🏭

Machine Condition Monitor

Leave this running next to a motor, pump, fan or gearbox and watch its frequency signature over time. The monitor takes a periodic FFT snapshot, tracks the overall level plus up to 10 frequency bands you choose on a live trend chart, raises a visual and optional audio alert when a band rises past your custom warn / danger threshold, keeps a timestamped session log you can annotate, and exports everything to CSV.

This reads airborne SOUND, not accelerometer vibration — and the levels are RELATIVE, not ISO mm/s. A device microphone hears the sound your machine radiates into the air, so it reveals the machine's frequency signature and how it changes over time — genuinely useful for spotting "something changed" and matching peaks to your calculated bearing, gear-mesh, imbalance and blade-pass frequencies. The thresholds here are custom RELATIVE changes in dBFS (digital full scale), not ISO 10816 / ISO 2372 severity zones, which require a calibrated accelerometer reading true mm/s velocity. Treat this as a companion for trend-spotting, not a certified condition-monitoring instrument. AGC, noise suppression and echo cancellation are requested off. Everything stays in your browser; nothing is recorded or uploaded.

Idle — add the frequency bands you want to watch, then press Start to allow your microphone.

Green = overall level. Coloured lines = tracked bands. Dashed lines = captured baselines. Vertical axis = relative level (dBFS, uncalibrated); horizontal axis = time.

Snapshots0
Elapsed0:00
Overall now

Tracked frequency bands

Add up to 10 bands to watch. Use the related calculators to find the frequencies that matter for your machine: shaft 1X = RPM ÷ 60 Hz, bearing fault frequencies (BPFO/BPFI/BSF/FTF), gear-mesh (teeth × shaft Hz), and fan blade-pass (blades × RPM ÷ 60). Thresholds are a relative rise above the baseline you capture — they are not ISO severity.

Run the machine in its normal/healthy state, then press Set baselines. Each band's most-recent level becomes its reference, and alerts fire when a band later rises above that baseline by your warn / danger amount.

Session log

Timestamped snapshots, threshold alerts and your own notes — useful for documenting a maintenance check. Times use your device clock. Add a note to record what you did (e.g. “swapped the belt”).

TimeTypeOverallDetail
No log entries yet. Start monitoring to record timestamped snapshots, alerts and your notes.

CSV is a local download. The most recent 40 rows are shown above; the full log (up to 5,000 rows) is exported. Values are relative dBFS, not certified measurements.

How It Works

Healthy rotating machinery has a stable frequency signature: the spectrum of the sound (and vibration) it produces stays roughly the same from day to day. When a fault develops — a bearing starts to spall, an imbalance grows, a gear tooth chips, a mount loosens — energy appears or grows at the characteristic frequencies of that fault. The earliest and most reliable warning is usually not the absolute level but the change: a band that used to be quiet getting noisier relative to its own history.

This monitor leans into exactly that. Every few seconds it grabs a fresh window of microphone audio, runs a Fast Fourier Transform, and records two things: the overall broadband level (RMS of the time-domain frame, in dBFS) and the energy inside each frequency band you asked it to track. Those values are plotted on a bounded trend chart so you can see drift, steps and spikes at a glance, and every reading is written to a timestamped session log.

You define what “normal” means by running the machine in its healthy state and pressing Set baselines. Each tracked band's current level becomes its reference. From then on, the monitor compares each new snapshot to that baseline and raises a warn or danger alert — a coloured flash plus an optional beep — when a band rises above its baseline by the number of decibels you chose. Because both numbers are relative to the same starting point measured with the same microphone, the comparison is meaningful even though the absolute level is uncalibrated.

What is trustworthy and what is not

What survives the lack of calibration is the part you actually care about for trend monitoring: which frequencies are present, and how their relative levels change over time with the same mic, position and gain. What does not survive is any claim about absolute magnitude. The vertical axis is relative dBFS, not mm/s velocity or g acceleration, so you cannot read a number off this chart and look it up in an ISO 10816 / ISO 2372 severity table — those standards are defined for a calibrated accelerometer mounted on the bearing housing. The thresholds here are custom relative changes you invent for your own machine, not certified zones. Use this to decide “has something changed, and where in the spectrum?” — then confirm anything serious with a real accelerometer-based analyzer. Consumer microphones also roll off at the frequency extremes, so very low shaft speeds and deep sub-bass may read weak.

Frequently Asked Questions

Is this a real vibration condition-monitoring instrument?
No. It listens to the airborne sound your machine radiates through a device microphone, not to true vibration from an accelerometer mounted on the machine. That reveals the frequency signature and, crucially, how it changes over time — which is genuinely useful for spotting that something has changed. But it is not a calibrated condition-monitoring instrument, and it cannot replace an accelerometer-based analyzer for diagnosis or sign-off.
Why are the thresholds in dB and not mm/s like ISO 10816?
Because a browser microphone is uncalibrated, the levels here are relative dBFS (digital full scale), not mm/s velocity or g acceleration. ISO 10816 and ISO 2372 severity zones (Good / Acceptable / Unsatisfactory / Unacceptable) are defined for a calibrated accelerometer reading true velocity at a specific location, so they cannot be applied to this signal. The warn/danger thresholds here are custom relative changes you set for your own machine — meaningful for trend-spotting, but not certified severity.
What frequencies should I track, and how do I find them?
Track the frequencies your machine should produce. Shaft running speed (1X) is RPM ÷ 60 in Hz. Bearing fault frequencies (BPFO, BPFI, BSF, FTF) come from the bearing geometry — use the Bearing Fault Frequency Calculator with numbers from the bearing datasheet. Gear mesh is teeth × shaft Hz, and fan blade-pass is blades × RPM ÷ 60. Add a band at each, set a sensible half-width (the default is ±3% of the centre, with a minimum of ±2 Hz), and label it so the trend and log are readable.
How do baselines and alerts work?
Run the machine in its normal, healthy state and press Set baselines. Each tracked band's most-recent level is captured as its reference. After that, every snapshot is compared to the baseline, and the monitor flags a band as Warn or Danger when it rises above its baseline by the dB amount you specified — with a coloured flash and, if you enable it, a short beep. Alerts only fire for bands that have a baseline set, and the audible tone is throttled so it doesn't repeat constantly.
Why must auto-gain and noise suppression be off?
The tool requests the raw microphone signal with automatic gain control, noise suppression, and echo cancellation switched off, because those features actively reshape the level and spectrum — pumping gain up and down and carving out frequencies — which would corrupt a trend. We can only request them off; if your operating system forces processing that can't be disabled, the trend may look artificially flat or clean. Turn off mic enhancements in your system settings for the best result.
Is my audio recorded or uploaded, and where does the log live?
No audio is ever recorded or transmitted. The microphone signal is analyzed in real time entirely in your browser to compute each snapshot, and the source is connected only to an analyser node — never to your speakers, so there's no playback or feedback. The session log lives only in this page's memory; a CSV export is a local download you trigger yourself. The microphone is released when you press Stop or close the tab.