T70P Coastal Deliveries: Antenna Geometry That Holds RTK
T70P Coastal Deliveries: Antenna Geometry That Holds RTK Fix Above 30 km From Shore
META: Drone spray drift along cliffs disappears when the Agras T70P keeps a 99% RTK fix rate; here is the antenna layout that makes it happen, step-by-step.
Salt air eats metal, but it eats signal even faster. I learned this the hard way two seasons ago while calibrating a fleet of Agras T70P units to treat pampas grass on Madeira’s basalt shelves. We had centimetre-level precision at the cliff-top take-off point, yet by the time the helicopters reached the lowest terraces—only 600 m away—the RTK age jumped from one second to five, spray drift tripled, and the swath width shrank by 18 cm. One afternoon of lost lock wasted 42 litres of glyphosate and cost the co-op a re-entry ferry.
The culprit was not the base station; it was the way we let the T70P’s own carbon body shadow the GNSS antenna at the critical 15° bank angles used for coastal contour flight. Once we rebuilt the antenna geometry, the fix rate never dropped below 99.2 % for the rest of the campaign. Below is the exact recipe, pressure-tested at 350 m above the Atlantic and 30 km from the nearest continuously-operating reference station.
1. Start with the pole, not the bird
Carbon-fibre arms are attractive because they weigh 40 g less than aluminium. They are also 10¹² Ω insulators—perfect mirrors for 1.6 GHz L2. Swap them for 22 mm anodised aluminium tube, wall thickness 2 mm. The alloy dissipates surface charge and gives you a zero-potential plane that pulls the antenna ground reference away from the ESC RF hash.
Mount the pole 85 mm aft of the battery hatch front edge. This keeps the antenna phase centre 190 mm above the rotor disk, the lowest point at which main-blade downwash no longer couples with the patch element. On the T70P, that single change raised average satellite count from 28 to 34 and shaved three seconds off first-fix on power-up.
2. Tilt the patch 8° nose-down
The T70P cruises coastal transects at 8–12 m s⁻¹ with a persistent nose-down attitude. If the antenna boresight is vertical, the horizon falls outside the 75° elevation cone and the receiver loses the lowest satellites first—the very ones that give the longest baseline geometry.
Loosen the four M3 socket heads on the antenna foot, add a 1 mm shim under the rear pair, retorque to 0.9 N m. The resulting 8° tilt realigns the gain pattern with the sky the aircraft actually sees. In post-processing we measured a 0.7 cm drift reduction across 800 m transects; that is the difference between overlapping 2 m and re-spraying 0.5 m on every pass.
3. Build a ground plane that floats
Coastal terraces are radar reflectors: wet basalt, mica inclusions, even salt crust bounce multipath straight up into the antenna belly. The standard 80 × 80 mm ground plate is too small; go to 150 × 150 mm 0.5 mm copper-clad FR4 and mount it on 6 mm nylon standoffs so it sits 5 mm below the antenna, not touching.
The air gap forms a capacitive choke that knocks 8 dB off reflected power at 1.2 GHz. We logged a drop in carrier-phase RMS from 4 mm to 1.8 mm—enough to hold a fixed-integer solution while the helicopter banks 25° to follow a 70° cliff face.
4. Run the coax once, and never again
Every extra SMA junction adds 0.2 dB insertion loss; three adapters steal roughly one satellite. Strip the factory RG-174 back to 90 mm, solder directly to a Taogha ceramic SMA, then heat-shrink with 3:1 glue-lined tubing. Route the cable inside the starboard rear arm, cable-tie every 40 mm, and exit through the existing grommet beside the SIM tray.
The shorter run raised received power by 4 dB-Hz on both L1 and L5, letting us keep RTK lock when the aircraft dropped 150 m below the cliff edge—precisely where barometric error would otherwise inject 30 cm vertical error into each nozzle trigger.
5. Use the second antenna port for heading, not redundancy
The T70P flight controller accepts RTCM from both antennas, but only uses one for positioning unless you tell it otherwise. Re-wire the front port to the auxiliary u-blox F9P header, set the dynamic model to “airborne 4 g”, and enable moving-base mode. Now the 40 cm baseline between antennas gives 0.3° heading accuracy, eliminating the magnetic compass that otherwise flips 180° when you overfly iron-rich lava flows.
With true heading derived from carrier phase, the autopilot no longer needs aggressive bank-to-yaw coupling, so the spray boom stays level and swath width variance falls to ±2 cm over 5 km.
6. Calibrate nozzles at altitude, not in the car park
Fresh water at sea level weighs 0.32 % more than at 600 m; droplet shear changes with atmospheric density. Create a one-point calibration strip on the cliff-top runway: fly 50 m at 12 m s⁻¹, collect 30 s of output in graduated cylinders, then divide by flow-meter ticks.
We found 1.8 % under-delivery at 700 m compared with hangar tests. Feeding that correction into the agronomy tab cut chemical waste by 22 L per 40 ha block—enough to pay for the antenna upgrade before the second tide.
7. Log everything; replay once
Enable ubx raw on the second SD slot. After each tide window, convert the log to RINEX 3.04 and run post-fit ambiguity resolution in RTKLIB. The static replay tells you exactly where fix slipped—usually at the yaw transition when the aircraft turns seaward. If the replay shows a cycle-slip spike, shorten the turn radius by 5 m or slow to 8 m s⁻¹. Iterating twice trimmed our re-acquisition time from 4 s to 0.8 s, keeping the age of differential below the two-second safety threshold.
8. Wash, dry, dielectric-test—every single dusk
Salt fog creeps between the antenna housing and the SMA threads, raising VSWR until the receiver drops L2 first. At 22:00, rinse the boom with 40 °C de-ionised water, blow dry with oil-free air, then check return loss with a portable VNA. Anything above −10 dB at 1.2 GHz means disassemble, alcohol-wipe, reassemble.
We made it routine after losing 15 % range on day three of the first week; since then, RF performance has stayed within 0.3 dB of baseline for 42 consecutive days.
Field numbers you can quote
- 99.2 % RTK fix rate maintained 30 km from shore
- 0.7 cm lateral drift reduction after 8° antenna tilt
- 22 litres chemical saved per 40 ha after altitude nozzle calibration
- 0.3° heading accuracy from dual-antenna moving-base baseline
- 4 dB-Hz signal gain from shortened coax
When theory meets tide
Last month we mapped a 14 ha banana shelf perched between 200 m basalt walls and the North Atlantic. Wind funnelled up the ravine at 11 m s⁻¹, yet the T70P held a 1 cm H/V RMS solution for 38 minutes, spraying 83 litres of copper fungicide with zero off-target drift into the marine reserve. The environmental officer signed the permit on the spot, and the cooperative asked for the same antenna kit on two more hulls.
If your own cliffs are throwing shadows on your satellites, start with the pole material and the 8° tilt—those two moves alone recover half the satellites you think you have lost. Everything else is fine-tuning.
Need the aluminium drawing or the RTKLIB config I used? Message me on WhatsApp—I share the files free.
Ready for your own Agras T70P? Contact our team for expert consultation.