Modules vs SoCs vs SiPs: Choosing the Right BLE Package for Your Product
The most expensive BLE component decision isn’t the one you get wrong. It’s the one you make by default. A team picks a $12 module because it’s familiar, ships 50,000 units, and realizes they’ve left $300K on the table annually. Another team picks a $2 SoC to save on BOM, then burns $40K and four months on failed FCC testing and RF layout rework. Both teams had engineers who understood BLE. Neither had a clear framework for choosing how BLE should live on their board.
The problem is that vendor marketing has thoroughly blurred the lines between SoCs, SiPs, and modules. A “module” from one vendor is a SiP with an antenna from another. A “system in package BLE” component from one catalog is barely distinguishable from a bare SoC in another. If you’re in the architecture phase of a project and trying to make a BLE module vs SoC decision, or wondering whether a SiP belongs in the conversation at all, you need precise definitions and a repeatable decision process, not a datasheet beauty contest.
That’s what this piece provides. It serves as the foundational reference for the BLE Chip Form Factors pillar, and every other guide in this series links back here.
What Each Package Actually Contains
The three categories differ in one fundamental way: where the integration boundary sits. Here’s what’s inside each, and what’s left to you.
SoC (System on Chip)
A bare IC: BLE radio, microcontroller core, Flash/RAM, and peripherals on a single die. You design everything around it, from matching network and crystal placement to antenna, PCB layout, and RF tuning. You own RF performance, and you own the entire certification process.
Examples: nRF52840, EFR32BG22, CC2652R.
SiP (System in Package)
Multiple dies and/or discrete passives co-packaged into a single solderable component. Typically integrates the SoC die plus crystal plus matching network, sometimes with additional filtering. You get a BLE-ready component with minimal external parts, but you still choose and integrate the antenna.
Examples: nRF5340 QKAA (SiP variant), BGM220S, CC2652RB (crystal-less, quasi-SiP approach).
Module
A fully integrated, pre-certified sub-assembly: SoC + crystal + matching network + antenna (PCB trace or chip) + shielding, all on a small daughter PCB that solders onto your main board. FCC/CE/IC regulatory certification is typically included.
Examples: BMD-340, ANNA-B412, ISP1907-LL.
┌─────────────────────────────────────────────────────────┐
│ INTEGRATION LEVEL │
│ │
│ SoC SiP Module │
│ ┌─────┐ ┌──────────┐ ┌──────────────┐ │
│ │ Die │ │ Die+Xtal │ │ Die+Xtal+Ant │ │
│ │ │ │ +Match │ │ +Match+Shield│ │
│ │ │ │ │ │ +Cert │ │
│ └─────┘ └──────────┘ └──────────────┘ │
│ │
│ You add: You add: You add: │
│ • Crystal • Antenna • Power │
│ • Match network • Power • Host MCU (if │
│ • Antenna • Host MCU (if applicable) │
│ • RF layout applicable) │
│ • Certification │
│ │
│ ◄── More control More convenience ──► │
│ ◄── Lower BOM cost Faster time-to-market ──► │
└─────────────────────────────────────────────────────────┘The Tradeoff Matrix Engineers Actually Need
Definitions are necessary but insufficient. Here’s how the three options compare across the dimensions that drive BLE hardware selection decisions.
┌──────────────────┬────────────┬────────────┬────────────┐
│ Criterion │ SoC │ SiP │ Module │
├──────────────────┼────────────┼────────────┼────────────┤
│ Unit BOM cost │ Lowest │ Mid │ Highest │
│ (at volume) │ $1–3 │ $2–5 │ $5–15 │
├──────────────────┼────────────┼────────────┼────────────┤
│ Board area │ Largest* │ Smallest │ Largest │
│ │ (ext parts)│ │ │
├──────────────────┼────────────┼────────────┼────────────┤
│ RF expertise │ High │ Low-Med │ Low │
│ required │ │ │ │
├──────────────────┼────────────┼────────────┼────────────┤
│ Certification │ You own │ Partial │ Included │
│ burden │ 100% │ (antenna │ (usually) │
│ │ │ dependent)│ │
├──────────────────┼────────────┼────────────┼────────────┤
│ Time to first │ Months │ Weeks │ Days-Wks │
│ working proto │ │ │ │
├──────────────────┼────────────┼────────────┼────────────┤
│ Supply chain │ Flexible │ Limited │ Vendor- │
│ flexibility │ │ sources │ locked │
├──────────────────┼────────────┼────────────┼────────────┤
│ Antenna control │ Full │ You pick │ Pre- │
│ │ │ │ selected │
├──────────────────┼────────────┼────────────┼────────────┤
│ NRE / tooling │ Higher │ Mid │ Lowest │
│ investment │ │ │ │
├──────────────────┼────────────┼────────────┼────────────┤
│ Ideal volume │ 100K+/yr │ 10K-100K │ <10K/yr │
│ (rule of thumb) │ │ /yr │ │
└──────────────────┴────────────┴────────────┴────────────┘
* SoC die is small, but total footprint with external
components (crystal, matching network, antenna keep-out)
can exceed a SiP.A few cells deserve explanation.
The SoC board area paradox. A bare nRF52840 in QFN is 6×6 mm. But once you add a 32 MHz crystal, a 32.768 kHz crystal, the LC matching network, decoupling caps, and an antenna keep-out zone, total board real estate can easily hit 150–200 mm². A SiP integrating those passives can come in under 50 mm². If board area is your primary constraint, the SoC may actually be your worst option.
“Partial certification” for SiPs. A SiP doesn’t include an antenna, so there’s no finished intentional radiator for the FCC to test. You’ll need to certify the final product with your chosen antenna. However, many SiP vendors provide pre-tested reference antenna designs with characterized performance data, which significantly reduces your RF risk compared to a ground-up SoC design.
Volume thresholds are heuristics, not laws. A team with strong RF talent might justify a SoC at 20K/year. A team on a tight deadline might use modules at 200K/year and still make the right call. These numbers reflect where the economics typically tip, not where they must.
A Decision Flowchart for BLE Hardware Selection
START
│
▼
┌─────────────────┐
│ Do you have RF │
│ layout expertise│
│ in-house? │
└────────┬────────┘
Yes │ No
┌──────┘ └──────┐
▼ ▼
┌──────────────────┐ ┌─────────────────┐
│ Expected annual │ │ Is board area │
│ volume > 50K? │ │ highly │
│ │ │ constrained? │
└───────┬──────────┘ └────────┬────────┘
Yes │ No Yes │ No
┌──────┘ └────┐ ┌─────┘ └─────┐
▼ ▼ ▼ ▼
┌──────┐ ┌──────────┐ ┌─────┐ ┌──────────┐
│ SoC │ │SiP or │ │ SiP │ │ Module │
│ │ │Module │ │ │ │ │
└──────┘ │(evaluate │ └─────┘ └──────────┘
│ timeline)│
└──────────┘This flowchart is a starting point, not gospel. Real projects involve interacting constraints. You might have RF expertise but a 10-week timeline that makes a SoC approach unviable. Use it to anchor the conversation, then adjust.
The first gate, RF layout expertise, is there for a reason. It’s the single most important question, and underestimating it is the number-one cause of BLE hardware schedule slips. “We have a guy who did a WiFi layout once” does not count. BLE matching network tuning, antenna placement relative to ground planes and battery packs, and harmonics suppression are specific skills. If you’re honest about not having them, a module or SiP saves you from learning expensive lessons on your customers’ timeline.
Pitfalls That Cost Real Money (and How to Avoid Them)
Choosing a SoC to save $2 without pricing in the RF rework
A failed FCC test costs $10K–$30K in retesting fees and typically adds 6–12 weeks to the schedule. At volumes under 50K/year, the $3–$8 module premium per unit is often cheaper than the risk-weighted cost of a single certification failure. Do the math before chasing BOM savings.
Assuming all modules include usable antenna certification
Some modules ship with a chip antenna on the PCB, and their FCC grant covers that antenna, but only in the vendor’s reference layout. If your ground plane geometry, enclosure material, or nearby copper pour differs significantly, you may radiate out of spec and effectively void the grant. Always read the certification grant conditions, not just the datasheet. If the grant says “as tested in reference design,” verify your layout matches or budget for intentional radiator testing.
Skipping the SiP category entirely
Many teams think in binary: module for prototyping, SoC for production. This misses the middle path. A system in package BLE component can capture 60–80% of the SoC’s cost savings while taking on only about 20% of the RF design risk. The main work you’re adding versus a module is antenna selection and placement, which is meaningful but far less daunting than designing a matching network from scratch. If you’re in the 10K–100K volume range, SiPs deserve serious evaluation.
Not planning the migration path upfront
If you prototype with a module, choose one based on a SoC or SiP you could realistically migrate to later. Pin-compatible migration paths exist. The BMD-340 module, for example, is based on the nRF5340, so firmware developed on the module transfers directly when you move to the bare SoC. Choosing a module without considering the downstream migration can mean rewriting board support packages and re-qualifying firmware months later.
The “prototype with module, produce with SoC” strategy works, if you budget for it
This is a legitimate and common approach. Validate firmware, BLE stack behavior, and system-level performance on a module. Once the product spec is frozen and volumes justify it, port to a SoC layout. Budget 2–3 months for RF layout, matching network tuning, antenna characterization, and FCC/CE certification. This timeline surprises teams who expect a drop-in swap. It isn’t one.
When the Categories Get Blurry
Vendor terminology is inconsistent, and the lines are genuinely fuzzy at the margins. The CC2652RB eliminates the external crystal using a BAW resonator. Is it a SoC or a quasi-SiP? Some “modules” are mechanically just SiPs with an antenna trace and a shield can on top. Nordic, Silicon Labs, and TI all use slightly different naming conventions.
Don’t get hung up on what a vendor calls the part. Instead, ask one question: what am I responsible for? Specifically:
- Do I need to design a matching network? (If yes, you’re working with a SoC.)
- Do I need to select and integrate an antenna? (If yes, you’re in SoC or SiP territory.)
- Can I change the antenna without affecting regulatory certification? (If no, you’re functionally in module territory regardless of what the datasheet says.)
What matters is understanding what’s inside the package and what engineering effort sits on your side of the boundary.
Building This Into Your Component Selection Process
The BLE module vs SoC decision is fundamentally about where you want the integration boundary: inside your team’s capabilities, or inside the component. Neither answer is wrong. The wrong move is choosing by default, by habit, or by optimizing a single variable (usually BOM cost) without accounting for NRE, timeline, certification risk, and team bandwidth.
Here’s a practical approach:
- Start with the flowchart to establish a baseline recommendation.
- Validate against the tradeoff matrix. Check whether any single criterion (board area, supply chain, timeline) overrides the default.
- Confirm the migration path. Even if you’re building an early-stage product, know where you’d go at 10× volume.
- Price the risk, not just the part. Include certification costs, potential rework, and schedule impact in your form factor comparison.
This decision isn’t permanent. It can and often should evolve as your product matures, your volumes grow, and your team’s RF capabilities develop. The best hardware teams revisit it at every major product revision.
For deeper dives on specific aspects of this decision, explore our guides on antenna selection for BLE products, the certification process walkthrough, and our SoC comparison guide for current-generation BLE chips.
Hubble Network connects BLE devices directly to satellites — so whichever package you choose, your product can transmit data from anywhere on Earth. See how it works →