|
Figure 1 shows a 30 second recording made by Mauno Ritola
12.29 on 5th May, 2009.
The recording lasts 30 seconds and during its first half there
is strong interference from some kind of industrial switchpower.
The recording is repeated twice in figure 1.
|
|
|
|
Fig. 1. Switchpower QRM. No blanker.
Download this file fq-xformer-stop.zip (128239900 bytes) and unpack the .wav file (which is about twice as large.) The station on 1188 kHz is a suitable candidate for comparison of SDR software. It is a 10 kW relay station for Deutsche Welle located in St Petersburg, 280 km south of Maunos location. With a conventional AM detector without blanker it should sound like this: am-nobln.mp3 During the first 15 seconds when the interference is on it is impossible to hear what is being said, one can just have a feeling there is a voice hidden in the noise. Listen to 1188 kHz with perseus.exe, Winrad, Linrad and other software you may have and compare how well you can copy this station when the interference is on. One can improve AM reception by use of SAM (Synchronous AM detection.) In Linrad this corresponds to the processing mode Coh3. The synchronous detection splits the detected singnal in two parts. One that is in phase with the carrier and one that is 90 degrees out of phase. The out of phase signal contains noise only and is just disregarded. That means that 50% of the noise energy is thrown away with a 3 dB improvement in S/N as a result. In Linrads Coh3 mode the signal sounds like this: coh-nobln.mp3 It is obvious that the difference is very small. The synchronous detection does not improve enough to make it possible to understand what is being said while the interference is on. The lower left quarter of figure 1 shows oscilloscope tracings of the entire wideband signal. The upper white trace is power vs time. The next two traces, yellow and magenta show I and Q for a signal that does not contain the spectral components that are red in the main spectrum. The bottom traces show I and Q at twice their relative amplitudes for a signal formed by only those points in the main spectrum that are coloured red. It is obvious from the oscilloscope tracing that the pulses are quite long. That is the consequence in the time domain of the rather narrow frequency spectrum of these pulses. The strongest maximum in the spectrum at about 1100 kHz is only about 200 kHz wide. That is only 10% of the total bandwidth and therefore we should expect the pulse to last 10 times longer than normal in the time domain. That is indeed what they do. Linrad has two blankers. The smart blanker looks for pulses that have the same shape as the reference pulses with which Linrad was calibrated. Pulses with any similarity to that are not present in this recording and the smart blanker does not do anything at all. The dumb blanker simply gates out the signal at times when the power (the white oscilloscope trace) goes above a threshold value. Figure 2 shows the waterfall diagram with the dumb blanker enabled. |
|
|
Fig. 2. Switchpower QRM. Blanker enabled.
The effect of the blanker is obvious in the oscilloscope tracings. Totally 20% of the data points are removed which means that about 1 dB of the signal is lost. The interference pulses largely repeat at 6 kHz so the spectrum consists of spectral lines separated by 6 kHz. At the peaks the level is 45 dB in the high resolution graph with the blanker off, but with the blanker on, the level is about 27 dB. Between the peaks, the noise floor is 23 dB without the blanker and 22 dB with the blanker. On the average (in linear power scale) the noise goes from about 40 dB to about 23 dB so S/N should improve by about 17 dB. In reality it is not quite as good because the blanker also adds some noise because it gates other signals on-off, those signals that are not red in the main spectrum. Nevertheless, the dumb blanker gives a significant advantage. Listen to this file which is the loudspeaker output from the conventional AM detector: am-bln.mp3 Synchronous detection does not make a significant difference. With the blanker running one can easily hear what is being said while the interference is on. The blanker makes a qualitative difference. It is far from perfect however and one can clearly hear when the interference stops. The blanker introduces hum that sounds like 100 Hz. Presumably the amount of clipping waries at this rate. The hum that the blanker introduces can be seen in figure 3. The best way to get rid of this hum would be to apply a notch in the loudspeaker output. There is no such facility in Linrad, but one can use the notches in the baseband and remove the hum sidebands one by one as can be seen in the figure. The loudspeaker output then becomes like this: bln-notch.mp3 |
|
|
Fig. 3.The hum sidebands that the blanker introduces are
at 120 Hz and 240 Hz.
They can be removed with notch filters in the baseband as can be seen here
in the yellow curve that shows the baseband filter.
This particular pulse interference with pulses that repeat at a high and nearly fixed rate can only be removed by blankers that operate at a considerably larger bandwitdth than the pulse repetition frequency. Conventional transceivers can not resolve these pulses, one will need a bandwidth in the order of 100 kHz. On the other hand, because of the frequency response of the pulses themselves, a bandwidth much above 200 kHz will not improve much. The phenomenon of "keying clicks" created by the blanking process is very easy to see with this interference. Since it has a dominating 6 kHz pulse repetition frequency, the keying clicks form sidebands symmetrically at 6 kHz around every signal present in the gated signal. This is very obvious in figure 4 where the station at 1188 kHz is de-selected during the second half of the second repetition of the file. |
|
|
Fig. 4. The sidebands created by the blanking process disappear from
the 1188 carrier when the station is de-selected and therefore coloured red.
From figure 4 we can see that the pulse repetition frequency is much more 12 kHz than 6 kHz. Maybe it is mainly 6 kHz, but two signals that are phase shifted 180 degrees every time the mains power reverses sign? Note that the "keying clicks" extend over the entire bandwidth. If not all strong signals were removed first as indicated by red points in the main spectrum, the keying clicks from all the signals would have degraded blanker performance and if some of them were strong, the blanker would have became useless. The smart blanker does not have any keying clicks. In principle it would be possible to make the smart blanker adaptive so it could recognise pulse shapes that occur often and subtract the correct one out of several different pulse shapes. Such an improved blanker should completely remove the interference presented on this page so one would not hear when the interference stops. It would not be trivial however and nothing I give high priority. SM 5 BSZ Home page Linrad home page |