Why phase matters more than people think
Every stereo recording assumes the two drivers push and pull in the same direction at the same time. When they do, content mixed to the center — lead vocals, bass, kick drum — arrives at both ears in matching polarity and your brain fuses it into a solid image between them. When one driver is inverted, that fusion fails in a specific and recognizable way.
A headphone driver wired with reversed polarity produces a distinct signature: the stereo image loses its center, vocals that should sit between your ears smear into a diffuse ring around the head, and low frequencies that both channels share partially cancel, sounding thin rather than merely quiet. Nothing goes silent — both sides still play at full level — which is why the fault can survive casual listening for months. The usual causes are mechanical rather than electronic: a repaired or re-terminated cable with two conductors swapped at one plug, a detachable cable seated wrong or assembled incorrectly at one connector, or a counterfeit unit built with looser assembly control than the original. Our phase check plays 200 Hz to both sides in matching polarity (signal A), then inverts one side (signal B). On a healthy stereo path A sounds solid and centered while B sounds hollow; if A is the hollow one, polarity is reversed somewhere in the chain.
One caveat we repeat on the tool itself: the comparison depends on an ordinary stereo path. Mono summing, spatial or "3D" processing, and some virtual-surround modes flatten the A/B contrast without any hardware fault. Similar-sounding A and B results are a reason to check mono settings and retest with a known stereo source — they are not, on their own, a diagnosis of internal wiring.
One side quieter, louder, or missing
When one earbud or one headphone side plays quieter than the other, the cause is usually one of four things, roughly in order of how often they turn up. First, on in-ear monitors, the mesh screen over the nozzle collects earwax and skin oil; a partially blocked screen drops level and dulls treble on that side only, and it happens gradually enough to pass for driver wear. Second, cable damage concentrates at the strain reliefs — the flex points at the plug and where the cable enters each earcup; a fault there is usually intermittent and changes when you bend the cable. Third, the operating-system balance slider may be off center, and on phones it often lives in accessibility settings rather than the volume menu. Fourth, a mono-audio accessibility toggle left on sums both channels into both ears, which masks a real imbalance and defeats channel and phase testing entirely.
The order of operations matters: rule out software before inspecting hardware, because software costs nothing to check. Then work outward — clean the nozzle screens per the manufacturer's guidance, flex the cable near each strain relief while a tone plays, and swap in another cable or source to isolate the failing link.
- Balance slider centered? On phones, check the accessibility section as well as the volume controls — several platforms keep a second slider there, next to the mono-audio toggle.
- If crackle or dropout changes when a plug or cable moves, the connection is the suspect: compare another cable, port or source before blaming the driver.
- For Bluetooth dropouts, retest close to the source with other 2.4 GHz traffic reduced before concluding that the battery or headphones are faulty.
- After sweat or rain, follow the manufacturer's drying guidance before retesting; do not charge a wet device.
Bass and treble sweeps: the listener is part of the instrument
Our sweeps run 150→20 Hz downward and 8→20 kHz upward, and your tap timestamps where the tone disappeared. That reading mixes two filters in series — the transducer and your ears — and we want to be direct about the second one.
A failed 17 kHz tone does not, by itself, mean your headphones are broken — and it usually does not mean that at all. High-frequency hearing declines with age for nearly everyone: ceilings near 20 kHz are typical of young children, many adults hear little above 15–16 kHz, and ceilings in the 12–14 kHz range are common from middle age onward. None of that is pathology, and a browser sweep cannot distinguish it from a transducer limit. What you can do is hold one variable constant: run the same sweep, at the same moderate volume, on the same ears with a second pair of headphones. If the stop point moves by several kilohertz, the hardware was the limit; if it barely moves, the listener probably was. Either way, treat the number as a repeatable comparison point on one setup — not as an audiogram, and not as a hearing diagnosis.
The bass end has its own confound: seal. An in-ear tip that does not close the canal, or an over-ear pad lifted by eyeglass temples, leaks low frequencies before the driver gets any credit for them. Refit, keep playback moderate, and repeat before comparing numbers across headphones. And if a sweep result worries you about your hearing rather than your hardware, that is a question for a hearing professional with calibrated equipment, not for a web page.
Bluetooth: codecs, call mode, and latency
Bluetooth audio quality depends on which codec and which profile the link negotiates, and a web page can observe neither — the browser sees only a generic output device. For music playback over the A2DP profile, the codec ladder runs from SBC, the mandatory baseline every device supports, through AAC (the codec Apple devices prefer), to aptX and aptX Adaptive on many Android phones and LDAC at up to 990 kbps. Higher rungs buy more bitrate headroom, not automatically better sound. The moment an application engages the headset's microphone, many systems drop the link from A2DP to the hands-free profile, whose voice channel was historically 8 kHz mono and later 16 kHz wideband — which is why music collapses to telephone quality when a call or voice chat starts. That is a profile switch, not a hardware failure, and it reverses when the microphone is released.
Latency follows the same hierarchy. A typical SBC link adds roughly 200–300 ms end to end — noticeable as lip-sync error and disqualifying for rhythm games — while low-latency codecs advertise well under 100 ms and a wired connection can bring the same acoustic loop under 40 ms on gaming-grade hardware. Our latency check measures the entire round trip — browser, output device, air, microphone, input path — so use it to compare runs on the same setup, or wired against wireless on the same computer. It cannot name the codec in use, and it cannot isolate headphone-only delay from the rest of the chain.
The practical implication for this bench: run the phase and channel checks over the connection you actually use, but if a Bluetooth result looks wrong, retest wired or on another source before concluding anything about the headphones themselves. Codec transcoding and aggressive spatial processing live in the link, not in the drivers.
Interpreting your readings
- Bass floor — a lower repeatable limit means you heard farther into the 150→20 Hz sweep. Seal, playback level, room noise and hearing all influence the result, so compare like with like: same ears, same volume, same room.
- Treble ceiling — the stop point combines the output chain and the listener. It is not an objective headphone response measurement or a clinical hearing result; the sweep section above covers how to tell the two apart.
- Latency — five clicks travel through the full browser-to-microphone loop and the bench reports the median. Compare the same setup against itself, or wired against wireless, rather than treating the number as headphone-only delay.
- Phase and channels — listening comparisons, not automated verdicts. Both ears playing the melody cleanly, and B sounding wider than A, means the path behaved the way a stereo path should on that run.
The individual readings are the result — there is no single score, because grading an untested parameter would be inventing data. The methods page lists every threshold and every limit we accept.
Checking used or unfamiliar headphones
No web audio test can authenticate a product, but counterfeit and gray-market headphones tend to stumble on the same three bench checks, so we suggest running them as a pattern. First, channel balance: cheaply matched drivers often show an audible level or tone difference between left and right on the same melody. Second, the phase check: sloppy assembly is a classic source of an inverted driver, which the 200 Hz A/B comparison makes audible as a hollow, centerless image. Third, the frequency sweeps: a unit that rolls off far earlier than the same listener measured on a known-genuine pair suggests a different, cheaper transducer inside. Read the pattern, not any single result — one failed check has many innocent explanations, but a newly purchased or used pair failing all three, when a reference pair passes them on the same device and the same ears, is a reason for skepticism.
Treat the pattern as a heuristic, never as authentication. Genuine units can fail a check for reasons this page has already listed — blocked screens, damaged cables, software settings — and a competent fake can pass all three. For identity, warranty or authenticity, use the manufacturer's official serial-number and service checks and inspect the physical product; keep the listening report as diagnostic context only.