Skip to content

Module alternatives — is there a better-certified way to build this board?

Status: working document, not normative. Research pass, 2026-08-11. Question it answers: "BM20C and WM02C are just two modules using the nRF54LM20A and the nRF7002. There are many more. Is there one with a more permissive FCC certification, or less work overall?" Reads onto: Module Grant Conditions.md (why this question got asked) · NRF Redesign.md (the fallback analysis this extends) · FCC Respin.md §1 (the antenna keepout, which one option deletes entirely).

⚠️ Grant text could not be verified for any module below. fccid.io, fcc.report and apps.fcc.gov all return 403 to this machine, so every certification claim here comes from a vendor page or an integration manual, not from the grant itself. Treat the vendor claims as leads, not facts — reading the actual grant is precisely the step whose absence caused the WM02C problem. Steve Koster looks these up daily and can settle any row in this document in minutes.

0. The short answer

Your premise is right and the conclusion doesn't follow. There are many more modules — twelve for the nRF54LM20A, twenty-one for the nRF7002. Three findings, in descending order of how much they change:

  1. No nRF54LM20A module from any vendor holds an FCC grant yet. Raytac's says "intended," Ezurio's is "early preview," Fanstel's is in testing. The BM20C isn't uncertified because you picked a small Taiwanese vendor — it's uncertified because the whole part is too new, and Fanstel may well be first across the line.
  2. The 20 cm clause is not a Fanstel quirk. It is the industry default. u-blox — the best-documented module vendor in the business — carries word-for-word the same restriction on the NORA-B2, and tells integrators plainly that a body-worn design needs a permissive change. No module switch avoids the permissive change.

    ⚠️ Corrected 2026-08-12. This item used to read "the 20 cm / no-co-location clause" and treated the two as one restriction. They are not. The clause quoted at §4 below — from u-blox, and word for word on the WM02C grant — ends "except in accordance with FCC multi-transmitter guidelines." Co-location is permitted via the multi-transmitter procedures; the 20 cm exposure category is the unqualified part. This document quoted the exception and then argued past it. See Module Grant Conditions.md §0.

  3. What a module switch genuinely buys you is a vendor who has written down how they will help. u-blox documents the Change-in-ID process and commits in print to supplying the approval letter. Fanstel's willingness is an open question you'd have to ask. That is a real difference, and it is the only lever a swap actually pulls — plus, on the WiFi side, a u.FL part that deletes half of FCC Respin.md §1.

The decision this actually surfaces is not Fanstel vs. someone else. It is nRF54LM20A vs. nRF54L15 — a part with no certified module anywhere, against a mature certified market. §5.

1. Every nRF54LM20A module that exists

From Nordic's own third-party list, cross-checked against vendor pages.

Vendor Part Antenna Size (mm) GPIO FCC status
Fanstel BM20C Chip 10.0 × 15.0 × 2.0 66 In testing — "this month," per 2026-08-04
Fanstel BM20E u.FL 10.0 × 15.0 × 2.0 66 Same filing family
Fanstel BM20M PCB 10.0 × 14.0 × 2.0 66 Same
Fanstel BM20XE u.FL 10.0 × 21.0 × 2.0 66 Same
Raytac AN54LM-20A Chip 9.9 × 13.3 × 2.0 66 "Intended" — not granted
Raytac AN54LM-P20A PCB 9.9 × 13.3 × 2.0 66 "Intended" · est. Q3 2026
Raytac AN54LM-U20A u.FL 9.9 × 13.6 × 2.05 66 "Intended"
Ezurio BL54LM20A (453-00585) MHF4 11 × 8 × 1.8 40 "Under development: early preview"
Ezurio BL54LM20A (453-00586) Chip 11 × 8 × 1.8 40 Same
Abluetech PTR54LM20A PCB 18.0 × 13.0 × 1.8 66 Unknown
Kaga FEI EJ4M20BA1 PCB 5.1 × 11.3 × 1.5 34 Unknown
MinewSemi ME54BS13 PCB 16.5 × 12.0 × 2.4 64 Unknown — datasheet updated 2026-03

Nothing here is certified. Raytac lists nine regions and marks every one intended; Ezurio's product page is a sign-up form for preliminary documentation. Rutronik's 2025 round-up of "pre-certified nRF54 modules" lists four vendors and not one nRF54LM20A part — every entry is nRF54L15 or nRF54H20.

So the BM20C is not the weak link you might reasonably have assumed it was. Fanstel has silicon in production, modules shipping to you, and a filing in progress. Raytac is a quarter behind on availability alone. Switching vendors on the BLE side trades a known filing date for an unknown one.

1.1 Two things that are genuinely better elsewhere

  • Ezurio BL54LM20A is 11 × 8 × 1.8 mm against the BM20C's 10 × 15 × 2.0 — meaningfully smaller on a 40 × 42 board that FCC Respin.md §1.3 is about to shrink. It costs 26 GPIO (40 vs 66), which needs checking against firmware_spec_v3_nrf.md §2 before it means anything.
  • Ezurio and u-blox publish real integration manuals. §4 is what that's worth.

2. Every nRF7002 module that exists

Twenty-one, all dual-band 2.4 + 5 GHz. The ones that matter:

Vendor Part Antenna Size (mm) FCC status
Fanstel WM02C Chip 12 × 19.1 × 2.0 Granted — X8WWM02C, mobile / 20 cm; co-location permitted per multi-transmitter procedures
Fanstel WM02E u.FL 12 × 16.8 × 2.0 Same grant family
Fanstel WM01C / WM01M Chip / PCB 12 × 18.2 / 12 × 17.2 Smaller siblings, separate grant
Raytac AN7002Q Chip 10.8 × 17.1 × 2.1 Listed granted · WFA CID approved
Raytac AN7002Q-U u.FL 10.8 × 16.4 × 2.1 Listed granted
Raytac AN7002Q-P PCB 10.8 × 17.1 × 2.1 Listed granted
IndieSemiC ISC-nRF7002-A PCB 18 × 12 × 2.5 Unknown
MinewSemi ME16WS03 u.FL 23.2 × 16 × 3.6 Unknown

Raytac's AN7002Q is a credible drop-in alternative — smaller than the WM02C in both axes, dual-band, certified, and available now through DigiKey. Whether its grant is better is unverifiable from here and is exactly the wrong thing to assume twice.

2.1 The thing no module switch can fix

The 2.4 GHz SAR problem is physics, not paperwork. Every module in that table is an nRF7002 running the same radio at the same power into a similar antenna. Whatever any of them declares, it will be an order of magnitude over the 5 mm exclusion threshold, because the threshold at 5 mm is 9.6 mW and a WiFi radio is not a 9.6 mW device.

There is no nRF7002 module with a friendlier grant on this point, because the grant is describing the same physics. The routes out of the SAR problem are still the three in FCC Certification.md §5 — declare lower power, test SAR, or don't ship WiFi — and vendor choice is orthogonal to all three.

2.2 The combined modules, and why they don't work here

Six vendors sell a single module carrying both WiFi and BLE — Raytac AN02C53C / AN02U53U, Fanstel WT02E40E / WT02C40C, MinewSemi MS12SF18, Abluetech PTR5302, IndieSemiC ISC-nRF5340-nRF7002. On the face of it this is the ideal answer: one module, one grant, and the co-location clause becomes the vendor's problem rather than yours. It would also delete the whole opposite-edges argument in FCC Respin.md §1.

Every one of them pairs the nRF7002 with an nRF5340, and the nRF5340 does not do channel sounding. Channel sounding is the entire reason for the 2026-08-07 pivot — it is what replaced proximity_engine with a direct distance measurement. A combined module costs the product its ranging.

No vendor makes an nRF54L + nRF7002 combined module. Checked against Nordic's own list; the nRF54L parts are too new to have been designed into a combo yet. This is the option to re-check in six months — it is the one that would genuinely collapse the problem, and it is a plausible product for Raytac or Fanstel to build.

3. What the industry grant text actually says

u-blox NORA-B2 (nRF54L15, FCC ID XPYNORAB2) publishes a 55-page system integration manual with a full regulatory chapter. It is the clearest statement available of what a module grant in this class permits, and it is worth reading in place of guessing about any other vendor.

§6.4.1, United States compliance statement:

"The internal / external antenna(s) used for this module must provide a separation distance of at least 20 cm from all persons and must not be co-located or operating in conjunction with any other antenna or transmitter."

Word for word the X8WWM02C condition. This is boilerplate for modular grants, not something Fanstel did to you.

§6.4.3.2, FCC RF exposure — and this is the interesting one:

"NORA-B2 modules are approved for installation into mobile and/or portable host platforms and must not be co-located or operating in conjunction with any other antenna or transmitter — except in accordance with FCC multi-transmitter guidelines. To ensure that the max output power … remains below the SAR Test Exclusion Threshold …, an OEM integrator … must ensure a separation distance of 15 mm between the user and the antenna."

§6.4.6, antenna requirements:

"the u-blox grant also requires a separation distance of at least 20 cm from the antenna(s) to all persons. The antenna(s) must not be co-located with any other antenna or transmitter (simultaneous transmission) as well. If this cannot be met, a Permissive Change … must be made to the grant."

3.1 Three things this settles

A BLE module can be approved for portable host platforms, and u-blox says theirs is. That is the property the BM20C needs and the question Module Grant Conditions.md §6 tells you to ask Fanstel while their filing is open. It is achievable — this is proof — and it is a choice the vendor makes at filing time.

But "portable-approved" still comes with a distance, and u-blox's is 15 mm. A watch back is under 5 mm. So even the best-documented portable-approved BLE module in the market does not, as granted, cover a wrist device at 5 mm. (The physics is fine — FCC Certification.md §5 puts BLE at +8 dBm at a ratio of 2.5 against a threshold of 3.0 at 5 mm. u-blox's 15 mm is their declaration, not the limit.)

And co-location needs a permissive change regardless of vendor, stated flatly by u-blox in §6.4.6. Two radios on one 40 × 42 mm board is co-location under anyone's grant.

The honest conclusion: there is no module you can buy that makes the permissive change go away. A wrist-worn dual-radio product is outside the standard modular grant no matter whose module is on the board. What varies is how much the vendor helps.

4. Where vendors actually differ: the change-of-ID

This is the one axis where the choice is real, and Module Grant Conditions.md §5 identifies it as the route worth taking.

u-blox, §6.4.5, "Obtaining own FCC/ISED certification ID":

"Integrators who want to base their own certification on the u-blox certification can do so via a process called 'Change in ID' (FCC) / 'Multiple listing' (ISED). With this, the integrator becomes the grantee of a copy of the u-blox FCC/ISED certification. u-blox will support with an approval letter that shall be filed as a Cover Letter exhibit with the application."

That is the letter Steve Koster told you to ask Fanstel for, promised in advance, in a public document, as standard practice. With Fanstel it is a favour you have to negotiate with a vendor who has no contractual obligation to grant it.

Rough ranking on this axis, from documentation quality alone:

Vendor Change-of-ID posture
u-blox Documented, committed in print, named process
Ezurio (ex-Laird) Publishes per-module regulatory guides; grantee codes SQG / TFB; long OEM track record
Panasonic, Murata Large vendors, formal processes, less public documentation
Raytac High-volume module house, extensive certification lists, no public change-of-ID policy found
Fanstel Unknown. Responsive by email; nothing published

Fanstel being unknown is not the same as unwilling — they have answered every email so far within days, and the WM02C grant is a change-of-ID-able asset. But it is a dependency you currently have no evidence about, on the critical path, and asking is free. It is already item 2 of the three additions to drafts/fanstel-followup.md.

5. The decision this actually surfaces

Researching module vendors turned up something bigger than a module vendor.

The nRF54LM20A has no certified module from anyone. The nRF54L15 has a mature certified market:

🎯 X8WBC15 "with SAR" — the one row that matters

Found 2026-08-11 on page 1 of datasheets/BC15M+Product+Specifications.pdf, a file already in this workspace since 2026-08-09, and confirmed independently on Fanstel's own BC15 product page. The certifications block reads, in full:

• FCC ID: X8WBC15 with SAR. • ISED ID: 4100A-BC15 · TELEC ID: 201-250527 · CE and RCM: Compliant

Fanstel already paid for the SAR testing on this module. A grantee does not run SAR on a device certified for 20 cm mobile use — there would be nothing to measure. "With SAR" is how a vendor says this part is authorised against a body, which is the exact property a wrist-worn device needs and the exact property X8WWM02C lacks.

If the grant matches the datasheet, a BC15-based watch needs no permissive change, no change of ID, no SAR test, and no cooperation from Fanstel. It is the only part found in this entire research pass with that property.

⚠️ Verify against the grant before betting anything on it. Neither the BC15 nor the WM02C datasheet states an exposure condition anywhere — the WM02C's 20 cm appears only on its grant, not in its 36-page specification. Datasheet silence proved nothing there and a datasheet claim proves nothing here. X8WBC15 is one lookup for Steve Koster.

✅ Checked, 2026-08-11 — and it splits into good news and a correction

You pulled the exhibits yourself the same evening, and they are now in the vault: files/X8WBC15_rf_exposure_report.pdf and files/X8WBC15_MPE_report.pdf, both International Standards Laboratory report ISL-25LR090FMPE. That settles the row above without waiting on Steve. Three findings, in descending order of how much they matter.

1. The good news is real, and it is the part that mattered. The RF Exposure Report states in its own words: "According to §2.1093 this is a Portable device." §2.1093 is the portable rule — a radiator used within 20 cm of the body. That is the exposure category a wrist device needs, and it is exactly what X8WWM02C does not have. The premise of this section survives: on exposure category, the BC15 is in a different and better position than the WM02C, and it is in that position today.

2. The correction: "with SAR" does not mean a SAR test was performed. No SAR measurement exists in either exhibit. What Fanstel filed is a SAR test exclusion, by calculation, under the same KDB 447498 §4.3.1 formula this workspace has been running by hand:

8.01 dBm conducted + 1.50 dBi gain = 9.51 dBm EIRP = 8.94 mW
(8.94 / 5 mm) × √2.480 GHz = 2.82   ≤ 3.0   →  EXCLUDED

So the correct reading of the datasheet's "FCC ID: X8WBC15 with SAR" is "this filing carries an RF exposure evaluation for portable use", not "this module has been through a phantom and a robot." The practical consequence is the same in the happy case — no SAR test is needed for a BLE-only watch — but the reasoning is different, and the difference is finding 3.

3. The headroom is 0.18, and it is 0.18 because Fanstel declared a tune-up tolerance of zero. This is the thing to carry forward, because it corrects an error in our own arithmetic as well. FCC Certification.md §5 computes the BLE ratio as 2.5 from +8 dBm plus 1 dB of tune-up — and omits antenna gain entirely. Fanstel's lab includes 1.50 dBi of gain and declares 0 dB of tune-up. Include both, and the margin is gone:

Declared Ratio Verdict
8.01 dBm + 1.5 dBi gain, 0 dB tune-up (as filed) 2.82 excluded, by 6%
+ 0.5 dB tune-up 3.15 not excluded
+ 1.0 dB tune-up (the figure FCC Certification.md §5 assumes) 3.46 not excluded

A BLE-only build is SAR-excluded only at a declared tune-up tolerance of essentially zero, and a tight declared tolerance is paid for in production yield. That claim is currently load-bearing in FCC Certification.md §5, launch-plan.md §5, BLK-15 and §5.1 below, and all four state it with more confidence than 0.18 of margin supports. It still clears — this is a caution, not a reversal.

Two things the exhibits do not settle, both still one lookup for Steve:

  • Co-location. These are exposure exhibits, not the grant. What X8WBC15's grant says about co-location is unanswered — expect the same "except in accordance with FCC multi-transmitter procedures" wording X8WWM02C carries, and for a BLE-only build with one radio it may not bite at all, which is §5.1's point.
  • The two exhibits disagree with each other. The MPE report, same report number, opens "This is a Mobile device, the MPE is required" and evaluates at 20 cm using 6 dBi of antenna gain and 1 dB of tune-up — against the exposure report's 1.50 dBi and 0 dB. A filing that covers both categories is normal and is good for us; a filing that uses two different antenna gains for the same module is a question, and the 6 dBi figure is the one that would matter if anyone ever re-ran the portable arithmetic against it. At 6 dBi the portable ratio is over 8 and nothing is excluded.

Note the part number carefully. BC15 and BM15 are different modules with different FCC IDs — X8WBC15 and X8WBM15C. launch-plan.md §5 already flagged this naming risk; it is now load-bearing, because only the BC15 datasheet says "with SAR." The BM15 page makes no such claim. Ordering one letter wrong buys the module without the property.

Vendor Part FCC Notes
Fanstel BC15C / BC15E / BC15M / BC15P X8WBC15 — §2.1093 portable, SAR-excluded at 5 mm (verified 2026-08-11) Production. 256 KB / 1524 KB · channel sounding · +8.01 dBm · 6.0 × 11.0 mm (BC15C) down to 6.0 × 6.3 mm (BC15P)
Fanstel BM15C / BM15E / BM15M X8WBM15C In production, in stock. 256 KB / 1524 KB. No "with SAR" claim — different part, different grant
u-blox NORA-B20 series XPYNORAB2 Portable-approved, 15 mm, documented change-of-ID
Panasonic PAN B611-1x Granted "Certified for CE RED, FCC, ISED, MIC"
Ezurio BL54L15 series Granted 14 × 10 × 1.6 mm
Raytac AN54LQ-15 / -P15 / -U15 Granted Chip / PCB / u.FL
Murata Type 2NR Q1 2026 7.0 × 7.4 mm SiP
InsightSIP ISP2454-MX Unknown 8 × 8 × 1 mm — smallest by a distance
MinewSemi ME54BS01 series Unknown —

launch-plan.md §5 and BLK-15 already carry the nRF54L15 fallback, framed as a schedule contingency — what you fall back to if the BM20C grant slips. This research says it is also a certification strategy, and a considerably stronger one:

  • A grant exists today, from several vendors, one of which (X8WBM15C) is the vendor you are already talking to and is in stock.
  • u-blox's NORA-B20 is portable-approved with a documented change-of-ID path — the two properties this whole problem needs, from one part, in writing, before you ask anyone anything.
  • Channel sounding survives — settled 2026-08-09 from the nRF54L15 datasheet, and it is why the fallback was ever acceptable.
  • The cost is unchanged and still real: RAM halves, 256 KB against 512 KB, which is the exact constraint that drove the pivot off it in the first place. And on Fanstel's part specifically, no WiFi — though that is a Fanstel BM15C property, not an nRF54L15 property, and a u-blox or Raytac nRF54L15 alongside a Raytac AN7002Q is a combination nobody has costed.

That last point is worth stating plainly, because no document currently says it: the "BM15C means no WiFi" constraint in launch-plan.md §5 and BLK-15 is about Fanstel's module, not about the silicon. Whether an nRF54L15 from another vendor can host the nRF7002 over QSPI is a real question with a real answer, and it decides whether the fallback still costs getOnWifi / getOffWifi and WATCH_REMOVED.

And the reason given for that constraint does not survive inspection. launch-plan.md §5 says the BC15C "cannot drive the WM02C." The BC15 datasheet's pin table maps a full QSPI bus — QSPI CLK, D0–D3, CSN on P2.00–P2.05 — which is the interface the nRF7002 uses. Its Ver 1.03 revision note (March 2026) reads "Remove WiFi combo pins," meaning Fanstel withdrew the WiFi-combo documentation from a module that still exposes the bus.

Do not read that as "it works." The likely real constraint is RAM, not pins: Nordic's nRF7002 host driver is normally paired with an nRF5340 / nRF9160 / nRF54H, and the Wi-Fi stack's footprint against 256 KB is the plausible reason Fanstel stopped advertising the combination. That is the same 256 KB that drove the pivot off this part. But the constraint is a software footprint question with a checkable answer, not the hardware impossibility §5 asserts, and the two have very different implications for whether a BLE-only board is permanently BLE-only.

5.1 What BLE-only actually saves

Costed against the project's own figures — FCC Certification.md §5 (~$6k filing vs ~$20k with SAR), launch-plan.md §5 (the timing table) and §12.5 ($15k → $35k for the permissive-change scenario).

Removed outright

Why it goes
SAR testing 1–2 weeks of chamber time and the largest single line item. BLE is exclusion-qualified at 5 mm — ratio 2.82 vs 3.0 on Fanstel's own filing, and only at a near-zero declared tune-up tolerance (§5's correction)
The whole permissive-change apparatus No change of ID, no C2PC, no Fanstel approval letter, no KDB 178919 risk. X8WBC15 is portable-granted — confirmed 2026-08-11, leaving only the co-location clause unread, which one radio may moot anyway
Co-location One transmitter. No multi-transmitter procedures, no summing question, no coex-arbitration argument to win
5 GHz scope §2.1's missing-15E question becomes moot
The firmware grant condition CONFIG_NRF700X_* gain and band-edge backoff (FCC Certification.md §3) leaves firmware_spec_v3_nrf.md §1.1
Most of the Part 15B mode matrix FCC Lab Quote Responses.md §2 drops 802.11 b/g/n/ax, a/n/ac/ax, and the coex interface
BLK-15 entirely A grant that exists today replaces one that is being filed. This is the critical-path item, not a cost item
Half of FCC Respin.md §1 One module to place, free to overhang, no opposite-edges constraint, no 30 mm separation problem, no QSPI length-matching, no VDDIO load switch, ~$5 of BOM
Failure-and-retest risk launch-plan.md §5: each failure adds 3–6 weeks, and SAR is the likeliest to fail

Unchanged

Part 15B SDoC on the finished product is the irreducible core and it does not shrink much. The unintentional-radiator sources are the 12 SK6805s, the vibration motor, the buzzer, the servo, the linear charger and USB-C — every one still present. Also unchanged: the 2–6 week lab booking lead time (still the binding constraint), the anchor's servo and mains-PSU scope, the battery-only test mode, production-representative samples, the enclosure work, and all of FCC Respin.md §2 — the power tree is untouched by any of this.

The number

WiFi build BLE-only on BC15
Certification ~$20k lab + change of ID + C2PC → §12.5 territory ~$6k
Booking → signed report 6–10 weeks ~4–7 weeks
Depends on Fanstel Yes — approval letter, grant conditions, filing scope No
Blocked on a pending grant Yes (BLK-15) No

Roughly a 3–4× reduction on the certification line, and the removal of every external dependency on it. The schedule saving is smaller than the cost saving, because booking lead time dominates and doesn't move.

Two BLE-only builds, not one

They are not equivalent and the difference is the whole trade:

BM20C, WM02C depopulated BC15 build
RAM / flash 512 KB / 2 MB 256 KB / 1524 KB
FCC grant Pending — BLK-15 stays alive Exists today
Exposure category Unknown — the §6 question §2.1093 portable, SAR-excluded at 5 mm — verified from the filing, 2026-08-11
Board No respin needed to drop a part New footprint — but 6.0 × 11.0 mm against 10 × 15
Channel sounding Yes Yes

The first keeps the RAM and keeps the blocker. The second clears the blocker and pays the 256 KB that started the pivot.

6. If you change anything, change the WiFi antenna variant

Independent of every certification question, and cheap:

FCC Respin.md §1.2 spends its whole argument on a constraint that a u.FL part deletes. The current plan puts the WM02C on the west edge with its 7.4 mm chip antenna overhanging, which shrinks the PCB to ~32.6 × 42 mm and probably changes the enclosure — §1.3 calls that "the item most likely to cost time somewhere else."

A u.FL WiFi module has no mounting restriction and no keepout, because its radiator is on a cable. FCC Certification.md §2 already records this: the WM02E is "No mounting restriction." Options:

  • Fanstel WM02E — same grant family as the WM02C, u.FL, 12 × 16.8 mm, 2.3 mm shorter, pads interchange with the footprint already drawn.
  • Raytac AN7002Q-U — u.FL, 10.8 × 16.4 mm, smaller again, but a different footprint and an unverified grant.

The trade is a u.FL cable and an external antenna inside the case, against keeping the 40 × 42 outline, keeping the enclosure, and removing one of the two mandatory items in FCC Respin.md §1. It also gives the antenna somewhere to go that isn't 1.15 mm from the BM20C — which is the co-location geometry problem, improved by physically moving the radiator away.

This does nothing for SAR. It is a layout and schedule win only. But it is available without waiting for anyone to answer an email.

7. Recommendation

Do not switch modules to solve the FCC problem. It doesn't solve it. The permissive change is required by the product's shape — wrist-worn, two radios — not by the parts. Switching now would forfeit a filing already in progress, modules already shipping to you, and a footprint already drawn.

Three things worth doing, in order:

  1. Ask Fanstel the two questions in Module Grant Conditions.md §6 — will the BM20C be certified for portable host platforms, and will you grant a Change in ID. §3.1 proves the first is achievable and §4 proves the second is routine. Both answers are free, and both determine whether any of this document matters. If Fanstel says yes twice, the current BOM is fine and this research was insurance.
  2. Cost the u.FL WiFi variant — WM02E, same grant family, same pads. It removes a mandatory respin item at the cost of a cable, and it needs no answer from anyone. (§6)
  3. ~~Re-frame the nRF54L15 fallback as a strategy rather than a contingency, and settle the question §5 ends on: whether a non-Fanstel nRF54L15 can host an nRF7002, because that decides whether the fallback still costs WiFi.~~ Settled 2026-08-12 in §8.1 — it cannot, and not for any Fanstel-specific reason: nRF7002 station mode needs 222 KB of host RAM against the nRF54L15's 256 KB total, and Nordic state that BLE/WiFi coex is unsupported on that part. The re-framing still stands, and §8 replaces this item — it works out what a portable-certifiable WiFi watch would actually be built from, since the answer is no longer "an nRF54L15 hosting an nRF7002."

And revisit combined nRF54L + nRF7002 modules in six months. Nobody makes one, six vendors make the nRF5340 equivalent, and it is the part that would make the co-location problem somebody else's. (§2.2)

8. The BC/BM split, and whether the WiFi has to be an nRF7002

Added 2026-08-12, answering two questions from the user, who had spotted the structural problem before this document had:

"The BC15C is certified as portable and the BM15C is not. The BC series is meant for low power, wrist applications; the BM series is the opposite. They didn't continue the BC series — there's no BC20C. That means in Fanstel's lineup there is no chip which is BOTH certified as portable AND has the RAM required to connect to a WiFi chip."

That is correct, and it is a sharper statement of the problem than anything above it. §5 treated portable-vs-mobile as a filing accident to be asked about. It is closer to a product-line decision: Fanstel certifies the wrist-oriented parts for the wrist, and the BC series stops at the nRF54L15.

8.1 The RAM claim, verified — and it is worse than "tight"

Nordic's own figures for an nRF7002 station-mode application, measured on an nRF5340 host:

Configuration RAM Flash
Wi-Fi station mode 222 KB 497 KB
Scan-only (minimal) 74 KB 104 KB

The nRF54L15 has 256 KB of RAM total, and Ezurio publish an FAQ explaining that applications get less than that because part is reserved for the FLPR RISC-V core. 222 KB of WiFi against ~256 KB shared with the BLE stack, channel sounding, the enforcement state machine, the LED driver and NVS does not fit. This is not a tuning exercise.

And Nordic say so directly. On DevZone, asked about nRF54 + nRF7002: coexistence is not supported on the nRF54L15, a user porting the ble_coex sample hit RAM overflow, and Nordic's recommendation was to use an nRF5340 for a simultaneous BLE/WiFi design. So the constraint is two constraints — capacity and an unsupported configuration.

This settles the question Module Alternatives.md §5 and §7.3 left open — "whether a non-Fanstel nRF54L15 can host an nRF7002, because that decides whether the fallback still costs WiFi." It can't. Not from any vendor, because the limit is Nordic's silicon and Nordic's stack, not Fanstel's module. §5's note that the BC15 exposes a full QSPI bus is true and irrelevant: the bus was never the binding constraint.

8.2 The reframe: the nRF7002 is a companion IC, and that is why it costs 222 KB

The nRF7002 has no host of its own. The WiFi MAC, the supplicant and TCP/IP all run on your MCU. That architecture is why a WiFi radio the size of a fingernail consumes more RAM than the entire rest of this product.

Most WiFi modules are not built that way. The offload class — "Internet-on-a-chip" — carries its own MCU and runs the whole stack on-module, exposing sockets or AT commands over SPI/UART. Host cost is kilobytes.

⚠️ Leads, not facts — same caution as §1 and §2. These come from vendor pages, not grants:

Part Interface Why it's on the list
TI CC3200MOD / CC3235MOD SPI / UART Dedicated network-processor MCU on-module; embedded TCP/IP and TLS/SSL; FCC / IC / CE / TELEC
Inventek ISM43362 / ISM43340 SPI or UART Integrated TCP/IP, FCC-certified, long OEM history
Microchip ATWINC1500 / 1510 SPI The classic minimal-host part; Microchip publish a SAR-exemption note for their certified modules
ST SPWF04Sx UART / SPI Full stack, AT commands
Espressif ESP32-C3 / ESP8266 UART / SPI Self-contained. You already have ESP32-C3 firmware experience and a working codebase

With an offload module, BC15C becomes viable as the host — portable-certified, channel sounding, 256 KB — because the WiFi subsystem stops competing for it. That is the combination Fanstel's lineup appeared not to offer.

8.3 ⚠️ But switching WiFi modules does not buy the exposure category, and this is the part that matters

No WiFi module is going to be portable-certified at usable power, from anyone. §2.1 already established why and it applies to every part in the table above: the SAR-exclusion ceiling at 5 mm is 9.6 mW, and a WiFi radio is not a 9.6 mW device. Checked against the obvious candidate — the ESP32-C3-MINI-1 grant carries the same 20 cm separation condition and is scoped for mobile and fixed applications. Identical boilerplate, identical trap.

So:

Any WiFi radio in a wrist-worn device is outside its module grant while the device is worn, whichever module you choose. There are exactly two ways out, and part selection is not one of them: a permissive change carrying a SAR test, or guaranteeing the radio never transmits while the device is worn.

That promotes wornness-gating from a cost optimisation to the load-bearing architectural decision. It is not a clever way to save the SAR line item — it is the only architecture in which a WiFi watch is certifiable without one, regardless of vendor. FCC Respin.md §4.1 should be read as a design requirement rather than an idea to explore.

8.4 The architecture this points at

Putting §8.1–§8.3 together with FCC Respin.md §4.1:

BC15C (or u-blox NORA-B2) + an offload WiFi module + a firmware wornness gate.

Property Why it works
BLE radio Portable-certified today — X8WBC15, §2.1093, SAR-excluded at 5 mm (2.82 vs 3.0)
WiFi radio Never transmits while worn → evaluated at ≥ 20 cm under MPE, which it passes by a wide margin
Host RAM Offload module costs kilobytes, not 222 KB
BLE/WiFi coex Not required — the two radios are mutually exclusive by design, which sidesteps the configuration Nordic says isn't supported
Channel sounding Survives — nRF54L15 has it (confirmed 2026-08-09)
WATCH_REMOVED Survives — it is by definition an off-wrist message
BLK-15 Dies. No dependency on the BM20C grant, no change of ID, no permissive change

What it costs, stated honestly: a new BLE module and footprint, a new WiFi module and driver, a second respin, everything in the 256 KB budget that 512 KB was bought for, and getOnWifi/getOffWifi — which the 2026-08-09 log had already offered to the phone's GPS. It also puts a firmware-enforced exposure condition into the grant, joining firmware_spec_v3_nrf.md §1.1 — a class of requirement that fails silently.

This is not a recommendation to switch. It is the answer to "does a certifiable WiFi watch exist that doesn't depend on Fanstel," and it does. Its value right now is as the thing you know exists while you wait for Fanstel's answer to drafts/fanstel-exposure-category.md. If the BM20C comes back portable-certified, none of §8.4 is needed.

8.4.1 "Don't modules have a list of approved companion modules?" — no, and here is what you were remembering

There is no companion-module whitelist in FCC modular approval. Nothing says the BC15C may be paired with these parts. Two real things sit near that idea and get conflated with it:

  1. Approved antenna lists are real. A modular grant covers specific antenna types and gain levels, and an integrator must use one of them. That is a genuine per-module list, and it is why swapping a module's antenna is a permissive change. It says nothing about other radios.
  2. Co-location must be evaluated, and each grant must permit it. When two transmitters share a host they are co-located by definition. The requirement is procedural — the integrator performs the FCC multi-transmitter evaluation — but each module's grant does have to allow co-location at all. That is the grain of truth behind the memory.

Both grants in question allow it, and both say so in nearly identical words. X8WWM02C: "must not be co-located … except in accordance with FCC multi-transmitter procedures." And the ESP32-C3-MINI-1 grant 2AC7Z-ESPC3MINI1, read in full 2026-08-12, is even more explicitly procedural:

"Co-location of this module with other transmitters that operate simultaneously is required to be evaluated using the FCC multi-transmitter procedures."

Two grants from two unrelated vendors and two different TCBs, carrying the same construction. Co-location is a procedure, not a permission slip — which is the 2026-08-12 correction confirmed from a second independent source.

8.4.2 The worked architecture — and the one thing it rests on

Recommended co-processor: ESP32-C3-MINI-1 (FCC ID 2AC7Z-ESPC3MINI1), running Espressif's stock ESP-AT firmware over UART.

Why this part over the other offload candidates:

Reason Detail
Grant read in full Single Modular, 15C. Co-location explicitly evaluable. Unlike every other candidate in §8.2, this is not a vendor-page lead
Lower TX power than the WM02C 0.071 W conducted against the WM02C's 0.17 W — better MPE margin at 20 cm, and a smaller number to defend
No co-processor firmware to maintain Stock ESP-AT gives WiFi + TCP/UDP over AT commands. Nothing to build, and the RF behaviour is exactly what Espressif certified
You already know this silicon Every existing Impulse firmware is ESP32-C3
Not bigger than what it replaces 13.2 × 16.6 mm vs the WM02C's 12 × 19.1 — slightly less area, 0.4 mm taller
Host cost is a UART driver Which is what makes BC15C's 256 KB workable at all

Worth pricing the -1U u.FL variant in the same breath — it would delete the chip-antenna keepout that FCC Respin.md §1 spends its whole length on, the same argument §6 makes for the WM02E. Unverified whether the -1U shares the grant.

⚠️ Its BLE radio must stay disabled. The module is certified for Wi-Fi and Bluetooth; enabling both would put a third transmitter in the co-location evaluation for no benefit.

What compliance would then require, in full:

# Obligation Status under this architecture
1 Each module used inside its own grant conditions BC15C ✅ portable, SAR-excluded at 5 mm. ESP32-C3 ⚠️ depends entirely on the wornness gate
2 Certified antennas only, per each approved antenna list ✅ both use their integrated module antennas unmodified
3 Each vendor's keepout / layout integration instructions ✅ mechanical, same class of work as FCC Respin.md §1
4 Co-location permitted by both grants ✅ both explicitly permit it subject to the multi-transmitter procedures
5 Multi-transmitter RF exposure evaluation ✅ satisfiable by construction — worn: BLE alone, 2.82 < 3.0. Off-wrist: both, evaluated at ≥ 20 cm under MPE, passing by more than an order of magnitude
6 Composite investigation — host tested with all transmitters installed, simultaneous-transmission spurious ⚠️ real lab time. KDB 996369 D04 recommends it; put it in the quote rather than discovering it
7 Part 15B SDoC on the finished product ⚠️ unchanged and unavoidable — LEDs, motor, buzzer, servo, charger, USB-C

Six of seven are satisfiable on paper. Everything rests on row 1.

⛔ The load-bearing assumption, stated plainly so it cannot be skimmed

The ESP32-C3-MINI-1 grant says: "This device is to be used only for mobile and fixed applications … Separate approval is required for all other operating configurations, including portable configurations with respect to 2.1093 and different antenna configurations."

A wristwatch is a portable configuration. The wornness argument says the WiFi radio isn't in one, because exposure only exists while transmitting. That is a coherent reading and it is not what the grant text says on its face — the grant speaks of configurations, not transmissions.

This is precisely the shape of the mistake made on 2026-08-11, when a grant clause was read for what it ought to mean rather than what it said. Do not design around this until a TCB confirms it. It is question one for Steve Koster, and it decides whether §8.4 is an architecture or a dead end.

If the answer is no, this architecture collapses to the same place as the BM20C one — a permissive change with a SAR test — and the honest conclusion is that BLE-only is the only WiFi-free-of-doubt build, which is §5.1.

8.5 Question B — are any BM20C alternatives portable-certified?

No, and not because of the exposure category: none of them is certified at all. §1 found twelve nRF54LM20A modules — Raytac's grants are marked "intended", Ezurio's is "early preview", Fanstel's is in testing. There is no certified nRF54LM20A module from any vendor, so there is nothing to be portable-certified.

On the nRF54L15 side there is a second portable-approved option, and it is already in §5's table: u-blox NORA-B2, whose System Integration Manual §6.4 states the modules are "approved for installation into mobile and/or portable host platforms" — and which documents the change-of-ID process and commits in print to supplying the approval letter (§4). Two portable-approved BLE modules are known: Fanstel BC15C and u-blox NORA-B2. Both are nRF54L15. Neither can host an nRF7002, which is §8.1.

⚠️ Both claims come from vendor documentation, not grants. The BC15's is corroborated by its actual RF exposure exhibit (files/X8WBC15_rf_exposure_report.pdf); the NORA-B2's is not, and XPYNORAB2 has never been looked up. That is one question for Steve and it is worth asking in the same breath as the others.

Sources

Added 2026-08-12 for §8: