Appearance
8.4 — NFC and RFID: How Tap-to-Pay Physically Works
Hold a card near a reader and it responds — with no battery, no button and no connection. The card has been sitting in a wallet for three years and it works instantly.
The card is powered by the reader. That is the whole trick, and everything else about the technology follows from the fact that the energy and the data travel the same way.
1. Near field, not radio
Chapter 7.6 described radiation — fields that detach from the antenna and travel away. NFC does not use that.
The boundary is at roughly
r=\frac{\lambda}{2\pi}
At NFC's 13.56 MHz, \lambda=22 m, so the near field extends to about 3.5 m. Everything at tap distance is deep inside it.
In the near field the magnetic and electric fields are not yet a propagating wave. They are stored energy around the coil, and the field strength falls as 1/r^3 rather than the 1/r of a radiating wave.
Two consequences follow, and both are design features rather than limitations:
The range is inherently tiny. 1/r^3 means doubling the distance reduces the field eightfold. A card that works at 4 cm is completely dead at 15 cm, and this is what makes NFC safe for payment — an attacker cannot read your card from across the room, because the physics does not permit it.
Energy can be transferred efficiently. In the near field, the reader's coil and the card's coil are two windings of a loosely coupled transformer. This is Chapter 9.1's transformer, operating at 13.56 MHz with air between the windings.
2. How the card is powered
The reader drives its coil with a 13.56 MHz signal, producing an alternating magnetic field. The card's coil sits in that field, and by Faraday's law a voltage is induced:
V=-N\frac{d\Phi}{dt}
The card's coil is tuned to resonate at 13.56 MHz with a small capacitor, and the resonance (Chapter 1.6) multiplies the induced voltage by the circuit's Q — typically 20 to 40.
Worked example. A field inducing 0.3 V per turn, with a 5-turn coil and Q=30:
V=5\times0.3\times30=45\ \text{V}
which is then rectified (Chapter 2.2) and regulated down to the 1.8 or 3.3 V the chip needs. A shunt regulator dumps the excess, which is why a card held very close to a strong reader is not destroyed.
Available power at the card: 5 to 50 mW depending on distance and coupling. That is enough to run a small processor doing cryptography, which is exactly what a payment card is.
The coupling coefficient between the two coils is typically 0.01 to 0.1 — a very poor transformer, which is why the range is centimetres and why coil alignment matters.
3. How the card talks back
The card has no transmitter. It modulates the reader's own field.
Load modulation
The card switches a resistor across its coil. When the resistor is connected, the card draws more current from the field, which — because the two coils are coupled — changes the impedance the reader's own coil presents.
The reader detects that change in its own drive current. So the card communicates by varying how much power it consumes, and the reader listens by watching its own load.
\text{modulation depth} \approx 1\ \text{to } 10\%
Small, and the reader's receiver has to detect a few percent change on a signal it is transmitting at full power — which is why reader design is harder than it looks.
A subcarrier makes it practical. The card switches its load at 847.5 kHz — exactly 13.56/16 MHz — and modulates that. The reader then filters around 847.5 kHz, well away from its own carrier, which recovers the signal from under the transmission. This is the same reasoning as the intermediate frequency in Chapter 7.2.
Reader to card
The reader modulates its own carrier's amplitude.
Type A uses 100% modulation — the field is switched off entirely for short intervals. The card must survive those gaps on its stored charge, which is why every card has a substantial reservoir capacitor.
Type B uses 10% modulation, so the field never disappears and power delivery is continuous. More robust, and it needs a more sensitive detector in the card.
Bit rates: 106, 212, 424 or 848 kbit/s, negotiated during the initial exchange.
4. The standards, and what each is for
| Standard | Frequency | Range | Use |
|---|---|---|---|
| ISO 14443 | 13.56 MHz | up to 10 cm | payment cards, passports |
| ISO 15693 | 13.56 MHz | up to 1 m | library books, asset tags |
| NFC (ISO 18092) | 13.56 MHz | up to 10 cm | phones |
| LF RFID | 125–134 kHz | up to 10 cm | animal chips, immobilisers |
| UHF RFID | 860–960 MHz | up to 10 m | supply chain, tolls |
NFC is essentially ISO 14443 with extra modes, which is why a phone can read a payment card and act as one.
The three NFC modes:
- Reader/writer — the phone powers and reads a passive tag.
- Card emulation — the phone pretends to be a card, which is how phone payment works.
- Peer to peer — two phones exchange data. Largely abandoned in favour of using NFC only to bootstrap a Bluetooth or Wi-Fi connection, which is far faster.
LF RFID at 125 kHz deserves a note because its frequency choice is deliberate: low frequencies penetrate water and tissue, which is why animal identification chips and car immobiliser transponders use it and not 13.56 MHz.
UHF RFID at 900 MHz is a genuinely different technology. It operates in the far field, using backscatter — the tag reflects the reader's radio wave with a modulated reflection coefficient. Range of metres, and hundreds of tags can be read per second, which is why it is used for pallet-level inventory and motorway tolling rather than payment.
5. A contactless payment, step by step
The whole exchange takes about 300 to 500 ms, and here is what happens in it.
1 — Field detection. The reader pulses its field periodically, looking for the impedance change that a card's presence causes. A reader in standby is polling several times per second.
2 — Anticollision. Several cards might be present, and the reader must talk to exactly one.
The protocol is a binary tree search. The reader asks all cards to send their identifier; if two reply simultaneously, the bits collide and the reader detects the position of the first collision. It then asks only for cards whose identifier has a 0 at that position, and repeats.
With n cards, this converges in O(n\log n) steps, and it is why a wallet with two payment cards produces a "card clash" error rather than a random charge — the reader detects the collision and refuses to guess.
3 — Application selection. The reader asks which payment applications the card supports and picks one. A card may hold several — a debit application, a credit application, a transit application.
4 — Data exchange. The card returns the account number, expiry and cardholder data, plus a certificate.
5 — Offline data authentication. This is the step that makes cloning hard.
The card holds a private key that never leaves the chip. The reader sends a random challenge; the card signs it. The reader verifies the signature using the card's certificate, which is itself signed by the issuer, whose key is signed by the payment scheme's root key stored in the terminal.
A cloned card cannot produce a valid signature, because copying the data does not copy the private key — the chip is designed so the key cannot be read out.
Three variants exist, in increasing strength: SDA signs static data only and is clonable for offline use; DDA signs a fresh challenge and is not; CDA signs the transaction data too. CDA is the current requirement.
6 — Transaction authorisation. The card generates an application cryptogram over the transaction amount, currency, date and a counter, using a key shared with the issuer. The issuer can verify that this specific card authorised this specific transaction, and the counter prevents replay.
7 — Decision. Below the contactless limit, the terminal may approve offline. Above it, or periodically at random, the transaction goes online for the issuer to authorise, and increasingly a PIN or a phone biometric is required.
The reason for the limit is exactly this trade: offline approval is fast and works without connectivity, and it accepts a bounded fraud risk. The limit is where the issuer judges the risk acceptable.
6. Phone payment, and why the card number is not on the phone
Phone payment adds tokenisation, and it is a genuine security improvement over the plastic card.
When you add a card to a phone, the phone sends the real number to the payment network, which returns a device account number — a different 16-digit number, valid only on that device.
The real card number is never stored on the phone and is never transmitted in a transaction.
Each payment also generates a one-time cryptogram from a key held in the phone's secure element or trusted execution environment.
What that buys, stated plainly:
- A compromised merchant learns a token, not your card number. The token is useless elsewhere.
- Losing the phone means revoking one token, not reissuing a card.
- Biometric authentication happens on the device and the biometric never leaves it — the phone asserts that the user authenticated, and that assertion is what is signed.
Where the key is stored differs by design. A dedicated secure element is a separate tamper-resistant chip — Apple's approach. Host card emulation keeps the key in software with cloud-derived limited-use keys — Android's original approach. The secure element is stronger; HCE is more flexible.
7. Range, and why it is deliberately small
H\propto\frac{1}{r^3}
A field of 5 A/m at 1 cm is 0.6 A/m at 2 cm and 0.02 A/m at 6 cm.
ISO 14443 requires operation between 1.5 and 7.5 A/m, which pins the working range to a few centimetres and is the specification's way of enforcing physical proximity.
Can it be read from further away? With a large antenna and a sensitive receiver, eavesdropping on an active exchange has been demonstrated at a few metres. But that requires the transaction to be happening — an attacker cannot power a card from that distance.
Skimming — powering and reading a card covertly — has been demonstrated at up to 50 cm with a large purpose-built coil, and it requires equipment that is neither small nor discreet.
And what it would yield is limited. The card number and expiry, which are printed on the card anyway. Not the CVV, not the private key, and not a usable cryptogram for a future transaction, since each cryptogram is bound to its transaction and its counter.
So the shielded wallet's real value is modest, and the strongest protections are cryptographic rather than physical.
8. Where else NFC appears
Transit. Oyster, Suica, Octopus and their equivalents. These need sub-300 ms transactions to keep gates flowing, which is why they use simpler, faster protocols than payment cards — and why some of the older ones use weaker cryptography that has since been broken.
Access control. The classic 125 kHz proximity card transmits a fixed identifier with no cryptography at all, and can be cloned with a device costing a few pounds. Modern systems use 13.56 MHz with challenge-response, and the transition is still incomplete in many buildings.
Passports. An e-passport chip holds the printed data plus a facial image, signed by the issuing country. Access requires optically reading the machine-readable zone first — the key is derived from the passport number, date of birth and expiry, so the passport must be physically open to be read.
Pairing. Tapping a phone to a speaker exchanges Bluetooth pairing information over NFC, avoiding the pairing menu entirely.
Product tags. A tag in packaging linking to authentication or instructions.
Medical. A continuous glucose monitor worn on the arm is read by tapping a phone to it — NFC powers the sensor's memory readout, so the sensor itself needs only enough battery for the measurement, not for a radio.
9. Why it feels instantaneous
The whole exchange fits in a few hundred milliseconds because each part is small:
| Step | Time |
|---|---|
| Field detect and card power-up | 5 ms |
| Anticollision | 10 ms |
| Application selection | 20 ms |
| Read card data | 50 ms |
| Cryptographic authentication | 100 ms |
| Generate cryptogram | 50 ms |
| Total on the card | ~250 ms |
Online authorisation, when required, adds 500 to 2000 ms — which is why a transaction that goes online feels noticeably slower, and why the offline limit exists.
The cryptography is the largest single item, and it is being done by a processor running on tens of milliwatts scavenged from a magnetic field. That is the part worth appreciating: an RSA or elliptic-curve signature, computed by a chip with no battery, in a hundred milliseconds, from energy it collected on the way in.
Chapter 8.5 leaves short range entirely and covers the two systems that carry almost all of the world's long-distance traffic — one 36,000 km up, and one on the seabed.
Every formula above, built from scratch
None of the results in this chapter are worth memorising, because each one can be rebuilt in under a minute from something simpler. What follows is that rebuilding, one result at a time, so the formula and the reason for it sit on the same page as the explanation that needed them.
NFC and RFID
Near field boundary:
r=\frac{\lambda}{2\pi}
At 13.56 MHz, \lambda = 22 m, so the near field extends about 3.5 m — every tap distance is deep inside it.
Field falls as 1/r^3 in the near field, against 1/r for a radiating wave. Doubling the distance reduces the field eightfold, which is what makes the range inherently short and therefore safe.
Induced voltage:
V=-N\frac{d\Phi}{dt}, \qquad V_{resonant}=Q\cdot V_{induced}
With Q\approx30 and 5 turns at 0.3 V/turn, the card develops around 45 V before regulation.
Coupling coefficient between reader and card coils: 0.01 to 0.1 — a deliberately poor transformer.
Load modulation depth: 1 to 10%, carried on an 847.5 kHz subcarrier (13.56/16 MHz) so the reader can filter it away from its own carrier.
Operating field: ISO 14443 requires 1.5 to 7.5 A/m, which pins the range to a few centimetres.
Anticollision: binary tree search, converging in O(n\log n) for n cards present.
Transaction time budget: field detect 5 ms, anticollision 10 ms, application selection 20 ms, data read 50 ms, authentication 100 ms, cryptogram 50 ms — about 250 ms on the card, plus 500–2000 ms if it goes online.
What the next chapter fixes
Short-range radio covers a room. Chapter 8.5 covers the two technologies that carry traffic between continents, and shows why one of them wins on reach and the other on essentially every other measure.