4 Maintenance Manual
1.2.1.2 Aliasing
A further requirement is that the bandwidth of the measured signal be controlled
so that its frequency does not exceed a value of 1/2 the sampling rate. This fre-
quency amplitude control is done with an aliasing filter. A conservative design
stops the analysis well before the frequency limit is reached.
Aliasing shows up as a generation of random or non-random dot patterns in the
sampled data points, and occurs when the signal can take a number of excur-
sions between samples. This aliasing in an FFT will produce a number of com-
ponents in the spectrum that are really not there. It is also desirable to eliminate
the high frequency components from the signal generator portion of the system.
A smoothing filter is used there also.
1.2.1.3 Noise Reduction
The effects of ambient or hearing aid noise can be reduced by the use of time
domain signal averaging because the sampling process is exactly synchronized
with the signal generator.
Noise reduction averaging is done in steps of 2, 4, 8 and 16. The process is done
by use of an averaging buffer. The data is added to the data already in the buffer
used to create the last spectral display in a ratio of 1/2, 1/4, etc., depending on
the averaging called for. The result of the addition is then divided to get a prop-
erly scaled number. The effect of the non-synchronous noise is thus reduced
because of the averaging process.
Averaging does not slow down the display process, but does slow down the ef-
fect that an acoustical change will cause on the displayed waveform. Changes in
the phase of signals will also show up if averaging is being used. The change of
phase is accomplished by movement of the hearing aid in a sound field while the
measurement is in process. The effect of phase changes is a dropout of the signal
and an eventual recovery to the correct level when the motion has ceased.
Modern hearing aids now employ "interesting" signal processing techniques that
apparently destroy the phase relationship between input and output signals as
a means of preventing feedback. Time domain averaging does not work with
this class of device. Further, this class of aid will reduce the gain when a steady
state test signal is applied. To enable a reasonable analysis of this modern class
of hearing aid, a newer form of composite signal has been developed. It is called
"digital speech" and consists of a series of composite signal bursts to imitate
speech. The analysis of the hearing aid output has been modified as well, by do-
ing averaging in the frequency domain rather than time. Curve display smooth-
ing can also be applied to remove rough spectra that sometimes result with some
of the newer aids.
To Next Page
Loading...