Not shown are frequency calibration scales added with the modifications.
|
|
Not shown are frequency calibration scales added with the modifications. |
Introduction
In the March 1995 issue of Amateur Radio,I described an adjustable audio filter system which can be added to the receiver. One feature of the system is a rejection notch which has a 3 dB band rejection width of about 100 Hz. This is fine to reject an interfering carrier or CW signal. However, I pointed out in the text that some means to increase the width of the rejection band width might be a useful addition to reject wider bandwidth interfering signals.
The 100 Hz rejection notch remains unchanged, but I have added a simple modification to enable the high pass (HP) and low pass (LP) filter sections to be set up for a variable width rejection band. This is an alternative to their normal function of setting the limits of the pass band. The modification is simple in that it is achieved by the addition of one switch and one resistor and changing the value of one other resistor.
HP and LP Filters In Parallel
To understand how the
system works, refer to figures 1 and 2. To achieve the bandpass
characteristic of figure 1, the signal is first fed through the HP
filter which sets the low frequency cut-off and then through the LP
filter which sets the high frequency cut-off. Observe that the HP
cut-off is lower than the LP cut off.
 |
To achieve the band reject
characteristic of figure 2, the two filters are connected in
parallel with inputs both connected to the input signal and outputs
combined after filtering. Observe that, for this operation, the HP
cut-off is now set to a higher frequency than the LP cut-off.
 |
The block diagram for the
whole system, as published in the previous article, is repeated in
figure 3 but with the additional switch S4 to enable the filters to
be switched from their normal sequential connection to the parallel
connection. Detail of the circuit changes is given in figure 4.
Resistor R33 and capacitor C18 form the original post anti-alias
filter. Addition of resistor R42 to this circuit enables it to be
also used as a mixer to combine the output of the LP filter via R33
with the output of the HP filter via R42.
 |
The maximum setting of the
cut-off frequency in the HP filter was originally 1.2 kHz. This has
been altered to 2 kHz so that the upper frequency roll over of the
rejection trough could be extended a little higher. Resistor R17 in
the HP Clock circuit has been changed from 10 k to 4.7 k ohms to
achieve this alteration.
 |
Operation
To operate in the discussed band reject mode, switch S1 is set to the HP/LP position, notch switch S2 is left off, and the new switch S4 is set to the band reject position. The LP adjust potentiometer is set for a cutoff at the low frequency side of the rejected band and the HP adjust potentiometer is set for a cut-off at the high frequency side of the rejected band. It is desirable to have a frequency calibration on the adjustment controls otherwise it is difficult to know exactly what one is doing. Pointer knobs are coupled to each of the three filter adjustment controls in the filter box. As originally assembled, no calibration to indicate the setting of the pointers was provided. A back plate of paper is now glued to the box and this is marked to show the approximate frequency cut-off indicated by the pointing of each of the three knobs.
It is not intended that a condition be set up whereby the LP cut-off is set higher than the HP cut-off (as in bandpass operation). This crosses over the two pass bands of the two filters causing what might initially appear to be an all-pass condition. Actually, it is not quite all-pass, as the phase of the output from one filter does not track with the other and, when the outputs are combined, troughs occur in the frequency response curve due to signal cancellation. The problem does not occur when the parallel filters are set up for wide band rejection as there is no frequency at which both filters together provide an appreciable output.
For all other modes of operation, such as wide bandpass, narrow bandpass, and notch (as described in my previous article), switch S4 is returned to the normal position.
Some Observations
It is interesting to observe the effects on speech intelligibility when part of the middle of the speech frequency band is taken out. Inserting the 100 Hz bandwidth notch has no effect and you can't detect that it has been switched in. Taking out a large slice of the band alters the speech quality, as one would expect. However, I have observed that, if the rejected band is between around 500 Hz to 1.5 kHz, quite good intelligibility and tonal balance is retained. Loss of intelligibility and change of tonal balance seems to really occur when frequencies are cut below 500 Hz or above 1.5 kHz. It seems that if speech is troubled by interference concentrated within the frequency range of 500 to 1500 Hz, the interference can be reduced, without loss of intelligibility, by rejecting this part of the band.
Another consideration is a speech signal received in the presence of broadband noise which spreads right across the audio spectrum. The level of noise can be reduced by restricting the audio bandwidth up to a point, but intelligibility is reduced when low frequencies are cut above 200 Hz or high frequencies are cut below 2.5 kHz. As an alternative, one might consider cutting between 500 and 1500 Hz and I have tried this on a number of noisy signals. Whilst the effect is not dramatic, it can give a few dB of signal to noise improvement whilst still retaining reasonable intelligibility and tonal balance.
On a slightly different subject, there are various ways of processing speech into a transmitter to improve the effective speech power. Speech clipping and speech compression are two well used techniques. I now wonder whether a reduction of frequency components in the 500 to 1500 Hz range would also be worthwhile. This would allow an increase in power of frequency components which are more critical in determining intelligibility and tonal balance. This gets away from the subject in hand concerning filtering of received audio but it is an interesting idea leading from the band reject tests on received speech.
Conclusion
A simple modification has been added to the audio filter described in a previous issue of Amateur Radio. Switching in parallel operation of the HP and LP filters allows them to be used in an adjustable band reject mode.
Some tests using the band
reject mode seem to indicate that a band of frequencies in the
range of 500 to 1500 Hz can be taken out of a speech signal whilst
still retaining a reasonable intelligibility and tonal balance.
This is a characteristic of speech which can be useful in improving
intelligibility in the presence of an interfering signal or noise
in this part of the audio spectrum.
Addendum
I refer to my observation that speech frequencies in the 150 to 1500 Hz range can be cut out without any noticeable reduction in speech quality. I now know that in the December 1977 issue of QST, Dr R.W.Harris and J.C.Gorski described a method of narrow band method of voice transmission by not transmitting frequencies in the 600 to 1500 Hz range. By moving the 1500 to 2400 Hz speech frequencies down into this region and restoring them on receiving, a total bandwidth of only 1500 Hz was needed. I reviewed this idea in an article in "Amateur Radio", January 1999 and headed "Narrow Band Voice Transmission".
.