IoT and remote telemetry
Devices that are installed once and expected to work for a decade with nobody visiting. What ends them is rarely the radio — it is a leakage path nobody measured and an antenna the enclosure detuned.
Evidence
What we can substantiate
Everything below is work we have actually done, described at a level that respects the client’s confidentiality. If you need detail, ask under NDA.
- Sleep current measured on the assembled board Datasheet sleep figures describe the die under ideal conditions. Real boards run five to fifty times higher because of regulator quiescent current, a pull-up on an idle-high line, a sensor that never enters standby, or a floating input drawing through its protection diodes. Finding out which is one of the most valuable days in a battery project, and it cannot be done from a spreadsheet.
- Antenna tuned in the enclosure, with the battery in place A ground plane, a battery pack, a wall thickness and a mounting screw all detune an antenna, and a design tuned on the bench loses several decibels once assembled. Tuning happens with the real mechanics, and the matching network keeps a footprint so it can be adjusted rather than redesigned.
- Converter efficiency checked at the sleep load A regulator rated 92% at 100 mA can be under 50% at 10 µA because its own control circuitry draws a fixed amount. In a device that sleeps 99.9% of the time, light-load behavior matters far more than the headline figure.
- Duty cycle planned against the regional limit Sub-GHz airtime is regulated, and an SF12 uplink in EU 868 can force minutes of enforced silence under the 1% duty cycle. Airtime is budgeted at design time, not discovered when the fleet scales.
- Secure update path decided at schematic stage Signed images, on-device verification, rollback protection and a recovery slot need flash capacity, boot ROM options and key storage that are fixed the moment the schematic is. This is the CRA requirement that cannot be retrofitted, and it is the one most often missing.
- Cold behavior verified, not assumed Cell capacity falls and internal resistance rises together in the cold, so a radio pulse that was fine in summer browns out the microcontroller in January. Verified at the minimum specified temperature, at end-of-life cell impedance.
Common questions
Straight answers
Why do IoT devices miss their battery targets so badly?
Because the calculation is done on datasheet sleep currents and the assembled board does something else. A regulator with high quiescent current, a floating input, a sensor that never enters standby, or a converter that is 50% efficient at microamps will each cost more than the radio does. We measure the real board rather than trusting the arithmetic, and the first prototype almost always reveals something. Our battery estimator models all four of those effects if you want to see the shape of it first.
Does the Cyber Resilience Act apply to a simple sensor?
Almost certainly, if it connects to anything and is sold in the EU. The scope is "products with digital elements" and indirect connections count — a wired sensor reporting to a gateway is in scope. The expensive requirement is the secure update path, because it depends on hardware decisions made in the first schematic: flash for two images plus recovery, verified boot, and key storage that is not a constant in the binary. Our CRA scoper walks the classification in eight questions.
LoRaWAN, NB-IoT or LTE-M?
It is a coverage and power question, not a technology preference. LoRaWAN wins on battery life and cost per node but needs gateways you or someone else must own and maintain. NB-IoT and LTE-M use existing carrier coverage with no infrastructure of your own, at higher energy per message and a subscription per device. If the deployment is dense and private, LoRaWAN usually wins; if it is sparse and geographically scattered, cellular usually does. We will run the numbers for your actual deployment rather than recommending by habit.
Got a board to design — or one that won’t boot?
You talk to the engineer who would do the work. Reply within one business day, and we’ll sign your NDA before you go into detail.