Post-processing 3: How to detect if a measurement wrong?

If you have mesured using averaging (repeating the measurement signal several times)

Go to Setup->Load... and select the setup named

Speech_All Parameters (STI RASTI %ALcons etc.)

Download the measurement from

www.winmls.com/2004/measurements/WrongResult.wmb

Open the measurement using File->Open….

The right plot in WinMLS will look something as shown below.

If you zoom in on the impulse response in the Energy-Time Curve plot you will se that it consists of 14 peaks and not one peak as one should expect.

In addition, the start was not correctly detected so two of the peaks are at the last part of the impulse response at the very left of the figure shown above.

In a total it is therefore 16 peaks. The measurement was performed using 16 averages, this is suspicious.

If you go to Measurement->Listen, you will hear that the impulse response sounds unnatural, like a machine gun with 16 shots, which also indicates that something is wrong.

One reason for displaying the Energy-Time-Curve in this setup is that the experienced user can clearly see from the plot that something is wrong in such cases as this.

This measurement was performed using an external source. In this case there might be problems with that the sampling frequency is not exactly the same on the playback and recording device, but in this case this does not seem to be the problem. This is because the peaks are separated so many milliseconds apart and the distance between the peeks is not constant.

The error is most likely that during the measurements there was gaps caused by the operative system. To improve this, see this website:

http://www.musicxp.net/tuning_tips.php

WinMLS has no chance to detect this in the current version when doing external measurements. In Measurement->System Delay..., you can in the upper list box select Yes, with loop-back and end-check and it will be detected, but then the impulse response will not be positioned at the beginning but randomly.

In a future version of WinMLS we are planning to add a "remove system delay" check box. Then you would get an error message if such a gap occurs and it would be 100 % proof even for remote measurements.