An AM signal disturbed by several spurs.
(Dec 17 2010)

The recording

The wideband recording 4.wav is a part of a much longer recording made by Michele IZ2EAS with perseus.exe. The signal of interest is on 3345 kHz (Channel Africa in Portughese). There are several carriers that disturb reception.

The file can be downloaded here (79612748 bytes) It unpacks to 4.wav (153262166 bytes.) This file is produced from a much longer Perseus file by pressing 'S' in Linrad to save a selected time slice. That produces a file in Linrads proprieritary format .raw which is subsequently converted to a 24 bit .wav file by use of the file converter .raw to .wav which is available in the main menu of Linrad. For the wav file to be accepted by Perseus a Linrad version of 3.19 or later is needed. Older versions do not write the .wav header as required by Perseus.


Figure 1 shows the 4.wav file played in perseus.exe. The waterfall speed is adjusted for the repeating playback to be visible about 1.5 times.

Fig 1. Perseus.exe playing the 4.wav file.

The best possible result with perseus.exe is obtained with xxxxxxxxxxxxxxxxxxxxxxx settings and sounds like this: LINK TO MP3 FILE.

The conventional AM detector.

When listening in Linrad with a conventional AM detector and without any notches, the loudspeaker output sounds like this: am.mp3 (358008 bytes) It is not so easy to hear what is being said even for someone understanding Portugese.

The conventional AM detector is available in all SDRs as well as in old analog receivers. The loudspeaker output will not differ much between different receivers even though some operators might prefer a narrower bandwidth.

Figure 2 shows Linrad in the equivalent mode of what is shown in figure 1 for perseus.exe. The main waterfalls are fairly similar although Linread is set to show it at a higher resolution. The baseband windows differ quite a bit in that Linrad is quite a bit more information rich. That does not affect the conventional AM detector in any way.

Fig 2. Linrad playing the 4.wav file in simple AM mode.

Baseband notch filters.

Linrad allows the user to set up to 9 notch filters in the baseband. Figure 3 shows the linrad screen when running the 4.wav file in Coh3 mode with the notch filters set as follows:

No     Offset      Width
1        0           3
2      -29           3
3     -157           3
4      227           3
5     -285           3
6     -410           3
7     -540           3
8     -794           3
9       99           3

Fig 3. Linrad playing the 4.wav file in Coh3 mode with 9 notches.

The carrier has deep QSB so it is better to replace it with a constant level signal, i.e. to run linrad in a coherent mode and to apply a filter that removes the carrier which would otherwise affect the AGC. That is why one of the notches is placed at offset zero.

The audio output cohn.mp3 (358008 bytes) is quite a bit better than the audio output from a conventional AM detector without notches.

By use of stereo head phones one might hear a little better in Coh2 mode. Do not listen to this file with loudspeakers: coh2.mp3 (358008 bytes> The notches are the same.

Wideband notches in Linrad.

The baseband notches in Linrad are not automated. It would be possible and not very difficult but there has not been any user interest in this feature...

There is however another mechanism for spur removal in Linrad. It operates in the frequency domain and it is highly efficient and can remove hundreds of spurs, probably several thousands although I have never encountered a data file with many enough spurs to know where the limitations might be.

The Linrad spur removal may be set to remove all spurs. That means all sinewaves that are stable in frequency as well as in amplitude. In AM mode all carriers will disappear and the user would have to listen in USB or LSB mode for easy processing. It will also be possible to use Coh modes when the frequency is set precisely to the frequency of the no longer present carrier.

Another option is to set manual spur removal, (Enable AFC/SPUR/DECODE = 1 in the parameter setup.) Then the user can press 'E' (eliminate) with the mouse placed on the spur that should be removed. This is a quick and easy procedure. The screen would lok something like figure 4 and the loudspeaker output would be something like this with the conventional AM detector: am-manspur.mp3 (358008 bytes) It might be better to use a Coh mode to avoid the carrier amplitude from influencing the loudspeaker sound level. In Coh3 mode with the wideband notches set as in figure 4 the loudspeaker output sounds like this: coh3-manspur.mp3 (358008 bytes) For operators using stereo head phones Coh2 mode is probably better on stations suffering from selective fading that distorts the phase relation between the AM sidebands.

Fig 4. Linrad playing the 4.wav file in Coh3 mode with wideband notches. The small white marks on the main spectrum frequency scale show where the notches are placed. There are 16 notches but he carrier is not removed.


Spurs seem to be a fairly unusual problem for DXers. Linrad has tools to fight them, but those tools are not particularly user friendly. It would be possible to put the algorithms that Linrad uses for auto spur removal in the wideband spectrum into the baseband spectrum and thereby efficiently remove any spur that might be present. The lack of interest indicates that the spur problem is not very important and in light of that conclusion it seems reasonable to not spend more time on it.

To SM 5 BSZ Main Page