A Fix on a Dev Board Is Not a Product

Almost any GNSS module will lock onto satellites on a bench with an evaluation antenna near a window. That demo takes an afternoon. The product is a different animal: a sealed enclosure that detunes the antenna, a battery that has to last months, and users who report "it says I'm two blocks away" without a log file. Tracking projects rarely fail on software. They fail on antenna placement, on a power budget nobody wrote down, and on a receiver chosen by price alone.

Pick the Receiver by Constellation and Band, Not by Price

GPS is a habit of speech; the hardware you buy is a GNSS receiver tracking several constellations at once. Adding GLONASS, Galileo, and BeiDou roughly triples visible satellites, which matters enormously in urban canyons and under tree canopy. A single-constellation receiver in downtown Chicago can drop to four or five satellites and errors of tens of meters where a multi-constellation receiver holds a usable fix.

The bigger jump is dual-band. Single-band L1 receivers deliver 2 to 4 meters CEP in open sky and degrade badly with multipath. Dual-band L1/L5 receivers reject multipath and hold 1 to 2 meters in the same environment, at two to four times the module cost and noticeably more current. Ask what accuracy the customer pays for: asset recovery is fine at 5 meters, lane-level telematics is not.

  • Single-band, multi-constellation. The default for consumer and asset trackers: low cost, low power, adequate accuracy.
  • Dual-band L1/L5. Worth the premium where multipath dominates or sub-2-meter accuracy is a spec.
  • RTK-capable. Centimeter accuracy, but only with a correction stream and an antenna-grade RF path. A system decision, not a module swap.
  • Dead reckoning. An integrated IMU keeps estimating position in tunnels and garages. Essential for vehicle products, pointless for a static asset tag.

The Antenna Is Where Trackers Die

More field failures trace to the antenna than to every other subsystem combined. GNSS signals arrive at roughly -130 dBm, below the thermal noise floor, and the enclosure either preserves that margin or destroys it.

A ceramic patch antenna gives the best gain but needs a ground plane: a 25 x 25 mm patch on a 40 x 40 mm (1.6 x 1.6 in) ground plane behaves, while the same patch on a 20 mm sliver of copper loses several dB and takes far longer to acquire. Chip antennas fit tighter products but are sensitive to keepout violations, nearby metal, and the plastic above them. Either way the antenna must face the sky in the normal mounting orientation, so industrial design and RF design get settled together. The discipline in antenna design for wireless products applies with less margin for error.

Three rules: no metal, batteries, or metallized paint above the antenna; keep switching regulators, the modem, and high-speed digital traces away from the RF front end; and add an external LNA plus SAW filter when a cellular transmitter shares the enclosure, since a modem burst at +23 dBm inches from a -130 dBm receiver desensitizes it in practice.

Time to First Fix Drives the Power Budget

A receiver in continuous tracking draws 25 to 45 mA and flattens a 2,000 mAh cell in days, so every long-life tracker duty-cycles. The cost of duty-cycling is time to first fix: a cold start with no almanac takes 30 to 60 seconds at full current, a hot start with valid ephemeris takes 1 to 5 seconds. That difference is the product's entire battery life.

That is why the backup supply for the receiver's RTC and ephemeris memory is not optional: keep it powered between wakeups and cold starts become warm ones. Assisted GNSS cuts acquisition further. Do the arithmetic explicitly: fixes per day, seconds per fix, milliamps during the fix, plus board sleep current. A tracker reporting four times a day with warm starts can run a year on a small cell; the same tracker reporting every fifteen minutes cannot. The firmware techniques that make this work are the ones covered in low-power firmware and sleep modes, and the cell sizing question is treated in battery pack design.

Getting the Position Off the Device

A tracker that cannot tell anyone where it is is a paperweight, and backhaul usually costs more over the product's life than the GNSS hardware. Cellular LPWAN gives national coverage at roughly $1 to $5 per device per month; the tradeoffs are in LTE-M vs NB-IoT. A private LoRaWAN network eliminates recurring fees but only works where you own the gateways. Whatever the link, buffer positions on the device: coverage gaps are normal, and a tracker that discards untransmitted fixes produces gap-riddled route history customers read as a defect.

Certification and Field Testing

A GNSS receiver alone is passive, but adding a transmitter means FCC authorization, and cellular products also need carrier network certification, a separate program with its own queue and cost. Budget both, and read FCC certification for electronic products before freezing the schedule.

Then test the way the product will be used. Bench numbers mean nothing. Drive a real route with a reference receiver logging alongside, park in a garage, mount the unit in the orientations customers will use, and plot the error distribution rather than a best case. When the sky is blocked entirely, the fallbacks in GPS-denied navigation are the same ones a ground tracker needs.

Design a System, Not a Stack of Modules

Accuracy, battery life, enclosure shape, mounting orientation, and monthly data cost are one coupled problem, not a receiver plus a modem plus a battery integrated late. Fix the required accuracy and reporting interval first, then let those two numbers pick the receiver, antenna, cell, and network.

Get the RF Right the First Time

Projects House designs GNSS tracking products end to end: receiver and antenna selection, RF layout, power budgeting, firmware, backhaul, and certification. Tell us the accuracy, reporting interval, and battery life you need through our contact form.