We are offering 20% off custom cables, preterminated cables and accessories with promo code PIROUZ20. Furutech 15% off, Audioquest 5% off. 20% off sitewide — code PIROUZ20 ★
Cyrus Lab

Precision audio tools, in your browser

Instruments we use in the workshop, rebuilt for the web. Everything runs locally on your device: no audio is uploaded. Only small anonymised analytics counts and an ABX score you choose to submit ever leave your browser.

Cable Conditioning Engine

A calibrated signal generator for controlled cable conditioning (burn-in). Choose a program, set a level, press play, and leave it running. The session persists in this browser, so you can close the tab and resume later.

Before you press play:
  • Start with your amplifier volume at minimum and bring it up slowly. Quiet at the listening seat is plenty for conditioning.
  • These signals can damage speakers or headphones at high level. Full-range sweeps and noise stress tweeters far more than music does.
  • Never run these signals into headphones while wearing them. Disconnect headphones or leave them on a desk.
  • Make sure the correct output device is selected and nothing else is connected that you do not want driven.
Idle.
Elapsed
00:00:00
Remaining
24:00:00

Watch it in the Listening Room below for radial, lissajous and particle modes, music and fullscreen.

RMS level-inf dBFS
Crest factor0.0 dB
Spectral tiltn/a
Program phaseStandby
About this tool

The sweep is a logarithmic sine from 5 Hz to 40 kHz (the top is clamped to 90% of your playback Nyquist, so at 44.1 kHz it reaches about 19.8 kHz), quantised so the loop point lands on an exact integer number of cycles and repeats with no click. The pink noise program uses a Paul Kellet style IIR filter over white noise for roughly equal energy per octave, crossfaded at the seam for the same reason. Red noise integrates white noise through a leaky integrator for a bass-weighted spectrum. The MLS program is a maximum-length sequence from a linear feedback shift register, one exact sequence period per loop. The AM sweep adds a 0.5 Hz tremolo quantised to the loop length, the warble steps through 1/3-octave bands with about 5 seconds per band, and the multitone sums 20 discrete tones with frequencies quantised to the loop. The burst program plays a 0.4 second noise burst every 4 seconds and is calibrated so the burst itself meets the level setting, so treat it as louder than the number suggests. Cascade duty-cycles sweep and noise with rest blocks and inverts polarity on alternating blocks, so the program has no net DC component across a full cycle. Level is calibrated in dBFS RMS with a hard cap that keeps peaks below -6 dBFS. The design rationale is in our white paper.

Proof of Conditioning

Record a 30-second in-room measurement before you start conditioning, then again after the program completes, and see what changed. This is a relative in-room response captured with your own microphone, not a lab-grade measurement. Use the same mic position and playback volume both times.

No baseline recorded yet.
About this tool

Each recording plays 30 seconds of the same shaped pink noise the conditioning engine uses while your microphone listens. The captured audio is averaged into a Hann-windowed FFT spectrum and reduced to log-spaced bands from 20 Hz to 20 kHz. The baseline is stored only in this browser; the only thing ever sent to us is a single anonymised count when you generate a certificate (program and hours, nothing else). Because the microphone, room and speaker all colour the result, compare only measurements taken with the same setup, and treat the chart as a relative before/after view rather than an absolute frequency response.

The Double-Blind Panel (ABX)

Can you actually hear the difference between two files? A and B are played as a randomised X. You say which one X is, and we compute the exact binomial probability of getting your score by guessing. Files are decoded locally in your browser; nothing is uploaded.

Loading samples...
Trials0
Correct0
p-value (exact)n/a

About this tool

Each trial, X is randomly A or B with equal probability. After n trials with k correct identifications, the p-value is the exact two-sided binomial probability of scoring at least that far from chance if you were guessing. p below 0.05 counts as significant evidence of an audible difference; p below 0.01 is strong evidence. Anything else means no audible difference is proven at that trial count, which is itself a useful result. With your consent via the Finish button, an anonymised score (trials, correct, p-value, sample source; nothing else) is sent to our analytics so we can see how audible these differences are in the wild. The methodology is documented in the companion white paper.

System Diagnostics

One-tap diagnostic signals, generated fresh each time. Keep the volume modest; several of these are noise-based and sound louder than the meter suggests.

A five-step walkthrough of the essentials: polarity, channel identification, balance, subwoofer integration and room modes.

Channel, balance & polarity

Three 880 Hz beeps from the left speaker only. If you hear them on the right, your channels are swapped.

Three 1320 Hz beeps from the right speaker only. The higher pitch also helps confirm which side is which.

An 880 Hz tone in both channels, in phase. It should image as a tight phantom centre, midway between the speakers, with equal loudness on each side.

100 Hz in both channels, in phase. Bass should sound full and solid, locked to the centre.

The same 100 Hz with the right channel inverted. With correct wiring this should sound thin, hollow and diffuse, with much less bass than the in-phase tone. If it sounds fuller, one speaker is wired backwards.

A slow 8-second sweep through the woofer band. Listen for rattles, buzzes or cabinet resonances that only appear at certain frequencies. Keep the level modest.

Ready.

Width & imaging

Pink noise switching between speakers every second. Each side should sound identical in tone and level, and the image should snap cleanly between speakers with no centre leakage.

Noise panning smoothly from hard left to hard right over 6 seconds. The image should glide in a straight line at constant loudness, with no jump in the middle or dip at the ends.

Noise bursts stepping through 25, 50, 75 and 100% left and right. Each step should land at a distinct, stable position; blurred or wandering positions suggest imaging or room problems.

Mono noise stepping in fine increments from left through centre to right. The centre step should sit exactly between the speakers; if it pulls to one side, adjust toe-in or speaker position and try again.

Pink noise with the right channel inverted. Should sound extremely wide, diffuse and hard to locate, wrapping around the sides. A centred or narrow sound here suggests a channel or phase integrity problem.

Depth

Alternating 2-second blocks: identical noise in both channels (narrow, upfront) versus independent noise in each channel (wide, recessed). The contrast between the two is a rough gauge of how your system and room render front-to-back depth.

Height (perceptual exercise)

Honest caveat: there is no height channel here. These band-limited noise variants rely on spectral cues that many listeners perceive as vertical position. Treat it as an ear exercise, not a measurement.

Noise shaped to the low-mid band. Most listeners localise this low in the stage, near floor level.

Noise shaped to the presence band. Tends to localise at tweeter height, straight ahead.

Noise shaped to the brilliance band. Often perceived above the speakers; compare with the low band for the full effect.

Speed & attack

Short transient bursts at 2, 4, 6 and 8 per second. Attacks should stay separate and crisp as the tempo rises; smearing or blurring points at slow drivers, room splash, or an over-damped amplifier.

One isolated 20 ms transient. Should be a single clean crack with immediate silence after it; any echo, buzz or tail is coming from the room or the speakers, not the signal.

Bass texture & room

Warbling tones stepping through 1/3-octave bands from 20 to 200 Hz, 2 seconds each. Every band should speak at a similar level; a band that booms or vanishes is your room, not the cable.

A slow 20-300 Hz sweep with 2-second dwells at 40, 80 and 160 Hz. Frequencies that suddenly swell, sustain or rattle the room are your standing-wave hotspots.

Band-limited noise sweeping 50-400 Hz. With a well-integrated subwoofer the level should stay even all the way up; a dip near the crossover point means the phase or crossover setting needs work.

Level & dynamics

A 1 kHz tone alternating between channels while the right side drops in 0.5 dB steps. Note the round where you first hear the image pull left; that is roughly your balance sensitivity.

A 1 kHz tone stepping down in 6 dB increments to -30 dB. The step where it disappears into room noise marks your system's usable dynamic floor at that listening level.

Ready.

Measure Your Room

Play a logarithmic sweep through your system while your microphone listens. Cyrus Lab calculates a relative in-room frequency response and a decay waterfall, then flags strong bass resonances.

Before you measure:
  • Keep the microphone at the listening position and the room quiet. A phone or laptop microphone is enough for a relative comparison.
  • Start with the volume low, then use a normal listening level. Do not wear headphones during the sweep.
  • This is a RELATIVE in-room measurement, not lab-grade absolute SPL. Microphone, speaker and room response are all included.
Ready. Allow about 24 seconds for the level check, sweep and decay capture.
How the measurement works

The tool deconvolves the microphone recording against the sweep it emitted, using frequency-domain division with regularisation. It aligns the direct arrival, estimates the impulse response, averages the response into 1/12-octave bands, and measures decay over short time slices. The recording is processed and saved only in this browser. Audio is never uploaded.

Live RTA + Level

A live microphone spectrum for watching room response and bass build-up. Microphone processing starts only when you switch it on and stops when you switch it off or leave the page.

Microphone is off.
Wideband microphone level -inf dBFS
Unweighted, calibration offset +0.0 dB

dBFS is digital level relative to full scale. Calibrate against a known level by entering the difference here. The result remains dependent on your microphone and device gain, and is not an absolute SPL certification.

Showcase

The Listening Room

A real-time picture of sound. Follow the Conditioning Engine, play something from the bundled CC0 music library, or drop in a file from your device. Nothing is uploaded, ever.

Waiting for audio. Start the conditioning engine or pick another source.
About the Listening Room

The visualiser reads the same analyser taps that drive the Lab's meters, so what you see is exactly what is being generated. The music library is CC0 (public domain dedication) and is bundled with the site, not streamed from anywhere. Your own files are decoded locally and never uploaded. Rendering pauses automatically when the visualiser is off screen or the tab is hidden.

Music credits: "C-Beams" and "Floodlight" by John Bartmann, and "Monster Parade" by Loyalty Freak Music, all CC0 via Wikimedia Commons.