Recorded check
{{ summaryValue }}
{{ summaryLine }}
{{ badge }}

Example session — synthetic evidence

This browser reports no touch support. No touch reading is assumed. Mouse and pen cannot test the glass.

{{ stageCue }}
Optional cell IDs, for example R1C1, R1C8. Mark only a notch, camera cutout, curved edge or bezel cover you cannot touch. Applied to the next coverage pass; excluded cells leave its denominator.

{{ exclusionMessage }}

This field needs finger touch. {{ guidance }} Use whole screen for edges, then move the controls and repeat to check the area beneath them. A resize stops capture; Start creates a new pass.

Browser and OS gestures or palm rejection may hide touches. Fullscreen requires your action; the fallback covers only the visible browser viewport. Stop keeps evidence; Reset clears it.

{{ fullStatus }}

{{ ignoredType }} ignored. Finger touch only.

Whole columns 3–16; standard 8. Rows follow field aspect ratio (3–40). Tool heuristic. Blank uses the standard.

{{ resolved.messages.columns }}

1000–120000 ms; standard 30000. Tool heuristic. Blank uses the standard.

{{ resolved.messages.window_ms }}

100–5000 ms after the last contact releases; standard 1000. Tool heuristic. Blank uses the standard.

{{ resolved.messages.settle_ms }}

0.01–1; standard 0.15. Flag a larger jump, without filling missing cells. Tool heuristic. Blank uses the standard.

{{ resolved.messages.gap }}

10–1000 ms; standard 150, inclusive. Tool heuristic. Blank uses the standard.

{{ resolved.messages.recontact_ms }}

0.005–0.25; standard 0.04, inclusive. Tool heuristic. Blank uses the standard.

{{ resolved.messages.recontact_distance }}

Kept in this tab only, included in your report.
Describe the protector, case, cable or repeated symptom. Never added to the URL.
{{ reportText }}

Map unavailable. Cell positions remain in the report.

No coverage pass recorded.

FactValueMeaningCopy
{{ row.fact }}{{ row.value }}{{ row.meaning }}
Window / eventStatus / outcomeStarts / contactsStart msEnd msPosition x / yDuration msCopy
{{ row.label }}{{ row.status }}{{ row.count }}{{ num(row.start_ms,3) }}{{ num(row.end_ms,3) }}{{ row.position }}{{ num(row.duration_ms,3) }}

No hands-off window or contact recorded.

Kind / cellPassRuleTime msFrom x / yTo x / yDiagonal distanceInterval msCopy
{{ row.kind+' · '+row.cell }}{{ row.pass }}{{ row.threshold }}{{ num(row.time_ms,3) }}{{ num(row.from_x,3)+' / '+num(row.from_y,3) }}{{ num(row.x,3)+' / '+num(row.y,3) }}{{ row.distance_text }}{{ row.interval_text }}

No recorded stroke candidate. Untested areas remain untested.

SampleInterval msMethodPointer IDTime msCopy
{{ row.index }}{{ num(row.interval_ms,3) }}{{ row.method }}{{ row.pointerId }}{{ num(row.time_ms,3) }}

No positive timing intervals recorded. Use a steady finger drag in Sampling.

Chart unavailable. Timing rows remain available.

No positive timing intervals. Drag steadily in Sampling to record browser-delivered timing.

Similar tools

A phone can show a clear picture yet miss part of a swipe, lose contact with a second finger, or register a tap while nobody is touching it. Checking a new or used phone or tablet therefore involves more than opening an app and tapping a few buttons. Slow sweeps, simultaneous finger contacts and a quiet observation period reveal different aspects of touch behavior and provide specific symptoms to discuss before a purchase or repair.

The display produces the picture. The touch-sensing system, often called the digitizer, detects contact positions. In a capacitive touchscreen, a finger changes electrical capacitance, which the touch controller uses to locate the contact. A normal picture with repeatable missed or misplaced input suggests investigating the touch-input path. Visible image defects while touches still register suggest investigating image output. These are clues about where to look; they cannot identify a failed part, and both kinds of problem can occur together.

Three practical checks help distinguish common touch symptoms:

  • A coverage check follows slow, overlapping sweeps through rows, columns and accessible edges. A dead zone is an area that repeatedly fails to respond. An unvisited area or a gap during one fast swipe needs another attempt before it can support that suspicion.
  • A multi-touch check keeps several fingers down together. The number recognized demonstrates simultaneous input during that session. The capacity reported by the browser is a separate declaration and may not describe the exact physical limit of the touchscreen.
  • A hands-off check watches for unwanted contact while the glass is untouched. These unwanted inputs are commonly called ghost touches. A meaningful observation requires keeping fingers, palms and objects away from the surface throughout the observation period.

Touch sampling rate describes how frequently touch positions are sampled. Display refresh rate describes how frequently the picture updates, while touch-to-response latency describes the delay before an action produces a response. A browser receives input after processing by the operating system and browser. Events may arrive in groups or follow display-frame timing, so their timestamps can support an estimate of browser-delivered sample frequency. They do not establish the panel’s advertised sampling rate or a calibrated latency reading.

Conditions around the screen can change the result. Moisture, debris, a protector or a case can affect touch response or access to an edge. Charger interference can also disturb capacitive sensing and produce unstable or false contacts. A repeat check with clean, dry glass and charging accessories disconnected helps establish whether the same symptom remains. Restarting and checking the effect of cases or protectors are useful follow-up checks when response is intermittent.

Battery level and power-saving settings are useful context for timing comparisons. Power-saving behavior can change performance and display refresh behavior on supported devices. If the battery is low, recharging and repeating under comparable conditions can help separate those conditions from the original observation. Browser events cannot establish that battery charge caused a ghost touch.

Edge swipes and other system gestures may open navigation controls or switch apps instead of reaching a web page. Palm rejection can suppress contacts that the device considers accidental. Browser controls, cutouts and overlays also leave some areas inaccessible. A cancelled drag or an empty patch near an edge therefore needs a repeat check at the same physical location. Reproducing the symptom in ordinary device use or an available manufacturer diagnostic provides more useful repair evidence than a single browser flag.

How to Use This Tool:

Open the test on the phone, tablet or touch laptop being checked and use a finger. Mouse and pen input do not count as touch evidence. Keep one orientation during a pass and leave the numerical Advanced fields empty to use the standard settings.

  1. Choose Use whole screen before starting when you want to reach more of the screen. It requests fullscreen where available and otherwise expands to the visible browser viewport. Browser or system controls can remain outside the testable area.
  2. Enter genuinely inaccessible cells in Not reachable on my device, if needed. Cell IDs such as R1C1 identify rows and columns from the top-left corner. Exclude a cutout or covered area you cannot touch, rather than an area that merely failed to respond. Entries that do not belong to the next grid are reported and ignored.
  3. Select Coverage, press Start and move slowly through each reachable cell. Received positions mark the cells they enter. Use Finish step when you have completed the sweep to record completion and advance.
    If an Advanced number is invalid, the field explains the accepted range and the standard value is used. Correct the value or clear the field before starting. Simply switching steps, choosing Skip step or pressing Stop does not mark the current step finished.
  4. Add fingers one at a time during Multi-touch and keep them down together. Compare the observed maximum with the browser-reported capacity. Lift your fingers and choose Finish step after the check.
  5. Place the device flat during Hands off, choose Arm hands-off and release all contacts. With standard settings, the quiet window starts after a one-second settle delay and lasts 30 seconds. A new contact during settling restarts the wait. Leave the glass untouched for the full observation period, then choose Finish step. In Ghost-touch log, check that the window is marked Completed.
  6. Drag steadily with one finger during Sampling to collect consecutive timing samples. Choose Finish step to end the guided sequence. Stop is available whenever you need to end capture while keeping the observations already recorded.
  7. Review the Defect report and revisit locations identified in Coverage map or Stroke findings. In the expanded view, Move controls stops capture and relocates the controls so you can select Coverage and start another pass over the area they previously covered. Save the report before clearing the session.
    A change in field size or orientation, loss of focus, a hidden tab or leaving the expanded surface stops capture. Return to the test and press Start for a new pass. If an evidence-limit message appears, save the report and use Reset before recording again.

Interpreting Results:

The summary describes the evidence recorded during the session. No session recorded means capture has not started. After recording, the other outcomes have specific meanings:

  • Observations to retest means at least one stroke candidate or contact start during a hands-off window was recorded. Inspect its location and circumstances, then repeat that check. A flagged session may also contain unfinished checks.
  • Partial evidence means there are no selected flags, but at least one retained pass has unregistered reachable cells, a step has no recorded completion, no hands-off window completed, or no positive sampling interval was collected.
  • Recorded checks complete means those evidence requirements were met without a selected flag. It does not certify the touchscreen or establish that every physical point on the glass works.

On the coverage map, Registered means at least one qualifying position reached that cell. Never registered identifies an area to revisit. Excluded records your decision that the area was inaccessible, and a stroke-candidate marker identifies a recorded gap, recontact or cancellation. Successful contact somewhere inside a cell does not establish response at every point within it.

Each pass remains separate. A later successful sweep does not fill missing cells in an earlier pass, so an older incomplete pass can keep the session marked partial. Match the physical location when comparing orientations because the row and column labels may change.

For multi-touch, Reached reported, Below reported and Above reported compare the observed maximum with the browser’s declaration. An unavailable declaration remains unknown. A below-reported count does not by itself create a flag or prevent Recorded checks complete; inspect the actual count and repeat with the intended fingers held down together.

A Completed hands-off window with zero starts means no new contacts were recorded during that window. Zero starts in a Cut short or In progress window do not establish a completed quiet observation. Any finger contact during an active window can enter the log, including an intentional one.

Changing numerical Advanced settings stops active capture. The current stroke thresholds also re-evaluate retained stroke evidence, so a changed finding count may reflect changed rules. Keep the same thresholds when comparing repeats and retain the report’s settings alongside its observations. The plain-text report and CSV numeric fields retain the recorded numerical precision even when a table shows rounded values or a tolerance label; those extra digits do not establish measurement accuracy.

A session labelled Example session contains synthetic evidence and says nothing about your device. For real observations, repeated symptoms at the same physical location after basic troubleshooting are useful evidence to take into a repair discussion.

Technical Details:

A browser touch observation records a contact position, a timestamp and whether the contact started, moved or ended. Contact identifiers connect positions within a continuous touch. Coverage depends on received positions, simultaneous-contact counts depend on overlapping contacts, and timing estimates depend on differences between timestamps.

Formula core: grid coverage

A coverage pass fixes the field dimensions, grid and exclusions. The number of rows follows the field’s aspect ratio. Reachable cells are the grid cells remaining after unique exclusions, and never-registered cells are the reachable cells without a qualifying position.

R = min ( 40 , max ( 3 , round ( C×H W ) ) ) N=C×R−E U=N−A

Here, C is the column count, R is the row count, W and H are the field width and height in CSS pixels, and E is the number of unique excluded cells. N is the reachable count, A is the registered count and U is the never-registered count. Row rounding uses the nearest whole number, with halfway values rounded upward, before applying the 3-to-40 row limit.

For an illustrative 8-column by 16-row pass, two exclusions leave 126 reachable cells. If 120 register a touch position, six remain unregistered. The result is 120 of 126 reachable cells, with six locations to revisit; it is not a screen-health percentage.

A strip of five cells has received positions only in its first and last cells. A dashed connection crosses three cells without recorded positions.
A connection between two received positions does not establish coverage between them. Only cells containing qualifying positions register.

Qualifying positions are nonduplicate finger-touch starts, moves and in-field releases collected during Coverage. Cancellation positions do not register cells. No path is filled between samples. Excluded cells remain outside both the registered count and the reachable denominator, even if a position reaches them.

Rows run from top to bottom and columns from left to right. A position exactly on the right or bottom boundary belongs to the final column or row. Missing-cell coordinates describe cell centres as fractions of field width and height. An outside release or cancellation can remain in the evidence without marking an outside position as a cell.

Normalized distance and stroke rules

Spatial comparisons use positions scaled separately to field width and height. Within the field, x runs from 0 at the left to 1 at the right, and y runs from 0 at the top to 1 at the bottom. The distance between two positions is divided by the diagonal of that normalized unit square.

d = (x2−x1) 2 + (y2−y1) 2 2

The result d is a dimensionless normalized coordinate distance. The visible “field diagonal” labels refer to this convention. On a rectangular field, it is not the movement in CSS pixels divided by the actual rectangle’s diagonal, and it does not represent millimetres or a fraction of the physical screen diagonal.

The standard settings are heuristics for collecting and reviewing observations. They are not manufacturer specifications or diagnostic thresholds. Accepted override ranges include both endpoints; only the column count requires a whole number.

Setting Standard Minimum Maximum
Grid columns 8 3 16
Hands-off window, ms 30,000 1,000 120,000
Settle delay, ms 1,000 100 5,000
Spatial gap, normalized distance 0.15 0.01 1
Recontact interval, ms 150 10 1,000
Recontact distance, normalized distance 0.04 0.005 0.25

Three received patterns produce stroke candidates during Coverage:

  • Spatial gap requires a move whose separation from the preceding position exceeds the nominal gap by more than 10−12 normalized distance units. With nominal gap g, the exact comparison is d − g > 10−12. The standard nominal gap is 0.15. Fast movement or sparse browser delivery can produce this pattern without establishing a dead line.
  • Recontact requires a contact that moved and then released, followed by a new contact in the same pass. The new start must be at or after the release. Both the elapsed interval and separation must fit their bounds: at most the nominal interval plus 10−9 ms, and at most the nominal distance plus 10−12. The standard nominal bounds are 150 ms and 0.04. A deliberate lift and nearby replacement can satisfy this rule, and the new contact need not have the previous identifier.
  • Moving contact cancelled records cancellation or lost contact tracking after a positive positional change. There is no additional speed or elapsed-time cutoff. When the terminal position is outside the field, the candidate is located at the last observed inside cell while the terminal coordinates remain available.

The tiny numerical allowances accommodate rounding in stored coordinates and timestamp subtraction. They are not physical tolerances. Calculations retain full precision, and findings show extra digits when ordinary rounding would obscure their relationship to a rule. An accepted recontact just above a nominal bound is labelled Equal to [nominal] within tolerance. Contact counts, coverage counts and hands-off window boundaries do not use these allowances.

Simultaneous contacts and hands-off boundaries

The observed contact maximum is the largest number of active finger contacts received together across the recorded session, including steps other than Multi-touch. Pen and mouse input do not contribute. The browser-reported capacity does not cap the observed count. Depending on the platform, that declaration can be a guaranteed supported count rather than an exact ceiling.

A hands-off observation starts only after the contacts involved in arming have ended and the settle delay has elapsed. With no contact held, the delay runs from arming. A new contact during settling postpones the start. The recorded window uses its scheduled start and end, even if the browser checks the timer slightly later.

A new contact counts from the inclusive start up to the exclusive observed end. The observed end cannot extend beyond the planned end:

S ≤ tstart < Eobserved ≤ Eplanned

S is the window start, tstart is the contact’s start time, and E identifies the observed or planned end. All use the same browser time basis in milliseconds. A start exactly at S counts; a new start exactly at the observed end does not. A release exactly at that end can still close an already counted contact.

Contact duration runs to a qualifying release or cancellation, or to the observation cutoff if no ending was received. The log distinguishes those outcomes. Stopping or interrupting a window before its planned end preserves its shorter observation period and marks it cut short.

Formula core: browser timing estimate

Consecutive timestamps support an interval only when they belong to the same continuous finger contact during Sampling and the later timestamp is greater. Release, cancellation and a new contact break the sequence. Only touch starts and moves contribute timing samples.

Δti = ti − ti−1 > 0 m = median ( Δt1 , … , Δtn ) festimate = 1000m

The intervals and median m are in milliseconds. Dividing 1,000 by that median gives the estimated frequency in hertz. After sorting the positive intervals, the median is the middle value for an odd count or the midpoint of the two middle values for an even count. The estimate is the reciprocal of the median interval, rather than an average of individual reciprocal rates.

For an illustrative continuous contact with samples at 0, 8 and 24 ms, the positive intervals are 8 and 16 ms. Their median is 12 ms, giving an estimated browser-delivered frequency of about 83.33 Hz. These example timestamps explain the arithmetic; they are not a device benchmark.

Consecutive samples with exactly the same timestamp and position within one contact are duplicates and do not add another sample or interval. A different position at the same timestamp remains a sample but contributes no zero-length interval. With no positive interval, the timing estimate is unavailable.

Some browsers supply historical positions grouped with a delivered movement event. These are called coalesced samples. A nonempty supplied group is used in place of its parent movement event; otherwise the delivered event supplies the sample. The recorded method can therefore be delivered, coalesced or mixed. Merely exposing coalescing support does not establish which samples were actually supplied, and a separate high-frequency update stream is not combined with the movement stream.

The interval distribution counts exact received interval values, without rounding them into histogram bins. Display labels use readable precision while calculations retain the original numerical values. Additional decimal places do not establish greater physical accuracy.

Browser scheduling, grouping of events and display-frame alignment can change the available intervals. Compare repeats using the same device, browser, orientation and power-saving conditions, and check whether the sampling method changed. There is no hardware pass band for the estimated frequency, and it does not establish panel sampling rate, display refresh rate, USB polling rate or calibrated touch latency.

Privacy Notes:

Touch observations are processed in the current browser tab. The test does not upload touch evidence, device details, unreachable-cell entries, the optional device model or notes, and it does not save them automatically for a later visit. Only valid, nonstandard numerical Advanced overrides enter the page URL. Sharing that URL does not share the recorded session.

Reset clears observations, completion records, the device model and notes. It retains the numerical settings and unreachable-cell entries, so review those before a fresh session. Reloading or closing the page loses the recorded evidence unless you have deliberately saved a report.

The report includes browser and platform information, reported touch capacity, screen and viewport sizes in CSS pixels, and any model or notes you enter. Review that context before sharing the report.

Finger capture uses ordinary browser touch input and does not request camera, microphone or location access. Fullscreen requires your action and may be unavailable. Chart resources can load from the network, so local handling of test evidence should not be read as a promise that the page makes no network requests.

References: