BLE module or chip-down for a new product

You have a product that needs Bluetooth Low Energy. The prototype runs on a development kit, the firmware works, and now someone has to decide what actually goes on the production board: a finished radio module with its own FCC ID, or the bare SoC with a crystal, matching network and antenna you lay out yourself. The forums call this “module versus chip-down,” and the question comes up every week because the honest answer depends on things the datasheet does not show you: what the regulator will ask for, who owns the antenna, and how long the part will stay in production.

We design both ways. This note is the decision as we walk through it with a client, so you can walk through it yourself before the layout starts.

What a module actually buys you

A Bluetooth LE module is the SoC, its crystals, decoupling, RF matching and usually an antenna, on a small board with a metal can, sold as one part. Nordic lists third-party modules for the nRF52 and nRF54 families from vendors such as Ezurio, Raytac, u-blox, Insight SiP and Murata. Espressif sells its own modules around every ESP32 variant. ST sells the STM32WB5MMG around its STM32WB SoC. In each case the module vendor has already done three things you would otherwise do yourself.

The RF layout is done. The crystal load, the matching network and the antenna are tuned and tested as a unit. On a chip-down design those are yours, and a 2.4 GHz trace antenna that looks right in KiCad can still be detuned by a nearby battery, a metal enclosure or a hand. Getting it right means a network analyzer, a tuning pass on real boards, and time in a chamber.

The radio is certified as a modular transmitter. Under 47 CFR 15.212 a module can be granted its own FCC ID if it meets a list of conditions: its own shielding, its own power supply regulation, buffered data inputs, an antenna that is either permanently attached or uses a unique connector, and stand-alone testing outside any host. The grant belongs to the module, and a host product that follows the module vendor’s integration instructions can carry a “Contains FCC ID” label instead of going through intentional-radiator testing itself. The FCC’s guidance for both the module maker and the host integrator lives in KDB 996369; the D04 document is the one written for the host manufacturer, and it is worth reading before you commit to any module.

The Bluetooth qualification is done. Bluetooth is a trademark, and the SIG requires every product that uses it to be qualified before sale. Since July 2024 the process runs under the Qualification Program Reference Document v3, where the old Qualified Design ID became the Design Number, as Ezurio explains in its summary of the change. A product built on an unmodified qualified module follows what the SIG’s quick-start guide calls Option 1: reference the module’s Design Number, no test selections, no testing. You still submit your own product, and you still pay the product qualification fee, but you do not run the radio test suite.

What a module does not buy you

This is where the forum threads go wrong, usually in the optimistic direction.

Your product still has to pass as a whole. A modular grant covers the intentional radiator. Your board is also an unintentional radiator, with a switching regulator, a display and a microcontroller clocking away next to the radio, and that part of Part 15 is still yours. Emissions testing of the finished product is not skipped by using a module. It is shorter and lower-risk than certifying a radio from scratch, but it is not zero.

The grant has conditions, and you can break it. If you change the antenna to something not on the module’s approved list, run the module outside the conditions in its grant, put another transmitter next to it, or make it body-worn when the grant assumed a distance, you may need your own testing or a permissive change. The vendor’s integration guide tells you exactly what the grant covers. Read it before the schematic is drawn, not after the first article is back.

The Bluetooth listing is still your listing. Option 1 is fast, but it is a submission in your company’s name with your product’s model number, and the SIG expects it done on or before the day you start selling. It is a line on the project plan, not something the module vendor does for you.

A module is one more supplier. The SoC vendor has a long-lifetime program and a second source in the sense that several module makers buy the same die. The module maker is a single company with its own product roadmap, its own minimum orders and its own end-of-life notices. When a module goes obsolete you do not just change a part number; you change the RF layout, the certification and possibly the enclosure cutout. We cover the general recovery playbook in our note on what to do when a microcontroller goes end-of-life, and modules deserve the same review at design time.

When chip-down is the right answer

Chip-down wins on three things: size, per-unit cost, and control.

Size. A module with a trace antenna has a fixed footprint and a keep-out where the antenna needs clear board. If the product is a ring, a tag or anything where every millimeter is fought over, the bare SoC in a wafer-level package with a chip antenna gets smaller than any module will.

Per-unit cost. At volume, the difference between a module and the loose parts is real, and it is the number most people start with. It is also the number that matters least until the first production run is sold. Certification, antenna tuning and the extra spin or two of the board are paid once, up front, whether the product sells or not.

Control. With the SoC on your board you choose the antenna type, the crystal, the power path and the sleep behavior. Ezurio’s own comparison of the nRF52, nRF53 and nRF54 generations is a good example of how far the SoCs have moved on sleep current and features such as Channel Sounding. A module lags the SoC it is built on by however long its vendor takes to spin and re-certify, and it locks in whatever antenna its vendor chose.

The cost of that control is a full intentional-radiator certification in every market you sell into, and a Bluetooth qualification under Option 2b with real testing, plus the antenna work above. The SoC vendor’s reference design gets you close, but a reference design is certified on the reference board, not on yours.

The questions we ask before choosing

Question	Leans module	Leans chip-down
Is there an RF engineer and a chamber in the plan?	No	Yes
Will the product ship in more than one country in the first year?	Yes	Either
Is board area the hard constraint?	No	Yes
Is the expected lifetime volume modest or unknown?	Yes	No
Does the product have a second radio, or is it worn on the body?	Check the grant	Full test anyway
Will the design need another spin within a year?	Yes	Either

A pattern we see a lot and think is sound: ship the first version on a module, with the module’s footprint on a daughter or on a clearly bounded region of the main board, and revisit chip-down once the product has a sales history. The firmware does not care. A Zephyr or nRF Connect SDK application written against an nRF54L15 module runs on a chip-down nRF54L15 with a board file change, not a rewrite. The same is true of ESP-IDF across Espressif modules and bare chips. Design the firmware so that the radio is a board choice, and the module-or-chip decision stops being a one-way door.

How we approach it

We treat this as a certification and supply question first and a BOM question second. If you send us the sketch of the product, the countries it ships to, a rough volume and the enclosure material, we can tell you which way we would go and what the compliance path looks like for each, before any copper is drawn.