The technology

How PhoneProof checks a phone.

Most of the scan is measured, not ticked off. The phone tests itself: it plays, shows or senses something, records how its own hardware responds, and turns that into a plain sentence and a score. Where a check needs your eyes or your hands, the app says so.

Preview of the app during a check: the loudspeaker is being measured, then the full-screen touch test, where every square turns green except one that did not respond.

One idea behind every check.

Each part is asked to do its job while another part of the same phone watches. What the phone observes is compared with what a working part should do.

  1. Play, show or sense

    The speaker plays a tone, the screen shows a colour, the motor vibrates, or you turn the phone in your hand.

  2. Measure the response

    The microphone, touch sensor, motion sensor or camera records what actually happened, in native code on the phone.

  3. Compare

    The readings are checked against what a working part does. For sound, repeated runs show whether the result can be trusted.

  4. Explain and score

    You get a plain sentence for each result, what it means for resale, what to do next, and one score out of 100.

If a check cannot run properly, for example because the room is too loud or a permission was refused, it is shown as not checked. It never counts as a fault.

Speakers and microphones

The phone listens to its own speaker.

The loudspeaker plays a short tone that rises from low to high, and the phone’s own microphone records it. A healthy speaker reproduces the whole range evenly. A muffled, blocked or rattling speaker shows up as missing highs, a weak signal or distortion.

The earpiece gets the same test with the nearest microphone. On phones with two loudspeakers, the left and right channels play one at a time. Each microphone then records a moment of silence to check its noise level.

What the speaker plays: one rising tone

Low High

What the microphone hears, low to high

louder quieter 100 Hz 1 kHz 10 kHz pitch (frequency)
Example, not real data: a healthy speaker against a muffled one.

A little more technical

The tone is a logarithmic (exponential) sine sweep from 100 Hz to 16 kHz. Native code deconvolves the recording to recover the speaker’s impulse response and the round-trip delay, then reports the frequency response in 23 one-third-octave bands, normalised to 1 kHz. Distortion comes from the harmonic responses the same sweep reveals. The sound score combines signal-to-noise ratio, distortion and how even the response is between 300 Hz and 8 kHz. Each speaker runs three times; if the runs disagree, the result is flagged as low confidence. Background noise is checked first, and if the system applies voice processing to the microphone, the result is marked as not measurable. Microphone noise is a relative level, not a calibrated sound level, so it can raise a warning but never a fail.

Screen and touch

Touch is mapped square by square.

You slide a finger over a grid until every square turns green. The phone’s touch sensor reports every touch it sees, so a square that never responds while the squares around it do is a dead zone, and the app tells you where it is. A second step checks that several fingers register at once.

A phone cannot see its own display, so the colour and burn-in screens rely on your eyes: solid colours, grey and patterns fill the screen and you tap any spot that looks wrong.

Example: one dead square on the touch map, and the colour screens you check by eye.

A little more technical

Touches are recorded natively with the touch sensor’s own timestamps, never with app timers. The result reports coverage, the squares that responded and the touch sampling rate (readings per second, from the median gap between samples). Strips the system keeps for its own edge gestures are shown but not counted. A dead area is a critical finding: it caps the screen score however good the other checks are. The colour and burn-in results are marked as checked by you, and the same spot marked on several colours counts as one defect.

Cameras and flash

Every lens has to react.

Each lens opens a live picture. The app changes the exposure and moves the focus, and checks that the picture gets brighter and darker and comes into sharp focus. Then you confirm the picture looks clear, because dirt, scratches and haze on a lens cannot be measured reliably.

For the flash, the camera watches the light: with the flash on, the picture should clearly brighten.

Sharpness while the focus moves

in focus near far focus position sharpness

Picture brightness, flash off and on

on off time brightness
Example, not real data: a clear focus peak, and a flash that brightens the picture on every cycle.

A little more technical

Native code analyses the live frames: picture brightness as the exposure changes, and a sharpness measure across a focus sweep. A lens that will not open or stream, or exposure or focus that never moves on a usable scene, is a measured fault. Each lens the system lets apps use is tested; some Android phones keep extra lenses for their own camera app, and the result says so instead of failing them. For the flash, exposure is locked and the rear camera compares brightness with the torch off and on over a few cycles. No change at all usually means the lens was covered, so it is a warning, not a fail.

Sensors and buttons

Sensors check each other.

The vibration motor buzzes and the phone’s own motion sensor feels it. You turn the phone and the motion sensors must respond and agree with each other. Most of the others need one simple action from you.

  • Motion and compass. Turn and tilt the phone.
  • Proximity. Cover the top of the phone.
  • Light sensor. Cover it, then uncover it. On iPhone it is read through auto-brightness, which can confirm it works but cannot prove a fault.
  • Barometer. A few seconds of air pressure, which should be believable and steady.
  • Buttons. Press volume up and down. On Android the power button is checked by turning the screen off and on. iOS does not let apps see the side button.
  • Face or fingerprint. The system’s own unlock prompt. The app only learns whether it worked.

What the motion sensor feels

motor on phone resting resting time
Example, not real data: the vibration shows up as fast movement only while the motor runs.

A little more technical

Sensor sampling and analysis run in native code; the app only receives the finished figures. For vibration, the high-frequency energy in the accelerometer signal is compared with and without the motor running; no clear response is a warning, never a fail, because silent mode or low power mode can turn vibration down. For motion, the app checks the gravity reading while the phone is still, compares the tilt predicted by the gyroscope with the tilt measured by the accelerometer in one-second segments, and checks that the magnetic field is plausible (metal nearby can disturb it, so that is a warning at most). Proximity, light and barometer each need a real change or a believable reading within a short window.

Battery and charging

Every battery number says where it came from.

The battery’s level, state and temperature come from the system. On Android, the phone measures charging power while you leave the charger plugged in for about 20 seconds. Android 14 and later also share the charge-cycle count.

iPhone keeps battery health and cycle count from apps. You can type them in from Settings, and they are marked as entered by you. On iPhone the app confirms that charging starts, but it cannot measure the speed, so charging is not scored there.

Charging power on Android, sampled for about 20 seconds

0 s about 20 s charging power
  • MeasuredCharging power (Android)
  • Reported by the systemLevel, state, temperature; cycle count on Android 14 and later
  • Entered by youBattery health and cycle count on iPhone, typed in from Settings
Example, not real data. The labels show the three kinds of figure you will see in a result.

A little more technical

On Android, native code samples the current into the battery and the battery voltage and averages the power over the window. If the figures the phone reports look implausible, the result says so and is flagged as low confidence. Slow charging is a warning, never a fail: it is usually the charger, the cable, wireless charging or a battery that is already above 80%, where phones charge more slowly. Battery health appears only where the system shares it (rare on Android) or where you type it in; a battery the system reports as dead, or one worn well below its original capacity, caps the battery score.

Signal

The GPS is timed, the radios are checked.

The app looks for GPS satellites and times how long the phone takes to get its first location fix. Indoors the signal is much weaker, so a slow fix there is a warning with a suggestion to try by a window.

For Wi-Fi, Bluetooth, NFC and the mobile network, the app checks that each one is present, switched on and, where it applies, connected. These are status checks, not speed tests. A radio that is switched off or a missing SIM is never counted as a fault.

Looking for satellites until the first fix

time to first fix searching fix
  • Wi-Fipresent, connected
  • Bluetoothpresent, on
  • NFCpresent, on
  • MobileSIM, network, level
Example: satellites found one by one until the phone fixes its location.

A little more technical

On Android, the app listens to the satellites the phone can see and reports how many are in view and used, their mean signal strength (C/N0, in dB-Hz) and the time to first fix. iPhone shares no satellite details with apps, so there it times one precise location request and reports the accuracy the system gives. Only precise location uses GPS; if only approximate location is allowed, the check is shown as not run. On Android the Wi-Fi signal strength (RSSI) is shown too; iOS shares only whether Wi-Fi is present and connected, and only part of the mobile network details.

Storage and processor

Speed is reported, not scored.

The app writes a temporary file, reads it back to time the storage, then deletes it. The processor runs a fixed task on one core, then on all cores, to see whether it slows down as it warms up.

These figures are for information only. They never change the score. Total and free storage and memory are shown too.

All cores, second by second

first second last second work done
  1. Write a test file
  2. Read it back
  3. Delete it
Example, not real data: a processor that keeps most of its speed as it warms.

A little more technical

Storage speed uses sequential writes and reads of a test file of up to 64 MB (smaller when space is low), plus small random reads. The system may still cache some reads; when it does, the result says so. The processor task’s answer is verified, so a wrong result is never used. “Speed kept under load” compares the last second of the all-core run with the first, alongside the thermal state the system reports. These checks run last in the scan, so the heat they make cannot affect the battery and charging readings.

The score

Six areas, weighted, one number.

Each check gives a result from 0 to 100. The checks in an area are averaged into that area’s score: screen, sound, battery, cameras, sensors and signal. The six areas are then combined, with battery, screen and sound counting the most.

Example: the phone from the landing page. Bar length shows how much an area counts; the fill shows its score. The screen is capped by a dead touch area. Weights are still being tuned on real phones.
  • Not checked is not failed.

    A check that was skipped, does not apply to this phone or could not run is left out of the average. If a whole area could not be checked, the other areas share its weight.

  • One dead part cannot hide.

    A critical finding, like a dead touch area, a main camera that will not open or a worn-out battery, caps its area at 40 however good the rest is.

  • No number from too little.

    If too little of the phone could be measured, you see every result but no overall score.

  • Information stays information.

    Storage, memory, processor and headphone checks are shown but never change the score.

What a phone cannot measure about itself.

Some things no app can read, and some only you can judge. PhoneProof says so instead of guessing.

  • Genuine or replaced parts

    No app can read a phone’s part history.

  • How its own screen looks

    The phone cannot see its display. Colours and burn-in rely on your eyes, and the result is marked that way.

  • Dirt or scratches on a lens

    The cameras are measured for exposure and focus. You confirm the picture looks clear.

  • Battery health on iPhone

    iOS keeps health and cycle count from apps. You can type them in from Settings, marked as entered by you.

  • Charging speed on iPhone

    The app confirms charging starts. iOS does not share the speed.

  • Exact loudness

    Phone microphones are not calibrated instruments, so sound levels are compared, not given in absolute decibels.

  • Network speed or range

    Radios are checked for being present, on and connected, not for how fast they are.

  • Stolen, locked or financed

    PhoneProof does not check theft, carrier lock or payment status.

Everything runs on the phone.

  • Every measurement is made and analysed on the phone itself.

  • Audio and camera frames are analysed on the phone and never saved or sent.

  • No account, no ads, no analytics. No internet connection is needed.

Read the privacy policy

See it for yourself.

About six minutes for the full check. Free, no account, no ads.

Get the app