🎤 SEM Mic Array Setup Sheet

For event organizers running electronic judging on tuninglabs.net · v1.1 · 2026-05-19

← Back to Mic Array

1Before the event — desk prep

Do this before you leave for the venue. 15–20 min the day prior.

Log in at tuninglabs.net/portal/tl-comp-organizer.html with your organizer email + password. (If you don't have an account yet, register on the same page.)
From My Competitions, click the 🎤 Mic Array button next to your event. This drops you straight into the Curves tab with your event pre-selected.
Alternatively from any page: Shop Tools → 🎤 Mic Array → 🎚️ Curves tab, then pick your event from the dropdown.
Add the target curve to judge by:
- Library preset (Harman / B&K / IASCA RTA flat / MECA RTA flat / Audiofrog / Dolby X) — fastest path
- Shop House — pick from your or a linked shop's saved curves
- Inline — paste your own freq dB pairs
Use speaker_position = "all" for one curve covering every speaker, OR add per-position curves (FL, FR, SUB, CEN, etc.) for finer judging.
If specific contestants need a different target (unusual layout, motorcycle vs car), add Per-Contestant Overrides in the same tab.

The sparkline next to each curve previews its shape. Verify the curve you picked looks right before you save.

2At the event — hardware setup

Bring: laptop, USB mic array (or interface + matched mics), tape measure, contestant roster.

Plug all mics into the USB array / interface. Power the interface on.
Open the portal in Chrome or Edge. Safari may hide device labels even after permission.
Go to Mic Array → Mic Catalog tab.
Click 🔍 Detect Connected Mics. Grant the one-time mic permission when the browser asks.
Verify every mic you plugged in shows up in the catalog. The system identifies common mics (UMIK-1/2/X, EMM-6, OmniMic, Earthworks, Focusrite, MOTU, etc.) and pre-fills brand/model.
If a mic doesn't appear: replug the cable, click Detect again. If still missing, check OS sound settings to confirm the mic is listed there.

Detect requires microphone permission in the browser. If you previously denied it, click the lock icon in the URL bar → Site settings → allow Microphone → reload.

3Build the array template

One array template per rig. You can clone it later if your rig changes.

Open the Array Builder tab.
Click + New Array. Name it descriptively — e.g. 2026 Spring Sound Off — 4-mic UMIK-X.
Click 🔍 Auto-fill from connected to populate slots automatically, OR click + Add Slot and assign each mic manually.
Up to 16 USB channels per array.
Save. The template stays valid for the entire event.

4Per-contestant — place each mic + measure

When a contestant's car is in the judging bay, do this for each entry.

Switch to the 📐 Measurements tab → + New Measurement.
Context type: SEM Contestant. Pick your event, then pick the contestant from the roster.
Pick the array you built in step 3.
Set the Reference Point for this car (default: driver headrest center).
Place every mic in the array at a distinct named position following the academy's 6-position ear-ellipsoid pattern:
- Driver headrest center, Passenger headrest center
- Driver A-pillar (ear-height), Passenger A-pillar (ear-height)
- Driver shoulder (±6 in lateral from headrest), Passenger shoulder
- Add rear-seat positions only if the event judges rear-seat imaging.
For each mic slot, fill the row:
- Type the position label — free-form, editable per measurement.
- Measure with the tape from the reference point to that mic's capsule and enter inches in the Distance (in) column.
- XYZ axis fields are optional — fill them only if you're capturing 3D placement.
Do not pre-average. The platform captures from all mics simultaneously and produces the spatial average automatically. Your job is placement + distances, nothing else.
Click Save Placement. The "🎯 Live Judging Capture" card now appears at the bottom.

How many mics? 4–8 mics is the sweet spot for SEM judging. The Geddes-style 6-point ear-ellipsoid is the academy-cited baseline. 1–3 mics is acceptable for quick checks but reduces spatial confidence. Above 8 mics gives diminishing returns inside a typical cabin.

Position labels live on the measurement, not the array template. They're editable per use because car geometry changes contestant-to-contestant.

4.5🌬️ Ambient noise capture (5 seconds, optional)

Captures the room's sonic signature so it can be subtracted from the actual test. Skip in a quiet room; use when there's crowd noise, HVAC, or other systems playing nearby.

Before the contestant plays anything, click 🌬️ Capture Ambient (5s) in the judging modal.
The system listens for 5 seconds with no playback expected — operator + contestant stay silent. No tone, no music, no engine revs.
The platform stores the ambient spectrum and shows a small sparkline confirmation so you can see the room's noise floor.
During the 90-second test, the platform subtracts this ambient in the power domain band-by-band, so the measurement reflects only the contestant's system — not the room.
If the room is genuinely quiet (no PA, no other rigs running, no HVAC bleed), click Skip — quiet room and proceed straight to Step 5.

How the math works: power-domain spectral subtraction. If the room has a –40 dB hum at 100 Hz and the contestant plays at –10 dB at 100 Hz, the score uses ~–10.04 dB (the hum is mathematically removed without distorting the contestant's signal). Loud bands stay loud; only the noise floor moves.

5Live capture + score

90 seconds. Platform handles the averaging.

In the placement editor, click 🎯 Live Judging Capture (90s).
Pick the playback source:
- Tone — pink noise or sweep from the contestant's head unit (cleanest result)
- Music — the event's reference track played from the contestant's system
Contestant starts playback. Operator clicks Start. 90-second timer runs.
The platform captures from all mics simultaneously and produces the spatial average using RMS / power averaging (AES TC-AA standard) with per-mic SPL alignment before the average.
When the timer ends, the score appears:
- Curve accuracy (out of 100) vs the target curve
- Rattle / outlier flags per mic, if any
Click Show per-band detail to see green/amber/red deviation per band.
Coherence weighting (8+ mics): each mic gets a per-band quality weight. A mic sitting next to a panel rattle gets quietly down-weighted at that band only, not across the whole capture — so one bad mic at one frequency doesn't poison the rest of its data. This is the pattern Smaart and Dirac use. No operator action needed.
Spatial consistency metric: alongside the score, you'll see a cross-position variance number. Lower = system sounds the same at every mic. Higher = the system has hot spots and dead spots. Useful as tuning feedback; available as an optional sub-score for judging.

Scoring formula: 100 − (MAD × 5), clamped 0–100. 0 dB MAD = 100. 10 dB MAD = 50. Target precedence: per-contestant override → event default → IASCA flat fallback.

ℹ️Why the platform uses RMS / power averaging

We use power averaging because equal-weight dB averaging underweights peaks and overweights nulls (logarithmic bias) — a 6 dB null between two mics would look like a real cabin null when it's just placement spread. Power averaging is what REW, Dirac Live, Audyssey, and Trinnov all use under the hood. It's the AES TC-AA standard for multi-position spatial averaging and what Floyd Toole and Sean Olive endorse in the Harman research.

For 8 or more mics, the platform automatically upgrades to coherence-weighted RMS averaging — the per-band, confidence-weighted version Smaart uses. Below 8 mics we stay on plain RMS / power averaging because the statistics get thin and the weights become unreliable. Either way, a single-mic capture degenerates cleanly to just that mic's response — no special case.

6Re-measure / dispute resolution

If a contestant disputes the score, re-pull the same placement record. Position labels and inches are preserved.
Re-capture, paste new bands, click Save — the old measurement is replaced by the new one.
Delete the latest measurement (button below the textarea) to revert to the prior one.

📚Quick reference — library presets

Preset	Use case	Notes
Harman in-car	SQ tonality judging	Olive/Welti research target. Bass shelf, gentle treble roll-off.
B&K 1974	Classic room reference	Strong bass tilt, gentle treble roll-off.
IASCA RTA flat	IASCA RTA/SPL Challenge	Flat across 31 bands. Variance from flat = penalty.
MECA RTA flat	MECA RTA Freq Out	Flat across 31 bands. Same scoring pattern as IASCA.
Audiofrog / Wehmeyer	Pro install / tuning judging	Harman variant with slight midbass emphasis.
Dolby X-Curve	Cinema-style	SMPTE/ISO 2969. Flat to 2 kHz, –3 dB/octave roll-off above.

⚠️Troubleshooting

Symptom	Fix
Detect button does nothing	Browser blocked mic permission. Click the URL-bar lock icon → allow Microphone → reload.
Mic shows up with no name	Permission granted, but browser didn't expose the label. Safari does this — switch to Chrome or Edge.
Same mic appears twice	It won't — upsert is keyed on device ID. If two rows exist, they're truly two different USB devices.
Auto-fill from connected drops mics	Max 16 slots per array. Excess detected mics are skipped.
Score shows "no measurement"	You saved the placement but haven't pasted the band data yet. Click Save Measurement & Score on the placement editor.
Target source says "fallback flat"	No event-level "all" curve set and no per-contestant override. Go back to Step 1 and set an event curve.
Score lower than expected	Click "Show per-band detail" — large red Δ values point to the bands hurting the score. Often a level offset between measurement and target (system is louder/quieter than the target curve expects).
Score shows a yellow ⚠ outlier warning — one of my mics flagged	One mic's response deviates >6 dB from the median in >20% of bands. Check that mic for: (1) covered/blocked port, (2) loose connection, (3) sitting near a panel rattle. Re-position and re-capture, or accept the flag — the platform does not auto-drop outliers (auto-drop is indefensible at a judged event).
Score got worse when I added more mics	Probably picking up crowd / HVAC noise via the additional positions. Run the Ambient Capture (Step 4.5) before the next attempt so the room signature is subtracted out. Also check the outlier-flag warning — one of the new mics may be sitting on a panel rattle.
Cross-position variance is very high (>4 dB avg)	System sounds very different at different mic positions. For tuning: this points to imaging / staging weakness — work on time alignment, level matching, and seat-to-seat balance. For judging: it's a real characteristic of the install — don't "fix" it by removing mic positions. The spread is the truth.

Tuning Labs SEM Mic Array Setup Sheet · v1.1 · 2026-05-19
Live: tuninglabs.net/portal/sem-mic-setup-sheet.html · Print for clipboard use, or open on a tablet at the event.