 0.09(1 − ρ/ρs)/(2hs + 0.09). Use wet resistivities: clean crushed rock is a few thousand Ω·m wet; wet concrete only 20 to 100 Ω·m, about the same as soil. Asphalt is very resistive but cracks and patches weaken it.
Other standards set the limit differently. IEC 61936-1 and EN 50522 use the body impedance and heart-current curves of IEC/TS 60479-1 to give a permissible touch voltage that depends only on fault duration, then add foot and footwear resistance. The numbers differ from IEEE 80, so use the standard that applies to the project.
Where the fault current goes: split factor and return paths
Only the current that flows from the grid into the earth raises the GPR. IEEE 80 calls it the grid current IG = Cp · Df · Sf · 3I0, where Sf is the split (current division) factor, Df the decrement factor for DC offset (it depends on X/R and the fault duration) and Cp an allowance for growth.
- Local or remote source. For a fault inside a substation, current from that substation's own grounded-wye transformers returns to their neutrals through the grid conductors. It never enters the earth. Only the current from remote sources does. For a feeder fault outside a delta-wye distribution station the opposite happens: the station transformer supplies the fault, and current coming back through the earth must enter this grid to reach the transformer neutral. IEEE 80 asks for the worst case of both.
- Shield wires and neutrals take current two ways. The faulted phase current induces a current in a parallel return conductor (the coupling factor kc = Zm/Zs, about 0.1 to 0.3 for overhead shield wires, much higher for a concentric neutral wrapped around the conductor). The rest divides between the grid and the ladder of tower or pole grounds along the line in proportion to their impedances.
- Multigrounded neutrals help most. A 4-wire multigrounded system brings the neutral back to the source and grounds it at many points; the concentric neutral of an underground cable is coupled so tightly that only a small fraction of a padmount fault enters its rods. On a 3-wire (unigrounded) system there is no neutral to share the current, and the earth carries much more.
- The return conductor transfers voltage. The towers, poles or customer neutrals next to the site rise to a large share of the GPR during the fault.
The Fault & split tab takes every circuit at the site and works the split factor two ways. Endrenyi ladder with coupling: each return conductor is a ladder of span impedances (Carson self impedance with earth return) and grounds; on the lines that bring fault current, the share Zm/Zs induced by the phases returns on the conductor and the rest divides between its ladder and the site; the lines are solved one at a time with the others as passive branches and the results add. IEEE 80 Annex C equivalent impedance (after Garrett, Myers and Patel): each return conductor is reduced to its driving-point impedance with no coupling, all are put in parallel to give Zeq, and Sf = |Zeq/(Zeq + Rg)| for the remote current. The impedances come from your data, not from the Annex C table. The recommended (design) value is the largest of the credible cases: all circuits in service, each group short one circuit, tower and pole grounds at the high end, and customer neutrals not credited.
Under the Annex C assumptions (one 7 No. 10 Alumoweld shield wire, 1/0 ACSR neutrals, 100 Ω·m) the equivalent impedances computed here come within 1 to 3 % of Annex C Table C.1 for 100 and 200 Ω grounds and 9 to 19 % above it for 15 and 25 Ω grounds (higher impedance, so a higher, safer Sf). For the Annex C.3 example the ladder with coupling gives 0.29 against the EPRI program's 0.26. A detailed fault-current distribution study remains the basis for final design.
Which fault location is worst
Only the current that enters the earth through this grid raises its GPR. The worst location is the one that puts the most current into the earth here for its clearing time. It is not always the one with the largest fault current.
- Fault at this site's bus. The current from remote sources has to get back to them: part on shield wires, neutrals and cable screens, the rest through the earth (IG = Sf · 3I0). Current from this site's own grounded-wye transformers or generators circulates in the grid conductors back to their neutrals and raises no GPR. This usually governs at transmission and switching stations.
- Fault out on a line or feeder fed from this site. When this site has a grounded source (a transformer neutral or a generator), the current it feeds to a fault outside comes back on the faulted circuit's return conductor and through the earth, and the earth share enters this grid to reach the neutral. With no return conductor (a 3-wire feeder, a line without a shield wire) all of it comes back through the earth. Feeder faults are often cleared more slowly than bus faults, which lowers the allowed voltages, so at a distribution station a feeder fault can govern.
- Fault at the first towers or poles outside. With the shield wire or neutral bonded to the grid, the faulted structure is one span from this grid, and the current from remote sources goes into this grid much as it does for a bus fault. The bus fault covers it.
- Fault on the other side of a transformer at this site. The current flows between the fault and that transformer's neutral through the grid conductors, and raises no GPR here.
- Switching station with no transformer. A fault outside draws no current back to a neutral here: it returns to the source transformer elsewhere. Return conductors bonded to this grid can carry part of it through the grid, so a fault just outside acts much like a bus fault, and metal leaving the site can transfer potential. The bus fault usually governs.
The Fault & split tab lists the study's fault as location 1 and, when this site has local sources, adds a fault out on each line they feed. Add others from the list under the table, for example a feeder fault with its current from the short-circuit study and its own relay clearing time. A fault out on a circuit is computed with its largest current (a fault just outside) and its largest earth share (a fault at the far end of the route), so it bounds every point along the route. The one-line view, the split bar and the ground-surface profile on that tab show the location you pick.
The rod, the soil and the neutral
Two different things are both called the soil's resistance, and they behave differently.
- Spreading resistance, at each electrode. Current leaving a rod has to pass through shells of soil around it. Close in, the shells are small, so they hold most of the resistance: for a 10 ft rod about half of it lies within a foot of the rod and about 90 % within 15 ft. Farther out the shells are so large that they add almost nothing, so the total levels off (Dwight's formula). A grid spreads its current over a large area, so it has a much lower resistance.
- The earth along a line's route. Once the current is in the earth, carrying it along the route costs Carson's earth-return resistance, π² f × 10⁻⁴ Ω/km: about 0.059 Ω/km or 0.095 Ω/mile at 60 Hz, whatever the soil. It adds up with length (100 miles is about 9.5 Ω) because at power frequency the return current stays under the line, within an equivalent depth of 658 √(ρ/f) m. With direct current it would spread without limit and this part would vanish. The reactance of the conductor–earth loop is larger than either resistance.
- Three zones from the rod outward. The spreading zone, from the rod to about 25 to 30 ft, holds about 95 % of the rod's resistance (a single 8 ft by 5/8 in rod is about 0.4 × ρ ohms, with ρ in Ω·m). The transition runs from there out to about Carson's depth De ≈ 658 √(ρ/f) m: about 2,800 ft (half a mile) at 100 Ω·m, 1,500 ft at 30 Ω·m and 1.7 miles at 1,000 Ω·m. Across it the current keeps spreading but adds very little. Beyond it the current follows the line, and the earth adds about 0.095 Ω per mile. For an earth path of a few hundred feet, use the rod's resistance alone; for miles, use the rod or grid resistance at each end plus 0.095 Ω per mile.
- Two rods. Two joined rods come close to half of one rod's resistance but never reach it, because each sits inside the other's shells. For two 8 ft rods: about 58 % 6 ft apart, 57 % one rod length apart, 54 % two rod lengths apart, and 50 % only when very far apart. Spacing beyond about two rod lengths gains little, and each further rod helps less than the one before.
- Neutral or soil. A fault at a pole ground tied to the neutral has two ways back to the station: the neutral, or the rod, the earth and the station ground in series. The rod's spreading resistance dominates the soil path, so the neutral carries most of the current. The current the faulted phase induces in the neutral adds to its share.
The Fault & split tab draws these with this study's values: the rod's resistance shells, how much of the resistance lies within a distance, the three zones, two rods, the two paths back to the station, and what grows with the line's length and what does not.
System grounding
| System | Ground-fault current | Where it returns | What it means for grounding |
|---|---|---|---|
| Solidly grounded, 4-wire multigrounded | High, often thousands of amperes | Largely on the neutral, which is grounded along the line and connected to every station and customer neutral | High fault current but a low split factor. Neutral-to-earth voltages and transferred voltage along the neutral matter. |
| Solidly grounded, 3-wire (unigrounded) | High | Through the earth, or on cable shields and shield wires where they exist | Large share of the fault current in the earth: higher GPR at equipment grounds. Each electrode matters more. |
| Low-resistance grounded | Limited by the neutral resistor, typically 200 to 1,000 A | Earth and equipment grounding conductors back to the resistor | Low GPR. Common in industrial medium-voltage systems; faults are tripped. |
| High-resistance grounded | A few amperes | Mostly capacitive and through the resistor | Negligible GPR; the system can run with one ground fault while it is located. |
| Ungrounded (delta) | Charging current only, a few amperes | Through the capacitance of the unfaulted phases | Small GPR for the first fault, but the unfaulted phases rise to line voltage to ground and a second fault becomes a phase-to-phase fault through two grounds. Arcing faults can cause overvoltage. |
What changes the result
| Change | Grid resistance | GPR | Touch | Step | Notes |
|---|---|---|---|---|---|
| More grid conductors (smaller meshes) | A little lower | A little lower | Much lower | Lower | The most direct way to lower mesh voltage. Concentrate conductors near the edge, where meshes run hottest. |
| Rods around the perimeter | Lower | Lower | Lower | Lower at the edge | Rods discharge current deep, flattening the surface gradient at the edge where step and corner-mesh voltages peak. |
| Rods placed by hand (custom layout) | Lower for each rod, less as they crowd together | Lower | Lower where they are | Lower where they are | Rods at the corners and along the edge, where the meshes run hottest, help touch and step most. Rods in the middle of a large grid mostly lower the resistance. |
| Outer conductor beyond the fence (grid lines extended to it) | A little lower | A little lower | Fence touch lower (by how much depends on where the fence was: about a quarter in the medium substation example) | Lower at the fence | The usual fix for fence touch: a perimeter conductor about 3 ft (1 m) outside the fence, with the rods inside it, so a person touching the fence stands over the grid. |
| Longer (deep-driven) rods | Lower, a lot if they reach wetter or more conductive soil | Lower | Modest in uniform soil | Modest | Most effective in layered soil with a conductive lower layer (not modelled here). |
| Larger area | Lower, roughly with 1/√A | Lower | Depends on spacing | Depends | Extending the grid beyond the fence also fixes fence touch. |
| Deeper grid | Slightly lower | Slightly lower | Small, either way | Lower | The IEEE 80 equations show touch falling with depth; the numerical model shows a minimum near 0.5 m and a rise below that, as the surface sits further from the conductors. |
| Surface layer (crushed rock) | No change | No change | Allowed value much higher | Allowed value much higher | Keep it clean and thick; check outside the fence. |
| Faster clearing | No change | No change | Allowed value higher | Allowed value higher | Allowed voltage rises with 1/√t. |
| More return paths (shield wires, bonded neutrals) | No change | Lower (smaller split factor) | Lower | Lower | Watch the voltage transferred along those paths. |
| Higher soil resistivity | Proportionally higher | Higher | Higher | Higher | Touch and step are proportional to ρ for a given grid current; the split factor shifts toward the return conductors. |
Small installations: padmounts, vaults, poles and single rods
IEEE 80 is written for substations, and its mesh and step equations assume a grid of conductors in both directions with spacing above 2.5 m. A padmount transformer's ground ring and two rods, or a vault's ring, are too small for them, but the physics is the same. This tool's numerical model handles any of these electrodes. The touch check is within 1 m of the equipment footprint.
- On a 4-wire multigrounded system the cable concentric neutral and the transformer's secondary neutral carry most of the fault current, so the padmount rods see only a few per cent of it. Because a ring and two rods are 10 Ω or more, even that small share can raise the tank by hundreds to thousands of volts, and the result depends strongly on how many customer and secondary grounds are tied in. Switch the same padmount to a 3-wire system with a tape-shield cable and the share in the earth, and the GPR, rise sharply.
- A single 5/8 in by 8 ft rod in 100 Ω·m soil is about 40 Ω. For a building service, NEC 250.53(A)(2) requires a supplemental electrode for a single rod unless it is 25 Ω or less, and 250.53(A)(3) puts the two at least 6 ft (1.8 m) apart (check the edition that applies). Rods closer than about one rod length share the same soil and gain less; try it with two rods. Utility pole and equipment grounds fall under the NESC (IEEE C2), whose Rule 096 sets ground resistance expectations for supply systems.
- Utility construction standards set the actual padmount, vault and pole grounding details. Use the owner's standard; this tool shows why they work.
Measuring soil and grounds
- Soil resistivity (IEEE 81): the Wenner four-pin test gives apparent resistivity ρ = 2πaR for pin spacing a (with the full formula when the probes are deep). Repeating it at several spacings, on two perpendicular traverses, reveals layering. The Soil tab fits one to five horizontal layers to the readings by least squares on the layered-earth Hankel transform, recommends the fewest layers that fit within one percentage point of the best, and shows how well each layer is fixed. Take spacings out to at least the grid diagonal; deep layers beyond the largest spacing are guesses.
- Grid resistance: the fall-of-potential method with the current probe far away (several times the grid diagonal). In uniform soil the potential probe at 61.8 % of the distance gives the true value for a small electrode. Large grids need long leads or other methods.
- Surface material: test a wet sample of the actual rock. Fines and contamination can lower its resistivity by an order of magnitude.
What this tool calculates
- Tolerable voltages: IEEE 80-2013, Dalziel body current, 50 or 70 kg, Cs from the IEEE 80 empirical equation. Inside the surface layer area and outside it (native soil) separately.
- Grid resistance: Laurent-Niemann, Sverak and Schwarz equations (IEEE 80 clause 14), Dwight for a rod, and a numerical model of the actual electrode. In layered soil the equations use the uniform resistivity that gives the numerical model's grid resistance.
- Numerical model: the conductors and rods are divided into short segments that each leak a uniform current, and the potential is averaged along each segment (Galerkin moment method, symmetric matrix, Cholesky solution); the conductors are at one potential. Uniform soil uses one image for the earth surface; two-layer soil uses the image series of Sunde and Tagg, with rods split at the layer boundary. The surface potentials give touch voltage (GPR minus the surface) and step voltage (largest difference over 1 m) everywhere. See Validation below.
- Mesh and step voltage: the IEEE 80 Em and Es equations, reproducing Annex B.1 to B.4, with their range of validity flagged. For a drawn or imported grid the shape factors na, nb, nc, nd come from the grid's outline and the spacing is D = 2A/(LC − LP/2), exact for an equally spaced grid. Rods within a quarter of the mesh spacing (at least 0.5 m) of the outermost conductor count as perimeter rods (Kii = 1 and the longer effective rod length in LM); a custom layout counts when at least four rods, or all of them if fewer, are on the perimeter. An outer conductor at a different spacing is outside the equal-spacing assumption, so the numerical model governs.
- Split factor: the two methods described above, at the worst credible case; or enter Sf with its basis. Paths can be long (semi-infinite) or finite, such as a secondary run with a few customer grounds; cable sheaths can be bonded at both ends, cross-bonded or single-point bonded (no return path).
- CAD import: DXF, DWG (R13 to the 2018 format that AutoCAD 2018 to 2025 save) and DGN (V7 and V8) are read in the page with the readers of the Kewo CAD Viewer. Pick the layers or levels (split by colour where a layer holds several) that hold the grid conductors, rods and fence; the import snaps ends together, merges overlaps, splits at crossings, and reports gaps, dangling ends and pieces not joined to the main grid. The grid can be edited on the plan and written back as a DXF at the drawing's own coordinates.
- Where the limits apply: touch is checked where a person can reach grounded metal (over the grid inside the fence, or within 1 m of an equipment footprint). Step is checked against the surface-layer limit inside the rock area and against the native-soil limit outside it.
- Conductor size: the IEEE 80 equation with Table 1 material constants and the connection temperature limit.
- Not modelled: three or more soil layers in the field model (the fits are shown; the grid uses the best two-layer fit), separately grounded or floating metal (a fence on its own posts is approximated as in Rev 1), conductor longitudinal impedance (very large grids, high frequency), lightning, the effect of foundations and buildings, footwear resistance, and detailed telecommunication protection.
Scenarios
Distribution substation, 66/12 kV
Bus fault fed by two shielded lines; feeder neutrals share the current.
Grid 4 ft outside the fence
Rods inside the fence, an outer conductor 3 ft outside it: compare the fence touch voltage.
Same station, 3-wire feeders
No shield wire and no neutrals: watch the split factor go to 1.
Feeder fault outside the station
The station transformer feeds a remote fault; earth current returns through this grid.
Transmission substation, 230 kV
Large grid, 40 kA, local sources supply 30 %.
Padmount, 4-wire multigrounded
Ring and two rods; concentric neutrals carry most of the fault current.
Padmount, 3-wire
The same pad with a tape-shield cable and no neutral.
Underground vault
Ring and four rods around a 3 x 6 m vault under a concrete sidewalk.
Single ground rod
One 8 ft rod and the 25 Ω question.
Industrial, low-resistance grounded
A 400 A resistor keeps the GPR low.
Ungrounded 4.16 kV
Charging current only on the first fault.
IEEE 80 Annex B.1 to B.5
The standard's worked examples, for checking: B.4 is a drawn L-shaped grid, B.5 is in two-layer soil.
Validation
Every check below is run by the tool's test suite on each release; the numbers are this revision's results. Published values are cited, not reproduced.
| Check | Reference | Result |
|---|---|---|
| Numerical solver against an independent solver | A separate Python program (point matching, very fine segments, extrapolated), same physics | Grid resistance and touch within 0.1 % (IEEE 80 B.2 grid and the B.5 two-layer grid) |
| Two-layer image series | Boundary conditions: no current through the earth surface, continuous potential and current density at the layer boundary | Met to about 1 part in a million |
| Uniform soil: IEEE 80 Annex B.1 to B.4 computer results (EPRI program) | IEEE Std 80, Annex B | Grid resistance 1 to 2 % lower; touch voltage 1 to 7 % lower; step voltage within 6 %. The difference is with the older program, not the implementation (see the first row); the default verdict also checks the IEEE 80 equations where they apply |
| Two-layer soil: IEEE 80 Annex B.5 (300 Ω·m to 4.6 m over 100 Ω·m, nine rods) | IEEE Std 80, Annex B.5 | With the rods stopping at the layer boundary: grid resistance within 0.3 % and corner-mesh touch within 0.7 % of the published values. With the 9.1 m rods reaching into the lower layer, as drawn, the resistance is 16 % lower; the published results appear to treat the rods as ending at the boundary |
| IEEE 80 equations | Annex B.1 to B.4 worked examples | Reproduced to the printed rounding |
| Soil forward model | Direct numerical Hankel integration (SciPy) for one to five layers | Within 0.02 % |
| Soil fitting | Synthetic readings from known models, with and without scatter | Layers recovered; the recommended number of layers matches the model |
| Split factor | IEEE 80 Annex C Table C.1 and examples C.2 and C.3 | See the split-factor section above |
| CAD import | Made-up DXF drawings of the B.1 and B.4 grids (feet, state-plane coordinates, overlaps, gaps, rod blocks) | Conductor length, rod count, area and perimeter exact; the imported B.4 grid solves to the same resistance as the one built directly |
| Dense grids (segments laid across the junctions when cutting at every junction would need more than 1,200) | The same model cutting every conductor at every junction, at standard and fine detail | Grid resistance within 0.1 %; worst mesh touch within about 1.5 % and on the high side. Grid resistance falls steadily as the spacing is halved (150 × 110 m grid: 9 × 9, 17 × 17, 33 × 33) |
| Examples on the Soil, Grid and Fault & split tabs | The soil models each set of example readings was made from; the tool's own results with everything else held the same | Every example loads and solves. The two-layer sets fit back to their models within 3 % (both resistivities and the boundary depth), the three-layer sets are recommended as three layers, and each "what to look for" statement is checked against the numbers |
IEEE 80-2013 Annex H compares CDEGS, ETAP and WinIGS on five benchmark grids and a current-division case. That comparison will be added when the full 2013 text (the grid layouts) is available to the project.
Standards and references
Grounding design
- IEEE Std 80-2013, Guide for Safety in AC Substation Grounding (revision project P80 in progress)
- IEEE Std 81-2012, Measuring Earth Resistivity, Ground Impedance and Earth Surface Potentials
- IEEE Std 142-2007 (Green Book), Grounding of Industrial and Commercial Power Systems
- IEEE Std 837-2014, Qualifying Permanent Connections Used in Substation Grounding
- IEC 61936-1 and EN 50522, power installations above 1 kV AC and their earthing; IEC/TS 60479-1, effects of current on human beings
Codes, GPR and related practice
- IEEE C2, National Electrical Safety Code, Section 9 (grounding methods), including Rule 096
- NFPA 70, National Electrical Code, Article 250
- IEEE Std 367-2012, Determining the Electric Power Station GPR and Induced Voltage from a Power Fault
- IEEE Std 487-2015 and 487.1 to 487.5, protection of wire-line communication facilities serving electric supply locations
- IEEE Std 1048-2016, Protective Grounding of Power Lines (worker equipotential zones)
- CPUC General Orders 95 and 128 for California overhead and underground construction