SM 5 BSZ - Linux dsp radio, demo - 2 minutes of 144MHz EME at 2x20kHz bandwidth.
(April 22 2003)
The archive file eme.tbz 10 273 698 bytes is compressed with bzip2. Make a directory /home/demo Copy the eme.tbz file to your /home/demo directory, then uncompress it with bunzip2 to get eme.tar:
bunzip2 eme.tbz
Use tar to create the eme directory and its contents:
tar xvf eme.tar

The /home/demo/eme directory will contain eme_2min.raw. The directory will also contain obsolete parameter files that you may remove.

To run the recorded data through Linrad you need a file named adfile in your linrad directory. This file must contain the full name of the recorded file including the path.

The text file adfile may also contain a second filename on the same line as the one specifying the recording. If a second filename is given, Linrad will use it for save and restore of processing parameters that are used only when processing recorded files with this filename given as the second name.

The archive file emepar.tbz 2575 bytes is compressed with bzip2. Treat it like the eme.tbz file, you will get adfile and four sets of processing parameters in the /home/demo/eme directory. Copy the four lines of adfile into your existing adfile if you already have other recordings installed. Otherwise just copy adfile into your existing Linrad directory. Start Linrad and select "J=Select file from 'adfile'" from the main menu. You should get a list like in fig.1


Fig 1. Press J on the main menu to get four choices to run the /home/demo/eme/eme_2min.raw recording. The four different choices give four different parameter sets as described below.

The first line

Select 0 for the first line to read parameters from /home/demo/eme/par1_eme_2min.raw........ This is a setup that will run on a 640 x 480 screen or larger. The first fft bandwidth is 40 Hz and this setup is about minimum for the eme_2min.raw recording. Fig. 2 shows the screen while running the first parameter selection. For unknown reasons this parameter selection does not work with Linrad00-57 but it does run with Linrad01-00.

Fig 2. The 2min_eme.raw file played with the parameters supplied in /home/demo/eme/par1_eme_2min.raw......... The screen is 640 x 480.
The first set of parameters run fine on my 200MHz Pentium MMX even though the MMX instructions are not used. To run the eme_2min.raw recording on a Pentium 60 MHz computer it is necessary to change the "First FFT window power of sin" from 3 to 0. This means running without any window at all. With the modest dynamic range required this is perfectly ok. There are no strong signals in the passband during the entire recording. It is also necessary to change the "First mixer bandwidth reduction in powers of 2" from 5 to 7. This means one can not make the baseband filter bandwidth larger than about 100 Hz which is perfectly ok for EME signals. It is also possible to leave the mixer bandwidth reduction at 5 but then one has to select a bandwidth of about 100Hz or more to keep fft3 small enough for the limited processing power of a Pentium at 60 MHz.

The second line

Select 1 for the second line of fig.1 to read parameters from /home/demo/eme/par2_eme_2min.raw........ This is a setup with low processor demands but it has the second fft enabled. The MMX routines are not selected so this setup may run on the very fastest old Pentium computers that does not have multimedia instructions.

Fig. 3 shows the screen when running the second example. The mouse is clicked on K2GAL at 7.99 kHz like on all the screen dumps on this page. The waterfall resolution is similar in fig.2 and fig.3. Note that the interference pulses, horizontal lines in fig. 2 are absent in fig.3 because the noise blanker is running.


Fig 3. The 2min_eme.raw file played with the parameters supplied in /home/demo/eme/par2_eme_2min.raw......... The screen is 640 x 480.

The third line

Select 2 for the third line of fig.1 to read parameters from /home/demo/eme/par3_eme_2min.raw........ This is a "normal" setup for eme using MMX routines. A pentium MMX 200 MHz is fast enough

Fig.4 shows the screen when running the third setup example. The resolution for the second fft is 16 times narrower than the first fft which is about 50 Hz. Note that the keying sidebands are well visible in the high resolution graph (red dB scale) which shows the second fft with one pixel per bin. The waterfall is also produced from the second fft but the number of pixels on the screen is not large enough for a 32768 FFT so each pixel in the waterfall is based on the information from 32 frequencies.


Fig 4. The 2min_eme.raw file played with the parameters supplied in /home/demo/eme/par3_eme_2min.raw......... The screen is 1024 x 768.

The fourth line

Select 3 for the fourth line of fig.1 to read parameters from /home/demo/eme/par4_eme_2min.raw........ This setup is very similar to the previous one. The second FFT size is 16384.

Note that the spurs at about 8.8, 9.3 and 18.6 kHz are greatly reduced. The 9.3 kHz spur is stable enough in both frequency and amplitude to disappear completely in the very narrow notch centered on it by the spur cancellation process. The spur at about 1.3 kHz is actually a very large number of carriers which is the same as to say it has a complicated modulation on it. To fight this kind of interference, more sophisticated algorithms than the current spur removal of Linrad are needed.


Fig 5. The 2min_eme.raw file played with the parameters supplied in /home/demo/eme/par4_eme_2min.raw......... The screen is 1024 x 768.