How accurate are the predictions?

Predictions are a professional planning baseline, not a guarantee. They are good enough to decide the layout, size the amplifiers and price the job; the room writes the last decibel, and a meter on site confirms it.

How close to a measurement#

The audio engine predicts each speaker as a point source in a diffuse reverberant field. For the average level, the coverage and relative decisions, such as how many speakers, how far apart and whether the amplifier has headroom, expect it to agree within about ±2 to 3 dB of a careful SPL-meter measurement at mid and high frequencies.

Where it is strongest#

Comparisons. Which speaker is louder, where the dead spots are, whether a layout is even and whether the amplifier has headroom: use the tool with confidence to make those decisions.

Where to treat it as an estimate#

• Single points, steep off-axis angles and overlaps. Where real speakers overlap they comb-filter, and the level at one spot can vary by ±3 to 6 dB. Turn on the Interference layer to see that ripple.
• Bass in small rooms. Below the room’s modal, or Schroeder, frequency, real rooms have peaks and nulls of ±10 dB that no steady-state model resolves. Treat low-frequency figures there as approximate.

Get the best accuracy#

The model is only as good as what you tell it:

• Honest sensitivity. The library normalises every speaker to 1 W at 1 m from its datasheet.
• A realistic RT60, estimated from materials or measured. See Set room acoustics and RT60.
• The right heights: the ceiling and the listeners’ ears.
• The noise floor and the climate of the venue.

The model already includes the power the amplifier can actually deliver, bass reinforcement near walls, compression near a speaker’s rated power and the air absorption of high frequencies.

Flat-panel speakers#

Distributed-mode panels, such as the Flat Panel Audio DML500 and any speaker listed as DML or flat panel, behave differently from a cone speaker. The tool models them as true panel radiators: they lose only about 3 dB for each doubling of distance, against 6 dB for a point source, out to about 10 m, and then follow the normal law. Their 165° × 165° rating is a window, not a −6 dB cone: inside it, the level stays within about 3 dB of the axis, on a wall or in a ceiling.

Inside the engine#

For readers who want the methods:

• Reverberation is calculated with an Eyring correction, which stays exact in absorptive rooms where the simpler Sabine formula breaks down and agrees with Sabine in live ones. Air absorption at high frequencies follows ISO 9613-1.
• Directivity. Each speaker gets its own directivity factor from its real coverage angle, so a narrow speaker feeds less energy into the reverberant field than a wide one.
• STI is computed per IEC 60268-16:2011, from the full modulation transfer matrix, with the masking of the venue’s noise per octave band. One STI figure is used everywhere: in the results, the standards check and the report.
• High accuracy (reflections), in Layers, adds the first reflections from the walls, floor and ceiling to the map, which brings it closer to a meter reading near the boundaries.
• Speed. The engine caches each speaker’s field, so the map recalculates in milliseconds and follows a speaker while you drag it. The calculation grid defaults to Auto.

For rooms where the reverberation is not the same everywhere, the Ray-Traced Check, in beta, follows the real geometry.

Always verify on site#

Commission with an SPL meter, and with an STI or STIPA check for a speech system, then fine-tune. Two tools close the loop:

• Site calibration trims the whole model to your meter reading, so the handover documents show the room as built.
• Site tracking records the measured level, STI and RT60 next to the prediction for every area.

See Commissioning and tuning tools.