If you've shipped an IoT tracker, you already know the GNSS module is where the bill of materials and the field-failure log collide. A tracker that can't get a fix in a parking garage, or burns 200 mW idling on the GPS engine, is a returned unit. Sourcing the module from China can cut your unit cost meaningfully, but the cheaper part is where most of the silent failures live. Here's what actually matters.
Module selection: u-blox-class vs lower-cost Chinese GNSS
The reference point most engineers use is a u-blox MAX-M10 or NEO-M9 class module. They're well-documented, multi-constellation, and the AGPS/assisted-fix story is mature. The trade-off is price: a u-blox module can run $8-15 in modest volume, while a Chinese GNSS module built around a Quectel L76/L80, an Allystar, or an AT6558 (BDS/GPS/GLONASS) chipset lands closer to $2-5.
The cheaper modules are not automatically worse. They're worse when you don't validate them. The common gaps:
- Constellation support that's real on the datasheet but flaky in firmware. "Multi-constellation" can mean GPS + BeiDou only, or it can mean the GLONASS/Galileo channels exist but degrade lock under weak signal.
- TTFF claims measured under open-sky lab conditions. A datasheet cold-start TTFF of 30 s can become 90+ s on a real PCB with a marginal antenna.
- AGPS/AssistNow ecosystem. u-blox has a hosted assistance service. A no-name chipset may need you to roll your own ephemeris injection, or it works only with a specific cloud you don't want a dependency on.
The antenna is half the system
You can't evaluate a GNSS module in isolation from its antenna. Two real choices:
- Active vs passive. A passive antenna saves a few cents and ~10-20 mW but needs a clean, short RF trace and good ground plane. An active antenna (built-in LNA) tolerates a worse layout and longer cable, at the cost of power and BOM.
- Ceramic patch vs chip antenna. A 25×25 mm ceramic patch with a real ground plane outperforms a tiny chip antenna by several dB of gain — often the difference between a 90-second fix and no fix at all under tree cover. If your enclosure forces a chip antenna, budget for worse TTFF and plan field testing accordingly.
What to validate on samples
Don't accept a module on its datasheet. Before you commit to a production run, put 5-10 sample units through this:
- Cold-start TTFF, on your board, with your antenna — not the eval board. Power-cycle 20 times, log the distribution, not just the best case. A median under 35 s open-sky is reasonable; watch the worst-case tail.
- Sensitivity under attenuation. Feed the antenna through a step attenuator or test in a known-marginal spot (indoor near a window, urban canyon). Note where it loses lock.
- Position accuracy logged static over 10 minutes — CEP should settle inside a few meters; a wandering fix means a noisy front-end.
- Idle/tracking current measured with a current probe, against the datasheet. Cheaper modules sometimes hide a higher real-world draw.
- Re-marked chip check. Pull the can, inspect the die markings under magnification, and confirm the chipset matches what you ordered. Substituting a cheaper IC under the same module label is a known trick.
Where outside help pays off
The sourcing failure mode isn't picking the wrong module — it's getting a different module than you approved once you're at 5,000 units. The sample passes; the production lot ships with a downgraded chipset or a swapped antenna. Catching that needs someone on the ground who reads the datasheet and physically inspects the lot.
If you don't have an engineer in Shenzhen, an engineering-led agent like China Sourcing Agents does the on-site QC step — a pre-production sample check plus a pre-shipment inspection of 80-100% of units — which is exactly where the substituted-GNSS problem gets caught before it leaves the factory. For a tracker, getting that one check right is cheaper than a return wave.
Pick your module against your own antenna and your own enclosure, validate the TTFF distribution rather than the headline number, and treat the production lot as guilty until inspected.
Top comments (0)