RTC, Backup Batteries, and Timekeeping on Android SBCs

Engineering guidance for RTC backup power, Android clock policy, drift testing, and reliable event timestamps.

RTC, Backup Batteries, and Timekeeping on Android SBCs

A device can boot normally after months in storage and still fail every secure connection because its clock has returned to an old date. The display works, Ethernet has an address, and the server is reachable, yet certificates appear invalid.

Timekeeping is a system requirement for Android products. The RTC, backup supply, kernel driver, Android time policy, application timestamps, and service procedure all need to agree about what time means and how trustworthy it is.

Separate the clocks your software uses

Wall-clock time represents calendar time and can jump when corrected. Monotonic time measures elapsed duration and should be used for timeouts, retry intervals, and performance measurements.

An RTC preserves approximate calendar time while the main system is off. It does not guarantee that time is correct, and its output should be treated as uncertain after a depleted battery or oscillator fault.

Store timestamps with a clear convention, normally UTC, and apply time zones for display. Time zone changes and daylight-saving transitions should not reorder transactions or change elapsed-time calculations.

Clock or valueUseMain risk
RTC calendar timeStartup estimate after power lossDrift or lost backup power
Android wall clockHuman dates and event timestampsCorrection can jump forward or backward
Monotonic elapsed timeDeadlines and durationsDoes not provide calendar date
Server timestampShared ordering and reconciliationRequires connectivity and trusted server
Event sequence numberLocal ordering across uncertain timeMust be persisted consistently

Specify off-time and allowable drift

Define the longest expected unpowered period, storage temperature, service interval, and maximum acceptable clock error. A panel that is never off for more than a weekend differs from a spare unit stored for three years.

Oscillator error is often expressed in parts per million. As an illustration, 20 ppm corresponds to roughly 1.7 seconds per day before temperature, aging, and calibration effects are considered. That can accumulate into minutes during a long disconnected period.

Use the selected RTC’s specified behavior across temperature and voltage. A room-temperature nominal figure is not a field guarantee.

If precise timestamps are important, plan regular correction and record when the last trustworthy synchronization occurred. The application can then report both event time and confidence.

Choose the backup source deliberately

Coin cells, rechargeable cells, and supercapacitors have different life, leakage, service, and safety characteristics. Match the RTC’s backup input and charging circuit to the selected chemistry.

Do not connect a nonrechargeable cell to a charging path. Review the schematic, component options, and assembly instructions so a battery substitution cannot create an incompatible combination.

Calculate backup life using RTC current plus board leakage, protection components, contamination effects, and battery self-discharge. Very low RTC current does not help if another circuit leaks through the backup rail.

Backup choiceStrengthDesign concern
Primary coin cellLong off-time with low loadReplacement access and no charging
Rechargeable cellRepeated recharge capabilityCorrect charge profile and aging
SupercapacitorMany charge cyclesLeakage and limited long off-time
No backup sourceLower BOMStartup time depends on another source

Measure backup current on populated boards. Flux residue, alternate components, and test equipment can change the result. Document the measurement method and instrument loading.

Validate the RTC hardware and driver

Confirm address, bus pull-ups, interrupt wiring, oscillator requirements, backup switchover, and register interpretation. Check whether the device exposes an oscillator-stop or invalid-time flag.

The driver should preserve useful validity information rather than presenting an obviously wrong date as trustworthy. Test reading and setting time across the supported calendar range.

Verify startup ordering. The kernel may initialize system time from an RTC before Android’s automatic time service applies another source. Record which RTC is selected when the platform exposes more than one.

Test alarm wake separately from basic time retention. A clock that keeps time does not prove its interrupt reaches the wake controller in the selected low-power state.

Review Android time-source policy

Android can receive time suggestions from supported network or telephony paths, depending on version and product configuration. The priority is configurable in the platform, so inspect the shipped BSP rather than assuming a universal order.

A network connection does not mean time has already synchronized. Applications that immediately open TLS sessions after boot must handle the period before trustworthy time is available.

Do not bypass certificate validation to work around a bad clock. Repair startup time acquisition and provide a controlled error state.

The network failover guide explains why connectivity and service readiness are separate states. Time readiness belongs in that same application model.

Handle clock corrections without corrupting records

Wall time can move backward. A database key based only on the current second can collide, and a retry timer based on calendar time can wait too long or fire repeatedly.

Use unique identifiers and monotonic deadlines. For audit records, retain local sequence, boot identifier, event timestamp, and synchronization status where the business requirement warrants it.

When the device reconnects, reconcile using stable identifiers rather than assuming the oldest timestamp is always the oldest event. A queue created with a wrong clock may otherwise be replayed in a misleading order.

Keep operator time changes restricted and logged. A service technician may need to correct the clock, but that action should not silently rewrite historical records.

Test the uncomfortable transitions

Remove main power while preserving backup, then remove both on a sacrificial unit. Test a weak battery, a disconnected holder, cold startup, long storage simulation, and network unavailability.

Verify that the product detects invalid time and returns to useful operation after synchronization. A prominent service warning is better than unexplained cloud failures.

TestExpected resultRecord
Main power removedRTC continues within drift limitBefore/after reference time
Backup depletedInvalid time is detectedStatus flags and UI behavior
Network unavailableDefined offline operationTime confidence and queue behavior
Large correctionNo broken timeout or duplicate IDEvent ordering and timer behavior
Time zone changedDisplay changes only as intendedStored UTC and user display
Alarm wakeCorrect wake and resumeInterrupt and power-state trace

Include the shipping and warehouse process. An insulating battery tab can preserve shelf life, but the factory must verify its removal or activation at the intended stage.

Build a service policy

Specify battery replacement interval from evidence and environment, not just nominal capacity. Provide access, polarity marking, approved part number, and a post-replacement time check.

If the enclosure is sealed, decide whether battery life matches the product life or whether replacement requires a service procedure and seal renewal.

Monitor invalid-time events and repeated large corrections across the fleet. They can reveal battery problems, temperature sensitivity, or incorrect platform configuration.

Connect the record to lifecycle planning so an RTC or battery substitution triggers the relevant regression tests.

Frequently asked questions

Does Android need an RTC if it has Internet access?

Many products benefit from one because certificates, logs, and local operations begin before network time is available. The requirement depends on startup and offline behavior.

Can RTC time be used for retry deadlines?

Use a monotonic clock for elapsed-time deadlines. Calendar time can be corrected and can move backward or forward.

How long will a coin cell last?

That depends on capacity, self-discharge, temperature, RTC current, board leakage, and service conditions. Measure the complete backup circuit and apply conservative assumptions.

Release record

Keep the RTC part, oscillator, backup source, measured current, drift evidence, driver behavior, Android policy, and service procedure together. A reliable clock makes failures easier to diagnose and prevents time uncertainty from becoming a network or transaction fault.