Today W9VMT and I completed the test of the fftdsp41 program. It was very impressive.
I made a new hook up cable to go from the headphone jack on the speaker to the line input jack on the computer sound board. It had a stereo plug instead of using a one channel plug inserted so as to hit both contacts which worked far easier. I also used an additional 50 ohm variable attenuator in the line between the kenwood 830 at 28 mcs and the Measurments 900 signal generator.
The generator had about 25 cycles per minute drift. That actually was an advantage as the calculated drift due to doppler shift at 4ghz of a source entering the beamwith of a dish as 1.5 degrees from due south over the 4 minutes that the source is observed would be about 25 cycles per min. When the angle between the line between the observer and the center of the earth and the line between the source and the center of the earth is 1.5 degrees, and the observer is at 40 deg lat and the source is at 40 deg denclination, the doppler effect will cause a 95 cycle shift as the frequency observed will be the actual frequency of the source at the overhead position and the source frequency plus 95 cycles at 1.5 deg. The source takes about 4 minutes to transit the 1.5 deg so the shift averages 25 cycles per second.
The test consisted of operating the receiver in ssb position and zooming the FFTDSP41 program out so the entire 2.7 kcs could be observed.
The signal generator was set so we could just hear the signal in the stereo speakers from the sound card. That was also the point we could just hear the signal in the rig's speaker. The 20 db attenuation on the rig was in and additional attenuation was put in from the attuenator in the line.
The signal generator setting was noted and we then adjusted the output of the signal generator to the lowest point that a line was produced on the screen.
With integration equal to 1, we could take out the 8 db of the attenuation that was switched in at the time the line was produced on the screen and we could not hear the signal in the speakers with the 8 db out. We also used the readings on the attenuator on the signal generator to verify the results.
The fftdsp41 produced an amazing 10 db improvement in signal detection over our ears listening!!! And that was at an integration factor of 1.
The highest integration factor we could use to improve the detection was 4 as higher integration factors lost everything due to the drift. That resulted only in a 2 to 4 db improvement over an integration factor of 1.
Given the possiblitiy of doppler shift caused by rotation of the planet of the source adding to the doppler shift of the earth, the minimum integration factor ( .4 seconds ) will probably be used for a seti operation, unless that source is at a high denclination near polaris.
However, substantial improvement will be realized from fftdsp41, even at an integration factor of 1 (.4 sec. )
The only bug noted was that the small number that shows the integration factor below the click point in the screen would change from time to time when the integration factor was in fact not changed. The integration factor would be say 4 and it would run at 4, because we could time the additions to the screen at 2 seconds, but the number displayed before the click point would change, for no reason, to the number to which it had previously been set. The integration factor did not change, just the number on the screen reporting the integration factor.
That is not a serious bug, but I am reporting it as a bug because the program is in beta test.
Thursday, 21 March 1996 Update:
This evening Steve Henke, KB9KWD, and I tested both the setifox program and the fftdsp42D program. Version 42D is the latest version of FFTDSP from Mike's web page and the Setifox program is the latest version from Dan Fox.
Steve's R7000 was used with the attenuator in ( 20 db or so attenuation in the receiver ). Steve's HP8920A station monitor was used as the signal generator. It is very stable with calibrated output.
The R7000 was in USB position with the frequency at about 156 MCS.
The Setifox program could identify a carrier at -129 DBM on the HP output indicator. The carrier could be heard in the speaker at -121 DBM, an improvement of 8 DB for the program. We used the maximum value for "samples" in the CFG file which was 16384.
The FFTDSP program, version 42D, with integration set to 2 ( one second integration time ) identified a carrier by a easily read line on the screen at -135 DBM on the HP output. THe carrier again could be heard at -121 DBM on the HP output. THis was 14 db better than the noise.
I do not know if the FFTDSP version 42D is a version more adept at pulling signals out of the noise or if the results are better than earlier tests of version 41 of FFTDSP beacuse of a better output calibration on a signal generator.
We could not use SETIFOX to control the R 7000 as there was a problem with the homebrew CI-V control box.
The FFTDSP program has one other advantage over the SETIFOX program. The "waterfall" screen displays frequency shift of the signal and the setifox display does not. Display of the frequency shift is important if you want to decide whether a signal is from earth or not as the rotation of the earth causes a doppler shift and the frequency shifts if the signal is from near a star.
The SETIFOX program has an advantage as it has provision to control the receiver and record the frequency. IT also has automatic carrier detect with provision to do something about it.
I suspect that revsion 42D of FFTDSP is capable of about the reasonable limit of looking under the noise, given the constrants of doppler shift.
Saturday, 23 March 1996 Update:
I just realized why the setifox program did not go very far into the noise. We had a signal showing up at about 500 cycles no matter where the receiver was tuned, and with no antenna. That was apparently from some birdie in an if that remained the same no matter the receive frequency or band. It was there, on the setifox program, but not audible in the speaker no matter what we did.
The setifox program started "finding" the 500 cycle signal as soon as the signal from the signal generator dropped below the 500 cycle spurious signal. One problem with the setifox program is that the program finds one carrier and if the receiver has a very low signal in an if, it will find that until another signal is stronger.
We also figured out the Icom 7000 control problem mentioned above. Each command to a CI-V radio from the controller includes the address of the radio. See CI-V reference manual ("Manual") page 5-1 paragraph 5-1, basic message format. The address of the IC-R7000 is 08H and the address of the IC R7100 is 34H. see table 2-2 page 2-2 of the Manual. So each message to the 7000 differs from the same message to the 7100 as there is a different address byte.
entire website copyright © The SETI League, Inc.
this page last updated 23 November 2002
Top of Page