In a QST article in January, 1996, NASA invited amateurs the world over to participate in a test of the Mars Relay Radio System on board the Mars Global Surveyor on it's way to the Red Planet.1 The beacon was to be turned on for checkout purposes approximately 20-30 days after launch, when the satellite was 8 million Km from earth. The 1.3 watt signal and quadrifilar helix antenna would deliver only -177 dBm signal to a receiving antenna with 21 dBi gain, but amateurs were invited to attempt detection by narrowband DSP techniques anyway.
This report details the successful detection of the signals at the amateur radio station belonging to Randy Terrell, K9BCT, in south Florida, USA. Signals from the Mars Relay beacon were detected between 0806 and 1142 UTC on November 24, the first day of the test. DSP software written by the author was used to extract the weak signal for amplitude and frequency analysis. Graphs included in this report show the radiation pattern of the Mars Relay antenna as it swept past the earth once every 100 minutes. Frequency measurements showed a periodic variation at the same 100 minute spin rate.
Station owner: Randy Terrell, K9BCT
12021 NW 27th Ct.
Plantation, Fla. 33323 USA
North Latitude: 26 deg 09m 33.5s
West Longitude: 80 deg 18m 32.7s
Altitude: 10 m
Fig. 1 shows the equipment used for the test. The antenna was an array of eight 18 element yagis, with horizontal linear polarization. Although normally used on 432 MHz for amateur EME (earth-moon-earth) communications, computer modeling predicted an antenna gain of 25.4 dBi at 437.1 MHz. The GaAs Fet preamp was separated from the antennas by 0.8 dB coax loss. For an accurate frequency reference, a Cinox model H109 ovenized 1 MHz quartz oscillator was used. The oscillator was followed by a logic divider circuit to produce a broadband RF combline spectrum with components spaced every 1 KHz. This was injected weakly into the signal path to insure that an audio tone corresponding to a stable and accurately known RF frequency would always be in the passband along with the desired signal from the Mars Relay transmitter. The HP audio oscillator was used as a receiver tuning aid, to assist the operator to set the receiver to the desired frequency by ear.
Two computers were used simultaneously. One was used for signal capture, the other for signal processing. The audio output of the single-sideband receiver was fed to a Media Vision Pro AudioSpectrum 16, a 16 bit A/D converter card mounted inside the Macintosh Quadra 700 computer. Macromedia's SoundEdit 16 software was used to record one minute segments of audio on every even minute during the hour. These were saved on Iomega zip100 disks and transferred to a 66 MHz PowerPC Macintosh for signal processing.
Fig. 2 shows the DSP algorithms used. Four separate DSP routines were written to process the data files. The software was written and compiled in Zedcor FutureBasic for the Macintosh. The MARSFILT routine was first used to lowpass filter then downsample the data rate to 2780 samples/sec. The MARSCAL program was then used to determine the frequency of the audio tone created by the injected RF frequency standard. The program measured the phases of ten 0.1 second sequential Discrete Fourier Series to derive the frequency of the audio tone. This process was repeated at 5 second intervals during the 60 seconds of data, and a curve fit to the results was used to determine the average frequency of the tone and the drift rate of the receiver.
The MARSMIX and MARSFFT programs implement a band-selectable FFT system using a heterodyne technique described by Fred Taylor2. The MARSMIX program was used to derive quadrature signals for the complex FFT, as well as compensate for receiver drift and doppler drift of the Mars Relay signal. The routine then performed a filtering and downsampling to limit the output to baseband data and keep the subsequent FFT processing time to a minimum. The sine and cosine signals used for mixing were set to 600 Hz, with the chirp rate "c" manually chosen. It was calculated based upon measured receiver drift from the MARSCAL program plus the predicted doppler shift of the Mars Relay signal. Typical combined drift rates were between -.01 to -.04 Hz/sec. Although doppler rates were quite predictable, receiver drift was nonlinear. It might be flat for 30 seconds, then rise 0.5 Hz in the next 30 seconds. The entire MARSMIX program was therefore repeated at +/- .01 Hz/sec either side of the expected value of chirp, to search for the best fit for the final results. Further, since a signal halfway in frequency between FFT bins is down approximately 4 dB in amplitude, it was decided to repeat the entire MARSMIX program four times, stepping the mixing frequency at quarter bin spacings (600.00, 600.23, 600.45, and 600.68 Hz). Along with the additional values of chirp used, this kept the worst case coherence loss to around 0.3 dB and improved the amplitude accuracy of the measurement. Unfortunately, it also resulted in the MARSMIX program being executed 3X4=12 times, a total of 40 minutes processing time for each minute's data.
The MARSFFT program utilized a 1024 point complex FFT routine with equivalent bandwidth of approximately 0.9 Hz. The program averaged 55 sequential, non-overlapping FFT's during each minute's data. It then printed out the five strongest amplitude bins found for each of the 12 mixing variations. The strongest of all these was selected as the data point in the amplitude vs time graphs, and the frequency of the bin used in the frequency vs time graph. The amplitude chosen in this manner is believed to represent the best estimate of actual signal amplitude, by insuring the signal was most nearly centered in one of the FFT bins and most nearly matching the correct chirp. To allow a focus on the Mars Relay signal, the frequency range of the MARSFFT "high 5" printout routine could be restricted by the operator, with manual centering and windowing.
Fig. 3 shows the resultant amplitudes selected per the procedure described in the paragraph above. The Y-axis plots the amplitude of the FFT normalized to the average noise level in empty nearby FFT bins. Although it is labeled S/N, it is actually (S+N)/N. The noise floor was found to vary less than .25 dB over the duration of the test. Recording was started at 0734 UTC on Nov. 24, but the rotation of the spacecraft did not bring the main lobe of the antenna towards the earth until around 0800 UTC. For times earlier than about 0806 UTC, the received signal from the Mars Relay was too close in frequency to the reference oscillator tone at 437.096000 MHz to distinguish. This problem was not experienced again at any other time during the remainder of the short observing period, as the changing doppler shift carried the Mars Relay signal lower in frequency. Two data points were lost due to recording errors, at 0814 and 0842 UTC. Signal levels below about 2.0 dB are difficult to distinguish from noise peaks, and are not reliable.
The signal was not detected again until about 0930 UTC. The main lobe of this second "sweep" of the antenna past earth was captured in its entirety, as shown in Fig. 4. General similarity to the first plot can be seen, although there is variation in the detail. Peak signal amplitude was the same, within a fraction of a dB. Similarities to the data published by the Stanford team can be seen.3 The pattern is not smooth across the main lobe, but experiences peaks and dips of several dB, the worst being a 3 dB dip at 1008 UTC.
Fig. 5 shows a portion of the third sweep of the antenna past earth. It shows a gradual rise in amplitude over about 20 minutes, similar to the previous plot. Due to physical limitations on the receiving antenna array, it was not possible to track the satellite beyond 1143 UTC. There is a significant peak at 1124 UTC, which shows up to a much lesser extent in the two previous graphs. This too, can be seen in the data published by the Stanford team.3
|Transmit Antenna Gain||+1.7||dBic|
|Tx Pointing Loss (30 deg)||-0.3||dB|
|Path Loss (4.7E6 Km)||-218.7||dB|
|Receive Antenna Gain||+25.4||dBi linear|
|Rx Pointing Loss||-0.2||dB|
|Coax Loss ahead of Preamp||0.8||dB|
|Coax + Preamp + Receiver||120||deg, K|
|Sky Temperature||17||deg, K|
|Sidelobes + local noise||20||deg, K|
|Total Receive System||157||deg, K|
|Effective Receiver BW||0.9||Hz|
|Noise Level (kTB@157K)||-177.1||dBm|
|DSP Coherence Loss||-0.3||dB|
|Expected S/N Ratio||12.8||dB|
Table 1 shows the system calculations for the expected signal level, which should have been close to 13 dB S/N. As later reported by NASA and the Stanford Team, the Mars Relay beacon was actually in Mode 9 from 0730 to 1630 UTC on Nov. 24. This mode produced a cycle of 2 seconds of pure carrier followed by 14 seconds of RC1 modulation, then 2 seconds of pure carrier followed by 14 seconds of RC2 modulation. With this modulation, the carrier level as averaged over one minute dropped over 10 dB in amplitude.
Fig. 6 shows the 0816 UTC minute in detail, without averaging. For this plot, the 1024 point (1.1 second) FFT's were overlapped by 512 bits, to provide better detail on the rise and fall of the carrier. The 2 second carrier bursts can be seen peaking as high as 15.6 dB above the noise floor. This plot is unusually strong; most others examined were a few dB lower, closer to the predicted S/N level.
Fig. 7 shows the measured frequency of the Mars Relay beacon, corrected for geocentric doppler shift and the doppler shift of the receiving site. A program was written to take the Nov. 17 post-launch position and velocity estimates from NASA and compute the expected total doppler shift for the receiving site in Plantation, Florida. Techniques and coordinate transformations follow those outlined in Fundamentals of Astrodynamics by Bate, Mueller, and White4. The graph shows the beacon frequency rising 20 Hz or more as the satellite slowly rotated the antenna towards earth. Maximum rate of change was about +.02 Hz/sec. If change in the internal temperature of the spacecraft is assumed to be the cause, it would indicate that there was a temperature drop of approximately 5 centigrade degrees.5 From past experience, the frequency of the ovenized reference oscillator in the receiving station is normally accurate to within 5 Hz. Because temperature at the receiving site was not well controlled the morning of Nov. 24, it may have experienced a slightly greater uncertainty.
Signals from the Mars Relay beacon were detected between 0806 and 1142 UTC on November 24. DSP software written by the author used a heterodyne technique to provide a coarse cancellation of receiver drift and the changing spacecraft doppler shift. This allowed averaging a series of 1.1 second FFT's over 60 seconds of time to reduce the noise fluctuations to less than 2 dB. Unexpected modulation on the beacon reduced the average power level in the carrier by over 10 dB. Therefore, the resulting signal strength peaked only 5-6 dB above the noise, but was still usable to plot the shape of the radiation pattern of the Mars Relay antenna as it swept past the earth once every 100 minutes. Antenna patterns were not smooth, but showed peaks and dips of 2-3 dB in the main lobe. Frequency measurements made on the received signals showed a periodic variation at the 100 minute spin rate. The frequency rose approximately 20 Hz during the first half of each rotation cycle, perhaps a result of an ambient temperature change of about 5 degrees in the transmit oscillator.
I would like to thank Randy Terrell, K9BCT, for his assistance and the use of his moonbounce station which made this effort successful. I would also like to thank John Callas of JPL for coordinating this exciting test with the Mars Relay and providing necessary data about the transmitter characteristics and spacecraft location.
-Tom Kneisel Feb. 20, 1997
(1) DXing enroute to the Red Planet", by Michael R. Owen,W9IP, and John
L. Callas, QST Jan. 1996, pp. 44-45.
(2) Fred J. Taylor, Digital Filter Design Handbook, Marcel Dekker, Inc., 1983, p. 63.
(3) "MR CW Power (Preliminary)" graph for 1600 UTC, Nov. 25, by Twicken, Linscott, Cousins, 2 Dec. 1996. Internet URL=http://nova.stanford.edu/projects/relay/mr-cw.html
(4) Fundamentals of Astrodynamics by Roger R. Bate, Donald D. Mueller, and Jerry E. White, Dover Publications, Inc., 1971.
(5) Data provided by email from John Callas of JPL through the email reflector on June 24, 1996. It shows approximately 6.3 Hz/deg frequency variation with temperature on a transmitter identical to the Mars Relay transmitter.