How do you design an NFC antenna for your own board?
As a loop whose area sets the coupling, and a matching network that presents 50 ohms at 13.56 MHz. Trace width and spacing set the impedance and the quality factor, the bandwidth has to cover the sidebands of the protocols you read, and every piece of metal nearby moves the result.
The loop and what it reaches
Range follows the loop, and TI's antenna matching note gives the working rule for a reader antenna: expect a read range of about twice the diagonal of the antenna. That single sentence sizes most projects before any component is chosen, and it explains why a module the size of a stamp is a tap and a gate antenna is a metre.
The tag's coil counts as much as yours, which the antenna size page covers, and the shape decides how much of the area is usable, which the shape page covers. Between two ends you control, the cheaper one to change is usually the tag.
Matching is a measurement
The note's method is worth following in order. Calibrate the network analyser on the boards themselves, one with a 50 ohm load, one shorted, one open, so that the parasitics of your own printed circuit are part of the calibration rather than a surprise. Then measure the impedance of the bare antenna trace at 13.56 MHz, which is the starting point the match has to move to 50 ohms.
The example measures 0.4648 ohm and a reactance of 107.42 ohm, an inductance of about 1.26 µH, and the note observes that such a value sits at the edge of the Smith chart, which makes the match harder rather than impossible. A design that skips the measurement has no way of knowing which of those two cases it is in.
Quality factor against bandwidth
A three element match is the note's recommendation, for one reason: it lets the designer choose the quality factor rather than inherit it. Q is the ratio of the carrier to the bandwidth, and the note chooses 2 MHz of bandwidth to accommodate the upper and lower sidebands of the data rates of ISO/IEC 15693 and ISO/IEC 14443 A and B, which gives a Q of about 6.8 at 13.56 MHz.
The trade is direct. A higher Q couples more strongly and passes less bandwidth, so a reader tuned for reach can stop answering the fastest data rates; the trace width and the spacing between turns are what set that impedance and that Q on a printed antenna.
What the board and the case move
Keep ground planes away from the antenna traces, the note says, and the same applies to everything else that conducts: a battery behind the loop, a display, a screw, a metal enclosure. What they do to your antenna is what metal does to a tag, which the detuning page describes from the other side.
So measure the antenna where it will live. A match trimmed on a bare board and then closed into a case is a match for a board that no longer exists, and the failure appears as reduced range rather than as an error, which is the hardest kind to attribute.
Frequently asked questions
How much range can I expect?
About twice the diagonal of the antenna, as a working rule for a reader antenna. The tag coil and the surroundings move it from there.
Why a three element match?
Because it lets you choose the quality factor. A two element match fixes it, and with it the bandwidth you end up with.
What bandwidth do I need?
Enough for the sidebands of the protocols you read. The TI note works at 2 MHz, which gives a quality factor near 6.8 at 13.56 MHz.
Can I copy a reference design?
Only with its board. The match depends on your traces and your ground planes, so it is measured on the board that will ship.