Last week, at the ASNZ 2026 conference, I presented a novel and simple way to conduct STI/STIPA surveys without the need for expensive calibrated measurement equipment; a collaboration between ARD and Barton Sound Systems. Given the very positive feedback and interest in the method, I thought it was worth sharing here as well.
The principle is simple: play a standard STIPA signal through the PA system and record it using an ordinary audio recorder. The recordings can then be processed to calculate STI values, allowing us to assess the overall speech intelligibility, identify areas of poor performance, and compare results over time.
One can proceed with measurements at specific locations (as per the traditional measurement standards). The data comparing different low-cost equipment show a small positive bias but a manageable 95% accuracy for a survey or screening method.
One can also play with a continuous recording as one walks along a path covering all zones necessary and process it as a function of time, or of location.
The method isn’t intended to replace formal compliance testing with calibrated equipment, but to make objective speech intelligibility assessments more accessible, particularly for routine checks, troubleshooting and maintenance.
Of course, the key to a good survey also lies in the data processing and presentation.
To make things easier for PAVA manufacturers and installers, ARD has developed a small application that processes the recorded STIPA signals and generates STI maps directly onto PDF floor plans.
The processing therefore produces the deliverable itself: consistent, clean, and reusable, with the original recordings retained for future comparison.
Because the last thing anyone wants is another messy spreadsheet that speaks to no one.
If you manufacture, install or troubleshoot PAVA systems and are looking for a simple, cost-effective way to assess speech intelligibility, get in touch.
ARDsti could help your site engineers save time, reduce costs and produce consistent, professional survey results.
