Skip to content

FCC Certification — what the datasheets already answer, and what SAR actually requires

Status: working document, not normative. The normative home is launch-plan.md §5 (regulatory strategy) and §12.5 (the cost of getting it wrong). This file is the evidence behind those sections. Created: 2026-08-09, from reading datasheets/BM20C.pdf and datasheets/WM02C+Product+Specifications.pdf end to end against BLK-15 and AUG-11. Companion: NRF Redesign.md — the fallback-module analysis this feeds into. Downstream: FCC Lab Quote Responses.md — paste-ready answers for a lab's quote intake form, built on everything below. Reuse it unchanged across all three labs so the numbers stay comparable.

Actionable items live in Todos.md. This is the reasoning and the arithmetic.

1. The headline

BLK-15 and AUG-11 were both written on the assumption that Fanstel's Part 15B integration guidance is a document you have to request. It isn't. It's in both datasheets, and you already have both. What genuinely requires an email is much narrower: the BM20C's FCC grant number and grant conditions, which don't exist yet because the module isn't certified.

Second headline: the WM02C is already certified — FCC ID X8WWM02C, ISED 4100A-WM02C, TELEC 201-240555, printed on page 1 of a Ver 1.00 production datasheet. Half of BLK-15's regulatory risk was never real.

Third, and the one that costs money: the 2.4 GHz WiFi cannot be SAR-excluded at the power Fanstel's own table declares. Section 5 works the numbers. There are three ways out and one of them is free.

2. What each datasheet actually contains

BM20C.pdf — Draft Ver 0.82, July 2026, 31 pp.

It is a draft, and it says so on every page footer. The Model Summaries table on p.1 has an FCC ID row, an ISED ID row, and CE, RCM / TELEC / NCC rows. All five are empty. That is Fanstel confirming in print what they told you by email on 2026-08-04. Keep this file — a dated draft datasheet with blank certification rows is the artefact that justifies why the fallback plan exists.

Mounting guidance (p.17) — this is the Part 15B layout guidance AUG-11 was going to ask for:

  • The RFI shield is 10.3 mm long. That region is the "main body." Everything past it is the antenna area, which on the BM20C is 5.5 mm long.
  • The main body may sit over host ground plane. The antenna area must extend beyond the edge of the host board, or over a region with no ground plane and no signal traces.
  • "We don't recommend mounting BM20C module in the middle of a host PCB."
  • For best range, keep all external metal ≥ 30 mm from the antenna area.

Antenna and PCB layout notes (p.27):

  • Don't put a module with an internal antenna inside a metal case.
  • Avoid running any signal line below the module.
  • No ground plane below the antenna.
  • If possible, cut away the portion of the host PCB below the antenna.

FCC Label (p.29) is generic OEM boilerplate: label the finished enclosure with the FCC ID, a second label on the outer device if the ID isn't visible, the end product may need Part 15B unintentional-emissions testing, "intended for OEM integrator only." No ID, no grant conditions, nothing product-specific.

Incidental, and it closes a worry: the BM20C has both the 32 MHz and the 32.768 kHz crystals embedded inside the module (p.1 features, restated in "Suggestion for Battery Power Application"). The 32.768 kHz sleep crystal recorded in Todos.md as the one knowingly skipped optional item was never needed. That omission is not a deviation from anything and does not need to appear on the deviation list.

Also from p.1: max programmable TX power is +8 dBm (range −10 to +8 dBm, 1 dB steps). That number does most of the work in section 5.

WM02C+Product+Specifications.pdf — Ver 1.00, Sep 2024, 36 pp.

Production, not draft. Page 1:
FCC ID X8WWM02C
ISED ID 4100A-WM02C
Japan TELEC 201-240555

Mounting guidance (p.12) is the same shape as the BM20C's: RFI shield 11.7 mm, antenna area 7.4 mm on the WM02C, antenna area must overhang the board edge or sit over a keepout with no ground and no traces, ≥ 30 mm to external metal, don't mount mid-board.

A reference layout exists. Fanstel publish the EV-WM02C evaluation-board Gerbers (EVM-WM02C-VB-gb.rar) and its schematic at https://www.fanstel.com/nrf54-70-document. For a keepout check, a reference layout beats a prose description — pull them before doing the AUG-11 pass.

Corrected 2026-08-10: this section previously said there was no BM20C equivalent and to ask for it. There is. The same page carries BM20C library components for Altium/EAGLE/PADS (BMH_BM20-LIB.zip, and BM20-library260729.zip updated 2026-07-29), 3D STEP files, and an EV-BMH20C host-board schematic with Gerbers. It also carries the WM02C's certified gain and back-off CONFIG file by name — WM02C-Power-Setting_radiotest_prj.txt — which is the §3 grant condition in the form you should actually use it in, rather than transcribed out of a PDF.

The variant sizes, since these keep getting confused

Variant Antenna Size (mm) Mounting rules?
WM02C Integrated horizontal chip antenna 12.0 × 19.1 × 2.0 Yes — 7.4 mm antenna area must overhang
WM02V Integrated vertical chip antenna 12.0 × 25.3 × 2.0 Yes — 13.6 mm antenna area
WM02F High-performance PCB antenna 12.0 (15.1 in antenna area) × 24.3 × 2.0 Yes — 12.6 mm antenna area
WM02E u.FL connector, external antenna 12.0 × 16.8 × 2.0 "No mounting restriction"
WM02P Pads for host-board antenna 12.0 × 12.9 × 2.0 Place for minimum interference

All are 16 castellated pins on a 12.0 mm width, which is why the footprints interchange — the C is simply 2.3 mm longer than the E, and that extra 2.3 mm is the chip antenna. Todos.md (2026-08-09) resolved the ordering question correctly: it's a WM02C and the pads line up.

But the pads lining up is not the whole story, and this is where Board V1 went wrong. The WM02E has no mounting restriction because its radiator is off-module on a cable. The WM02C's radiator is on the module, 7.4 mm of it, and it inherits the full keepout rule. A layout drawn to a WM02E footprint has no reason to carry a keepout, and Board V1 doesn't — the antenna sits over the four-layer stackup. Same question applies to the BM20C's 5.5 mm antenna area, and to whether anything was placed in the 2.3 mm the WM02C occupies beyond the WM02E courtyard it was drawn as.

Confirmed by inspection 2026-08-09 and now tracked in Todos.md. Cheap on a prototype; the fix is an overhang and there are already no traces there.

2026-08-10 — the BM20C was checked, and it deviates identically. Read out of Impulse.kicad_pcb: both modules sit on the west edge with their antenna areas facing it and both stop 0.6 mm short of it, so both radiators sit over the full four-layer stackup. The two antennas are also 1.15 mm apart, against the 30 mm both datasheets recommend to external metal — and the other module's shield can is external metal. Full geometry, the arithmetic, and a placement rule the respin can be drawn to are in Antenna Placement.md. The short version: the two modules go on opposite edges, and only one of them can afford to overhang.

3. The grant condition nobody has written down: TX power is firmware

launch-plan.md §5 obligation 3 is "staying inside the module grant conditions — same antenna, gain, trace layout, separation." On the WM02C, part of the grant condition is source code, and it is currently recorded in no spec.

The WM02C was certified with specific gain and band-edge backoff settings. To stay inside the grant, the host application must use Nordic SDK ≥ 2.6.1 and set:

CONFIG_NRF700X_ANT_GAIN_2G=0
CONFIG_NRF700X_ANT_GAIN_5G_BAND1=0
CONFIG_NRF700X_ANT_GAIN_5G_BAND2=0
CONFIG_NRF700X_ANT_GAIN_5G_BAND3=0

plus per-protocol band-edge backoff values. Fanstel publish the full CONFIG set as a downloadable .txt alongside the datasheet — take theirs verbatim rather than transcribing from the PDF.

The conducted powers those settings produce, from the datasheet's own measurements:

Band Mode Freq BackOff Avg TX (dBm) Peak TX (dBm)
2.4 GHz 802.11b 2437 5 18.24 20.51
2.4 GHz 802.11g 2437 4 13.74 22.29
2.4 GHz 802.11n HT20 2437 4 13.60 23.11
2.4 GHz 802.11ax HE20 2437 6 13.97 22.14
5 GHz UNII-3 5200 0 10.94 20.36
5 GHz UNII-1 5300 0 9.74 19.78

Why this is a firmware requirement and not a note: ship the wrong backoff and the layout can be perfect and the product is still outside the module grant. That is the §12.5 scenario — Class II Permissive Change, grantee cooperation, full intentional-radiator retest, $15k → $35k. It belongs in firmware_spec_v3_nrf.md as a normative constraint on the WiFi driver, with its own parity row.

Note also the gap between average and peak — 18.24 vs 20.51 dBm on 802.11b. Section 5 uses average, because that's what the SAR exclusion calculation takes, but the distinction matters when you declare a maximum.

4. The obligations, restated with what's now known

# Obligation Status
1 Part 15B SDoC on the finished product Still owed. Needs a lab. Guidance documents are in hand, not outstanding.
2 RF exposure evaluation The real work. Section 5.
3 Stay inside module grant conditions Layout: rules known, Board V1 deviates, fix scheduled. Firmware: section 3, unrecorded until now. BM20C: conditions don't exist yet.

5. The SAR-exclusion calculation — what you actually do

⚠️ 2026-08-11 — this section asked the right question and stopped one step short. Read Module Grant Conditions.md before acting on any of it. Washington Labs replied with the actual X8WWM02C grant, and three things below are now wrong:

  1. The ratio was never the binding constraint. The grant carries a ≥ 20 cm separation condition, and the watch breaks it. No arithmetic result makes the module's approval cover a wrist device — that needs a permissive change, and only the grantee (Fanstel) can file one. Corrected 2026-08-12: this line also claimed a no-co-location condition. The grant's sentence ends "except in accordance with FCC multi-transmitter procedures", so co-location beside the BM20C is permitted — and the simultaneous-transmission summing worked through later in this section is that procedure. The exposure category is the sole unqualified problem.
  2. The power figure is wrong. The grant declares 0.17 W = +22.3 dBm conducted, not the datasheet's 18.24 dBm average. The 2.4 GHz ratio at 5 mm is 53.1, not 26.2.
  3. Route 1 is not free and route 3 is not proven. All three routes below now pass through a permissive change, except BLE-only — and that one is conditional on how Fanstel certifies the BM20C, which is being filed right now and which nobody has asked about. See Module Grant Conditions.md §6.

The formula, the input table and the ceilings below are all still correct and still worth reading. What changed is that clearing the threshold is necessary and not sufficient.

What it is

For a portable device used within 20 cm of the body, you must either test SAR (a lab, a phantom, a robot, 1–2 weeks, five figures) or show by calculation that you don't have to. The calculation is the whole of the work: a spreadsheet, no hardware, no lab, done in an afternoon. It is the difference between a ~$6k filing and a ~$20k one, which is why §5 says get it first.

The governing document is FCC KDB 447498 D01, "General RF Exposure Guidance" (current revision v06), §4.3.1.

The formula

For frequencies 100 MHz – 6 GHz at test separation distances ≤ 50 mm, a transmitter is excluded from SAR testing when:

   max power of channel, including tune-up tolerance (mW)
   ──────────────────────────────────────────────────────  ×  √f(GHz)   ≤   NT
            min. test separation distance (mm)

where NT = 3.0 for 1-g SAR (head and body) and NT = 7.5 for 10-g extremity SAR.

Rules that change the answer:

  • If the real separation is under 5 mm, you use 5 mm. A watch back is thinner than 5 mm, so 5 mm is the number for the watch, and you cannot improve it by making the case thicker until you exceed 5 mm of actual antenna-to-skin distance.
  • Power and distance are rounded to the nearest mW and mm before dividing; the result is rounded to one decimal place before comparing.
  • Tune-up tolerance is added, not ignored. You declare a production spread (±0.5 to ±1 dB is typical) and the calculation uses the top of it. Declaring a tight tolerance buys headroom and costs you production yield.
  • Where two radios can transmit simultaneously, the per-radio ratios are summed against the same threshold.

The four inputs, and who controls each

Input Watch value Who sets it
f 2.437 GHz (BLE and 2.4 GHz WiFi); 5.2–5.8 GHz (5 GHz WiFi) Fixed by the radio
Separation 5 mm (floored) Fixed by the form factor — a wrist device cannot improve this
Max power + tune-up Yours to choose Firmware, via §3's CONFIG values, and your declared tolerance
NT 3.0, or 7.5 if the wrist is accepted as an extremity The TCB, not you — ask

Only one input is a lever, and it is a firmware constant. That is the entire strategic content of this section.

Running it — BLE, BM20C

Max programmable +8 dBm = 6.3 mW. With a +1 dB tune-up tolerance: +9 dBm = 7.9 mW → 8 mW. √2.437 = 1.56.

(8 / 5) × 1.56 = 2.5   ≤ 3.0   →  EXCLUDED

Clears the strict 1-g threshold at the worst-case 5 mm separation. The ceiling at 5 mm is 9.6 mW = +9.8 dBm, so +8 dBm nominal tolerates roughly 1.8 dB of tune-up before it fails. Tight, but it clears — and it clears without any firmware power reduction.

⚠️ 2026-08-11 — this calculation omits antenna gain, and with gain included the tune-up headroom is gone. Fanstel's own filing for the nRF54L15 module (files/X8WBC15_rf_exposure_report.pdf, ISL-25LR090FMPE) runs the identical formula on comparable silicon and declares 8.01 dBm conducted + 1.50 dBi gain = 8.94 mW, giving 2.82 at 5 mm — with a declared tune-up tolerance of zero. Adding the +1 dB assumed above on top of 1.5 dBi of gain gives 3.46, which is not excluded.

Declared Ratio at 5 mm Verdict
8.01 dBm + 1.5 dBi, 0 dB tune-up (as Fanstel filed) 2.82 excluded, by 6%
+ 0.5 dB tune-up 3.15 not excluded
+ 1.0 dB tune-up (this section's assumption) 3.46 not excluded

The conclusion "a BLE-only watch is SAR-excluded by calculation" survives, but only at a near-zero declared tune-up tolerance, and a tight declared tolerance is paid for in production yield rather than being free. The +9.8 dBm ceiling above is a conducted figure and should be read as an EIRP ceiling instead. Worked through in Module Alternatives.md §5.

Running it — WiFi 2.4 GHz, WM02C

Worst case from the datasheet table is 802.11b at 18.24 dBm avg = 66.7 mW. With +1 dB: 19.24 dBm → 84 mW.

(84 / 5) × 1.56 = 26.2   ≫ 3.0,  and ≫ 7.5   →  NOT EXCLUDED

Not marginal. Nearly nine times the 1-g threshold, three and a half times the extremity threshold. It does not become excluded at any separation a watch can offer — even at 20 mm it is 6.6, still over 3.0.

The power ceilings that would exclude it:

Separation ≤ 3.0 (1-g) ≤ 7.5 (10-g extremity)
5 mm 9.6 mW = +9.8 dBm 24.0 mW = +13.8 dBm
10 mm 19.2 mW = +12.8 dBm 48.1 mW = +16.8 dBm

Running it — WiFi 5 GHz

Worst average is 10.94 dBm at 5200 MHz = 12.4 mW; with +1 dB → 16 mW. √5.2 = 2.28.

(16 / 5) × 2.28 = 7.3   > 3.0  (and only just under 7.5)   →  NOT EXCLUDED

Lower power, but √f punishes the band. The 1-g ceiling at 5 mm is 6.6 mW = +8.2 dBm.

Simultaneous transmission

BLE (2.5) + 2.4 GHz WiFi (26.2) = 28.7 if both can transmit at once. The WM02C has a BT-CoEx grant line (pin 12/43, BT-CoEx-Grant) and the BM20C exposes SPI_GRANT on P112, so the two radios arbitrate. If the coexistence arbitration genuinely prevents concurrent transmission, that is an argument against summing — but it is an argument you make to the TCB with evidence, not an assumption. Raise it early; it is free to ask and it changes the scope of the quote.

So: three routes, and one is free

  1. Cut the WiFi TX power in firmware to ≤ +9.8 dBm and declare that as the maximum. The lever is the same CONFIG_NRF700X_* backoff mechanism from §3, so it costs no hardware and no schedule — it costs WiFi range, roughly 8 dB of it, which is a real product decision that belongs in the v3 spec and not in a certification document. Whether the watch's getOnWifi / getOffWifi criteria survive an 8 dB cut is a question for whoever owns those criteria.
  2. Test SAR. Book it, budget it, absorb the 1–2 weeks and the failure risk. §5's timing table already carries this.
  3. Ship BLE-only. See below — this is the finding that reframes the fallback.

The BM15C fallback is also a certification shortcut

BLK-15 and NRF Redesign.md frame the BM15C fallback as a de-risking path with a product cost: no WiFi, so no getOnWifi / getOffWifi, no §5.5.3 anchor repair. True. But the SAR arithmetic adds something to the other side of that ledger:

A BLE-only watch is SAR-excluded by calculation alone. No SAR test, no phantom, no robot, no 1–2 weeks, no five-figure line item, no failure-and-retest risk. The same is true of a BM20C build with the WM02C simply not populated.

That does not make the fallback free — halved RAM is still the constraint that started the pivot. But when the labs quote the BM20C build as primary and the no-WiFi variant as a delta, the delta should come back negative on the exposure line, and if a quote doesn't reflect that, the lab has misunderstood the scope. Worth knowing before reading the numbers.

What you actually submit

The calculation goes into the equipment-authorization filing as an "RF Exposure Info" exhibit — a one-page sheet stating, per radio and per band: frequency, declared max conducted power, tune-up tolerance, minimum test separation, the computed ratio, the threshold, and pass/fail. You prepare it; the TCB reviews and accepts or rejects it. Prepare it before requesting quotes, so the quotes are scoped against a known answer rather than a contingency.

6. The anchor is a different calculation entirely

The anchor is wall-mounted and stationary, so it can claim ≥ 20 cm separation and leaves SAR's domain altogether. This is unaffected by the anchor carrying a backup battery (firmware_spec_v3_nrf.md §7, corrected 2026-08-09) — exposure category is set by separation, not by where the power comes from. The battery does change the emissions scope, because mains-disconnected operation is a normal mode the lab has to test; that is handled in FCC Lab Quote Responses.md §2 and §3. It's evaluated on ERP against the exemption thresholds in 47 CFR §1.1307(b)(3)(i) — for 1500–100,000 MHz, Table 1 gives a threshold of 19.2 R² with R the separation in metres, i.e. 0.768 at R = 0.20 m. Both radios clear that by a wide margin at any plausible antenna gain.

Confirm the units of Table 1 against the current rule text, or have the lab confirm them, before relying on this paragraph. The value is almost certainly watts — a 0.768 mW threshold would sit below the rule's own blanket 1 mW exemption and make the formula pointless — but "almost certainly" is the wrong standard for the one number this section rests on.

The anchor's real regulatory questions are elsewhere and §5 already names them: whether the servo drive affects conducted and radiated emissions, and whether the mains PSU pulls the product into additional conducted-emissions scope.

7. What still requires an email to Fanstel

Not integration guidance. Specifically:

  1. BM20C FCC grant number and current status, in writing. §5: get this before booking a lab slot.
  2. The BM20C grant conditions / test report, once issued. The labs need certified antenna type, peak gain, and maximum conducted power — the SAR calculation in §5 used the datasheet's +8 dBm, and the grant's declared maximum is the number that actually governs. None of it is in a draft datasheet.
  3. Confirmation that the part shipping to you is the part being certified — BM20C vs BM20BC, nRF54LM20A vs B.
  4. BM20C evaluation-board Gerbers or a layout library. The WM02C has a public link; the BM20C draft has none.
  5. BM15C ↔ BM20C footprint compatibility, in writing. The fallback currently rests on "to my knowledge."
  6. Shipping confirmation on the 2026-08-08 order.

Sources