A leaky feeder is one of the most common ways to provide underground radio coverage for voice, telemetry and low rate data, and it remains the dominant radio distribution method in many hard rock and coal mines, even as LTE, private 5G, Wi Fi mesh and fibre fed access points take on more of the high rate data load. It is a coaxial cable engineered to radiate along its length, so it behaves like an antenna that follows the tunnel instead of a point source that has to fight rock.
Designing one is not complicated, but it is unforgiving. The cable has two separate loss numbers that pull in different directions. Amplifiers fix one of them and add their own constraints. Mine geometry then breaks the tidy assumption that the cable runs in a straight line.
This article walks through the design as an engineer actually does it: the two loss mechanisms, the system gain budget, amplifier spacing, the downlink and uplink budgets, and how declines, stopes and junctions change the layout.
A note on numbers. The cable and amplifier figures in the worked examples below are representative values only. Radiating cable and amplifier performance varies significantly by frequency band, cable family, modulation and installation. Design against the actual datasheets, and confirm against commissioning measurements.
A normal coaxial cable is built to keep RF energy inside. A radiating, or leaky, cable is built to let a controlled amount of it out, through slots or a loose braid in the outer conductor. The cable continuously couples RF energy into the tunnel along its length, so coverage is created everywhere the cable runs.
That is the whole reason it suits underground work. A roof mounted antenna lights up the area it can see and stops at the first corner. A radiating cable carries coverage along the same path the cable follows. In a drift, a decline or a conveyor drive, the cable goes where the people and vehicles go, and the coverage goes with it.
The cost of that behaviour is that a leaky feeder is a distributed loss system. Signal continuously couples out of the cable while also attenuating along the run, so the design problem is keeping enough level at every point along a cable that may be several kilometres long.
Every radiating cable datasheet gives you two different loss figures. Keeping them separate in your head is most of the skill.
| Loss mechanism | Units | What it sets | Frequency behaviour |
|---|---|---|---|
| Longitudinal loss | dB per 100 m | How far signal travels down the cable, and therefore amplifier spacing | Rises strongly with frequency |
| Coupling loss | dB at a reference distance | Field strength a given distance off the cable, where the radio sits | Rises slowly with frequency |
Longitudinal loss is the along cable attenuation. It is the loss you would measure end to end on a normal coax. It sets how far the signal gets before it needs an amplifier. For a 7/8 inch radiating cable it might be around 2.7 dB per 100 m at 450 MHz and around 6.5 dB per 100 m for a smaller 1/2 inch cable at 900 MHz. Higher frequency and smaller cable both cost you here.
Coupling loss is the radial loss between the cable and a radio at a stated reference distance, commonly 2 m. It sets how strong the field is where the radio actually is. Datasheets usually quote it two ways: a median figure (CL50) and a worst case figure (CL95), the coupling loss met or bettered at 95 percent of measured locations around the cable. Design to CL95. The median will flatter your budget and then disappoint you at commissioning.
The trap is treating these as one number. Longitudinal loss is a per metre problem you solve with amplifier spacing. Coupling loss is a fixed penalty at every point that you solve with cable choice, frequency choice and how close the radio can realistically get to the cable.
A leaky feeder design is a cascade. Signal enters at the head end, loses level along every metre of cable, gets some of it back at each inline amplifier, and has to still be usable at the far end and at every point in between.
The structure of the budget:
Two rules keep a cascade healthy. Do not run amplifiers at full gain, because you lose the headroom to trim the design and you push the amplifiers toward intermodulation. And watch the noise figure: each amplifier contributes additional noise, and on the uplink that noise cascades all the way back to the head end, raising the effective system noise figure seen there.
Amplifier spacing is the first hard number in the design. It comes from the usable gain of the amplifier divided by the longitudinal loss of the cable.
Here is a worked downlink example for a 450 MHz system on 7/8 inch radiating cable.
| Quantity | Value | Note |
|---|---|---|
| Inline amplifier maximum gain | 28 dB | Typical mining line amplifier |
| Reserve gain held back for level control | 5 dB | Keeps the cascade off full tilt |
| Usable gain per amplifier | 23 dB | 28 minus 5 |
| Cable longitudinal loss at 450 MHz | 2.7 dB per 100 m | 7/8 inch cable, datasheet figure |
| Allowance for connectors, taps and bends per span | 3 dB | Engineering allowance |
| Cable length the usable gain supports | about 740 m | (23 minus 3) divided by 2.7, per 100 m |
| Spacing actually used | 500 m | Vendor cap for noise and level control |
Note that the calculation supports about 740 m, but the spacing actually used is 500 m. That gap is deliberate. Vendors cap spacing below the theoretical maximum to control the noise cascade, to keep level variation along the cable inside a sensible window, and to leave room for the cable to age and for splices to be added later. The first amplifier after the head end is often spaced shorter still, because the head end injects at a high level and the first span needs tighter control.
Treat the theoretical span as the ceiling and the vendor spacing as the design value.
Amplifier spacing keeps signal moving down the cable. The coverage budget checks that there is still enough field for the radio at the worst point, which is normally just before the next amplifier, where the cable level is lowest.
| Quantity | Value | Note |
|---|---|---|
| Downlink level on the cable just before the next amplifier | -4 dBm | After a full span from the previous amplifier |
| Coupling loss at 450 MHz, 95 percent confidence | 76 dB | At the 2 m datasheet reference distance |
| Extra loss from 2 m out to the radio | 6 dB | Distance into the drive, plus body and vehicle loss |
| Signal at the radio | -86 dBm | Sum of the above |
| Portable radio reference sensitivity | -110 dBm | Digital two way radio; varies with modulation and BER criterion |
| Fade margin | 24 dB | Headroom over sensitivity |
A 24 dB margin is comfortable. It needs to be, because the budget above assumes the radio is a few metres from the cable in clear air. The real worst cases are a handheld carried into a cross cut with no cable in it, a radio behind a vehicle or a load, or coupling loss that is worse than CL95 at a specific bad spot. The margin is what absorbs those.
If the margin goes negative, you have three levers: a lower frequency, which improves both loss numbers; a larger cable, which improves longitudinal loss; or shorter amplifier spacing, which lifts the worst case cable level. Frequency is usually the most powerful and the least available, because it is set by the radio fleet.
Frequency is the single most powerful input to a leaky feeder design, and usually the least free, because it is set by the existing radio fleet and the spectrum the mine holds.
Lower frequencies, in the VHF range, generally propagate better underground. They have lower longitudinal loss, so amplifier spacing can be longer, and they diffract more usefully around corners and into adjacent headings. The cost is larger antennas and less available bandwidth for data.
Higher frequencies, in the UHF and 900 MHz range, support wider data channels and smaller equipment, but they pay for it. Longitudinal loss rises, so amplifiers sit closer together, and passive coverage into cross cuts and around intersections falls off faster.
There is no single right answer. A voice and tracking network is well served by a lower band; a high rate data system may have no usable VHF option at all. Where the fleet allows a choice, run the loss numbers at both candidate bands before committing.
Most design attention goes to the downlink, because it is easy to picture. The uplink, from the handheld back to the head end, is often the real limit.
Three things make the uplink harder:
The result is that the noise floor at the head end, not the downlink field, frequently sets the usable range. A leaky feeder that sounds fine on the downlink can still fail to hear a handheld at the end of a long branch. Budget the uplink explicitly, cascade the noise figure back to the head end, and check the worst case uplink point separately. Do not assume a healthy downlink implies a healthy uplink.
Multi channel systems add another constraint. Where several carriers share the cable, the design needs intermodulation margin and amplifier linearity, particularly near the head end where the composite RF level is highest. An amplifier driven into compression there generates intermodulation products and raises the noise floor, desensitising the uplink it is meant to carry.
Because the two directions do not behave the same, many systems plan uplink and downlink gain separately and run asymmetric gain settings, sometimes with directional amplifiers and slope equalisation, to control noise accumulation and keep the return path stable.
A clean spacing calculation assumes the cable is a straight line. A mine is not.
Declines. Vehicle access between levels is by spiral decline, not by vertical raise. That is good news for a leaky feeder: the cable runs along the decline back and the coverage follows the ramp down naturally, the same as in any drive. Spacing logic does not change, but remember the decline adds real cable length between levels, and that length is in the budget.
Drives and drifts. The straightforward case. Main cable runs, amplifiers at the design spacing, taps where a branch is needed.
Cross cuts and stubs. Short headings off a main drive get little coverage unless you branch into them. Coverage around a corner and into a cross cut degrades rapidly, particularly at higher frequencies and in irregular geometry. Some signal does diffract around the corner, and tunnel waveguide effects can occasionally help, but a radiating cable a few metres away in the main drive cannot be relied on to cover a cross cut. Where a cross cut needs coverage, branch into it with a splitter or a tap and a short cable run. Where it does not, record that it is intentionally uncovered.
Stopes. A stope is a tall void created by extracting ore from below. It is an active, changing space, and you do not run a radiating cable through one. Coverage into a stope is genuinely hard. The practical approach is to feed the stope access drives well and accept that the open void itself has limited or no coverage, rather than pretending a nearby cable will fill it.
Junctions and intersections. Major intersections are RF shadow points. Coverage degrades sharply through an intersection, because signal diffracts around a rock corner far less effectively than it follows the cable. Plan branches, taps and sometimes a short overlapping cable run at important intersections so a radio does not drop out crossing them.
In gassy mines, typically underground coal, equipment in the explosive atmosphere must be intrinsically safe or otherwise protected for the hazardous area. That applies to inline amplifiers, power injectors and any active component on the cable. Intrinsically safe amplifier variants exist for exactly this reason, and they constrain the available gain and power.
Hard rock metalliferous mines are generally less dependent on intrinsically safe communications equipment than underground coal operations, so standard equipment is more common. Even so, diesel exhaust, blasting fumes, sulphide environments and localised hazardous zones can still apply, and the site always has electrical, fire and cabling requirements the installation has to meet. Confirm the hazardous area classification and the applicable regulations for the specific mine and jurisdiction before fixing the equipment list. It changes which amplifiers you can even consider.
A leaky feeder design is ready to build when these are all true:
The two loss numbers are from the real datasheet. Longitudinal loss and CL95 coupling loss, at the actual operating frequency, for the actual cable. Not a median figure, not a figure for a different frequency.
Amplifier spacing uses usable gain, not maximum gain. Reserve is held back. The first span from the head end is tighter than the rest.
The downlink budget closes at the worst point with margin left for cross cuts, vehicles and bad spots.
The uplink budget is calculated separately, with the noise figure cascaded back to the head end.
The geometry is in the layout. Decline lengths are counted, cross cuts are either branched or recorded as uncovered, stope voids are not assumed covered, and intersections have a plan.
The equipment suits the hazardous area classification of the mine.
noIM3’s tools follow this same path. The LCX Cable Selector exposes the longitudinal and coupling curves so cable choice is a real decision rather than a default. The Amplifier Spacing Calculator turns usable gain and cable loss into a spacing. The Leaky Feeder Designer runs the full cascade, downlink and uplink, on the network you draw, and produces the budgets, coverage prediction and review documentation for the design.
Underground communications is a safety system before it is a productivity system. When a leaky feeder has a weak span, the failure is not a dropped data packet. It is a radio that does not reach a person at the end of a decline.
The designs that hold up are the ones where the engineer kept the two loss numbers separate, set spacing on usable gain, budgeted the uplink as carefully as the downlink, and drew the cable along the mine that actually exists rather than the straight line that is easy to calculate. The ones that struggle are the ones designed on median datasheet figures and a tidy diagram.
Related Insights
There is no single best frequency for long range communication. The right choice depends on distance, terrain, antenna size, licensing, noise floor, and required availability. Here is how to pick.
Apr 27, 2026
A practical refresher on free space path loss, the log distance model, Fresnel zones, and the two ray ground reflection effect: the four numbers every field engineer should be able to estimate before climbing a tower.
Apr 22, 2026
RF interference is becoming more common as wireless systems multiply and spectrum becomes increasingly congested. Learn what causes it, why it’s worsening, and what engineers can do to mitigate it.
Mar 25, 2026
