The Mabuhay! SETI System


Welcome to our Search for Extraterrestrial Intelligence!


We have welcomed visitors since Jan. 31, 1999.

This page was updated on August 21, 2000 by Skip, Sarah, and Samantha Crilly.

Don't be alarmed. 
We are not secretly coming to the Planet Earth.

Samantha drew this project icon [Samantha is Skip's daughter]


Learn more about the Mabuhay! SETI system:

Update
Hypothesis and Speculations
System Description and Status
Mabuhay System Block Diagram
How the Radio Interferometer Works
Radio Astronomy using the Mabuhay! System
SETI Site Images
Mabuhay! SETI System FAQs
Acknowledgements
Seti is really cool
Seti is really cool, continued

Update

August 21, 2000
I installed two jack screws to remotely point the Liberty Lake 3 meter dish. The approximate Right Ascension axis has 20 degrees of coverage. The approximate Declination axis has about 10 degrees of coverage. More range is possible with mechanical adjustment to center the jack screws.

July 24, 2000
I've been reading and thinking about scintillation (variation of the intensity) of hypothetical SETI signals. The waves that might reach us will take many paths as they travel through our galaxy. These multipath effects are called reflection, diffraction, refraction and scattering. How big does a blocking object need to be to affect the signals we might receive? This blocking object's size is calculated by a Fresnel Zone formula. Waves that take an extra wavelength longer path to reach us, by "bending" (technically diffracting) around an obstructing object, just miss what is called the first Fresnel Zone radius. Let's say the waves have a wavelength of five centimeters. If a blocking object is one hundred light years distant, an object about one light second in diameter will have a substantial effect on the signal strength. Objects less than about 0.1 light second in diameter will have little effect at that distance. As the obstruction gets closer to Earth, it must be smaller to not block the signal to the antenna. Scattering objects in the galaxy are much larger than a light second (a light second is about seventy-five percent of the distance to the Moon). So we expect that we will notice effects due to blockage and scattering. Almost everything in the galaxy is moving around, so signals will vary in level over time. Astronomers notice these variations, particularly when they observe pulsars. How fast are the typical changes in signal level of a SETI-type signal? Estimates are on the order of a few seconds (in rare cases) to many hours, depending on how far away the source is, how close to the galactic equator the source is, and the wavelength of the hypothetical signal. There are very good analyses of scintillation effects at:
SETI Research at Cornell
Turbulence in the Interstellar Medium
Now You See Them, Now You Don't.

Imagine that we want to take advantage of the scintillation. Perhaps we can separate antennas to allow one antenna to be in a peak while others are in nulls, or at average levels. How far apart do two receive antennas have to be so that one antenna gets a strong signal while the other is in a null? At an instant in time, the received signal on each antenna is the result of all the waves reaching the antenna over all of the scattered paths in the antenna's "view". The source, scattering objects, and our antennas are all moving in space. We are travelling around the Sun at a speed of about 29.8 kilometers per second. The rotation of the Earth adds or subtracts up to about 0.464 kilometers per second to this. If we assume that a change in signal level happens only due to our motion in space through the scattered signal pattern, we may expect that we need to move about 1788 kilometers away (29.8 km/s times 60 seconds) to "counteract" the effect of one minute of null-to-peak scintillation time. In fact, the null and peak patterns in space are moving across the Earth's path, so there's much more to the effect. Closer antenna spacing may help us by moving a strong signal sooner at an antenna, or we may be unlucky if the two antennas lie in pattern nulls that move through space with us.

Are the null-to-peak spacings large compared to the Earth? One way to get a very large-sized null in a signal is to have a large blocking object in the signal path. If the object is moving faster than we are, then rapid (less than one minute) and large scale (compared to the Earth) scintillation may be explained in this way. In this case, antenna spacing will not help us much. A different effect occurs when the object is a collection of smaller scattering objects. In this case, the angular intensity pattern generated by the scattering objects has a peak-to-null spacing that depends inversely on the overall size of the collection of scattering objects. In other words, as the scattering area increases, the peak-to-null spacings (in distance) decreases. For example, if the scattering region is 0.001 light year across and is 1000 light years distant, the granularity of the pattern we will expect is on the order of (1000/0.001) times the wavelength, or 50 km at a 5 cm wavelength.

This seems to imply that a collection of antennas shouldn't be placed too close together if they are to have a chance of improving reception during times of significant scattering. The Mabuhay SETI system's farthest spacing is about 100 km. This may be too close to ameliorate one minute long scintillation. Another conclusion is that more antennas are better than fewer antennas. If we use one hundred antennas, we will expect one of them to have a signal ten times stronger than the average, an average perhaps too weak to hear. One idea is to use that "lucky" ten-dB-stronger antenna as a "seed" to selectively process the signals from all the other antennas. This scheme limits the inter-antenna communication required. And opens the possibility of not using microwave links to interconnect the antennas.

As more people obtain high speed Internet (I can't in my neighborhood), the net-linked system seems reasonable. On the other hand, eyebrows will be raised by the Internet Service Provider if 100 kbit files are transferred every second continuously over the net. The service will likely be curtailed. That's why we set up five microwave links here. But they are very hard to replicate. On the other hand, the PC board we are building makes the microwave link a little easier. Perhaps we need local-microwave-linked systems interconnected on the net. Individual contributing local systems could each be about 100 km in diameter, with perhaps ten antennas in each local system. The key is to produce an easily replicated client system.

I've been studying and trying Python running on the Linux PC. Thanks to Jim Brennan. I think Python is a wonderful way to quickly glue custom device drivers, C code for signal processing and free software together into applications.

I've mentioned in a previous update the concept of the "volume search rate". So I'd like to expand a little now on this. The volume search rate (VSR) is the speed we search space for a signal resulting from a given transmit power, transmit antenna size and distance from us. VSR is proportional to the number of simultaneous antenna beams we "listen with" times the diameter of the effective receive antenna used per beam. As we add antennas (neglecting the difficulty of signal processing to combine the signals), the number of beams we can "listen with" multiplies approximately equal to the number of antennas we add. This is due to a fundamental mathematical relationship attributable to Jean Baptiste Joseph Fourier. Basically, the beams get smaller as antennas are added, and we can't get more unique beams than antennas we add. The relationship between overall aperture distribution (affected by the number of antennas) and beams (different solid angles looking at the sky) is (very closely) approximated by the Fourier Transform. The diameter of the effective combined antenna, per beam, grows as the square root of the number of antennas because the overall area increases that fast. So the VSR grows as the number of beams times the effect due to the larger combined diameter. As a result, VSR increases as the number of antennas to the 3/2 power. All of this assumes equal sized antennas. VSR is a concept important in RADAR system design, although RADAR engineers may use a different term. I just made up VSR because it seemed to fit the calculation I was making. It's important to consider, because, with some math, one can conclude that a hundred single-beam 3 meter diameter antennas have the VSR of a single 3000 meter diameter antenna! Signal processing is a complicating factor, because each of the 100 beams needs a big FFT connected to it.

I've been looking for 5 GHz interference for a few weeks. So far, interference hasn't been seen. I use a 2 foot dish at 40 feet with 1 dB NF amp pointing at downtown Spokane, which is about 25 km away.

June 29, 2000
We've decided that there is too much interference at 2.7 GHz. The RADAR at Spokane International Airport unacceptably raises the noise temperature of the SCC receiver. I sometimes notice airport RADAR or some other 2.7-2.9 GHz RADAR at my house, which is not line of sight to the airport. I think the interference is coming from the out of band power of the radar, not the pulse itself. So last night Carman and I finished a new 4.5 to 5.8 GHz low noise synthesizer, and we will be re-doing its PC board in the next few weeks. Jeff and Carman will be building more copies of the synthesizers. We'll also have to put together some LNAs and dish feeds for the higher frequency.

Ray is working to get a motor installed on the SCC antenna elevation axis in the next few weeks. James Lamphere of LinuxTeams is designing a web-based interface and device driver to control the motor.

The FPGA-based FFT processor board and VHDL code is progressing well, thanks to the generosity of my employer and the enormous help I'm getting here from my friends Rich and Shanuj. We have the parts we need to load the PC boards when we get them back. Each PCI card will perform eight one-million point FFTs in hopefully less than 0.5 second. The system is real time, and loads memory while the previous data is measured. I made an error in a design file and fixed it, and the FFT now has better signal to noise ratio.
Ray put up a new tower at SCC and a 6 foot dish for the 5.7 GHz microwave link. It's working very well. I'm almost done with the 35 foot tower at my house to improve that link.

We are thinking about a major change in the Mabuhay!system design now that more people have megabit/s access to the Internet at their homes. We are thinking of distributing receivers, phased via the Internet rather than using the hard-to-install (but fun) microwave links. The Internet is still, at best, a thousand times slower than our links (can be) but it is much easier to increase the area of the combined aperture. I have some calculations to do on this, to confirm that the spatial volume search rate increases as much as I think it might.

Dec. 3, 1999
We seem to be getting some good diurnal effects from the cross-correlator. More days of data are needed to be sure about this. Here is what we are seeing.. Yesterday I tracked down the source of the pulses in the cross-correlator. A 74F169 high speed counter needed a power supply bypass capacitor near the IC. It looks like the 2^5 bit was incorrectly flipping when certain data pins changed. No unexplained pulses have been seen since the fix was implemented.
I just ordered a violin kit. Sarah wants to build her own violin. That should be fun.
Samantha is performing with the Alberta Ballet as a soldier in The Nutcracker. The ballet opens tonight at the Spokane Opera House.

Dec. 2, 1999
I put some fairly new cross-correlator data files on the page. At Julian 1506.1885 the Doppler shift was changed by about 0.3 Hz to see the effect. So far I have not seen anything change. I'm having quite a difficult time seeing some very convincing indication that the data files contain real objects. There are diurnal trends but they are a little weak right now.
Presently the algorithm I use is as follows:
- Before processing I remove very slow drift by subtracting the last 32 sample coherent average of each delay bin, i.e. add the last 32 samples of that bin, divide by 32, then subtract from the individual value. Do this for each value that is to be used. This slow drift may be OK as it is, so this may be unnecessary.
- Coherently (add keeping track of the sign in this case) average eight time samples of each of several delay bins. It may be good to try four samples or more than eight samples.
- Incoherently (sum of the squares in this case) average the different delay bins. I have tried many groups of delay bins, I think 23 plus or minus 4 different delays is good to look at.
The idea is to remove very slow and fast fluctuations from the numbers, remove changes that are due to +- 1 clock cycle drift in the timing of the delay samples, and then look for effects that show up on multiple days. Right now I'm thinking that small objects out there are producing diurnal effects, but the time of occurrence and the delay bin seems to drift around. In addition the data seems to straddle the object. In other words I see double peaks on each side of where the object is. The spacing of the double peaks seems to change from day to day. Sometimes there is a single peak in the middle between where the double peaks were. This is conjecture but it may be the next best strategy to producing a map of objects.
On Julian Day 1513, (I didn't include this data) as the telescopes were looking at the Milky Way galaxy, the correlator seemed to be failing. It appeared that values of eleven and above were getting 32 subtractedfrom them! So if you see a string of positive numbers and then a big negative number, it is probably a system problem. The problem has gone away but it will probably return. I haven't seen it in the data on this page.

Nov. 13, 1999
I put a Mabuhay System Block Diagram here.
Nov. 7, 1999
I put together a system to FFT the raw data coming from the SCC and LL antenna sites. A 131072 point FFT is performed on each 2.5 MHz bandwidth signal coming over the microwave links. The FFT bin resolution is about 22 Hz. Duty cycle is very low - about 40 milliseconds every fifteen seconds! Jim Brennan is working on ways to use the network and some unused computing resources to increase this duty cycle. I am able to supply 100 percent duty cycle over the network; I just can't process all the data fast enough with one workstation to keep up. I'm doing this initial very small FFT to experiment with data storage algorithms. Shanuj Sarin worked today on the 4 million point real-time FFT. The data comes out at a bit rate twelve times faster than the input bit rate! We will have to process its output, so I'm experimenting with ways to use the output he will produce. What data are we saving now? We simultaneously capture the SCC and LL baseband samples. We perform an FFT on the SCC signal, and then an FFT on the LL signal. The signal to noise ratio is calculated for each bin of each FFT. Then the SCC bins' complex values are added to the LL bins' complex values on a bin by bin basis. The signal to noise ratio for this "added" bin is calculated. (A picture to explain this would be nice I suppose?) The process is repeated for the difference between the SCC bins and the LL bins. The signal to noise ratio is calculated for each of these "difference" bins. If the signal to noise ratio of the added or subtracted bin exceeds twelve when the signal to noise ratio of the bin at each site exceeds three, the bin is recorded as having had a "hit". How did I come up with those numbers? Just guessed. We can refine the values later. An array is used to keep track of the number of hits per bin. A file is written to when a hit occurs that causes the hits in any bin to be greater than one. In other words - "repeat" bin hits. The time of the hit, signal to noise ratio and the the number of hits in that bin are saved to the file. Every ten minutes the hit array is cleared. We do this to keep the array from filling up with hits that aren't related to other hits. See Hypothesis and Speculations for an explanation of why we do this. This is all very preliminary.
I fixed another problem we were having with the cross-correlator. High dc offset driftiness was reduced by adding gain in front of one of the one-bit analog to digital converters. It looks much better now (but I have said that before).

Nov. 5, 1999
I found several big noise problems in the cross-correlator. I nulled out the cross-correlator dc offset, and the files look much better. During troubleshooting I captured a data file (148k) that had a huge dc offset. When you scroll through the data, notice the drift of the vertical lines from left to right. This drift can be used to measure the relative time drift in the clocks at the SCC and LL sites. Each column corresponds to 200 nanoseconds of delay. (Nov. 7, 1999 Note: I had changed this to 166.666 nanoseconds - so numbers calculated here are in error.) I measure a 3 microsecond drift in 115 minutes. The ratio is 4.3 x 10^-10. Each site's reference is multiplied up to the LO frequency. This is 1.17 Hz at 2.7 GHz; larger than one FFT bin. SCC uses GPS, so we can iteratively adjust the LL Rubidium to zero it. Or we will have to account for it somehow when we do multiple-site beam forming. The cross-correlator integration time is about 200 milliseconds so a 1 Hz error is acceptable there.
I nulled out the cross-correlator dc offset, and the files look much better.
I fixed the attenuation problem on the roof. As often happens, removing and bypassing an amplifier entirely fixed the problem. The live video feed from our 100 km distant site looks very nice now. Each morning I see frost on the ground. It goes away by about noon. Bob Conley is working on ways to protect the dishes up there from snow. Perhaps a smooth stretched-plastic boat cover and a twenty watt light bulb under it? We could check it out by looking at the video - but only at night!
The new cross-correlator data files are looking pretty good. I'll put them on the page in a few days.
We're making good progress on FFTs. I overlooked an exciting possibility on the FPGA-based system. Rich Maes and Shanuj Sarin figured out a way to possibly expand the number of real-time channels beyond 56 million. Perhaps to 80 million. We think we can gather the parts to do at least two PCI cards per Linux machine. I need to get busy and get more real-time microwave channels from all the antennas out there!

Oct. 28, 1999
Happy Birthday Steve!
The Linux device driver works extremely well! The cross-correlator is now being controlled by it. Some cross-correlator data files are appended to this page.
Oct. 28, 1999 (continued)
Last weekend I replaced the 5.7 GHz power amp at our North Idaho translator site. The test video came booming into my desk. The calculated capacity on this 110 km microwave link is now about 1 Gibibit/sec (1 Gibi = 1.073741824x10^9) :-). Something broke up on the roof and dropped the signal about forty dB. Fortunately I have lots of left over 5.7 GHz amplifiers, and when it stops raining I'll get on the roof and fix it.
I'm throwing together a system to do a small (256 k point) FFT on each of the 22 km separated antennas. Hopefully it will be working in few weeks. This small FFT won't be nearly as nice as the 56 million point FFT we are building (Rich, Shanuj, Bob and I are now burning midnight oil on that project), but it will be more sensitive than the cross-correlator by about a factor of a thousand. We are also working on a four million point real-time two channel FFT which we think we can easily build. It is a prototype of the 56 million point FFT and does not require a new PC board to be designed.
When we get this first small FFT running, we will start the search for narrowband simultaneous pulses at two sites. Hopefully we'll have the third site set up early next year.
I haven't tried FFTW yet; it looks very good.

Oct. 12, 1999
I'm writing a Linux device driver to control and access the cross-correlator. It will run in the background in KDE - X windows and be the prototype of the FFT computer device driver. I'm reading a great book on the subject: Linux Device Drivers by Alessandro Rubini. So far, I have a few things working: device registration, driver print to X windows, and driver sleep. Next step is to get data transfers to and from user space.

Oct. 6, 1999
Renee Frankel, a student at SCC, has started to analyze the cross-correlator data, looking for pulses and diurnal effects. Her husband Garth is writing Visual Basic programs for the system. By the way, Renee has a huge number of Seti@home units processed: something like 175!
Renee and my brother Jeff are helping me figure out ways to put the cross-correlator raw data on this web page. Then you too could look for pulses.
Thanks to Jim Brennan's help, I got Linux 2.2.13, Red Hat Linux 6.0 with KDE running on the radiotelescope computer. Wow! Is it nice! Very straightforward installation. I used Macmillan's edition of Linux-Mandrake 6.1.
Paul and Ginny Branham checked out the northern Idaho Mabuhay SETI site with John Richardson. The two tripods in the image will hold two 3.7 m dishes. Paul and Ginny are educators in Idaho. Paul built a really cool optical interferometer for classroom science demonstrations.
Shanuj, Bob and I are working on a quick implementation of a four million channel FFT. It's half-real time and has 2 Hz bandwidth per channel. We think we have found a way to build it without burning too much midnight oil. Rich is working on the 56 million channel system (0.5 Hz per channel real-time). We should learn a bunch by building the smaller system. This should help turn on the big FFT system.

Sept. 22, 1999:
The signal from SCC has been cleaned up a bunch. That should improve the fluctuating base line in the cross-correlator data. Also times are now all being sync'd to UTC within about one second. A microwave power amp (and/or its power supply) at one of our microwave radio sites failed. John and I pulled it out and bypassed it. We're getting a signal now but it's weak. Hopefully we'll get the power amp back in before the snow falls.
As were about to drive back down the mountain, a group of very athletic high school students arrived. During some small talk, John told one of the young women that we were building a radio telescope. Suddenly she confidently announced, "Oh, if you're John, then _you_ must be Skip!"
I retorted, "Wow! We need YOU to analyze our SETI data!"
The young woman, Fauna, I think, remembered our names a week after her science teacher, Rick Alm, mentioned to his Priest River Lamanna High class that he had received a letter from me. I'm planning to visit their school this Friday. Update: Oct. 6, 1999: I taught Astronomy all day Friday. Wow! I had a bunch of fun! What a great group of students and educators!

I added to this web page a description of the Mabuhay radio interferometer. It's here.

Aug. 18, 1999:
The long baseline radiotelescope crosscorrelator seems to be working better. Preliminary results are at: Radio Astronomy update

Aug. 8, 1999:
I'm having tons of fun writing software on the Linux machine to process the enormous amount of data from the cross-correlator. I'm not exactly sure about what I am seeing in the data. I'm pretty sure most of it is an artifact of the marginal microwave link from SCC. We're going to improve that. The rest of it might be real astronomical objects. So I'm real anxious to get the link improved so we can look at the objects, e.g. quasars and supernova remnants.

A very funny thing happened today. A dark green jet helicopter zoomed about a hundred feet over the three LL antennas, then climbed away fast! I have never seen anything like this in nineteen years of sky-watching. Someone told me that this same sort of low-flying helicopter event has been happening often over the LL antennas, starting about a month ago. The funny thing is...that's when I turned on the Mabuhay!SETI system. Please, do not e-mail me with your conspiracy theory. And, I want to make it very clear to everyone that I am accurately reporting (almost) every last SETI thing I am doing. It's all written right here on this page. So, please, there's no need to check up on us...oops, I mean, me. ;-)

July 19, 1999:
SETI with the Mabuhay! system has officially started!! (with reduced sensitivity)

The 22 km baseline interferometer with cross-correlator is running and capturing data around the clock. Effectively this system is searching the sky for correlated signals of any modulation type - as long as the signal is confined to a 2.5 MHz bandwidth. Integration time is about 0.2 seconds. The system is real-time and uses two microwave links to bring 2.5 MBit/s data to the processing center. Effectively we have a sensitivity approximately equivalent to a CW bandwidth of about 2 kHz. Any type of modulation, however, not just CW tones, will show up as a peak in the cross-correlation. Delay is swept to check for sources slightly off axis. Interferometer fine lobe resolution is about 10 arc seconds; the interference fringes have been stopped on the sky by offsetting an LO at one site by 11.4 Hz. Atomic and GPS clocks get the absolute LO frequencies to within about 0.1 Hz at 2.7 GHz. Site to site delay was measured using a GPS-derived one second pulse from each site. Data on quasars and other radio objects has been obtained from the GB6 catalog data base, and the peaks of the cross-correlation we are seeing are being compared against the objects in that data base. Radio astronomy sensitivity is about 4 Janskys. More detail will be added to this web page in coming weeks, including a block diagram of the interferometer, and the quasars we think we are seeing.

July 11, 1999:
The 3.0 meter and 4.6 meter antennas each have one polarization working with 2.5 MSPS baseband data and calibrated noise temperature values sent over two microwave links (SCC to Agilent Technologies and Liberty Lake to Agilent Technologies). 2.7 GHz antenna noise temperatures and 22 km spaced interferometer cross-correlation is being logged almost continuously. The FPGA FFT signal processor design is being improved to use four 1 million gate FPGAs. Speed improvements may allow each signal processing card to perform fourteen 4M point FFTs (real time) every two seconds (twice as many FFTs as previous planned). The number of simultaneous beams formed will be doubled if the faster design works. And more antenna feeds may be added to contribute to the beamforming calculations. Two processor cards are planned to be installed into PCI slots on two x86 systems running LINUX. The FPGA digital design is finished and works well. The design is being translated to code to load into the FPGAs. The PC board layout for the FFT/beamforming signal processor is starting. One or more of the northern Idaho to Agilent Technologies three-hop microwave links has been down for about a month. Troubleshooting is underway. The SCC to Agilent Technologies link bandwidth is planned to be increased to about 20 Mbps (from 2.5 Mbps) this year.

Hypothesis and Speculations

We seek relatively short, isolated pulses (each only a few seconds in duration). Our hypothesis is that short pulses sent by our neighbors in the Milky Way Galaxy are significantly separated in time, perhaps by days or weeks. The highly speculative thought driving this hypothesis is that a civilization may wish to announce its presence while simultaneously minimizing interference to technologically-emerging fragile civilizations. Announcing one's presence in a "non-interfering way" appears quite problematic. However, a highly improbable single pulse event arriving from a measured direction gives a clue as to where to subsequently look. All this leads to the question of how to follow up a "first pulse detection" given the clutter of human-produced electromagnetic artifacts that look like "first pulses".

Perhaps we can ameliorate this interference situation by doing the following:

-Very clear and/or well-regulated frequency bands are being sought.
-Multiple dish antennas are located at two of the sites.
-Site-to-site antenna separation will allow dish beamwidths to reject interfering signals, to about 5000 km.
-Site-to-site Doppler shifts will be used to reject interfering signals, perhaps to about 100,000 km.
-A broad beamwidth antenna located at the central site, driving narrow and wide bandwidth receivers, will check for dish off-axis interference.
-Accurate direction of arrival will be measured using monopulse radar signal reception techniques.
-Each pulse's direction of arrival will be archived.
-The same sky locations will be repeatedly searched.

Doing all this might help. However, I do not think it will help enough for the Mabuhay! SETI Project; our system is just too small.

Further work is certainly required here; signal processing ideas will likely evolve.

System Description and Status

Much work remains to get this system working; this description is preliminary.

Antennas

We have obtained, through generous donations, seven parabolic dish antennas. The antenna diameters are: 3 m, 3 m, 3.7 m, 3.7 m, 3.7 m, 4.6 m, and 6.1 m. The seven antennas are presently located at four sites in Washington and Idaho. Four of the antennas have been set up and are prepared to have feeds installed. The remaining antennas have mounts finished, with the exception of the 6.1 m. We are planning to motorize the pointing of one of the seven antennas. This steerable dish (which is 4.6 m dia.) will be pointed to "certain peculiar" places in the sky. The remaining non-steerable antennas will search about one degree of the sky around a fixed declination. For several hours a day, five of the antennas will be pointing close to the plane of the Milky Way galaxy. The 6.1 m antenna will be used sometime after the others are operational. We plan to monitor frequencies around 2.8 GHz at each site, using two orthogonal polarization feeds per antenna.

Low Noise Amplifiers

Several 2.8 GHz Low Noise Amplifiers have been built and tested. Agilent Technologies (split from Hewlett-Packard) ATF-36077 transistors are being used. Measured amplifier noise temperature is 34 Kelvins. A total of fourteen amplifiers need to be built and tested.

Receivers

A GPS-frequency-locked receiver connected to each feed will produce four million baseband samples per second, corresponding to a two megahertz segment of the input spectrum. We are carefully measuring the phase shift through each receive path, so that signal polarization may be measured. Remaining receiver work mostly involves loading parts on PC boards, packaging and testing.

Microwave Links

Baseband samples from each of the receivers are brought to a central site using microwave radio links. Data is sent over each link at a maximum rate of about twenty megabits per second. The microwave link center frequencies are chosen to be in various amateur radio bands. About 90% of the microwave links' installation and testing has been completed. 6/15/99 update: All links have been working. Every now and then (like right now) something bad happens and we have to go up the side of a mountain to fix it.

Central Processing Site

At a central processing site, seventy simultaneous real-time antenna beams will be formed. The beams each represent a pointing direction. For each of four million frequency channels, the power in each of these antenna beams will be measured. If the power exceeds a threshold, raw data will be archived. Signal processing is performed using a combination of FPGAs (Field Programmable Gate Arrays) and x86 Processor hardware and software.

Field Programmable Gate Array-based Signal Processing

We are presently designing two FPGAs that together perform a 56 million channel FFT (Fast Fourier Transform), a seventy-beam antenna beam-former and a four-baseline radio interferometry cross-correlator. The 56 million channel FFT is actually fourteen parallel-calculated four million point FFTs. The beam-former takes the outputs of a set of antenna FFT bins, (a bin=a frequency channel) and calculates a composite signal, similar to that of a phased array antenna. In this processing case, though, seventy phased array antenna beams simultaneously monitor every 0.5 Hz wide FFT bin. The effective area of the most sensitive synthesized antenna beam is equal to the combined effective area of all of the antennas. (This is about equal to the area of a nine meter dish.) The number of frequency channels monitored in each synthesized antenna beam is only four million because each of the twelve antenna feeds uses one of the fourteen parallel-calculated four million point FFTs. Two remaining four-million-point FFTs are used to measure off-axis interference picked up by a single wide-beamwidth antenna. All FFTs are performed in real time, i.e. all sampling and transforms are done without gaps in time.

Effectively, we hope to have an overall search system that is roughly equivalent to having about ten seven-meter dishes pointing in ten directions. Each pointing direction will have a four million channel receiver "listening" (with 0.5 Hz bandwidth per channel).

The signal processor consists of two circuit boards, each with two 100,000 gate FPGAs, and many fast SRAM ICs. (A circuit board will measure all antenna signals, for one polarization.) Each circuit board will be interfaced to a dedicated host x86 processor over a PCI bus. Given the number of FFT points and the Transform Time (56 million points in two seconds), this doesn't seem like much hardware. There are several reasons for this. The FFT is a two-dimensional algorithm that uses the x86 SDRAM for "row-column" storage. Small fast SRAMs are used for the one dimensional FFTs. The FFT algorithm is implemented using optimized word widths throughout the design, given that low Signal-to-Noise Ratios are present. The trigonmetric calculations in the FFT are implemented using a compact CORDIC algorithm. And, very dense field programmable gate arrays are being used. The hardware FFT system is roughly (performance-wise) equivalent to a few hundred x86 processors running conventional one dimensional FFTs.

All of the circuits to be placed in the FPGAs have been tested in gate-accurate simulation, and the design of the PC board for the FPGAs has begun. Nevertheless, finishing the signal processor in 1999 is a huge challenge.


Radio Astronomy using the Mabuhay! System

Radio interferometer cross-correlators and noise temperature measurement techniques will be used to verify that the antennas are all pointing at the same spot in the sky. Also we need to be sure that the receivers are tuned to the same frequency. Granted, there are other (less fun) ways to do this. Plus Radio Astronomy is a kick. :-) Drift scans of Cygnus A are seen in plots.

6/15/99 update: A real-time cross correlator is being built for the 22 km spaced SCC and Liberty Lake antennas. It should be working by mid July 99. Hopefully we will see fringes (an interference pattern) due to the distant spacing of the antennas. Delays need to be equalized to within 100 ns, and frequencies accurate to 1 Hz at 2.7 GHz. All the parts are laying around here to do this; about fifty wires need to be (properly) connected. And I need to write about a hundred lines of C code to control things and get the data into a nice file. If you are interested in details, the interferometer system is: local meridian transit, linearly polarized, one bit real sample per feed, five megasamples per second, (changed to 2.5 MSPS), (approx) 100 ms integration time per delay step, thirty-two 200 ns delay steps, frequency reference uses GPS-locked quartz and atomic clocks, and the system uses micowave link transmitted GPS time to measure and equalize the (approx. 63.4 microsec.) delay. Two microwave links are used (they are described in a section below). The long baseline interferometer may be working... The plot shows the cross-correlation as the interferometer sweeps the sky. One delay was chosen to plot. Other delays also show some interesting fluctuations. It is too early to know if the bump at x=0.875 is Cygnus A. More sky sweeps are being taken. The regular glitches at five minute intervals show the effect of the 290 K noise calibration source being switched in. We expect to see the correlation drop to zero when 50 ohm loads are placed simultaneously at the input of the LNAs. However, it's too early to say if the drop to zero is in fact due to the different sky / 50 ohm load correlation.

Update: July 9, 1999 : The cross-correlation pattern for Cygnus A did not repeat when re-checked. Troubleshooting is underway. Cygnus A as it affects antenna temperature was recently measured at Liberty Lake and is shown in this plot.

Update: July 11, 1999 : We now transmit antenna noise temperature for each antenna as well as the 2.5 MSPS data over the link. Cygnus A as it affects antenna temperature was measured at SCC and is shown in this plot. We plan to troubleshoot the multiple peaks the night of July 12.

Update July 12, 1999: Cygnus A at SCC for Julian Day =...1370

Cygnus A at SCC for Julian Day =...1371

The double peak is there for a second day. This will be worked on today. The other glitches went away. Others may have appeared (not in plot). We may have an intermittent switch in the 290 K cal system. The double peak seems to drift to the left by the expected 0.00274 day per day. Samples are taken five minutes apart, so it's a little hard to be sure about this.

Cygnus A at SCC for Julian Day =...1370 and 1371

Update: July 15, 1999 It looks like the SCC and LL antennas are finally pointing at the same spot in the sky. The SCC dish was moved to about 6 degrees from vertical and measurement re-run. The following shows the alignment of the SCC and LL antennas by (approx.) interleaving the temperature data of the two antennas on the same plot. The bump at about x=0.87 is Cygnus A. It appears that the coma lobe to the left of the main lobe exists similarly on both antennas. Perhaps it is not a coma lobe? I'll have to re-check the Cambridge 3C catalog.

Perhaps the cross-correlation between the antennas will show daily features now...

Cygnus A at SCC and LL with data interleaved for Julian Day = 2451374.

The SCC antenna was re-pointed again about 1 degree south. The double peak of the SCC antenna seems to be returning. The focus of the SCC feed may be a little off. Or perhaps the original pointing (indicated 6 degree south of zenith) is correct. Cygnus A at SCC and LL with data interleaved for Julian Day = 2451376. It doesn't look too bad.

Update: Aug. 5, 1999: Ray re-pointed the SCC antenna to 7.5 deg S of zenith (indicated) the evening of Aug. 3. The Aug. 3-4 SCC data was corrupted during the Cygnus A pass due to some unknown microwave link problem. The data from last night though was OK. So here it is. It looks quite good. The previous SCC data indicating a double peak of 10 degree rise in the noise temp for Cygnus A is very questionable. It repeated for two days while the SCC antenna was pointed away from Cygnus A. Another possibility is that both SCC and LL antennas are now tweaked to the same mis-focus and mis-pointing. If this is the case the sensitivity would be much less than what is possible. Further experiments with the LL antenna will be done to disprove this hypothesis. Cross-correlation data shows many interesting peaks but no diurnal effect has been found yet. I haven't processed and compared the last few nights of data yet.

Update: Aug. 7, 1999: We verified the SCC/LL pointing again last night. JD 2451397.6 SCC/LL comparisons.

We decided to point the LL antenna way off to check the cross-correlation peaks with expected zero correlation. To do this the LL antenna was steered 25 deg S of the zenith from its previous direction of 7.5 deg S of zenith. The re-pointing was done at Julian Date 2451398.309.

Update: Aug. 9, 1999: Cross-correlation seemed minimal after the antenna was "mis-pointed", as expected. Last night's set of files cc35.txt and ta35.txt didn't get written to for some reason. The antenna was pointed back to the original 7.5 deg S of zenith around 2451400.063. File writing will be checked today into cc36.txt and ta36.txt. The processing computer clock drifts very bad. So yesterday I made some changes to the LL program. GPS time from the GPS receiver at the LL site is now being sent over the LL-Agilent Technologies link and placed into the antenna temperature file. The link transmission is not sync'd to the one second GPS clock so the reported time is probably accurate to only about 1 second.

Update: Aug. 18, 1999: There is quite a bit of baseline drift, due to a marginal microwave link. To make matters worse, we are calibrating the noise temp too often and too long, so a high percentage of the time we are not measuring the sky. Another problem is we send serial data over the links at too many times, potentially corrupting the cross-correlator data. To fix this we need to improve the SCC-Agilent Technologies microwave link, and modify the radiotelescope programs at the sites. We are working on all this. The following is a plot that seems to indicate several objects in the Milky way. The telescope resolution is calculated and measured to be about 5 arc seconds in Right Ascension (the antennas are close to East-West spaced).

This plot shows the cross-correlation on two successive days, looking into the Milky Way in Cygnus. The apparent object seen at 20.235 RA hrs may be ADG073.8+01.0 which is in The Master List at 36 10 14.8 decl. However, we think we are looking at 40 deg. decl. The object in the list has a flux of 13.0000 Janskys at 2695 MHz. We may be seeing that object on the side of the main lobe. There are no other objects at these RAs in the Master List stronger than 100 milliJanskys from 35 to 45 decl.

It appears that the sensitivity of the telescope may be well below 1 Jansky. More calculations and measurements are required to confirm this. The vertical axis of the plot is the cross-correlation for "bin 4" which is a particular step in the 200 ns step digital delay line. The blue lines show when the radiotelescopes were looking at the reference 290 K noise source. Minimal cross-correlation should be expected during this time. However serial data is sent during the cal periods, so data can be corrupted. Plus, the baseline drift problem needs to be improved.

SETI Site Images

This image was obtained from a microwave-radio-linked video camera. At some time (about) every day, (soon to be automated) an image is captured, in whatever state everything is in. 6/15/99 update: There is too much work to do, so I'm not going to play any more with the remote video cameras. Three SETI dishes (3 m, 3 m 3.7 m) are in the background.

Other images on this page do not automagically update. Some may soon, though...

6/15/99: No.

Microwave links have been tested by transmitting live video using FM. The image shown below is frame-captured video sent 2 km (the link frequency is around 2.4 GHz; we are QSYing (changing frequency) to 3.4 GHz because of severe QRN (static) from microwave ovens at 2.45 GHz.

(update 2/8/99: 3.4 GHz is installed and works very well.) At this receiver site we set up two 3 meter dishes and one 3.7 meter dish.

Another site is shown below.

This image is derived from frame-captured video sent 110 km using three microwave links (freqs. one at ~3.4 GHz, another at ~5.7 GHz, and another at ~5.7 GHz). These links were a huge challenge to install. However, we had lots of fun and met many nice people who helped us a bunch.

Two dish antenna tripods are set up and may be seen down the slope. The white spot slightly to the left and below the "1" is one of the two 3.7 meter dishes at this site. The numbers and other artifacts in the image are produced by the video camera, which has a dead Real Time Clock battery but now conveniently displays the number of days the camera power has been on (twenty days in this case).

Our third SETI receiver site is shown in another image.

The 4.6 meter antenna on the left in the image is dedicated to SETI. The tower on the right holds a small microwave link dish (installed after the picture was taken) to send seven megabit per second data to the central processing site, about twenty-two km distant.


Mabuhay! SETI System FAQs:

Will the Mabuhay! SETI System receive anything unusual?

I do not expect to receive anything unusual using this system because I think we need more dish antennas, more microwave links, and more FPGAs. Plus I think that intelligent civilizations in our galaxy are being very careful to not interfere with our communication systems, radars, missile defense systems, etc. I do not expect a strong, long-lasting CW carrier, for example. This is certainly speculation, but is just my guess. This Mabuhay! SETI project is basically an attempt at a "null experiment": cancel out everything as best we can and look at what is left. I expect nothing. And, if anything unusual does appear in our data, I am going to be very skeptical about its possible extraterrestrial origin.

Why have you chosen to listen at frequencies "around 2.8 GHz"?

If civilizations are trying to not interfere with us, then there may be no magic frequency. Maybe they pulse their transmitted energy on different frequencies.

If this is true, then the best frequencies for us to listen on may be the ones with the fewest human-produced photons. I've been reading as much as I can about US and International frequency allocations. I found that the 2.69 GHz to 2.9 GHz frequency range has only two well-defined allocations: protected radio astronomy below 2.7 GHz; and Airport Surveillance Radar (ASR) and other ground-based radar, using frequencies between 2.7 GHz and 2.9 GHz. Unfortunately for SETI, many militaries have numerous Airborne Self-Protection Jammer (ASPJ) systems. Many of these will presumably jam the 2.7 to 2.9 GHz radar band using various CW and modulated signals. These potential jamming signals make the 2.7 to 2.9 GHz SETI search a little difficult. To make matters worse, the pulses from the ASR radars are very powerful, and will certainly produce noticeable reflections off of satellites. However, I expect these reflections to be quite broadband, and perhaps easily identified. I have learned that CW Doppler radar is not permitted in the 2.7 to 2.9 GHz band; that helps. I don't know if anyone has the capability to jam 2.7 to 2.9 GHz frequencies from near or deep space. My guess is no. Pilots fly the airplanes, and ASR radars are there to help them, but I think we would have a mess on our hands if a satellite-based ASR jamming system were to go awry.

This all sounds quite bad, but overall, I think the 2.69 GHz to 2.90 GHz band seems fairly free of "unknown" interference, (except for maybe the airborne jammers) and is particularly free of isolated CW carriers. Harmonics and other spurious signals are, of course, always an issue. Presently I examine local interference with an automated spectrum logging system comprising an Agilent Technologies 8566B Spectrum Analyzer. I have placed a small dish/BPF/LNA on the roof here, and plan to add other types of antennas to the system. In general, I think I need to choose some frequency range, get to know it well, then do the null experiment.

If you want more information on interference in various frequency bands, there are many resources on the web:

A US frequency allocation chart is at http://www.ntia.doc.gov/osmhome/allochrt.html.

A Canadian frequency allocation chart is at http://strategis.ic.gc.ca/SSG/sf01608e.html.

In addition I found numerous very useful spectrum allocation papers from the NTIA:

http://www.ntia.doc.gov/osmhome/reports/pub9835/Raspexec.htm(Radio Astronomy allocations).

http://www.ntia.doc.gov/osmhome/reports.html(Spectrum Engineering Reports)

http://www.its.bldrdoc.gov/pub/pubs.html(papers plus links to very extensive Denver and San Diego spectrum studies)

Why are you working on this project?

I am working on this very fun project because I love astronomy and I love to build radios, antennas and computers. In addition, I do hope someone receives something interesting, someday. Maybe a tiny idea from this will help the big searches that we need to do. Or maybe we can just point this and listen when and if someone ever picks up something interesting. That would be fun.

When will this Mabuhay! SETI system be working?

Hopefully in this century! I started building it in early 1996.

Do you believe in UFOs?

No. When I can touch and examine a UFO, i.e. hold a piece of a hypothetical space ship, I might start to believe in them. I admit that UFO "sightings" certainly have had a very emotional and in some cases, physical effect on many people.

How is your pulse receiver any different from a camera snapping UFO pictures?

The Mabuhay! SETI Project is an attempt to build a machine that unemotionally grabs artifacts from the universe, if they can be grabbed. You can build a similar machine and obtain similar results in a reasonable time.

If any UFO is a real flying machine, do you think they are from outer space?

I strongly believe that I need to accept and work with the Physics that I know about now. Anything else is speculation and/or the work of Theoretical Physicists, which is very interesting to me, but doesn't let me perform a SETI experiment today. Using known Physics, I know star travel is very difficult. A starship's energy consumption becomes almost unimaginable. On a starship, clocks, distances, and mass change as we watch the ship go out on its mission. In fact, the mass approaches infinity as the starship approaches light speed. Albert Einstein's Theory of Relativity explains this uncanny starship travel and predicts the difficulty of travelling at close to light speed. Sending radio messages is much, much easier, and is far more likely to be expected from our neighbors in the universe. So I'm expecting radio messages to be sent to us, not flying machines.

I have written a DOS program that graphically shows you the spacetime and mass of a starship as it approaches light speed. In addition, the program shows you why the Theory of Relativity explains the strange effects. Click here to download the zipped space.zip file. After you unzip the file, run the .EXE file in a DOS window, press the up and down arrow keys to change starship speed, exit with the space bar or the escape key. There is also a text file included that explains the program and why the starship gets so massive as it approaches light speed.

My good friend Joe Dubner has translated the DOS program into a very nice Windows 95/98 program. Click here to download the zipped SpaceW.zip file (~600kB).

What does "Mabuhay!" mean?

"Mabuhay" is a friendly greeting (translated "Long Live.."), spoken in the Philippines. My American parents raised me in that beautiful country. I chose the name "Mabuhay" because of its friendly meaning. (Plus, ET civilizations have to "live long" to communicate with other ETs.)

My interest in radio communication started in the Philippines. My Dad showed me how to build radios and antennas, and my Mom made sure I could find all the radio and antenna parts I needed. That's how this all started; getting radio parts at Spark Radio and Electrical Supply in Quiapo, a district in Manila.


Acknowledgements

Many thanks go to many people for more than I have written here: (I risk temporarily forgetting someone's help.)

Bob Conley (FPGAs), Ray Johnson (too much to list), Ray Johnson's SCC students (ditto), Mike Walters (GPS), Shiela Powell (PCBs), Jim Brennan (Linux), Shanuj Sarin (FPGAs), Rich Maes (digital design), Steve Kay (FFT PC board), Phil Wilshire (RT Linux), John, Robbie, Lissa, and Jena Richardson (flying), Pat Brown (microwave), Jules Gindraux (microwave), Ken Olsen (PCBs), Joe Dubner (microwave), Jim Ehrhardt (microwave), Dick Skinner (dish), Ken Lyons (PCI), Tom Faulkner (GPS), Ken Thompson (camera) Martin Howser (dishes), Bob Besser (microwave), Jake Laete (microwave), Roy Anderson (radio telescope instrumentation), Jerry Long (liquid Nitrogen), Ed Mitchell (ham radio), Gary Tong (SETI a long time ago), George Moore (1-bit ADCs), Jim Barrett (dishes), Brooks Shera (GPS), Bob Gray (lots of good ideas) at The Small SETI Observatory, John Dreher (lots of good ideas), everyone at The SETI Institute, everyone at The Dominion Radio Astrophysical Observatory, Barney Oliver (who sparked my interest in SETI, and whom I miss), my brother Jeff (Astronomy), my daughters Sarah (lots of good ideas) and Samantha (lots of good ideas), and my Mom and Dad. And thanks to the (approx.) fourteen people who moved the six meter dish!

Thanks to H. Paul Shuch and The SETI League.

Go to 
 the SETI League


Thanks to everyone at LinuxTeams.


LinuxTeamstm

A Division of L.A. Boone & Company, Inc.


And thanks to everyone at Agilent Technologies (split from Hewlett-Packard Co.).

Go to Agilent Technologies


Seti is really cool

by Sarah Crilly

I help my Dad all the time when he works on Seti. I went to a Seti site and climbed a steep mossey mountain ( it was slippery, too) to get to a microwave radio site. Later a bear visited the same spot. At home, there are cables all over the house for Seti. I helped pull, measure and cut them so they would be just perfect. At Dad's work, he uses his cubicle for Seti, too. Over the past few years, Dad has accumulated several cubicles. He says his wires grow like vines and just spread into cubicle after cubicle after cubicle- eating everything in their path! Anyway, I love doing Seti and learning new things. Even my school helped find ways to transmit and receive messages. Dad and I, (along with all the other people who've contributed) would like to thank my sixth grade class.


Seti is really cool, continued

by Samantha Crilly

My sister wrote Seti is really cool so I Samantha wanted to write Seti is really cool, continued because my sister decided to write nothing about me!!!!!!SARAH!!! Well anyway I would like to thank my 3rd grade class for contributing to my Dad's collection of stuff laying in the backyard. They built the "wave disrupters" for "energy and matter" unit in science. Mrs. Barkley was fascinated with what my Dad said and his lectures. So I am in 5th grade now and I am in a performance called, Oliver! I also might be performing in the Olympics in Sydney, Australia for Ballet and Jazz. If you want tickets for Oliver call, (509) 325-Seat. The performances are February 10-13, 1999 at 7:00pm in Spokane,Washington.

So nothing much is going on. We haven't picked anything up yet. Don't be alarmed, aliens are not secretly coming to the Planet Earth. We are going as fast as we can to set up everything!

(Skip's note 2/27/99: I helped Sarah and Samantha route eight radio telescope cables today. Samantha pulled and aligned the heavy cables. I couldn't fit under the deck, so Sarah "army crawled" where the spiders and mice live, pushing the cables through a small hole in the wall. Then we pushed them through a hole in the floor and into the receiver room.)

Some cross-correlator data files

1396314 Nov 27 12:00 cc1509_0.txt
1395980 Nov 28 12:00 cc1510_0.txt

73 de K7ETI

Thank you for visiting our web page.

(My email address is skip_crilly@agilent.com)


This page hosted by   Get your own Free Home Page