SpectraSense

Free. SpectraSense opens with a free Squaresmile account — on a phone, tablet or laptop.

SpectraSense listens to a microphone or an audio interface and draws the spectrum from 20 Hz to 20 kHz. Its job is ringing out: while you push a wedge, an in-ear mix or the PA up to the edge, it watches for the narrow, steady tone that feedback is and flags it while it is still building — with the frequency, the note, the nearest graphic-EQ band and a suggested cut. Every ring it finds goes into a Found list you can copy as text. Everything runs in the browser; no audio is recorded or sent anywhere.

SpectraSense is a ring-out assistant, not a feedback suppressor: it never touches your desk, and it is made for the ring-out before the show, not for running under music during it.

Who it is for

• Monitor engineers ringing out wedges and in-ears.
• FOH engineers bringing a system up in a new room.
• Bands without a tech who want to know which frequency is howling.

Start

1. Tap Start listening and allow the microphone.
2. Pick the input under Input: the device's own microphone, a measurement mic or a USB interface fed from the desk.
3. Bring the system up. When a tone starts to ring, a red marker appears on the spectrum and a card shows the frequency, the note (e.g. D7 −7¢), the nearest GEQ band, a suggested cut, and whether the ring is growing, ringing or already gone.
4. Cut that frequency on the mix's EQ — and bring the fader up again for the next one.

The phone, the tablet and the laptop all work; the screen stays awake while SpectraSense is listening. On a phone the answer card sits at the top, above the display (held sideways, it floats over the display), with the frequency in large type, so you can read it at arm's length with the other hand on a fader.

Ringing out a wedge, step by step

1. Music off, unused mics muted. Only the mic and the mix you are ringing out stay open.
2. Hold the phone where the singer's mouth will be — at the mic, not against the wedge — with the bottom of the phone (where its microphone is) pointing at the wedge. A phone microphone overloads long before a wedge does.
3. Type the label for this mix — Wedge 3, IEM 2, Sidefill — so the rings you find carry it.
4. Push the send slowly until it just starts to ring. SpectraSense flags it; pull the fader back.
5. Cut as suggested on that mix's output EQ, then go up again for the next ring. If the same frequency comes back, cut it a little deeper.
6. Stop when several frequencies ring at once, back the fader off a few dB, and move on to the next mix.
7. Copy as text when you are done and paste the list into a message or the advance sheet for the next show in that room.

Low hoots read smaller on a phone. Phone microphones roll off the low end, so a hoot around 100–300 Hz can look weaker on the display than it sounds in the room. Trust your ears there — or use a measurement mic or a USB interface.

The suggested cut

Each ring comes with a starting point, the way most engineers ring out:

• Start at −3 dB, narrow — about 1/10 octave (Q ≈ 14) on a parametric EQ. Cut only as much as it takes; deep, wide cuts take the voice with them.
• Deeper if it returns. When the same frequency rings again, the suggestion goes a further 3 dB down, to −9 dB at most. If it still rings after that, the mic or the wedge is too loud for the room — turn it down or move it.
• On a graphic EQ the card names the nearest band. When the ring sits between two bands, it names the neighbour too — pull both a little rather than one a lot.

It is a suggestion, not a setting: SpectraSense never changes anything on your desk.

The Found list

Every ring SpectraSense confirms is kept in the Found list: the time, the frequency, the note, the GEQ band, the suggested cut and the label of the mix you were ringing out. The live card fades a few seconds after the ring stops, and the answer card then keeps showing the last ring ("Last ring") until another one comes; the list keeps every ring, so you can pull the fader first and read afterwards.

• The list stays when you stop, freeze or reload the page, until you clear it.
• Copy as text puts the whole list on the clipboard, one ring per line.
• The list is kept on this device only, in the browser's storage. It is not synced to your account and never uploaded.

How it decides something is feedback

Research on howling detection (van Waterschoot & Moonen, Fifty years of acoustic feedback control, 2011) showed that no single test is reliable: "it lasts" alone is wrong more than half the time. SpectraSense flags a frequency only when three things are true at once for 0.3 seconds:

• It is a narrow peak standing clearly above the spectrum around it (the Sensitivity setting, 14 dB by default).
• It is a lone tone. A musical note has overtones at 2×, 3×, 4× its frequency; feedback does not. Notes are ignored (you can switch this off).
• It is steady or growing, not decaying like a piano note or wobbling like a vibrato.

Once a ring is confirmed it stays flagged for as long as it is there. After it stops, its marker fades for a few seconds (four by default), so you have time to read it.

The rest of the screen

• Peak hold — a thin line that keeps the loudest level each frequency reached for a few seconds, so a short ring between two glances still shows. Clear hold resets it.
• Freeze (or the space bar) — stops the display so you can read it; tap or drag on the spectrum to read the frequency, level and note under your finger.
• Waterfall — the same frequencies across, time downwards, to watch a ring build up as a fader goes up.
• How fast it reacts (in Settings) — four controls for how fast things come and go: Smoothing (how fast the live line follows), Peak hold (how long the peak line keeps a level), Peak fall (how fast it drops after that) and Feedback marker stays (how long a red marker lingers after the ring stops).
• Console analyser look (in Settings) — tilts the display by +3 dB per octave so pink noise and most music look flat, the way a third-octave analyser on a console shows them. Without it, a normal mix looks bass-heavy here, because every frequency slice is equally narrow.
• Start at 40 Hz (in Settings) — hides the rumble below 40 Hz that a phone microphone picks up once voice processing is off.
• Settings — also Sensitivity, the note check, the A4 reference (440 Hz by default), FFT size, the dB range and waterfall length. They are saved and follow your account to other devices.
• The input is the exception: each device remembers its own choice, so picking the measurement mic on the laptop does not change what your phone listens to.

Good to know

• Feedback is detected from 60 Hz to 16 kHz; the display shows more.
• On an interface with several inputs, SpectraSense hears the first one — patch the measurement mic or the desk feed there.
• Levels are in dBFS — relative to the input's full scale, not sound pressure. Use RoomSense for SPL and room measurements.
• A held pure tone — a test oscillator, a whistle, a flute or a synth pad with few overtones — looks exactly like steady feedback and will be flagged. That is the nature of the job, and why SpectraSense is for the ring-out, not for the show.
• No signal means nothing above −70 dBFS for three seconds: check the input selection and its gain. Input clipping means the input is overloaded: move the phone further from the wedge or turn the gain down — a clipped input adds overtones and false peaks of its own.
• If the browser insists on voice processing (echo cancellation or automatic gain), SpectraSense says so — readings can be off; a USB interface usually avoids it.