Troubleshooting

If you are experiencing problems when using iSTI, please refer to this section. If this does not solve your problem, and the information here (and elsewhere on this website including the iSTI FAQ) does not provide the solution, then contact us for support: iSTI@embeddedacoustics.com. Answering your question may take up to 5 business days, usually less. Support calls from iSTI Professional users are treated with priority. Please note that background information on the Speech Transmission Index itself, as well as general aspects of acoustic measurements, is beyond the scope of support offered to iSTI customers.

Using the app

The iPhone and iPod Touch offer little real estate for buttons and for showing results. Whenever a lot of information needs to be shown, such as in the detailed measurements view, your screen tends to become cluttered. There is a description of all buttons and controls for iSTI Professional in the iSTI professional walkthrough. If you are using iSTI Standard or Lite, you will be missing some of the options and controls shown in this walkthrough.

If you encounter obvious bugs or crashes, we recommend that you check the app store for new (free) updates first. If the problem persists, then we appreciate it if you let us know, so that we can develop a solution.

Measurement results

Whenever you run into measurement results that you do not trust, or are obviously incorrect, check your setup by carrying out a back-to-back measurement: connect your STIPA test signal source straight to your iOS device's input, with as little inbetween as possible. Note: be careful to match the signal levels. If you feed the headset signal straigh into the microphone input, then you will certainly overload this input, and possibly damage your device.

If back-to-back measurements give low STI values...

If your connection between source and input is nothing more than cables and attenuators, then you should find an STI of 0.97 or higher. If so, then iSTI is functioning well. If you are still experiencing problems, you should look elsewhere in the transmission chain.

If back-to-back measurements produce STI results that are lower than 0.97, check for the following:

  • Does the signal level correspond to a comfortable listening level? If iSTI reports an A-weighted sound level above 75 dBA or below 45 dBA, then the level-dependent features of the STI model may kick in, resulting in a (justified) reduction of the STI. You can test this by disabling level dependency from the settings menu. Note: do not forget to enable level dependency afterwards, or your STI results will be non-compliant.
  • Is your sample rate ratio calibrated? If you are using an external source for the STIPA test signal, then its sample rate may differ slightly from your measuring device. Calibrate this using the corresponding option from the settings menu.
  • Is your input overloaded? Check for signs of overloading in the measurement details view (swipe to the right in the main measurement view). Look particularly at the higher frequency bands ( 2 - 8 kHz). If you are seeing transmission index (TI) values below 0.90 combined with values of the relative frequency spectrum (Rel) above +3 dB, then your input is probably overloaded. Try doing a back-to-back measurement at a lower signal level.
  • Is your input signal too low? If so, then (electronic) noise may be influencing your measurement. Try increasing the signal level.
Back-to-back values are OK, but real measurements are not...

If you are consistently finding STI-values close to 1 in back-to-back measurements, but are still having trouble getting consistent results in real measurements, then you may be dealing with inherent limitations of the Speech Transmission Index model. Please note that the STI is not designed for dealing with the following conditions:

  • Strongly varying or fluctuating noise. During your 20 second measurement period, the ambient noise should be stable. If the environment features variations in ambient noise, try to include these into your schedule as a separate variable, doing separate measurements for each condition.
  • Impulsive noise. Strong noise pulses, such as slamming doors and electronic switching clicks, will ruin a measurement. If such an event occurs, you should discard the measurement and start a new one.
  • Multiple sharp peaks and troughs in the frequency transfer function. The STI was not designed to deal with channels that show extreme frequency-dependent behavior. This is rarely a problem, since such channels hardly occur, unless designed on purpose - and there is really no legitimate reason to design such a channel, since the effect on sound quality will be obvious.
  • Center clipping. Old-fashioned amplifiers may feature mismatched transistors, which clip the middle part of the speech waveform. The STI does not accurately reflect the effect of this type of signal degradation on speech intelligibility. This condition is exceedingly rare.
  • Narrow-band voice coders. These devices drastically reduce the bandwidth required for transmitting speech, and are applied in radio communications and mobile telephony. The STIPA test signal is not treated by these devices in the same way as natural speech; measured STI values may be unrealistically high or low. The only good news is that they tend to be consistent: you can usually measure differences in intelligibility between test conditions, even if the STI cannot be interpreted using the standardized scale.

STI values are high even without test signal...

You would probably expect to get an STI value of zero when you start a measurement without a test signal. Unfortunately, this is not what happens. In a noise-free setting, you may find a near-random set of STI values. If you inspect the measurement details, you will quickly conclude that the measurement is invalid: multiple m-values above 1, unusual patterns in the frequency transfer function, etc.

The standards for measuring the Speech Transmission Index require that you make sure that your setup is valid. Clearly, the absence of a test signal renders the channel invalid. The STI model is not fool proof in the sense that it detects unstable or invalid transmission chains. This is your responsibility.

Embedded Acoustics is currently field testing its own extensive algorithms for detecting unstable and invalid measurements. These will be included in future releases of iSTI Professional. Existing users will receive a free update. Until then, some simple checks are carried out to flag invalid STI measurements. STI values that are considered less reliable by iSTI are shown in different colors: black (instead of white) in the measurement view, red (instead of blue) in the measurement browser. This is done to assist you in choosing whether to use or discard measurements.

Copyright © 2017 Embedded Acoustics. All Rights Reserved.

Our general conditions are found here (English, Dutch). KvK / Chamber of Commerce The Hague nr 50227092.