Nfc test

NFC test board (for rev B / B1)

30 × 38 mm, 2 layers. It answers one question before B1 is trusted: does B1's NFC antenna read on a phone with a CR2032 right behind it, and with what trim? This is experiment 4.

It goes in the same order as B1, so it can't steer B1's layout. B1 copies this antenna and its trim pads exactly. Whatever trim value this board finds goes straight onto B1's pads.

What's on it

• Top 30 × 26 mm: B1's outline and B1's antenna block, from ../lib/antenna.py, which both boards import so they can't drift apart:
  • a 6-turn PCB coil, 0.25 mm track with a 0.2 mm gap, chamfered to clear the 3 mm corners, about 2.1 µH in air (NXP AN11276 formula);
  • its outer end drops to B.Cu at VO, crosses under the turns and comes back up at VI, as on B1, whose edge leaves no room for a tail;
  • U1 (NT3H2111, NTAG I2C plus 1k), C3 and the C1/C2 trim pads (not fitted) inside the bottom turns, where they sit on B1.
• Back: B1's CR2032 holder, MYOUNG BS-08-B2AA001. The holder has its own − spring, so the board under the cell is bare. A ferrite sheet can then lie on the holder floor under the cell, with a slot for the spring. Rev A's Keystone 3034 couldn't take a ferrite, because the cell sits directly on a PCB pad.
• Tail, 12 mm: D1 field LED, the pull-ups, JP1 and the J1 host header.

Regenerate

python3 hardware/lib/gen_lib.py               # NT3H2111 symbol, XQFN8 + BS-08 footprints
python3 hardware/nfc-test/tools/gen_sch.py    # tables -> schematic
python3 hardware/nfc-test/tools/make_pcb.py   # coil + placement + hand-drawn LA/LB, the router does the rest
python3 hardware/nfc-test/tools/fab.py        # -> production/

This follows the same procedure as rev A (../README.md).

How the antenna is drawn. The coil is LA track. Its inner end meets LB through net tie NT1; its outer end crosses under the turns on B.Cu (VO to VI) to reach the tag's LA pin. DRC doesn't check clearance within one net, so antenna.py builds the turns at a fixed pitch and asserts the 0.2 mm gap, and the vias' gap to the turns, itself.

State at commit:

• ERC: clean, apart from the library-table warnings.
• DRC: 0 errors, 0 unconnected, 0 parity issues.
• Repeatability: two runs give byte-identical boards.
• CPL: all 7 placements checked against JLC's EasyEDA footprints.

On the bench

1. No cell, no ferrite. Tap with an Android (NXP TagInfo) and an iPhone. Write a URL with NFC Tools. Note the read distance, and the distance at which D1 lights.
2. Find the resonance, if a NanoVNA and a small pickup loop are at hand. Couple the loop loosely and read the S11 dip, f0. Then the trim is C = 50 pF × ((f0 / 13.56 MHz)² − 1). Without a VNA, try 0, 4.7, 10, 15 and 22 pF C0G in C1, and keep the value with the longest read distance.
3. Fit the cell, no ferrite. Repeat steps 1–2. This shows how much the cell costs.
4. Fit the ferrite: WE-FSFS-class, 0.1–0.2 mm, cut to a 20 mm disc with a slot for the − spring, on the holder floor. Repeat steps 1–2.
5. Locked iPhone: with the screen on, a tap should open the URL with no app.

Record f0, the trim value, the read distance and the LED distance for each of steps 1, 3 and 4. The step 4 trim value goes on B1's pads.

Why the coil is under 2.75 µH. 13.56 MHz needs 2.75 µH with the tag's 50 pF (44–56 pF). The coil is short of that on purpose: ferrite raises L, the cell lowers it, and a parallel trim cap can only pull the resonance down.