Audio glossary
The terms that appear across every audio troubleshooting guide, defined in plain language — with the practical consequence of each, not just the definition.
Last reviewed on August 9, 2026
The terms that appear across every audio troubleshooting guide, defined in plain language — with the practical consequence of each, not just the definition.
Last reviewed on August 9, 2026
Audio troubleshooting is full of terms that get used loosely and mean something specific. This glossary defines the ones that appear across our guides, in plain language, with the practical consequence of each rather than only the definition.
A logarithmic ratio, not an absolute unit. On its own "dB" means nothing — it always compares two values. A 10 dB increase is ten times the energy and sounds roughly twice as loud. Two identical sources together add about 3 dB, not double. See the decibel levels guide.
Decibels of sound pressure level, referenced to the threshold of human hearing. This is what a sound level meter reports and what safe-exposure limits refer to. Our browser meter approximates it but cannot be calibrated.
Decibels relative to full scale, used inside digital systems. 0 dBFS is the maximum a digital signal can represent, so all values are negative. Recording peaks around −6 to −12 dBFS leaves headroom without approaching clipping.
A filter applied by proper sound level meters to match the ear's varying sensitivity across frequency. Unweighted measurements over-report low-frequency noise such as traffic rumble and air conditioning.
Amplification applied to a signal, usually at the input. Too little forces automatic systems to amplify room noise; too much causes clipping. Aim for peaks in the middle of the range on the microphone test meter.
What happens when a signal exceeds the maximum a system can represent: the peaks are flattened, producing harsh distortion. Clipping is permanent — it cannot be repaired after recording, only prevented.
The gap between your normal level and the clipping point. Headroom is what absorbs an unexpected loud moment without distortion.
The level of sound present when nothing intentional is happening — fans, traffic, electrical hiss. Everything you record sits on top of it, so lowering the floor improves recordings more reliably than any equipment change.
How far your voice sits above the noise floor. Getting closer to the microphone improves it just as effectively as silencing the room, and is usually easier.
Cycles per second — the physical measure of pitch. Human hearing spans roughly 20 Hz to 20 kHz at best, and the upper limit falls with age. Speech occupies about 250 Hz to 6 kHz.
How evenly a device reproduces or captures across the frequency range. A "flat" response is neutral. Our frequency sweep reveals gross irregularities but is not a measurement.
A gradual reduction at one end of the range. Laptop speakers roll off below about 150 Hz; that is physics, not a defect.
Narrowband speech carries roughly 300 Hz to 3.4 kHz — traditional telephone quality. Wideband roughly doubles the top end. Bluetooth headsets in calling mode use one or the other, which is why they sound like a phone. See why Bluetooth microphones sound bad.
Age-related loss of high-frequency hearing. Universal, gradual, and not damage. It is why most adults stop hearing above 15–16 kHz. See reading your hearing range result.
How many times per second the signal is measured. 44.1 kHz and 48 kHz are standard. A device reporting 8 or 16 kHz is in a narrowband calling mode, not a high-quality one.
How precisely each sample is stored. 16-bit is CD quality and fine for voice; 24-bit adds headroom that matters mainly while recording, not when listening.
The compression scheme used to transmit or store audio — SBC, AAC, aptX, LDAC, LC3, Opus. Bluetooth playback codecs affect listening quality only; none of them apply while the headset microphone is active.
A small store of audio waiting to be processed. Larger buffers are more resistant to dropouts but add delay. Almost every latency trade-off is really a buffer-size trade-off.
Converting between sample rates, for example when the OS runs at 48 kHz and an app at 44.1 kHz. Adds a little delay and CPU load; matching rates avoids it.
The delay between sound entering a system and leaving it. Under 20 ms feels instant; over 50 ms is noticeable on calls; Bluetooth typically adds 100–200 ms. Measure yours with the latency test.
Base latency is the browser's internal buffer. Output latency is its estimate of the time from handing off a sample to it leaving the speaker. Their sum is the floor of your system's delay, not the total.
Microphone in to speaker out, the whole path. This is what matters for live monitoring and for conversation rhythm on calls.
Variation in timing rather than its average. Steady delay is tolerable; inconsistent delay causes the glitches and dropouts people describe as "choppy".
Continuously adjusts input level to keep speech at a target volume. Helpful on calls, unhelpful when testing, because it hides real level problems and amplifies room noise during pauses.
Removes steady background sound. Aggressive settings cut the beginnings of words and can silence a quiet speaker entirely — a frequent cause of "my microphone works but nobody hears me".
Prevents your speaker output re-entering your microphone. Necessary when using speakers on a call; unnecessary and occasionally harmful when using headphones.
Mutes the signal below a threshold. Set too high, it clips off the start of every sentence. Discord's automatic input sensitivity is a noise gate, and a common culprit.
A newer, model-based form of noise suppression built into recent versions of macOS, Windows, and meeting apps. More effective than classic suppression, and equally capable of removing wanted sound.
Tip-Ring-Sleeve (two black rings) carries stereo output only. Tip-Ring-Ring-Sleeve (three black rings) also carries a microphone. Plugging a TRS headset into a combo jack gives you sound with no microphone — count the rings. See headset microphone not working.
The single 3.5 mm port on modern laptops that handles both output and input, requiring a TRRS plug. Desktop PCs usually have two separate jacks instead.
Two incompatible pin layouts for TRRS plugs. CTIA is the modern standard; older OMTP devices have microphone and ground reversed, producing distortion or silence on current hardware.
The Bluetooth listening and calling profiles. Only one can be active at a time, which is the entire reason wireless headsets sound poor on calls.
A Windows setting letting one application take sole control of an audio device, locking out everything else. Disabling it prevents a whole class of "another app is using the microphone" errors.
A USB device that works with the operating system's built-in drivers, no vendor software required. Class-compliant devices are far less likely to break after an OS update.
Term missing? Tell us and we will add it. For the guides that use these terms in context, start at the guides index.
← Back to all audio tests