Why Most First Hardware Products Should Use a Module Instead of a Raw Chip

Choosing a pre-certified BLE module over a raw chip for your first hardware product

A $2.50 Nordic nRF52 chip versus a $6 pre-certified module. Your EE friend says go chip-down: “It’s the same silicon, why pay 2.4x?” Three months later you’re staring down a $25K FCC test bill, a failed antenna match, and a CM who just swapped a 0402 cap for “an equivalent.”

The $3.50 you saved per unit just cost you a quarter.

That’s the BLE module vs chip decision, and for v1 the math almost always points one way: default to a module. Here’s why, with real numbers, real thresholds, and an honest section on when chip-down does win.

The Real Difference Between a Module and a Chip-Down Design

A wireless module is a postage-stamp-sized board with the SoC, crystal, decoupling caps, matching network, antenna, and RF shield already on it. The vendor designed the RF section, tuned it, and paid a test lab to certify the whole thing as an intentional radiator. You solder the module to your PCB and inherit that work.

Chip-down means you put the bare SoC on your board and design the rest yourself.

  CHIP-DOWN                    MODULE
  ┌──────────────────┐         ┌──────────────────┐
  │  SoC             │         │  ┌────────────┐  │
  │  + Crystal       │         │  │  Module    │  │
  │  + Caps/Inductors│   vs    │  │ (all-in-1) │  │
  │  + Matching Net  │         │  └────────────┘  │
  │  + Antenna       │         │                  │
  │  + RF Shield     │         │  Your PCB        │
  │  + RF Layout     │         │                  │
  └──────────────────┘         └──────────────────┘
   You own all of it.           Vendor owns the RF.

The silicon is identical. The extra $3.50 buys you finished RF design, RF testing, RF supply, and RF certification.

The Five Hidden Costs of Going Chip-Down on V1

1. FCC/IC/CE intentional radiator testing. A full intentional-radiator test for FCC Part 15 plus IC and CE runs $15K to $30K at labs like Element, UL, or TÜV. If you fail (and first-time chip-down designs often fail emissions or spurious limits), you re-spin the PCB and pay again. Modules carry modular approval under FCC KDB 996369, so your product inherits the cert. You’ll still do verification testing ($2K to $5K), but you skip the intentional-radiator gauntlet.

2. RF tuning and antenna design. A chip-down design needs an EE who owns a vector network analyzer and knows what an S11 plot is. That person costs $150 to $250 an hour. Tuning a PCB trace antenna with matching network takes 2 to 6 weeks if it goes well. If it goes poorly, three months. The usual culprits: wrong ground plane, a nearby battery detuning the antenna, plastic housing shifting resonance.

3. Schedule risk on every spin. Every RF problem means a new PCB revision: 3 to 4 weeks fab, assembly, retest. Two spins and you’ve lost a quarter.

4. Supply chain fragility. Chip-down means you source the SoC, the 32 MHz crystal, the matching components, the antenna, and the shield independently. When your CM substitutes a “compatible” 0402 inductor in the matching network, your range drops 40% and nobody notices until field returns. Modules ship as one part number with one datasheet and one cert.

5. Firmware bring-up. Module vendors ship validated radio stacks and known-good board files. Going chip-down, you’re debugging your own RF behavior, which is a different sport from debugging firmware. Logic analyzers won’t help you here.

Rough comparison for a typical 10K-unit v1:

                          CHIP-DOWN     MODULE
  BOM (per unit)          $2.50         $6.00
  Cert (one-time)         $25K          $3K (verification)
  RF engineer (3 mo)      $45K          $0
  Antenna tuning          $10K          $0
  Schedule risk           High          Low
  Break-even volume       ~75K units    —

At 10K units the module costs you $35K more in BOM but saves roughly $77K in NRE and gets you to market 2 to 4 months sooner. You break even on per-unit savings somewhere around 75K units, and that’s assuming nothing goes wrong with the chip-down design. It will.

When Chip-Down Actually Makes Sense

Let me be fair here, because the module vendors won’t be. Chip-down is the right call when:

  • Annual volume is above 50K to 100K units and the per-unit BOM savings actually move the P&L.
  • Form factor demands a custom antenna. Medical implants, ultra-thin wearables, anything where a 10x13 mm module won’t fit or where the antenna pattern has to be tuned to the housing.
  • You have an in-house RF team with a VNA, an anechoic chamber (or budget to rent time at one), and the patience to spin boards.
  • The product is locked and stable. If you’re still figuring out what the product is, don’t optimize the radio.

A simple decision tree:

  Annual volume > 75K?  ──No──► Use a module
         │
        Yes
         │
  In-house RF team? ─────No──► Use a module
         │
        Yes
         │
  Custom antenna required? ──No──► Probably still a module
         │
        Yes
         │
      Chip-down

Picking the Right Module

Once you’ve decided on a module, the choice space narrows fast. A quick framework:

Radio. BLE covers most v1 consumer and commercial products: wearables, sensors, asset trackers, anything that pairs to a phone. Add Wi-Fi if you need to talk to the cloud without a phone bridge. Pick Thread or Matter if you’re targeting Apple Home, Google Home, or Alexa ecosystems. Building a tracker that has to work without nearby infrastructure? Pick a module compatible with Hubble’s Bluetooth-to-satellite network, so a standard BLE radio can reach orbit.

Certifications. Confirm FCC, IC, and CE up front. If you’re shipping to Japan add MIC, to Korea add KC, to Brazil add Anatel. Module datasheets list this clearly; if a vendor is cagey, walk away.

MCU headroom. Pick a module whose host MCU has 2x the flash and RAM you think you need. v1 firmware always grows. nRF52840 (1 MB flash, 256 KB RAM) is the safe default for BLE; nRF52833 if you’re cost-sensitive.

Vendor ecosystem. Nordic-based modules from Fanstel, Raytac, and Insight SiP have the best tooling and the largest community. ESP32 modules from Espressif win on Wi-Fi+BLE combo and price. Silicon Labs is strongest for Thread and Matter. TI’s CC2340 is the new low-cost BLE contender.

Footprint stability. Pick a module family with multiple pin-compatible options (Raytac’s MDBT series, Fanstel’s BT840 family). That way you can swap up to more flash or down to a cheaper variant without a board re-spin. Hubble’s reference applications on GitHub cover Nordic, Silicon Labs, and TI families, so you can prototype on any of them.

The V2 Migration Path

Margin-anxious founders worry that shipping on a module locks them in. It doesn’t. The standard playbook:

  1. v1 ships on a module. You learn what the product actually is, what features matter, what the real BOM volume looks like.
  2. Once volume justifies it (typically 50K+ units/year), v2 goes chip-down using the same SoC family as the module. The firmware stack carries over: same Zephyr or SoftDevice build, same radio configuration, same OTA flow.
  3. You amortize the NRE for chip-down across a known-good product, not a hypothetical one.

Most teams I’ve seen ship 2 or 3 product generations on modules before chip-down ever makes sense, and plenty never bother because their volumes don’t justify the engineering spend.

Ship First, Optimize Second

The job of v1 is to reach the market and learn what’s wrong with your assumptions. Every dollar and week you spend on RF matching networks is a dollar and week you didn’t spend on the parts of the product that actually differentiate you: the sensor, the app, the user experience, the data.

Pre-certified modules let you trade roughly $3.50 of BOM for 2 to 4 months of schedule and $40K+ of NRE. On v1, that’s a trade you make every time.

Once your device is in the field, the hard problem shifts from RF to fleet: provisioning, OTA, telemetry, and getting data back from devices that wander outside Bluetooth range.

Building your first connected product? See how Hubble handles the connectivity layer once you ship → hubble.com

Hubble Network gives modules global connectivity out of the box—no gateways, SIMs, or cellular integration required. See how it works →