by Simon Byrne.
Obviously an audience needs to be able to understand what is spoken or sung. That is, the performance must be intelligible. Therefore well implemented PA systems combined with acceptable acoustics are crucial for this to be achieved. This becomes a major safety issue for evacuation PA systems in buildings such as concert halls, shopping centres, airports, or railway stations. The public must be able to understand emergency announcements or lives could be at risk. It is not just the loudness that affects intelligibility.
Signal distortions, frequency response, sound pressure level, multiple sources and reverb time all affect intelligibility.
You’d think that by simply talking though a PA system, a few “Check one TWO, one TWOs” with careful listening would quickly establish that things are intelligible or not. As it turns out, humans are quite poor at determining intelligibility.
This is because when you know what is spoken, or likely to be spoken, your brain fills in the unintelligible bits. So repeatedly doing “Check one TWOs” is close to useless because you know what is being said. Even if you are not forewarned as to what is being said, the brain has great capacity to construct the missing bits from a sentence.
And the end of this article is a Bitly link to a Youtube video called “Can you trust your ears?” By NTi, it demonstrates the remarkable effect by playing a distorted signal which cannot be understood. However, once you are told what it says, you subsequently understand it perfectly. The effect is amazing.
On a touring show or long theatre run, it is an interesting fact that the regular crew and performers are probably the worst people to comment on intelligibility. The performance is so ingrained into them that their brains replace missing auditory information.
I’ve certainly been to concerts where the vocals are lost in the mix, possibly because the front of house operator is plagued by this phenomenon and simply doesn’t realise that they are benefiting from knowing the show.
Also, the brain can be tricked into replacing what you hear with what you see!
The effect is called the McGurk effect. The McGurk effect is an interaction between hearing and vision in speech perception. The illusion occurs when the auditory component of one sound is paired with the visual component of another sound, leading to the perception of a third sound.
The visual information a person gets from seeing a person speak changes the way they hear the sound! If a person is getting poor quality information but good quality visual information, they may be more likely to experience the effect.
It is not the same for everyone either. People who are better at sensory integration have been shown to be more susceptible. Many people are affected differently by the McGurk effect based on many factors, including brain damage and other disorders (there is none of that in the production industry is there?). Interestingly, females are more susceptible to McGurk effect than males.
NTi have a stunning demonstration video of the McGurk effect on Youtube. A Bitly link is at the end. Check it out…it is worth it!
So an objective way to measure speech intelligibility had to be found. In the old days, they would get a large group of people into the space to be tested, and literally read a very long list of random words to them, and the people would then write down what they thought they heard. The results would be gathered, and the correct results counted which made up the score.
Very labour intensive and not so practical, so it wasn’t done much.
Fortunately, with modern technology and techniques, we have the Speech Transmission Index or STI. This is still quite complex to test because it requires 98 separate measurements to be obtained which takes a lot of time. Therefore an optimised version of STI for use with Public Address systems was developed, called STIPA.
STIPA has emerged as the most widely accepted method for objectively measuring the speech intelligibility of PA systems. STIPA gives you a single number between zero (totally unintelligible) and one (excellent intelligibility).
Measuring speech intelligibility using the STIPA method is not particularly difficult. With the proper tools and a minimal amount of practice, measurement results are quickly obtained that are the most accurate among all intelligibility test methods.
There are two ways to measure STIPA. The first is the Impulse response method, which you can do in Smaart. This method has some challenges, so the more accepted method is using a STIPA test source signal and a meter with STIPA capability, such as NTi’s XL2 sound level meter with the option installed.
The STIPA test signal is a noisy signal that contains all frequencies present in human speech, from approximately 80 Hz to 11 kHz.
In real speech, information is encoded through intensity fluctuations. Loss of modulation depth means loss of information, which translates into a reduction of intelligibility. This is how the STI works: the artificial test signal (STIPA) has a carefully designed pattern of modulation frequencies.
The STIPA analyser measures how the transmission through the PA system and room has reduced the modulation depth. This is measured in multiple frequency bands (125 Hz – 8 kHz), across a range of modulation frequencies (0.6 – 12.5 Hz).
The STIPA test signal is injected into the system by way of a generator which injects the signal as a line input, or a ‘talkbox’, which has its own built-in signal generator.
A talkbox is a specially calibrated loudspeaker that serves as a ‘human speaker replacement’ and is positioned in front of the microphone that feeds the PA system. The STIPA meter is located where the audience is likely to be (the listening positions).
It is important that the meter does not move during a test so it must be on a stand with the operator standing well back.
The meter records 15 seconds of the test signal and computes a result. For accuracy, several tests are taken and the average is derived. In large spaces, tests are taken every 6 – 10 metres and once again the results are averaged.
The final scores are ranked and given a score. 0 is obviously pathetically unintelligible through to 1 which is perfect. 1 is never achieved by the way, with greater than 0.76 being considered excellent and greater than 0.62 considered good in concert halls.
Using the Impulse Response method, Smaart provides figures for both STI and STIPA, however STIPA in this case is more properly termed STIPA(IR), since it is based on impulse response (IR) data rather than measured directly using the modulated noise method.
STIPA(IR) data is not considered to be as resilient as data gathered using
the modulated noise method, but it has the advantage of being much faster to acquire, plus having a full STI reading also.
Rational Acoustics (the makers of Smaart) consider it for informational purposes, for example, to facilitate comparison with readings from hand-held STIPA meters and is literally just a subset of the full STI measurement, calculated from exactly the same measurement data.
Another interesting way to access STIPA readings is by using the Sound Tools IOS app. It now has a STIPA in-app purchase. If you are interested in playing with it, they apparently offer a free seven day trial.
Speech Intelligibility can have serious legal implications where people’s safety is potentially at risk. It is in these installations that properly qualified acousticians, who have the skills and equipment (and insurances!), must take the STIPA measurements to ensure that the readings are accurate and comply with the legal standards.
STIPA is tool that gives users an objective measurement as to how effectively the spoken word is being understood. For live sound professionals, it is a useful exercise so as to understand how well PA systems combined with room acoustics are performing against a known set of parameters.
Can you trust your ears?
The McGurk Effect – What you hear is not what you see
STIPA Module for Soundtools on IOS
NTi Speech Intelligibility Application Note