Two antennas on a 40 × 42 mm board — the placement rule for the respin¶

Status: working document, not normative. The mandatory rules here are transcribed from the module datasheets; the arrangement proposed in §4 is a recommendation with its reasoning shown. Created: 2026-08-10, answering the 2026-08-09 log — "is there a minimum distance they must be apart from eachother? neither datasheet addresses directly… do they both hang off the edge?" Evidence: datasheets/BM20C.pdf p.17 and p.27, datasheets/WM02C+Product+Specifications.pdf p.12, and the Board V1 geometry read out of ~/Documents/KiCad/ImpulseBluetooth/Impulse.kicad_pcb on 2026-08-10. Companions: FCC Certification.md §2 (the same rules in their compliance context) · NRF Redesign.md §F.

The short answer to the log's question: yes, both antennas must hang off an edge — but not the same edge. Two 2.4 GHz radiators 1.15 mm apart is what Board V1 has, and it is roughly 30 mm short of Fanstel's own clearance figure.

1. What the datasheets actually say¶

The two documents are near-identical in this section, which is convenient: one rule set covers both modules.

	BM20C (p.17)	WM02C (p.12)
RFI shield ("main body")	10.3 mm	11.7 mm
Antenna area	5.5 mm	7.4 mm
Main body over host ground plane	Allowed	Allowed
Antenna area	"must be extended outside of the host board or in an area without ground or signal trace"	same wording
Mid-board mounting	"We don't recommend"	"We don't recommend"
External metal	"For the best Bluetooth range performance, keep all external metal at least 30mm from the antenna area."	"For the best WiFi range performance, keep all external metal at least 30mm from the antenna area."

BM20C p.27 adds three layout notes that sharpen the second row: avoid running any signal line below the module, no ground plane below the antenna, and "if possible, cut off the portion of main board PCB below antenna."

Two configurations are sanctioned, not one. Both datasheets carry the same figure, captioned "Best" and "Good":

Best — the antenna area extends past the host board edge. Nothing under it at all.
Good — the antenna area stays over the board, and the ground plane is pulled back. The figure annotates that clearance as 5 mm.

So an overhang is not the only compliant answer, and that matters here, because on this board two overhangs are expensive in a way one is not.

The 5 mm is read off a figure, not a sentence. It is unambiguous as a dimension and ambiguous as to which gap it measures — antenna-to-pour along the length, or the side clearance. Drawing 5 mm of clearance on every side of the antenna area satisfies either reading, costs nothing extra on this board, and is what §4 assumes.

The question neither document answers directly, and how they answer it anyway¶

Neither datasheet mentions a second module. But both state a clearance to external metal, and the other module's RFI shield is a 10–12 mm slab of grounded metal. Applied literally, Fanstel's own rule requires ≥ 30 mm between each antenna area and the other module's shield can.

That number is worth trusting more than its provenance suggests, because it agrees with the conventional figure from a completely different direction: λ/4 at 2.44 GHz is 30.7 mm. Fanstel's range recommendation and the standard co-location minimum land within a millimetre of each other.

Be precise about its status, though. The keepout rule is a mounting condition — deviate and you are arguably outside the module grant, which is the §12.5 scenario (launch-plan.md: $15k → $35k plus a full intentional-radiator retest). The 30 mm is prefaced "for the best … range performance" — a performance recommendation, not a stated grant condition. Treat it as a design target, not a compliance limit, and let Fanstel's own answer settle it: the question is already item 6 of drafts/fanstel-followup.md.

2. What Board V1 actually does¶

Read out of Impulse.kicad_pcb on 2026-08-10. Board outline x 83.00–123.00, y 84.50–126.50 — 40.00 × 42.00 mm.

	WM02C (`U8`)	BM20C (`U7`)
Body, global	x 83.60–102.70, y 89.50–101.50	x 83.60–98.40, y 102.65–112.65
Body size	19.10 × 12.00 mm	14.80 × 10.00 mm
Rotation	90°	0°
Pads span	x 93.49–101.19	x 90.70–98.15
Antenna area	x 83.60–91.00 (7.4 mm)	x 83.60–89.10 (5.5 mm)

Both antennas point west, and the deduction is solid rather than assumed. The castellations sit in the shield, never in the antenna area. On the WM02C, an east-facing antenna would put the shield at x 83.60–95.30 with pads running to 101.19 — pads outside the shield, which cannot happen. Same contradiction on the BM20C. So the pad-free end is the antenna end on both, and on both it faces the west board edge at x = 83.00.

Three findings¶

1. The WM02C antenna area does not overhang. Known, and now quantified. Its west end is at x 83.60, and the board edge is at x 83.00 — the antenna sits 0.6 mm inside the board, over the full four-layer stackup. To reach Best, the module has to move 8.0 mm west, which is off the current outline.

2. The BM20C has exactly the same defect, and nobody had checked. Its antenna area, x 83.60–89.10, is also entirely on-board, also 0.6 mm short of the same edge. Todos.md carried this as "check the BM20C too — same rule, 5.5 mm antenna area, and nobody has looked." Now looked at: it deviates identically. This does not change the fix, but it doubles it, and it removes any hope that the respin is a one-module nudge.

3. The two antennas are 1.15 mm apart. The module bodies are separated by y 101.50 → 102.65. Measured properly:

Pair	Distance	Against Fanstel's 30 mm
WM02C antenna area → BM20C shield	1.15 mm	3.8% of it
BM20C antenna area → WM02C shield	≈ 2.2 mm	7.4% of it
Antenna-area centre to antenna-area centre	≈ 12.2 mm	λ/10 at 2.44 GHz

This is the finding the log was reaching for, and it is worse than "no stated minimum." Two chip antennas stacked a millimetre apart are not merely under-isolated; they are close enough to detune each other's impedance match, shifting both resonances and costing efficiency on both radios.

The coexistence interface does not fix this, and it is worth being explicit about why. The REQ/GRANT/STATUS lines between the modules arbitrate in time — they stop the two radios transmitting at once. Near-field coupling between two antennas is passive. It is present when both radios are idle. Arbitration is the answer to collisions; separation is the answer to coupling, and only one of them is implemented on Board V1.

3. Can both antennas overhang the same edge?¶

No, and the arithmetic is short. The board's y dimension is 42 mm. Put both modules on the west edge and push them to opposite ends: the WM02C antenna area is 12 mm tall, the BM20C's is 10 mm, so the largest gap available between them is 42 − 12 − 10 = 20 mm, giving about 20 mm antenna-to-shield and ~31 mm centre to centre — and that is with both modules jammed into the corners, which the routing will not allow.

20 mm is not 30 mm, and the shortfall cannot be bought back on this edge. So the two modules go on different edges. That is the load-bearing conclusion of this document.

4. The rule, drawable¶

Four constraints, in priority order. The first two are mandatory; the third is Fanstel's recommendation; the fourth is mechanical.

Each antenna area is clear of copper on all four layers — either off the board edge (Best), or over a void with ≥ 5 mm of clearance to the nearest pour on every side (Good). No signal traces under either module, and no ground plane under either antenna.
Neither module mid-board. Both datasheets say so explicitly.
≥ 30 mm from each antenna area to the other module's shield, treating the other module as the external metal it is.
The board-plus-antenna envelope has to fit the enclosure, which is 45 mm across the axis in question (FCC Lab Quote Responses.md §4). Every millimetre of overhang is a millimetre the PCB has to give back.

The arrangement that satisfies them¶

West edge, north: the WM02C, antenna overhanging. East edge, south: the BM20C, antenna over a 5 mm keepout.

Worked against the current 40 × 42 outline, with the WM02C's on-board 11.7 mm at x 83.00–94.70 and y 84.50–96.50, and the BM20C's on-board 10.3 mm at x 112.70–123.00 and y 114.50–124.50:

Pair	Distance	Target
WM02C antenna → BM20C shield	≈ 34.7 mm	≥ 30 ✓
BM20C antenna → WM02C shield	≈ 33.5 mm	≥ 30 ✓
Antenna centre to antenna centre	≈ 54.8 mm	≈ 0.9 λ/2 ✓

The diagonal is what makes it work. Directly opposite each other across the 40 mm axis, the antenna-to-shield distances fall to ~29 mm and miss; offsetting them to opposite corners buys the last 5 mm out of the 42 mm axis for free.

Why the WM02C gets the overhang and the BM20C gets the keepout¶

Only one overhang fits inside the enclosure — 7.4 + 5.5 = 12.9 mm of overhang against roughly 3–5 mm of spare envelope means one of them has to take the Good configuration. It should be the BM20C, for three reasons:

The WM02C is the module with a grant. X8WWM02C exists, it was certified in a particular mounting configuration, and Best is the configuration least likely to be argued with. The BM20C's grant does not exist yet (BLK-15), so its layout is getting revisited when the grant conditions arrive regardless.
Its antenna area is smaller — a 5.5 mm antenna needs a 15.5 × 20 mm void with 5 mm margins, against 17.4 × 22 mm for the WM02C. On a 40 × 42 mm board that difference is real.
WiFi is the radio carrying the RF-exposure problem. 2.4 GHz WiFi is the band that fails the SAR exclusion at 26.2 (FCC Certification.md §5). If the declared power ends up being cut to stay excluded, WiFi will be operating with ~8 dB less to spend, and an antenna in Best rather than Good is free margin in exactly the place that needs it.

The cost, stated plainly: the PCB has to give back the 7.4 mm the WM02C hangs into, so the outline shrinks to about 32.6 × 42 mm if the enclosure holds. Splitting the modules to opposite edges frees the contiguous 19 × 23 mm block they currently occupy together on the west side, which is where that area comes from.

5. What would overturn this¶

Each of these is a real way the recommendation could be wrong, with what settles it.

Fanstel's answer on module-to-module separation — asked as item 6 of the Fanstel draft. If they give a number, it replaces the inferred 30 mm and this section is why. This is the one to wait for if the respin is not urgent.
The EV-WM02C reference Gerbers — EVM-WM02C-VB-gb.rar at fanstel.com/nrf54-70-document, alongside the EV-WM02C schematic. Fanstel's own evaluation layout is the best available evidence of what they consider an acceptable keepout, and it beats every inference here. Pull it and diff before drawing.
The BM20C library footprint. Fanstel publish Altium/EAGLE/PADS libraries (BM20-library260729.zip, updated 2026-07-29) and 3D STEP. Board V1's BM20C silkscreen is 14.80 mm long, but the datasheet's 10.3 mm shield plus 5.5 mm antenna is 15.80 mm — a 1 mm discrepancy in a dimension being used to place a keepout to 0.6 mm. Check the drawn footprint against Fanstel's library before trusting §2's antenna-area coordinates for the BM20C. The WM02C's 19.10 mm matches its datasheet exactly, so this is a BM20C-footprint question, not a systematic one.
Whether 5 mm is a side clearance or an end clearance. §1 takes the conservative reading. If Fanstel means only the end, the BM20C's void shrinks and more board comes back.
The enclosure. No case is specced yet (FCC Lab Quote Responses.md §4), so constraint 4 rests on the 45 mm figure given to the lab. A case designed around the antennas rather than the other way round could take both overhangs and make this whole trade disappear.