Why this guide exists
Fingerprint modules look interchangeable on a spec sheet and are utterly different on a bench. The 2025 smart-lock market made the stakes concrete: 19.8 million biometric modules were embedded in smart-lock devices, and fingerprint held 34.3% of that volume at about 6.79 million units — while face recognition took 54.6% at 10.82 million (2025 market reports). That split is moving, but the 34.3% still represents a very large set of doors, and it is the category with the best-documented unit economics. This guide is the seven-parameter sheet we send to every lock OEM before they ask for a price.
The 7-parameter specification table
| # | Parameter | State it as | Example value (AM160 platform) |
|---|---|---|---|
| 1 | DPI / array class | Sensor resolution in DPI and array size | 508 DPI class, BS-series capacitive arrays (208×288 … 252×350) |
| 2 | Match speed | 1:N verification time under full capacity | 0.3 s at 1,000 templates |
| 3 | Template capacity | Number of enrolled fingers stored on module | 100–1,000 templates; higher tiers on request |
| 4 | Interface | Host protocol and cable class | UART / SPI / USB — SDK examples for all three |
| 5 | Temperature class | Working range the design must survive | −40~85°C family-class verification |
| 6 | Liveness | Fake finger rejection, wet/dry finger handling | 99.99% company rejection rate, lab-verified platform tier |
| 7 | Customization openness | What the SDK lets you change without new code | Thresholds, capacity, behavior, strings — free SDK |
Values above are our published platform figures; per-model pages state them exactly, with on-request items marked as such.
1 — DPI, the parameter everyone skips
Ridges on an adult finger space about 300–500 µm apart; a 500-DPI-class array puts two pixels across each ridge. Drop below that class and dry fingers start matching each other by silhouette instead of by minutiae. This is the single most common root cause we see when an OEM reports "two users open the same door".
2 & 3 — Speed and capacity are one decision
A lock with 10 registered fingers feels instant on a 100-template engine; the same engine at 800 templates is slower by a measurable margin. The AM160-YYH660 keeps 1,000-template class with a 0.3 s figure, and the low-power YYH362L variant holds the same capacity on battery-duty ratios. If your product is a 16-template family lock, buy the smaller capacity and keep the margin; if it is a hotel or rental fleet, buy top tier.
4 — Interface decides your firmware cost
UART is the cheapest integration when your host already speaks serial; USB 2.0 is better for enrollment tools and service menus; SPI is the fastest board-level path. Our SDK ships all three for every module family, and the same demo routine runs unmodified across them — which is also why the interface row is where OEM custom wiring questions usually end up.
5 — Temperature is a product decision, not a datasheet footnote
Industrial locks sit behind a warm lobby; warehouse locks face a −20°C dock door at 4 a.m. The AM160 family is built for a −40~85°C working class, verified on family fixtures. Ask your supplier to tell you their fixture number if it matters for your climate — leaving temperature values unqualified is how a product earns a winter recall.
6 — Liveness is your fraud budget
Photo punches fool software cameras; a cheap molded silicone finger is the analogous trick for capacitive sensors. Dual-camera RGB+IR liveness is the face-side answer at 0.3 s; on the fingerprint side, fake-finger checks on the AS608/AM160 platform tiers achieve a 99.99% rejection rate in company lab verification. Declare a target in your spec; otherwise you buy what you get.
7 — Customization openness, measured in SDK lines
FAR thresholds, template capacity, reader behavior and menu strings should all be configuration items — not revision requests. Here "openness" is testable: install the free SDK and change one threshold. If you can ship that demo in a day, the module is OEM-ready; if you can't, it isn't — regardless of what the marketing line says.
OEM flow: the seven steps from spec to shelf
Spec sheet
One page: product, volumes, worst-case environment, host interface.
FAE match
Model list and datasheet package returned within 48h.
Mechanical
2D/3D drawings supplied with the sample; window fit and depth checked.
SDK demo
Free SDK and demo routine, run against your board in a day.
Sample
1–3 days to ship; mixed-sled samples welcome; 2-year warranty starts at first unit.
Beta & mass
Few hundred pcs branded, field-checked, then 10–25 days per <5,000 pcs run.
A worked BOM for a battery fingerprint lock
| Line item | Suggested class | Why |
|---|---|---|
| Fingerprint module | AM160-YYH362L (low-power) or YYH860 (⌀16 mm, full spec) | 360° capture, 1,000 templates, 0.3 s, −40~85°C class |
| Sensor array | BS-series capacitive, 508 DPI class | Dry and wet finger tolerance at the envelope |
| Host MCU | Any — interface via UART | Free SDK; no license or NDA gate |
| Battery rail | 4 × AA / Li class per duty cycle | L-class module keeps stand-by draw within family spec |
| Packaging | ESD-safe bag + foam, lot-labelled | 2-year warranty and lot traceability |
Volumes matter in reverse: an OEM buying 5,000 modules a year should ask about custom cable endpoints and silkscreen in the same conversation as the price. Silkscreen and box customization begin around a few hundred pieces, and firmware tuning carries no MOQ on the hardware itself — a commercial fact most suppliers keep for the negotiation call.
Fingerprint, face, or both?
Market direction is clear: face modules grew 3D-variant volume +41% while total module shipments ticked to 19.8 M, and new smart-lock designs are quoting face at 54.6% share. But hybrid honesty, not trend-chasing, is the advice here. Face costs more and demands camera-window engineering; fingerprint still wins on unit cost, battery and enclosure simplicity. A dual-die multimodal chip — biometric plus QR decode on one die — is the middle path for terminal-class products, and the AS801 is our 2026 generation of exactly that move.
