SOUND
TEST

Testing left and right speakers means playing a sound in one channel at a time and confirming it arrives from the side you asked for. Both halves of that sentence matter, and most people only check the first one.

A channel test can fail in two directions. One side can be silent, which is obvious. Or the two sides can be swapped, which is not obvious at all: every note is still there, at the right level, in the right order. Nothing sounds broken. The only thing that has moved is where the sound appears to be coming from, and unless you knew in advance which side you asked for, there is nothing to compare it against.

This is a piece about what a stereo channel actually is, why the assignment gets reversed so easily, what a tone sweep can and cannot tell you about your hearing, and how loud any of it should be played.

Key Takeaways

• A stereo pair carries no marking of its own. Which stream is left is decided by wiring, connectors and software, never by the audio itself.
• Swapped channels are the failure mode that sounds fine. Broadcasters solve it with a line-up tone that has a deliberate gap in the left channel.
• On a four-conductor 3.5 mm plug the left channel is the tip and the right channel is the first ring, so a rewired cable can reverse the pair silently.
• A silent high tone is very often a working speaker and an ageing listener. Above 12.5 kHz no standard supplies an age-adjusted normal to compare a result against.
• A tone played at full scale measures about 20 decibels hotter than correctly normalised broadcast programme, so turn the volume down before you start.

Nothing in a stereo signal says which side is left

A channel is a delivery position, not a property of a waveform. Recommendation ITU-R BS.775-4, the ITU's recommendation for multichannel stereophonic sound, sets out a listening arrangement in which "the left and right frontal loudspeakers are placed at the extremities of an arc" subtending 60 degrees at the reference listening point, and then names the signals that feed it: "the use of five reference recording/transmission signals for left (L), right (R), centre (C)" plus the two side and rear surround channels.

Read that carefully and you can see where the identity lives. The recommendation describes where the loudspeakers go and what the channels are called. It cannot describe anything intrinsic in the audio that marks one stream as the left one, because there is nothing there to mark. Two channels of stereo are two streams of samples. The label is an agreement held everywhere except in the signal.

That is why professionals do not rely on their ears for it. They send a channel-identification tone ahead of the programme. EBU Tech 3304 says plainly that "The primary uses of the alignment signal described are to identify the system and the channel allocations."

A home channel test is the domestic version of exactly that procedure. You are not really testing whether a loudspeaker can make a noise. You are testing whether the label and the loudspeaker agree. Four kinds of result are possible, and they prove very different things:

What you hear	What it most likely is	What it does not establish
One side silent	That side is not being fed: a balance control pushed over, or a plug or cable that is not seated	That the loudspeaker on that side has failed
Both sides play, but you did not decide in advance which side you asked for	Nothing you can diagnose, because a reversed pair sounds identical to a correct one	Either that the pair is correct or that it is reversed
Low and mid tones play, a tone in the teens of kilohertz does not	Age-related loss at the top of the range, in an adult listener	That the output path has failed above that frequency
Nothing from any application	Output-device selection or a mute, which is where the platform vendors start	That the hardware is at fault

Where the left and right assignment actually lives

On a four-conductor 3.5 mm plug, it is a physical position along the metal. Texas Instruments' datasheet for a headset interface chip specifies the two output pins as "Left headphone / line output. Connect to headset jack TIP (terminal 1)." and "Right headphone / line output. Connect to headset jack RING1 (terminal 2)." Left is the very end of the plug; right is the band next to it.

Android's accessory specification for 3.5 mm headsets requires the same order. It lists "CTIA pinout order (LRGM)" as required except in regions with a legal requirement for the alternative, and "OMTP pinout order (LRMG)" as optional. Both begin L, R. The two conventions diverge further along the plug: "For the OMTP pinout, switch the positions of the MIC and GND segments." So a mismatch between them moves the microphone and ground segments, not the left and right ones. Whatever else it does to a headset, it cannot reverse the pair.

Everywhere else, the assignment is even thinner. On passive speakers it is nothing but which cable you ran to which box, and neither end of that cable knows anything. On a laptop it is software: macOS keeps a stereo balance control in its sound output settings, and Apple's own instruction is simply "Adjust the balance: Drag the Balance slider." A balance slider pushed to one side produces a silent speaker that is in perfect working order.

The swapped stereo pair, and why broadcasters test for it by design

A swap survives because it damages nothing measurable. Frequency response, level and timing are all unchanged. What changes is the stereo image, and an image has no reference point unless you brought one with you.

Broadcasting treats this as a hazard to be engineered out rather than caught by ear. EBU R 123, which governs how audio tracks are laid out for programme exchange between broadcasters, puts the reason bluntly: "to avoid operational confusion during the international exchange of programmes, the allocation of audio channels must be unequivocal."

The tool for proving it is a tone with a hole in it. EBU Tech 3304 records that EBU Recommendation R 49-1999 "describes a line up tone that identifies the Left stereo, or channel 1, by means of interrupting a 1 kHz tone at the alignment level for 250 ms every 3 seconds", and adds that "This interruption enables phase relationships to be audibly checked, as well as checking and identifying the monophonic mix." One steady channel, one channel with a regular gap: that is the entire mechanism, and Tech 3304 calls the format "especially prevalent". The surround era kept the trick rather than replacing it. BLITS, the 5.1 identification sequence set out later in the same document, still carries a two-channel ident section in which "The right channel is a continuous tone burst of 5.1s duration." while the left is deliberately chopped up.

The lesson for a listener is the one the profession learned first: decide which side you are asking for before the sound starts. Checking afterwards proves nothing. And if you are on headphones, find the L and R markings on the cups first, because a headset worn back to front reverses every conclusion you are about to draw.

If the pair does turn out to be reversed, the fix follows the list of places the assignment can live. Swap the two speaker cables at the amplifier, or the two plugs at the source, whichever is easier to reach. Take out any adapter and retest without it, because an adapter adds a set of connections that nothing on the outside identifies. Then check the balance and any channel-mapping or accessibility setting in the operating system, since those override the hardware and survive a cable change.

A silent high tone is usually your ears, not your speakers

Tone tests answer different questions depending on where in the range they sit. A tone in the low hundreds of hertz, or a couple of kilohertz up, is a check on the equipment. Across the conventional audiometric range a 2026 study of 668 ears in adults aged 60 and over found, testing in a soundproof booth, that "all participants (and all ears) responded to all frequencies regardless of sex and age group", so silence down there points at the output path far more often than at the listener. A tone up in the teens of kilohertz is not that test at all. It is mostly a measurement of the listener.

Human hearing is conventionally described as running from 20 Hz to 20 kHz, but clinical practice does not test anything like that range. Standard pure-tone audiometry stops at 8 kHz. Above that you need extended high-frequency audiometry, and what thins out up there is not the calibration but the idea of a normal result. Earphones for the extended range are calibrated to "reference equivalent threshold sound pressure levels (RETSPLs) within the range of 8000 to 16,000 Hz", so a threshold at 16 kHz is a real measurement against a real reference.

What does not exist is an age-adjusted expectation to compare it with. EN ISO 7029:2017 predicts median thresholds only as far as 12.5 kHz, and that same 2026 study records flatly that "No thresholds are given for frequencies > 12.5 kHz". A second paper puts the consequence in one line: clinical extended high-frequency testing "lacks age-specific reference standards for diagnostic purposes due to age-related changes in hearing sensitivity". The criterion that does reach 16 kHz is ISO 389-5:2006, which "defines a hearing loss at EHF as a hearing level that exceeds 20 dB HL at least once in one or both ears at any frequency". That standard "is based on participants aged between 18 and 25 years", so it was never derived from an older listener.

What measurements exist are sobering. That study found that adults aged 60 and over "showed a > 93% audiometric response up to a frequency of 11.25 kHz", but that "median hearing threshold values at the highest frequency (14 kHz) were determined in fewer than 50% of the examined ears across most age categories".

A normative study of 134 healthy adults aged 18 to 70 pushed one step further, to 16 kHz, and its numbers show the age gradient plainly. Response held at every tested frequency except the top one. Even there the failures sat entirely in the study's oldest band: 35 of 44 ears responded at 16,000 Hz, which is the 79.55% the paper reports for that group. The thresholds say more than the response rates. Across the pooled 18 to 40 group the median at 16 kHz had already slipped to 30 dB HL, and for participants over 40 "thresholds ranged from 20 dB HL at 9000 Hz to 70 dB HL at 14,000 and 16,000 Hz". A 70 dB HL threshold means the tone has to be raised about seventy decibels above the audiometric reference before the median listener notices it, which at any sensible volume means it is not there. The 16 kHz half of that figure is close to being censored, too. The study's audiometer topped out at 75 dB HL there, five decibels above the median it recorded, so the real value may be worse.

None of this is a sign of damage. That normative study excluded anyone with hearing loss on a conventional audiogram or an abnormal tympanogram, along with anyone carrying more than a year of occupational noise exposure or a family history of hearing loss, so its thresholds describe healthy ageing rather than injury. Hearing loss proper is a separate and common thing: the NIDCD puts the broad figure at "about one in three people in the U.S. between the ages of 65 and 74 has hearing loss", and "Nearly half of those older than 75 have difficulty hearing."

So read a sweep asymmetrically. Hearing a tone proves the output path works at that frequency. Not hearing a high one proves very little, and the older you are the less it proves.

Silence on both sides is almost never the hardware

When nothing comes out at all, the odds are heavily against a broken speaker, and the clearest sign of that is where the platform vendors put their opening steps.

Microsoft's troubleshooting sequence for audio problems in Windows runs to thirteen numbered steps. It begins with "Check your speaker output in Windows settings" and works through cables and volume controls, audio enhancements, its own troubleshooter, muted or disabled devices and "Set default audio device" before it reaches "Update or reinstall audio drivers" at step nine. None of the thirteen ends in a verdict of failed hardware. One of the named traps is worth repeating because it is so easy to hit: an external monitor with no built-in speakers, sitting there as the default output device, sending every sound to a display that cannot play it.

The second layer is per-application volume, and it behaves differently from the system control. Microsoft's step two warns in passing that "Some speakers and apps have their own volume controls. Be sure to check them all." Apple states the rule that makes them dangerous: an application's own volume setting "can only be equal to or less than the computer's output volume", and such controls do not override what the sound settings say. A per-app control can only ever attenuate, and nothing about it announces itself.

That gives you a one-step discriminator, and it is worth running before anything else. Play audio from a different application. If the second application is audible, the output device and the system volume are both fine and the fault is inside the first application or its tab. If nothing is audible from anything, the fault is at the device layer and the per-app mixers are irrelevant. Two layers, one test, no guessing. The same split governs the input path when you check a microphone, where an operating-system privacy permission sits on top of the device selection. That permission is where the last of those thirteen Microsoft steps ends up.

How loud a test tone should be, and why the answer is not obvious

A test tone is not music, and the two sit about twenty units apart in two different standards.

Recommendation ITU-R BS.1770-5 defines how programme loudness is measured, and it fixes the meter against a sine wave. Feed a full-scale 1 kHz sine into a single channel, left, centre or right, and "the indicated loudness will equal -3.01 LKFS". Now compare that with real programme material: EBU R 128 requires "the Programme Loudness Level shall be normalised to a Target Level of -23.0 LUFS", and notes that LUFS and LKFS are the same unit under two names.

The distance between those two numbers is about 20 units, and the recommendation calls that unit "equivalent to a decibel", scoped exactly: "an increase in the level of a signal by 1 dB will cause the loudness reading to increase by 1 LKFS". A full-scale tone is therefore around a hundred times the power that standard's meter reads for correctly normalised broadcast audio at the same setting on the dial. Setting the volume by ear on broadcast material and then playing a tone at the same position is not a small error. Streaming is a separate case and R 128 says so in terms, handing it to a companion document: "that guidance for the normalisation of content for streaming is given in EBU R 128 s2". Do not assume the same twenty units apply there, which is one more reason to set the level by ear on each tone rather than by a number.

The reason the gap exists also explains why a tone feels relentless. The 3.01 figure is the sine's own average level relative to its peak: the recommendation notes that the constant in its equation cancels out the K-weighting gain at that frequency, so nothing is being added or taken away. Music spends most of its time well below its loudest moments. A steady tone spends all of its time at the top, at one frequency, with no let-up. NIOSH sets the hazard line low: "Noise is considered hazardous or loud when it reaches 85 dBA or higher", with a recommended exposure limit of 85 dBA averaged over an eight-hour workday.

The practical rule is dull and it works. Turn the volume most of the way down before the first tone, then bring it up until you can hear it comfortably, and repeat that for each tone rather than setting it once. The high ones are the ones you can least judge, and a tone you cannot hear can still be very loud. The same habit is worth carrying into any of the other quick hardware checks that run in a browser.

Frequently asked questions

How can I tell whether my left and right speakers are the right way round?

Decide which side you are asking for before you play anything, then play one side at a time and check the sound arrives where you expected. A swap changes nothing measurable, so there is no way to detect it from the sound alone after the fact. Professional practice does the same thing with a line-up tone: EBU Tech 3304 describes a 1 kHz tone in which the left channel is interrupted for 250 ms every 3 seconds, so the channel identifies itself rather than relying on the listener's memory.

Why can I hear the lower tones but not the very highest ones?

Almost always because of your hearing rather than your speakers. Extended high-frequency measurements show median thresholds climbing steeply with age: one normative study of adults aged 18 to 70 found that for participants over 40, "thresholds ranged from 20 dB HL at 9000 Hz to 70 dB HL at 14,000 and 16,000 Hz". Above 12.5 kHz no standard supplies an age-adjusted normal to compare against either, because EN ISO 7029:2017 stops there and the criterion that does reach 16 kHz was built on 18 to 25 year olds. If the low and mid tones play cleanly, the output path is working.

One side is silent. Does that mean the speaker is dead?

Not getting audio and being broken are two different conclusions, and a silent side only shows you the first. A stereo balance control pushed fully to one side produces exactly the same silence from perfectly healthy hardware, and macOS keeps such a slider in its sound output settings. Check the balance, then try the same test through different headphones or a different output device. If the silence follows the device it is hardware; if it stays with the computer when the device changes, it is a setting.

Everything is turned up and I still get no sound. What narrows it down fastest?

Play audio from a second application. If that one is audible, the output device and system volume are both fine and the problem is inside the first application, its tab, or its own volume control. If nothing plays from anything, the problem is the output device selection. Microsoft's own thirteen-step guide to Windows audio problems spends its first six steps on settings, cables and volume controls, audio enhancements, its own troubleshooter and the default output device, and does not reach drivers until step nine.

Is it safe to play a test tone at my normal listening volume?

Treat it as louder than it looks. A full-scale sine wave fed to one channel measures -3.01 LKFS on the standard loudness meter defined in ITU-R BS.1770-5, while broadcast programme is normalised to a target of -23.0 LUFS under EBU R 128, which is about 20 decibels quieter. NIOSH considers sound hazardous from 85 dBA upwards. Start with the volume well down and raise it, and be especially careful with tones you cannot hear, because inaudible to you is not the same as quiet.

Sources

• International Telecommunication Union, Recommendation ITU-R BS.775-4, Multichannel stereophonic sound system with and without accompanying picture, December 2022, recommends clauses 2 and 3 - retrieved 13 August 2026, https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.775-4-202212-I!!PDF-E.pdf
• European Broadcasting Union, EBU Tech 3304, Multichannel Audio Line-up Tones, May 2009, section 1 Introduction, section 2 on two-channel audio identification and section 4 on BLITS - retrieved 13 August 2026, https://tech.ebu.ch/docs/tech/tech3304.pdf
• European Broadcasting Union, EBU Recommendation R 123, Audio Track Allocation for International File Exchange, section 1 - retrieved 13 August 2026, https://tech.ebu.ch/docs/r/r123.pdf
• Texas Instruments, TPA6166A2 3.5-mm Jack Detect and Headset Interface IC datasheet, pin functions table - retrieved 13 August 2026, https://www.ti.com/lit/ds/symlink/tpa6166a2.pdf
• Android Open Source Project, 3.5 mm headset: Accessory specification, Mechanical requirements table - retrieved 13 August 2026, https://source.android.com/docs/core/interaction/accessories/headset/plug-headset-spec
• Gottfriedova N., Skerkova M., Mrazkova E. and others, Extended high-frequency audiometry in older adults: hearing thresholds, European Archives of Oto-Rhino-Laryngology, volume 283, 2026 - retrieved 13 August 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC13249744/
• Luengrungrus K., Thanawirattananit P. and Teeramatwanich W., Normative Data of Extended High Frequency Audiometry in Normal Hearing Subjects with Different Aged Groups, Audiology Research, volume 14, 2024 - retrieved 13 August 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC11674001/
• National Institute on Deafness and Other Communication Disorders, Age-Related Hearing Loss (Presbycusis) - retrieved 13 August 2026, https://www.nidcd.nih.gov/health/age-related-hearing-loss
• National Institute for Occupational Safety and Health, About Occupational Hearing Loss, Terms to know - retrieved 13 August 2026, https://www.cdc.gov/niosh/noise/about/index.html
• International Telecommunication Union, Recommendation ITU-R BS.1770-5, Algorithms to measure audio programme loudness and true-peak audio level, November 2023, Annex 1, the passage defining the LKFS designation - retrieved 13 August 2026, https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.1770-5-202311-I!!PDF-E.pdf
• European Broadcasting Union, EBU Recommendation R 128, Audio loudness normalisation and permitted maximum level, 2023 revision, recommends clauses h and r - retrieved 13 August 2026, https://tech.ebu.ch/docs/r/r128.pdf
• Apple, Change the sound output settings on Mac, macOS User Guide - retrieved 13 August 2026, https://support.apple.com/en-au/guide/mac-help/mchlp2256/mac
• Microsoft, Fix sound or audio problems in Windows, steps 1 to 13 - retrieved 13 August 2026, https://support.microsoft.com/en-us/windows/fix-sound-or-audio-problems-in-windows-73025246-b61c-40fb-671a-2535c7cd56c8