Keyboard and Mouse Events
A visible key or button event confirms that the browser received that event. Keyboard key follows the active layout, while code reports a physical position. Mouse button roles can be remapped. Keep the test area focused and read the text log alongside its visual indicators.
Events missing from the page can be intercepted by the system or browser. A reported interval describes browser-side timing; it does not reveal the exact physical press or the moment a screen pixel changes. Repeated keydown during a hold can be normal auto-repeat.
Controller States and Mapping
The gamepad test reads snapshots supplied by the Gamepad API. Buttons have reported values and pressed states; axes use a normalized range from −1 to 1. Standard mapping supports known positional names. Other layouts keep raw indices to avoid guessed labels.
A resting axis slightly away from zero is an observation to repeat, not a wear diagnosis. These readings do not calibrate the controller, change a game's deadzone or measure hardware polling rate. No detected controller can also reflect access, pairing, driver or browser limits.
Microphone Levels and Camera Settings
A microphone meter shows digital signal level from the permitted stream. RMS and dBFS are digital references, not calibrated sound pressure in dB SPL. Position, input gain and automatic processing can affect the reading. Speech and background noise can both move the meter.
A camera preview shows the permitted stream. Reported width, height and configured frame rate describe that track; they do not certify maximum sensor resolution, image quality or delivered FPS. The speaker test reports the requested channel and synthesized tone, while system routing determines what is audible.
Interpret Permission and Access Failures
A pending media request can mean the permission prompt has not been answered. It does not measure a camera or microphone delay. A denied state can reflect site permission, operating-system policy or a security restriction. A missing-device state means a matching input was unavailable to that request. An unavailable state can reflect access failure even when equipment exists. These outcomes need different next checks rather than one faulty-device verdict.
Read the message, inspect site permission and selected input, then repeat Start. Another application using the device can affect access in a particular setup; closing it supplies a comparison, not a universal repair. Changing the selected input stops this page's session and needs a new Start. Stop can cancel this page's interest in an unanswered request without erasing the browser's stored permission.
Treat Settings and Levels as Their Actual Measurements
A camera track reporting 1280 × 720 and 30 frames per second describes its configuration when those properties are available. The page does not count frames delivered over time or discover maximum sensor resolution. Missing properties remain unknown. Compare the input selected in each application before assuming that a low configured resolution describes the whole camera's capability or a separate meeting application's output.
The microphone meter displays RMS amplitude and its digital dBFS value. The manual gain calculator needs an observed sample peak and target peak instead. Copying RMS into a peak field mixes measurements. Distance, input gain and capture processing affect the meter; it is not calibrated acoustic pressure in dB SPL. A repeatable speaking position helps compare inputs, while the recording or call application's selected device still needs its own check.
Solid Fields and Moving Patterns
Solid colors let you inspect the actual display for persistent points. The page does not see the physical panel or count defective pixels. Dust and surface marks can resemble a pixel issue. A screenshot records the rendered image rather than physical panel output.
A moving pattern supports comparisons of trails or halos at the same speed and background. Browser scheduling, refresh configuration, scaling and eye tracking affect the view. It does not assign GtG, MPRT or input-latency measurements. A white patch also cannot measure panel kelvin or certify calibration.
Separate Browser Timing from Hardware Response
The double-click test times received presses against the chosen window. A 120 ms interval describes browser observations, not time from physical switch closure to light on the panel. Pointer coordinates use CSS pixels, not mouse sensor travel or DPI. Controller snapshots update at least 100 ms apart; that is a choice in this page, not a measured hardware polling interval. A short transition can be missed between readings.
The motion target depends on browser animation scheduling. Its requested speed is CSS pixels per second. Keep speed, background, scaling and display mode equal for a comparison. Rendering and background tasks can affect the visible sequence. A recorded trail is not a returned GtG, MPRT or refresh measurement. The input-lag checker suggests checks from selected observations and does not produce an end-to-end millisecond result.
Know the Network Destination and Data Flow
After consent, the network tools send requests to the endpoint shown on the page. Both take 20 small HTTP round trips. The speed test then completes four 4 MiB downloads and four 1 MiB uploads. Timing continues through consumed, validated response bodies or an upload acknowledgment. This describes the browser-to-server path during that run, including work the tool does not isolate. A different destination or busy server can change the result.
Throughput is completed body bytes × 8 ÷ elapsed seconds ÷ 1000000 for Mbps. Four 4 MiB bodies total 16777216 bytes; a hypothetical summed two seconds gives 67.108864 Mbps. This is example arithmetic, not a measurement from your device. The test does not measure loaded latency concurrently with transfers. Unlike local input logs and media previews, these checks exchange data with a server; ordinary request metadata can exist in hosting logs.
Read HTTP Variation and Failed Requests
Consider hypothetical samples of 40, 55, failed, 50 and 65 ms. The completed mean and median are 52.5 ms. The valid adjacent differences are 15 and 15 ms, giving 15 ms mean adjacent variation. The failed slot is neither replaced with zero nor bridged to create a 55-to-50 pair. Without a completed adjacent pair, variation remains unknown. Keep the raw samples beside the summary to understand what actually completed.
This variation differs from RTP packet-arrival jitter reported by a call application. A failed HTTP request is not a count of lost IP packets; access, protocol, server or timeout problems can cause it. Preserve destination, method, completion counts and run conditions with copied results. A game's ping and a call's media statistics describe their own paths. Use the relevant application's statistics instead of comparing incompatible numbers solely because both carry a jitter label.
Record a Repeatable Observation
Note the device, connection, browser, selected input and exact action. Keep the same scene or pattern, then repeat several times. Record what was received or visible, including missing information, instead of labeling an unexplained result healthy or faulty.
Change one available setting or connection at a time and keep its original value. Compare the same action again. A useful note is 'button 2 was reported on each press in this session', rather than 'the whole mouse passed'. A browser result need not match a separate game or app.
Stop Sessions and Clear the Test
Start microphone or camera capture deliberately and respond to the permission prompt. An ignored prompt can stay pending. Stop ends this page's capture tracks or request; a later stream for a stopped request is released. The prompt or stored site permission can remain, and other apps may still use the device.
Stop the tone and animation when finished. Use Escape or the exit control to leave a display pattern. Record any needed input observations, then use Reset to clear the event log before another test. A new media device selection stops its session and requires Start again.