Professional audio system optimisation is not a single act — it is a sequence of measurement disciplines applied at different phases of a project. The mistake most often made on-site is conflating those disciplines: using a frequency response trace to judge intelligibility, or using an STI score from one seat to characterise a space that seats eight hundred. The three tools that together cover the full scope of system measurement are Rational Acoustics Smaart, the NTI Audio XL2, and the NTI TalkBox. Each has a defined domain. Understanding where each tool's validity ends is as important as understanding what it measures.
Smaart is a dual-channel, FFT-based measurement platform. Its core function — transfer function measurement — works by comparing a reference signal (the input to the system) against the measured signal (the output, captured by a calibrated microphone) in the frequency domain. The result is a complex ratio: magnitude tells you the amplitude relationship between input and output at each frequency, and phase tells you the time relationship. From this complex transfer function, Smaart can derive the impulse response via inverse FFT, giving you the time-domain picture of how the system behaves — direct sound, early reflections, late reverberation, and the timing relationships between all of them.
The spectrograph view extends this into a running time-frequency plot, useful for watching how the system behaves as conditions change — a house filling with people, a mechanical noise floor rising, a feedback event. But the trace that separates competent Smaart operators from experienced ones is coherence, and it is the trace most often ignored or misread.
Coherence in a dual-FFT measurement is a statistical measure of the causal relationship between the reference and measurement signals. A coherence value of 1.0 means every component of the output is linearly predictable from the input — the system is behaving as a deterministic linear system at that frequency. Coherence drops when three things happen: the signal-to-noise ratio is low (the measurement microphone is picking up more noise than system output), there are uncorrelated signals arriving at the microphone from sources other than the one being measured (reflections from surfaces that arrive with varying delay), or the system itself is nonlinear. In practice in an installed audio environment, all three happen simultaneously and continuously.
The critical misuse of Smaart is trusting the transfer function trace in frequency regions where coherence is below approximately 0.9. Below that threshold, the transfer function values are increasingly contaminated by noise and incoherent energy, and the displayed curve is partly a measurement artefact rather than a true representation of system behaviour. Engineers who EQ to a low-coherence trace are EQing noise. This matters most in the low frequencies — below 200 Hz, coherence is almost always degraded in a reverberant space — and in the upper frequencies at long distances where SPL drops and the noise floor's proportional contribution rises. Smaart's delay finder function, which cross-correlates the reference and measurement signals to find the time-of-flight between source and microphone, is also coherence-sensitive: a delay lock made on a low-coherence signal will drift.
Smaart belongs to the alignment and system design verification phase of a project. It tells you what the electroacoustic system is doing: where the crossover points are, whether delay fills are time-aligned to the main system, what the phase relationship between arrayed enclosures looks like at the coverage boundary, and whether the DSP processing is introducing unexpected artefacts. It does not tell you whether what the system is doing translates into intelligible speech for the listener. That distinction is fundamental.
The NTI Audio XL2 is a handheld acoustic analyser. Its capabilities include calibrated SPL measurement with A, C, and Z weighting and time constants (Fast, Slow, Impulse), 1/3 octave real-time analysis, polarity detection using an impulsive reference, and RT60 measurement using the integrated decay method. All of these are standard functions found in instruments of this class. What separates the XL2 in the context of speech system commissioning is its implementation of STIPA — the Speech Transmission Index for Public Address systems, defined in IEC 60268-16.
STI is not a frequency response metric. It is an objective measure of how well a transmission channel — the entire chain from microphone input to listener's ear, including the room — preserves the amplitude modulation of speech. Human speech carries information through modulation: the rapid fluctuations in amplitude and spectral content that encode phonemes, syllables, and words. A room with high reverberation smears those modulations. Background noise fills the troughs between modulations, reducing the depth of modulation at the listener's ear. The STI method quantifies exactly this degradation.
The full STI method uses seven octave bands (125 Hz through 8 kHz) and fourteen modulation frequencies within each band, producing a 98-point modulation transfer function matrix. The STIPA variant — designed specifically for field use with PA systems — uses a fixed test signal that simultaneously transmits two modulation frequencies per octave band, covering all seven bands concurrently. This means a STIPA measurement can be completed in a few seconds rather than the several minutes required for full STI, making it practical for grid measurements across a large space. The XL2 processes the received signal, extracts the modulation transfer function values, applies the signal-to-noise ratio corrections and distortion corrections defined in the standard, and produces a single STI value between 0 and 1.
The NTI TalkBox is the calibrated companion stimulus source for the XL2. It is a self-powered, battery-operated loudspeaker that reproduces the STIPA test signal — a speech-shaped noise carrier amplitude-modulated at the fourteen STIPA modulation frequencies — at a defined SPL, with a known directivity pattern, from a fixed position. The significance of the TalkBox is that it provides a repeatable, standardised input to the acoustic measurement chain. Without a calibrated source, STI measurements made with music or speech programme material are difficult to compare between sites or between measurement sessions, because the characteristics of the source affect the result. The TalkBox removes that variable.
In the STIPA measurement workflow with XL2 and TalkBox, the TalkBox is placed at the talker position — at the podium, at the pulpit, at the gate agent's desk. It transmits the STIPA signal into the room acoustics and through the PA system simultaneously. The XL2, placed at successive listener positions across the coverage area, measures the STI at each point. The result is not a single number but a spatial distribution — a map of intelligibility across the space. Areas of low STI reveal where the system is failing: insufficient direct-to-reverberant ratio, excessive comb filtering from reflections, inadequate SPL to compete with the noise floor.
The workflow across a full project follows a clear sequence. Smaart governs the design and alignment phase: loudspeaker positions are verified, delays are set, level matching between zones is confirmed, and the electroacoustic response of the system is documented. The XL2 governs the verification phase: STIPA measurements confirm that the electroacoustic performance translates into actual intelligibility at the listener's ears under real acoustic conditions. The TalkBox provides the controlled stimulus for those STIPA measurements. These are not interchangeable roles.
Common errors in this workflow include three recurring patterns. The first is closing out a project based on frequency response alone — a flat transfer function in Smaart is a necessary but not sufficient condition for high intelligibility. A system with flat response in a space with RT60 of 3 seconds and a 68 dB(A) noise floor will have a flat frequency response and an STI of 0.35. The second error is making measurements in frequency regions where coherence is below 0.9, producing EQ decisions that are based on artefacts rather than system behaviour. The third, and perhaps most consequential in acceptance testing, is measuring STI at a single position and declaring the system compliant. IEC 60268-16 specifies that STI should be measured at multiple positions distributed across the listening area, and that the spatial uniformity of STI is a performance criterion. In a large airport concourse, the worst-case seat matters as much as the average.
In houses of worship, the combination of these three tools provides a complete commissioning framework. Smaart aligns the array, sets the delay fills, and verifies the system's electroacoustic behaviour. The TalkBox and XL2 then provide the objective evidence that the system achieves the intelligibility targets defined in the specification — not just in the front row, but across the full seating area including under-balcony zones where reverberant energy accumulates. In airports, where STIPA measurement must be conducted at the operational noise floor rather than in an empty building, the TalkBox provides the fixed-level calibrated source against which the competing noise can be properly characterised. In each case, the tools are most powerful when used in sequence, each one addressing the questions that the others cannot answer.