How to Build Your Own BLE Asset Tag Like the Samsara AT11

A Samsara AT11 costs $25 to $55 per tag, and that’s before the per-tag subscription kicks in. Strip the enclosure off one and you’ll find maybe $8 of silicon: a Nordic BLE SoC, an accelerometer, a coin cell, and a chip antenna. The margin isn’t hiding in the hardware. It’s in the firmware, the gateway stack, and the lock-in.

If you’re an embedded engineer or a hardware founder looking at a roadmap that needs 10,000+ tracking units, that math is worth running yourself. Here’s a credible reference design for a DIY BLE asset tag, what the firmware actually does, and how to route beacons into a gateway without reinventing the backend.

Why Build Instead of Buy an AT11?

The AT11 is a solid product. Ruggedized IP67 case, BLE 5.x, multi-year battery, motion sensing, tight integration with Samsara’s fleet dashboard. If you manage a yard of trailers and you just need to know where things are by Tuesday, buy it.

You should think about a Samsara AT11 alternative if any of these apply:

• Tracking is your product, not your internal ops.
• Your volume justifies NRE: roughly 10k units/year and up.
• You need custom payloads (temperature, tamper, specific sensor fusion) that off-the-shelf tags won’t expose.
• You want to own the firmware, the keys, and the data.

This guide isn’t for hobbyists making one tag for their bike. It’s for teams shipping a SKU.

Anatomy of a BLE Asset Tag

Industrial BLE tags are boringly similar under the hood. Here’s the reference block diagram:

   +-----------------------------------------------+
   |                BLE Asset Tag                  |
   |                                               |
   |  +----------+    +--------------+   +------+  |
   |  |  CR2477  |--->|  nRF52/54 SoC|<->| Accel|  |
   |  | Battery  |    |  (BLE 5.x)   |   |(I2C) |  |
   |  +----------+    +------+-------+   +------+  |
   |                         |                     |
   |                    +----+----+                |
   |                    | Chip Ant|                |
   |                    +---------+                |
   +-----------------------------------------------+
                              |
                              | BLE ADV (1M / Coded PHY)
                              v
                   +---------------------+
                   |  Hubble Gateway     |
                   | (scans, forwards)   |
                   +---------------------+

Five subsystems do all the work:

1. SoC/radio. A single-chip BLE MCU handles the advertising and any light logic.
2. Motion sensor. A 3-axis accelerometer with a wake-on-motion interrupt. The SoC sleeps until the accel pokes it.
3. Power. Usually a CR2477 or CR2032 coin cell. No regulator if you can avoid it.
4. Antenna. A chip antenna with a matching network, or a PCB trace antenna if you have the board real estate.
5. Enclosure. IP67 ABS, ultrasonically welded so it can’t be opened without destroying it.

Public FCC teardowns of the AT11 suggest a Nordic nRF52-class SoC paired with a ST or Bosch accelerometer. That’s the same recipe as RuuviTag (open hardware, worth studying) and most commercial tags from Minew, Estimote, and friends. Consumer tags like AirTag and Tile add UWB or network effects, but they’re not industrial.

Bill of Materials: A Credible Reference Design

A few trade-offs worth knowing:

• nRF54L15 vs nRF52833. The nRF54L15 has meaningfully better sleep current and cleaner Coded PHY support. If you’re designing a new SKU in 2025, start there. If your team already has an NCS codebase on nRF52, the inertia is real.
• BMA400 vs LIS2DW12. Both sip current. The BMA400 has a genuinely clever always-on motion engine that can pre-filter events, which cuts SoC wakeups. For tags that spend 99% of their life sitting on a trailer, that matters.
• CR2477 vs CR2032. CR2477 triples the capacity for about 2mm of extra thickness. Usually worth it for industrial form factors.

Firmware: What Actually Ships on the Tag

The radio is the easy part. The interesting firmware work is in advertising strategy, duty cycling, and OTA.

Stack choice. Zephyr with the nRF Connect SDK (NCS). Long-term support, a production-grade BLE stack, MCUboot for OTA, and decent power management. If you need a starting point, the Hubble device SDK and reference applications target exactly this kind of tag and will save you a week of plumbing.

Advertising strategy. Pick one:

• iBeacon or Eddystone-UID if you want broad gateway compatibility out of the box.
• Manufacturer-specific data if you want to pack battery %, motion state, temperature, and a sequence counter into the adv payload. This is what you want for a commercial product.

Typical cadence: advertise at 1000 ms while stationary, drop to 200 ms for 30 seconds after a motion event. That’s the single biggest lever on battery life.

Motion-triggered duty cycling. Wire the accelerometer’s INT1 line to a GPIO with wake-on-edge, keeping the SoC in System OFF until the interrupt fires. On wake, advertise fast for N seconds, then ramp back down. The accel itself should run in a low-power mode pulling <1µA.

Power budget, order of magnitude:

Sleep w/ 1 Hz adv:     ~8 µA average
CR2477 capacity:       1000 mAh
1000 mAh / 8 µA   =    ~14 years (theoretical ceiling)
Realistic (motion, OTA, temp): 3–5 years

The “realistic” number is what you put on a spec sheet. The theoretical number is what engineering review will demand you justify.

OTA. MCUboot dual-slot, delivered over a BLE connection initiated by the gateway. You’ll rarely update a field tag, but the first time you need to patch a security bug across 50,000 deployed units, you’ll be glad you built it in on day one.

Security. A commercial product needs rotating identifiers (Apple Find My style) so tags can’t be passively tracked. The alternative is AES-encrypted payloads keyed per tenant. Hubble’s SDK handles payload encryption and key management natively, which is one of the reasons to not build this piece yourself.

Connecting to a Gateway

Gateway choice decides your advertising format, encryption scheme, and realistic range.

The gateway’s job is narrow: scan, filter on your manufacturer ID, timestamp, and uplink to cloud over LTE or Wi-Fi. The payload contract between tag and gateway should include:

• Your 16-bit IEEE-assigned manufacturer ID
• Tag UUID (16 bytes)
• Sequence counter (for anti-replay)
• Sensor flags: motion, low battery, temperature band
• Optional MIC for integrity

  [Tag] --ADV--> [Hubble GW] --MQTT/HTTPS--> [Cloud]
                                                  |
                                                  v
                                         [Your Asset App]

If you go the Hubble route, a few specifics matter. Tune scan window and interval to match your adv cadence: a 100% scan duty cycle on the gateway radio pairs well with a 1 Hz tag. Coded PHY support is the difference between covering a 2-acre yard with one gateway vs four. The gateway handles LTE/Wi-Fi backhaul so you’re not building your own MQTT broker.

One field lesson worth internalizing: design your tag firmware gateway-agnostic in payload format, but test end-to-end with your target gateway before you tool the enclosure. Scan coverage is the #1 field failure mode for DIY asset tag projects, not battery life, not firmware bugs. A tag that doesn’t get heard is a paperweight. Put 50 tags in real conditions and watch what happens.

Hubble’s terrestrial network coverage docs are worth reading before you commit to an adv strategy: they’ll tell you what the gateway expects.

Build vs Buy: A Decision Framework

Run the numbers before you pick a side.

A 10-Week Prototyping Path

1. Week 1–2. Nordic nRF52 or nRF54 DK plus a LIS2DW12 or BMA400 breakout. Validate firmware and advertising behavior on the bench.
2. Week 3–6. Spin a dev PCB at JLCPCB or similar. Validate your power budget with a current profiler (Nordic PPK2 is the cheap standard). This is where most teams find their first 2x battery-drain bug.
3. Week 6–10. Pilot 50 tags against your target gateway in a real environment. Yard, warehouse, whatever your customer’s world looks like.
4. Certification. FCC, CE, IC, and a Bluetooth SIG Declaration ID. Budget roughly $8k all-in and 6 to 10 weeks. Plan for it early.
5. Tool the enclosure last. Injection molding is expensive and irreversible. Don’t commit until the firmware and RF performance are locked.

If you want to skip the gateway and backend buildout and focus on the tag itself, create an account at hubble.com and point your prototype at the network. You’ll know within a day whether your advertising strategy actually survives contact with the real world.

Hubble Network provides global BLE coverage via satellite, so your tags can report from anywhere without deploying gateways. See how it works →