The Big Ear Wow! Signal What We Know and Don't Know About It After 20 Years
Written by Dr. Jerry R. Ehman Original Draft Completed: September 1, 1997 (Send Comments to: ohioargus AT gmail.com )
Notes to the Reader The entries in the Table of Contents below are links within this document (i.e., bookmarks). Clicking on one takes you to the start of that section. This is helpful if you are not able to read the entire document in one sitting.
Table of Contents
The Wow! source radio emission entered the receiver of the Big Ear
radio telescope at about 11:16 p.m. Eastern Daylight Savings Time on
August 15, 1977.
Thus, at the time this article is being written it is just past the
20th anniversary of the detection of that now famous radio source.
What have we learned about that signal over the past 20 years? Could
it have come from an intelligent civilization beyond our solar
system, or could it have been just an emission generated by some
activity of our own civilization?
In 1973 the Big Ear radio
telescope was converted from measuring the location and strength of
wideband radio sources (the Ohio Sky Survey) to a similar study of
narrowband radio sources. Due to an unwise decision by the United
States Congress in 1972, we lost our funding from the National
Science Foundation (NSF) to support the Ohio Sky Survey. Eventually,
every person employed to work on the Ohio Sky Survey team (except the
Director, who was funded separately) lost his/her job; I was one of
those persons. We each found employment elsewhere. There was a strong
desire to continue to observe with the Big Ear but it had to be in a
project that was less human-resources intensive. The systematic
search for narrowband signals seemed to be the best way to use that
unique radio telescope. The Big Ear was well designed for a
systematic sky survey, as was clearly demonstrated by the success of
the Ohio Sky Survey (in which about 20,000 radio sources were
measured, about half of which had never been observed before). Also,
the combined observing time of all other narrowband observing
programs up to that time was very small. Use of the Big Ear would
quickly result in our achieving the record for the longest
continuously-running survey of narrowband radio emission (indeed, we
did achieve that record as described in the "Guinness Book of World
Records"), although we didn't purposely set out to achieve that
record.
The receiver and associated electronics were connected
under the leadership of Dr. Robert (Bob) S. Dixon, the Assistant
Director of the Ohio State University Radio Observatory. Bob and I wrote the
software for the IBM 1130 computer used to acquire and analyze the
data. Bob wrote most of the initial software to handle the data
acquisition and some basic analysis. I handled the rest of the
software, especially that involving some of the more involved
analysis of the data (including search strategies). Both of us had
other jobs so this was done in our spare time. After the data began
to come in regularly and we began a systematic survey of the 100
degrees of declination visible to the radio telescope, I took on the
task of looking at the computer printout on a regular basis.
A few days after the August 15, 1977 detection, I began my routine
review of the computer printout from the multi-day run that began on
August 15th. Several pages into the computer printout I was
astonished to see the string of numbers and characters "6EQUJ5" in
channel 2 of the printout. I immediately recognized this as the
pattern we would expect to see from a narrowband radio source of
small angular diameter in the sky. In the red pen I was using I
immediately circled those six characters and wrote the notation
"Wow!" in the left margin of the computer printout opposite them.
After I completed the review of the rest of the printout, I contacted
Bob Dixon and Dr. John D. Kraus, the Director of the Big Ear Radio
Observatory. They were astonished too. Then we began an analysis of
what has been called for 20 years the "Wow! source". Analyses have
continued even through recent years as ideas needed to be tested.
Let me describe the main features and some of the details
about the computer printout. This section will deal with the meaning
of the numbers and characters in the printout itself. A later section
will deal with other parameters related to the values on the computer
printout.
Each row of the computer printout represents
the results of the data collected during approximately 12 seconds of
sidereal (star or celestial) time. 10 seconds were used to obtain the
average intensity for each of 50 channels and approximately 2 seconds
were used by the computer to process the data and analyze it for
possible interesting phenomena. During each 10-second period of data
acquisition, one intensity was obtained each second for each channel
and then the 10 values obtained over the 10 seconds were averaged for
each channel. The left hand half of each row shows the intensity for
each of the 50 channels with channel 1 leftmost and channel 50
rightmost. Due to limitations of space on the computer printout, Bob
Dixon decided to use a single character to represent each intensity.
The average intensity over the 10-second integration period for each
channel was converted into an integer number or character by the
following 5-step process:
Step 1: the average intensity of 6 integration periods (1/6
of the current value plus 5/6 of the
previous value) was subtracted out to remove the baseline
intensity; The signal-strength sequence
"6EQUJ5" in channel 2 of the computer printout thus represents the
following sequence of signal-to-noise ratios (S/N): The strongest intensity received ("U") means that the
signal was 30.5 +/- 0.5 times stronger than the background noise
(note that the notation "+/-" means "plus or minus" representing a
range of values, in this case from 30.5 - 0.5 = 30.0 up to 30.5 + 0.5
= 31.0). Most of this background noise is generated within the
receiver itself, but some noise comes from the trees, grass and other
surroundings, and some from the celestial sky (the remnant of the
"Big Bang" explosion that is estimated to have occurred about 15
billion years ago).
Right Ascension and Declination
The next two groups of numbers on the
computer printout (just to the right of the center of the row) are
the right ascension and declination converted to epoch 1950.
Declination is the angular distance above or below the projection of
the earth's equator onto the celestial sky. Its range of values goes
from -90 degrees (at the south celestial pole) through zero (on the
celestial equator) up to +90 degrees (at the north celestial pole).
The Big Ear radio telescope can observe in the 100-degree range of
declination from approximately -36 degrees to approximately 64
degrees. Right ascension is analogous to longitude on the earth's
surface. It is measured in either degrees (0 to 360) or in hours,
minutes and seconds (00h00m00s up to but not including 24h00m00s).
The starting point (0 degrees = 0 hours) is currently in the
constellation of Pisces but is moving slowly although constantly (it
takes about 26,000 years to make a complete circuit; the major
component of this motion is called the "precession of the
equinoxes"). Because of this precession and other related but smaller
effects, astronomers convert the observed positions at any one
instant into one appropriate for a convenient point in time so that
locations can be more easily compared. The epoch (point in time) of
1950 was most commonly used during the middle to late part of the
20th century. Nowadays, the year 2000 is the epoch most likely
used.
For the strongest Wow! data point, the epoch 1950
right ascension shown on the computer printout was: 19h17m24s, while
the corresponding declination was: -27 degrees and 3 minutes of arc (-
27d03m). Thus puts the source in the direction of the constellation
Sagittarius (note, however, that the constellation gives just the
general direction and provides negligible useful information to an
astronomer).
It turns out that prior to the occurrence of
the Wow! signal, I made a mistake in the computer programming in
dealing with the correction of the R. A. coordinate for the offset of
the positive horn. I added the correction rather than subtracting it
as I should have. I corrected this error when it was discovered,
which, unfortunately, was after the Wow! source was detected. Later
in this article, I will compute the corrected value for R.A.
2nd L.O. Frequency (and the Corresponding
Frequency of Observation)
The computer printout shows
a 2nd L.O. frequency of 120.185 MHz for the strongest datapoint (the
one showing an intensity represented by the letter "U"; the 4th of 6
data points). What is meant by that frequency?
During the
planning stages of putting the receiver together, Bob Dixon decided
that observations would be conducted in a frequency band around
1420.4056 MHz (MHz means megahertz = millions of Hertz = millions of
cycles per second), the frequency of the neutral hydrogen line for
the case when there is no line-of-sight motion between our receiver
and the source of the neutral hydrogen line (transmitter). Since
hydrogen is the most abundant element in the universe, there is good
logic in guessing that an intelligent civilization desirous of
attracting attention to itself might broadcast a strong narrowband
beacon signal at or near the frequency of the neutral hydrogen line.
Bob surmised further that such a civilization might change its
transmitter frequency in such a way as to remove the effect of the
Doppler shift of frequencies that occurs when its transmitter is
either moving towards or moving away from the receiver. If the
transmitter frequency were adjusted to compensate for its motion with
respect to the center of our galaxy (called the "local standard of
rest" = LSR) and if our receiver frequency were separately adjusted
to compensate for its (and our) motion with respect to the same LSR,
then we should see their beacon signal right in the middle of our
receiver channels if it were strong enough and if it were in our
beam.
The 50-channel receiver we had available to us was
built by the National Radio Astronomy Observatory (NRAO) in Green
Bank, West Virginia. It was designed to operate so that the boundary
between channel 25 and channel 26 (i.e., exactly halfway through the
50 channels) occurred at 150 MHz. At the same time, we had an
intermediate-frequency (I.F.) amplifier that operated in a band
centered at 30 MHz, and we needed to use that amplifier as a part of
the chain of electronics to boost the minute signal so that the
subsequent electronics (including analog-to- digital (A/D)
converters) would have sufficient voltages that could be converted
into numbers to be recorded and analyzed by the computer. Thus, for
the case when there is no line-of-sight motion between us and our
LSR, we needed to have the neutral hydrogen line frequency of
1420.4056 MHz be eventually converted into 150 MHz with amplification
at 30 MHz occurring in between.
The plan was as
follows.
Step 1: Mix a 1st local oscillator (1st L.O.) signal
at 1450.4056 MHz with the weak desired neutral hydrogen line signal
at 1420.4056 MHz to yield an output signal at 30 MHz;
There was a minor glitch to the above plans, and it occurred in Step
1 above. It was discovered that the 1st L.O. was set to 1450.5056 MHz
(or 0.1000 MHz above the desired frequency). In order to compensate
for that offset of 0.1000 MHz, the 2nd L.O. would have to be set
0.1000 MHz lower than planned (e.g., at 119.9000 MHz instead of
120.0000 MHz).
The bottom line to the above discussion is
that the difference between the 2nd L.O. frequency and 119.9 MHz is
added to 1420.4056 MHz to obtain the frequency of observation at the
boundary between channel 25 and channel 26. Since each channel was
0.0100 MHz (10 kHz) wide, then 0.0100 MHz would have to be subtracted
off for each channel below the channel 25- 26 boundary.
The computer printout shows a 2nd L.O. frequency of 120.185 MHz at the
time of the strongest of the 6 data points. Subtracting 119.9 MHz
yields a difference of 0.285 MHz and adding this to 1420.4056 MHz
yields a frequency of observation at the channel 25-26 boundary of
1420.6906 MHz. It is necessary to move down 23.5 channels to get to
the middle of channel 2; thus we must subtract 0.235 MHz from the
center frequency to obtain the observing frequency for the center of
channel 2. That value is: 1420.4556 MHz.
In conclusion, we
can say that the frequency of observation of the Wow! source was
1420.4556 +/- 0.005 MHz (note that the error of +/- 0.005 MHz
represents one half of the width of channel 2, or any other
channel).
Galactic Latitude and Longitude
The next two groups of numbers on the
computer printout are the galactic latitude and galactic longitude
converted to epoch 1950. Galactic latitude is the angular distance
above or below the plane of our galaxy. It's range of values goes
from -90 degrees (at the south galactic pole) through zero (in the
plane of our galaxy) up to +90 degrees (at the north galactic pole).
Galactic longitude is analagous to longitude on the earth's surface.
It is measured in degrees (0 to 360) relative to a defined starting
point very near the direction of the center of our galaxy. Precession
of the equinoxes, among other apparent motions, affects the computed
galactic coordinates in a manner similar to the way right ascension
and declination are affected.
For the strongest data point
of Wow!, the computed epoch 1950 galactic latitude was -17.86 degrees
and the corresponding galactic longitude was 11.21 degrees. Thus, the
Wow! source direction was about 18 degrees below the plane of our
galaxy and a total of about 21 degrees from the direction of the
galactic center.
The computer was reading a sidereal (star-time)
clock. Eastern Standard Time (EST) was computed from the sidereal
time. Sidereal time covers 24 of its hours in about 23 hours and 56
minutes of our standard time. By the way, even though it was August
for these observations and, in Ohio, our civil time was Eastern
Daylight Savings Time (1 hour ahead of EST), we computed and printed
out EST to be consistent year around. Note that the Wow! source was
observed around 22:16:34 EST (about 10:16 p.m. EST or 11:16 p.m.
EDT). No one was at the telescope at that time. The receiver and
computer were doing their jobs unattended.
Analyses of Wow! to Correct
Errors
Even though anyone can read the values in the computer
printout and draw conclusions from them, there is information not
given in the computer printout that must be taken into account before
drawing certain conclusions. That additional information will be
provided in this section while describing some of the analyses done
by my colleagues and I.
Effect of Dual-Horn Feed System
The Big Ear used a dual-horn feed system.
A "feed horn" is a funnel-shaped metal structure (we used aluminum)
located between the flat and curved reflectors designed to collect
the energy focussed by the curved paraboloidal reflector located at
the south end of the radio telescope. The two feed horns were located
side by side in the focal region of the paraboloid 420 feet north of
the vertex of the paraboloid. The westernmost feed horn was located
about 8.79 feet west of the focal point. The easternmost feed horn
was located about 4.10 feet west of the focal point. Thus, the two
horns were separated by about 4.69 feet along an east-west line. The
receiver was configured as a Dicke switching receiver, switching from
one horn to the other horn and back again 79 times per second (79
Hz). The receiver measured the difference between the signals coming
from the two horns such that the signal coming from the westernmost
horn (the west horn) was subtracted from the signal coming from the
easternmost horn (the east horn). This difference signal was then
amplified, fed into the 50-channel detector, each channel digitized,
and the digital data fed to the computer for analysis. The west horn
was also called the negative horn while the east horn was called the
positive horn. Thus, the Dicke switching receiver subtracted the
negative horn signal from the positive horn signal. As the earth's
rotation swung the two beams across the celestial sky, a signal (with
positive energy) from a radio source was first seen by the west
(negative) horn and generated an inverted bell-curve-like shape on
the chart recorder. Within a minute or so after the negative horn
response was essentially complete (i.e., showed little energy from
the source), the same radio source began to be scanned by the east
(positive) horn and a non-inverted (right-side up) bell-curve-like
shape on the chart recorder was generated. Thus, for a strong radio
source of small angular diameter like a distant galaxy or quasar, we
see a negative (inverted) beam response followed by a positive beam
response shortly thereafter. However, this was not the case for the
Wow! source.
The computer printout for Wow! shows only one
detection instead of the two detections expected with the dual-horn
system. At the time (August 1977) the computer was not programmed to
identify whether the observed output was negative (from the negative
horn) or positive (from the positive horn). [Note. Later, the
computer was reprogrammed to overprint a minus sign on any printed
negative intensity (except a blank representing a signal-to-noise
ratio of 0 up to 1).] Unfortunately, this lack of knowledge about
which horn the Wow! signal entered leads to an ambiguity in the
calculated source position. Below the two possible right ascensions
are derived.
It would be a fair question to ask if the
analog chart record wouldn't resolve the discrepancy. Nice thought
but no such luck. An analog chart record was generated for the
continuum (wideband) receiver. That is, while the 50-channel receiver
was operating, a separate wideband (8 MHz wide) receiver was also
operating. It was called the "continuum receiver" because continuum
radio sources (like galaxies, quasars, nebulae, and stars) generate
radio waves over the entire radio spectrum (as well as in the optical
spectrum plus the rest of the electromagnetic spectrum). Its output
was digitized and available for analysis, but in addition, its output
(before digitization) was recorded on an analog stripchart recorder.
Although this continuum receiver easily shows continuum sources with
flux densities of about 0.5 janskys or more (where the radio emission
covers the entire radio band), a narrowband radio source like the
Wow! source would not be (and was not) detected. Let me illustrate.
Suppose a narrowband radio source generated enough energy in a 10 kHz
(0.01 MHz) band to be equivalent to a flux density of 50 janskys (but
only in that narrow band). What would be seen with a receiver 8 MHz
wide. The averaging process that would automatically occur (and is
unavoidable) would cause the continuum receiver to see a signal only
0.01/8 (or 1/800) of the strength seen in the narrowband channel. In
other words, the hypothetical 50 jansky narrowband source would
appear as a 50/800 = 0.0625 jansky wideband source, and that would be
undetectable. That is what happened to the Wow! source. Since it
appeared in only one 10 kHz channel, it contained little or no energy
in other channels. Hence, the average of strong energy in one
narrowband channel with negligible energy in the equivalent of 799
other channels yields a very low average energy, so low that it is
buried in the noise of the narrowband channel.
Determination of Corrected R.A. Assuming
Positive Horn Received Signal
Since there is an
ambiguity in the right ascension because we do not know in which beam
the source was observed, what are the two possible positions?
The computer printout shows the epoch 1950 right ascension (R.A.) of
the highest data point as 19 hours 17 minutes and 24 seconds of time
(or 19h17m24s, for short). The corresponding declination was -27
degrees and 3 minutes of arc (or -27d03m, for short). It is necessary
to understand that the printed R.A. is computed under the assumption
that the source was seen in the positive (east) beam and that each
R.A. represents the converted epoch 1950 value at the end of each 10-
second integration (averaging) period. Also remember that I had made
an error in applying the horn offset (horn squint) in R.A. so this
error must be corrected.
19h17m24s represents the end of
the 10-second integration period that yielded an intensity (signal-to-
noise ratio = S/N) of 30 (the letter "U"). However, it is better to
state the R.A. at the center of each 10-second integration interval
because it is more representative of the interval. Therefore,
subtracting 5 seconds from the computer printout positions yields
19h17m19s for the uncorrected R.A. of the largest data value.
Let's now deal with correcting the misapplication of the horn squint
(offset) in R.A. The computer acquisition and analysis program,
called N50CH, had built into it a horn squint in R.A. of minus
138/cosine(declination). This number means that at the equator
(declination = 0) the R.A. horn squint for the positive horn was
minus 138 seconds of R.A. At the declination of Wow! (-27d03m), this
horn squint would compute to minus 154.95 seconds of R.A. According
to Debbie Cree, a student who did a project and wrote a report in
1980 on the Big Ear under the supervision of John Kraus, the positive
horn was 4.10 feet west of the focus; hence, the Wow! source would
have achieved its maximum intensity in that positive horn 154.95
seconds of R.A. before it would have if that positive horn had been
located at the focus. Thus, the calculated R.A. would be too small by
that amount. I should have subtracted the negative horn squint in
order to create a larger R.A. Instead, I inadvertently added it.
Thus, in order to correct for this error, simply double the value of
154.95s and add it to the printed R.A. Since 2 * 154.95s = 309.90s =
5m9.90s (or approximately 5m10s), we do the following calculations to
the printed R.A. for the 6 data points.
In the table below
the first column presents the character used for the intensity, the
second column shows the original (incorrect) right ascension (epoch
1950) on the computer printout, the third column shows the corrected
epoch 1950 R.A. for the end of the integration interval (adding 5m10s
to the original R.A.), and the last column shows the corrected epoch
1950 R.A. for the middle of the integration interval (subtracting 5s
from the third-column results).
From the above table, using the middle of
the interval containing the largest data point, we have the R.A. of
the Wow! source near 19h22m29s under the assumption that it came in
the positive horn. A better position can be obtained if one fits the
antenna pattern to the Wow! data and determines the R.A. where the
peak of that pattern occurs. I did such an analysis. I fit two
different mathematical functions (as approximations to the antenna
pattern) to the Wow! data. One was the well-known bell curve (also
know as a Gaussian curve or normal curve). The second function was of
the form (sin(x)/x)^2, where the notation "^2" means raising to the
2nd power (squaring). These two functions are very similar from the
peak down to somewhat below half amplitude. Well below half amplitude
the second function displays multiple secondary peaks and valleys
while the Gaussian steadily drops toward a zero value. The second
function thus looks closer to what a strong source might look like
(i.e., having sidelobes). However, the Wow! source was not strong
enough to display sidelobes, so either function used as an
approximation to the real antenna pattern is a suitable fit.
In fitting the Wow! data to each of the two functions, each of the
six intensity values was increased by 0.5 to account for the
truncation error. That is, since the first intensity of 6 could have
been anywhere in the range from 6.0 up to but not including 7, the
value of 6 + 0.5 = 6.5 is the best estimate of the actual value.
Similarly, the value "U" representing a S/N of 30 is really some
value at or above 30.0 but below 31; hence I used 30 + 0.5 = 30.5 for
the best estimate of the untruncated value. Thus, the sequence
"6EQUJ5" represented the signal-to-noise (S/N) intensities: 6.5,
14.5, 26.5, 30.5, 19.5, and 5.5, respectively; an uncertainty of +/-
0.5 must be assigned to account for these truncation errors (note
that the system noise itself creates an error of 1.0 (at the 1-sigma
level by definition (which corresponds to a 68.26% confidence level)
or an error of 2.0 at a 95.44% confidence level).
The 5-second subtraction of R.A. for each data point, as described
above,
was also used, but the 5m9.9s corrected for the misapplication of the
horn squint was not used.. The best fit curves for the two functions
yielded the following positive horn R.A. at the peak:
Model 1: (Gaussian): 19h17m14.82s
Model 2: ( (sin(x)/x)^2): 19h17m14.66s
Applying the 5m9.9s correction for the misapplication
of the horn squint yield the following corrected values:
Model 1 (Gaussian): 19h22m24.72s
Model 2: ( sin(x)/x)^2): 19h22m24.56s
Thus, the two models agree within 0.16 seconds of
time. Using an average of these two models yields a corrected R.A. of
the Wow! source under the positive horn assumption of 19h22m24.64s
Note that the corrected value of 19h22m24.64s is 4.36s
smaller than the corrected R. A. of the 4th data point (the one with
the largest intensity). This makes sense when you view a plot of the
6 data point intensities vs. time. The peak of the best-fit curve
must be in between the 3rd and 4th data points but closer to the 4th
data point.
By the way, a calculation of the residuals for
each function showed that the Gaussian was a slightly better fit than
the (sin(x)/x)^2 model, although the differences were small (in fact,
for 3 of the 6 data values the Gaussian had the smaller residuals
while the reverse occurred for the other 3 of 6 data values).
Determination of Corrected R.A. Assuming
Negative Horn Received Signal
Now let's determine what
the R.A. would have been under the assumption that the signal came in
the negative horn. In the 1980 report by Debbie Cree, she quotes the
location of the east (positive) horn as 4.10 feet west of the focus,
and the location of the west (negative) horn as 8.79 feet west of the
focus. Thus, the difference in distance between the two horns is 4.69
feet (along an east-west line). The focal length of the paraboloidal
reflector is 420 feet. The horn center, the focal point and the
vertex of the paraboloid, all projected onto the ground plane, form a
right triangle. The focal length (420 feet) is the long leg, the horn
offset is the short leg at a right angle to the long leg, and the
hypotenuse is the line from the horn to the vertex. For each horn, we
desire to know the angle opposite the short leg. The difference
between those two angles equals the angle in the sky separating the
peaks of the two beams.
First, lets compute the two angles,
initially in arcminutes, then in seconds of time at the equator, and
finally, in seconds of time at the declination of the Wow! source (-
27.05 degrees). Call the two angles theta_pos and theta_neg.
Negative horn at 8.79 feet: theta_neg = (180/pi)*60*arctan(8.79/420)
= 71.9366 arcminutes.
Positive horn at 4.10 feet: theta_pos = (180/pi)*60*arctan(4.10/420) =
33.5579 arcminutes.
Note that the factor
arctan(offset/focal length) yields the angle in radians, the factor
(180/pi) converts the radians into degrees, and the factor 60
converts degrees into minutes of arc (i.e., arcminutes). Expressing
these results in seconds of R.A. at the equator by multiplying by 4
yields:
Negative horn: theta_neg = 287.75 seconds = 4 minutes 47.75 seconds.
Positive horn: theta_pos = 134.23 seconds = 2 minutes 14.23 seconds.
To convert an angle into
time or R.A. units away from the equator, one must divide by the
cosine of the declination. Using cos(-27.05 degrees) = 0.89061, we
have the following results for the Wow! source: Negative
horn: theta_neg = 323.09 seconds = 5 minutes 23.09 seconds.
Positive horn: theta_pos = 150.72 seconds = 2 minutes 30.72 seconds.
Now we compute the difference between these last
two results to obtain 172.37 seconds = 2 minutes 52.37 seconds as the
R.A. difference between the peaks of the positive and negative horns
for the Wow! source. Because the negative beam goes through a given
radio source before the positive beam does, and because the
calculation in the previous subsection computed the R.A. under the
assumption that the source came through the positive beam, it is
necessary to add this 172.37 second difference to obtain the R.A. for
the assumption of a negative beam detection. Using the best fit value
from the two mathematical functions shown above, that value is:
Negative beam R.A. for Wow! = 19h22m24.64s + 00h02m52.37s =
19h25m17.01s.
Estimated Errors in Computed R. A. and
Declination Values
Before estimating errors in the computed R. A. and declination, let's
restate those epoch 1950 values:
R.A. (positive horn assumption): 19h22m24.64s
R.A. (negative horn assumption): 19h25m17.01s
Declination: -27d03m
Let's deal with declination first, because it is the simplest. The
horn offset in declination (for each horn) was 1 degree (or 60
arcminutes), as accurately as we could measure it; this corresponded
to the centers of the horns being about 7 1/3 feet above ground. A
horn above ground makes less of an angle with respect to a horizontal
line from the center of the paraboloid to the point on the flat at
the same height above ground, and also a smaller angle of incidence
to the flat reflector than would a horn located at ground level.
Thus, the effect of the horn squint of 1 degree in declination means
that 1 degree needed to be subtracted from the declination setting (-
26d00m for the Wow! source) to obtain the squint-corrected
declination of -27d00m for the time of the observation. Applying the
precession and other corrections to convert to epoch 1950 yielded the
declination of -27d03m, the same as was shown on the computer
printout.
I estimate the error in the declination squint to
be about 1 arcminute. However, there is a much larger source of
error. Since Wow! was observed only one time (at only one
declination, of course), there was (and is) no way to estimate the
declination by comparing the source strength at other declinations.
Normally, as was routinely the case with the continuum sources in the
Ohio Sky Survey, observations at 20 arcminutes above and 20
arcminutes below the declination that gave the largest intensity
permitted a calculation of the declination where the peak intensity
would have been observed. [Note that the half-power beamwidth = HPBW
was 40 arcminutes; choosing one half of the HPBW (or 20 arcminutes)
to be the standard increment in moving the telescope in declination
yielded the fastest possible survey while still maintaining the
ability to accurately determine the declination of sources visible at
two or more adjacent declinations.] So for the Wow! source, seen at
only one declination, it is reasonable to assign an uncertainty
(error) in declination position of 20 arcminutes. By the way, since
the squint error and the error due to seeing the source at only one
declination are independent, the statistical procedure of taking the
square root of the sum of the squares of the independent errors
yields: square root (20*20 + 1*1) = square root (401) = 20.025
arcminutes. Since this is so close to 20 and since the component
error of 20 arcminutes itself was an estimate, it is OK to state that
the error in declination is 20 arcminutes.
Now let's deal
with the R.A. errors. First let's consider the error in the squint of
the two horns. In the above calculations I used the horn squint for
the positive horn as -138/cosine(declination). This value was based
on many measurements of sources with known R. A. in the Ohio Sky
Survey and was appropriate for the Wow! source measurements because
the positive horn was not moved between the period of the Ohio Sky
Survey and the occurrence of the Wow! signal.
However, about three years after the Wow! source occurrence, Debbie
Cree
measured the physical location of the positive and negative horns as
8.79 feet west and 4.10 feet west of the focus, respectively. As far
as we can remember, the positive horn was not moved during those
three years between the Wow! source occurrence and Debbie Cree's
measurements. However, her measurements do yield a slightly different
positive horn squint in R.A.
Recall from above, I
calculated that the 4.10 foot offset of the positive horn would yield
a R.A. squint of -134.23s at the equator or -150.72s at the Wow!
source declination. Compare these with the adopted value (from the
Ohio Sky Survey) of -138s at the equator or -154.95s at the Wow!
source declination. The difference between -150.72s and -154.95s is
4.23s. Having applied the R.A. squint in the wrong direction, I had
to double the squint and subtract to correct for the error. If I were
to use Debbie Cree's measurements and the squint derived from those
measurements, I would have to subtract twice 4.23s from my previously
stated R.A.s (both positive horn and negative horn) for the Wow!
source. Rather than adopt Debbie Cree's measurements and the
assumption that the focus is where she thought it was, I choose to
use the -138/cosine(declination) calculation but assign any
differences into the error. Thus, one component of the error in R.A.
will be taken as 2*4.23s = 8.46s.
A second component of
error occurs with uncertainty in the sidereal clock read by the
computer and used as the basis for all position measurements (except
declination) and for Eastern Standard Time (which was computed from
sidereal time). The clock that was in use during the SETI program had
been used throughout the Ohio Sky Survey where it had kept good time.
However, as it grew older, it became less reliable. Occasionally, we
would notice that it was off by as much as 2 seconds of time (very
large for a precision astronomical clock). Thus, I will assign an
error of 2 s for this second error.
A third component of
error is the measurement error due to the size of the beam in R. A.
At the equator the beam size (half-power beamwidth = HPBW) is 8
arcminutes. At the equator this converts to 32 seconds of R.A., and
at the Wow! declination it converts to 35.93s. I estimate that a
measurement error of 1 arcminute could arise for a source with the
strength of Wow!. Converting this into seconds of R.A. at Wow!'s
declination we have a value for this third error of 4.49s.
Thus, assigning independent errors of 8.46s, 2s, and 4.49s yields a
combined error of: square root (8.46*8.46 + 2*2 + 4.49*4.49) = 9.78s.
Because of the various uncertainties, I will call the total error 10s
and will round all R.A. valu es to the nearest second.
Summarizing, we have the corrected and final R.A.s and declination
for the Wow! source with their estimated errors as follows:
R.A. (positive horn): 19h22m25s +/- 10s
Conversion of Right Ascension and Declination
to Epoch 2000
The two values of right
ascension (for the two horns) and the value of declination for the
Wow! signal shown at the end of the last section were based on epoch
1950. Since it is near the year 2000, most astronomers are now
reporting the celestial coordinates of objects using the epoch 2000.
Thus, I will convert the above coordinates into epoch 2000 values.
Because of the size of the errors (+/- 10s in right ascension and +/-
20m in declination), I will simplify the computation to consider only
precession taking into account only the first order terms. Nutation
and aberration plus higher-order terms of precession would need to be
taken into account if our precision were better than 1 second of time
or a few seconds of arc.
The expressions I will use are as
follows:
Delta_R.A. is the expression for the additive
change in right ascension for one year of precession, measured in
seconds of time (or seconds of R.A.). Delta_dec is the expression for
the additive change in declination for one year of precession,
measured in seconds of arc. Trig functions of sine (sin), cosine
(cos) and tangent (tan) are used. The parameters "m" (measured in
seconds of R.A.) and "n" (measured in seconds of arc) are computed as
linear functions of T, the number of tropical centuries from the year
1900 involved in the change. Because we are going from epoch 1950 to
epoch 2000, I will use the average values of m and n for the average
epoch of 1975 (which is 0.75 tropical century from 1900). Doing
the computations for m and n, we have: Now computing
delta_R.A. we have for the two horns:
Since delta_dec involves right ascension, I
will compute delta_dec for both the positive horn and the negative
horn. The results are:
Now multiplying each of these by 50 years, the total
precessional corrections to be added to R.A. and declination,
respectively, are:
Positive horn: Negative horn:
Now adding
these corrections to the epoch 1950 positions, using the approximate
values because of the large error bars, we have as the epoch 2000
coordinates of Wow! the following:
R.A. (positive horn):
19h22m25s +/- 10s +3m6s = 19h25m31s +/- 10s
Galactic Latitude and Galactic
Longitude
Since the
computed R.A. for the positive horn on the computer printout was
wrong, and since I have obtained a corrected value for it as well as
for the R. A. for the negative horn, the printed galactic coordinates
need to be recomputed. I will do this by simply differences.
Looking at the computer printout, I record below the galactic
latitude and galactic longitude for the two printed rows having R.A.s
of 19h13m00s and 19h18m00s, respectively.
Thus, when R. A.
increases by 5m, the galactic latitude decreases by 1.03d and the
galactic longitude increases by 0.44d. Applying these rates linearly
(OK for the small changes in R.A.), the corrected and deduced R.A.s
for the two horns yield corrected galactic latitudes and longitudes
as shown in the table below.
Since Eastern Standard Time (EST) was computed
directly from the date and the sidereal time (read from the sidereal
clock), the error in applying the horn squint in R.A. did not affect
EST. However, from the best fit analysis referred to above, the
computed peak of the Wow! source occurred 4.36s prior to the time of
the 4th data point. Also, the EST on the printout referred to the end
of the integration interval rather than the middle of that interval.
Thus, we should subtract 4.36s to account for the peak of the source
and subtract another 5s to shift from the end to the middle of the
integration interval. Doing so results in the following EST for the
peak of the Wow! source: 22h16m10s - 4.36s -5s = 22h16m00.64s =
approximately 22h16m01s (or 10:16:01 pm). Since Eastern Daylight
Savings Time (EDT) was in effect at the time, the Wow! source peak
occurred at about 11:16:01 pm EDT.
In the above subsection
entitled "2nd L.O. Frequency (and Frequency of Observation)" under
the section "Computer Printout", the frequency band in which Wow!
occurred was calculated. Since the calculation of the observing
frequency (specifically, the setting of the 2nd L.O. frequency) was
based on the date and the sidereal clock, there is no need to redo
the calculation I did earlier; that is, the R.A. horn squint error
had no effect on the calculation of the observing frequency.
Vast Conclusions from "Half-Vast" Data
As an aside, the above discussions and calculations should
provide ample evidence that a person not familiar with all of the
special knowledge about a particular instrument should not try to
draw too many conclusions from printed data. Such data typically
contains certain assumptions about the equipment not necessarily
known to outsiders.
In the above subsection entitled "Determination of Corrected
R.A. Assuming Positive Horn Received Signal" under the section
entitled "Analyses of Wow! to Correct Errors", reference was made to
fitting two mathematical models (Gaussian and (sin(x)/x)^2) to the
Wow! data. I gave each of several variations of this fitting the
general name WOWFIT. Not only was the position of the peak found, the
half-power beamwidth (HPBW), the peak intensity, and a measure of the
goodness of fit called the "error sum of squares" (typically denoted
in statistics by the notation "SSE"). In the variation of WOWFIT
called WOWFIT6P, I allowed each of the 6 data points to be adjusted
either up or down by 1 unit or else remain unchanged. That meant 3
possible states for each of the 6 data points. This generated
3*3*3*3*3*3 = 3^6 = 729 cases for each of the two models. Before
making any adjustment to a data point, each of the original data
points had been incremented by 0.5 to account for the truncation
error caused by chopping off (truncating) the actual intensity value
to the integer portion so that a single character could be used on
the computer printout for each intensity for each channel. An
iteration (i.e., trial and error) procedure was used to obtain the
best-fit curve to the adjusted data because three parameters had to
be determined (location of the peak, amplitude (intensity) of the
peak, and HPBW), and a direct solution was not possible. Typically,
it took between 4 and 7 iterations to zero in on a solution.
The first case considered was the one where none of the six data
points was adjusted (except for the truncation error adjustment
applied in all cases to all six data points). For this case, the
Gaussian gave a slightly better fit (SSE = 7.525) than the
(sin(x)/x)^2 model (SSE= 10.542). The results of this case for the
Gaussian are as follows:
Location = 14.82s (corresponding
to a corrected epoch 1950 R. A. assuming the positive horn of
19h22m24.72s;
Amplitude = 30.76 (meaning the signal-to-noise
ratio at the peak (S/N) was 30.76); and
HPBW = 38.62s (at the
declination of Wow! (-27d03m); converting this to the equator
(declination = 0d) yields 34.395s = 8.599 arcminutes.
For comparison, the case that yielded the best fit allowing
adjustments
of the data was one in which the 2nd, 3rd and 6th data points were
each incremented by 1, while the 1st, 4th and 5th data points were
left unadjusted. The value of SSE for this case was only 0.321 (in
comparison with the value of 7.525 for the case where no adjustment
was made), meaning that almost a perfect fit was achieved). The
corresponding location, amplitude, and HPBW are, respectively:
Location = 14.28s, Amplitude = 30.53, and HPBW = 39.07s. My
conclusion here is that just a relatively minor change in 3 of the 6
data point values causes a significantly better fit, although the fit
of the original data was already excellent.
I should note
that the best fit using the (sin(x)/x)^2 model was somewhat worse
(SSE = 1.451) than the best fit with a Gaussian (SSE = 0.321).
In Which Horn Did Wow! Enter? Use of OY372 Data
for Antenna Pattern Fits
Data from June 16, 1994 on
the strong point source OY372 (flux density of 11.53 janskys (Jy))
were provided to me by Russ Childers (who has been conducting the
current LOBES narrowband survey and a concurrent repeat of the
wideband Ohio Sky Survey). Using both the negative horn and positive
horn responses of OY372, I made three comparisons of the antenna
patterns normalized to a peak amplitude of unity (1.0) at the
equator. I computed a cross-correlation factor (CCF), also known as a
correlation coefficient. If a CCF = 0, then there is no correlation
between the two sets of data. On the other hand, if the CCF = 1,
there is perfect direct correlation between the two sets of data
(i.e., the shape of the two curves is identical).
The following table shows the three comparisons made. The CCF is the
cross-correlation factor (correlation coefficient) and the SSE is the
"error sum of squares" (the sum of the squares of the differences
between corresponding data points):
All three CCFs are above 0.99 indicating almost
perfect correlations; graphs of the three beam patterns confirm the
conclusion that the beam patterns are almost identical. The negative
and positive horn beam patterns have virtually identical shapes
(although the positive horn had about a 10% greater amplitude and a
2.6% wider HPBW than the negative horn). The CCFs between Wow! and
the negative and positive horns are very close (99.05% and 99.19%,
respectively). Statistically, there is no significant difference
between those two CCFs. In other words, it is not possible, on the
basis of this OY372 data, using beamshape as a parameter, to
determine in which horn the Wow! signal entered.
There has been much
discussion at the Ohio State University Radio Observatory about the flux density of
the Wow! signal. Russ Childers used one method to compute it and
obtained the value of 212 Jy, while I used a second method and
obtained 54 Jy. Each method was independent of the other method, but
also each method had its own set of assumptions. In reviewing both
methods, I find no fault with Russ's method, but I feel that my
method is also correct. The ratio between 212 Jy and 54 Jy is over
3.9; that is much too large a discrepancy to be explained as simply
measurement error. There is some significant problem with one or both
methods, but we have not been able to resolve the discrepancy.
Comments need to be made about the interpretation of either
the 212 Jy or the 54 Jy figure. Since the Wow! signal was received in
only one channel of width 10 kHz (0.01 MHz), the flux density,
whatever its value, can only be interpreted as the average energy
(measured in units of 10^-26 watts) received by 1 square meter of
Big Ear antenna surface in a 1-Hz band somewhere within the 10 kHz
channel. The flux density has no meaning outside the 10 kHz channel
because it was a narrowband source seen only in that channel, not a
wideband (continuum) source.
Some persons have raised the topic of
sidelobes for the Wow! signal, so let me comment on that topic.
What are sidelobes? The antenna pattern response in the one
dimension of right ascension, for a point source located at the same
declination as the telescope is set, has the following properties. It
has a main beam that peaks when exactly on the source and falls off
to smaller intensities more or less symmetrically on either side as
the beam points further away from the source. The shape of this main
beam for the portion where the intensity goes from 100% of the peak
down to a bit below 50% of the peak (50% of the peak = half power)
can be represented quite well by a Gaussian curve (also known as a
normal curve or a bell-shaped curve) or almost as well by the
function (sin(x)/x)^2, as was shown by my WOWFIT analysis described
above. When we go well below 50% of the peak intensity, and
especially in the range of 10% and below, there is a significant
departure from the normal curve. A strong radio source shows minor
beams (i.e., bumps in intensity) on both sides of the main beam which
tend to be more or less symmetrical from one side to the other. The
first of these bumps on each side tends to be the highest, with
subsequent ones getting smaller the further out we go. These "bumps"
are called sidelobes (meaning minor lobes off to the side of the main
lobe or main beam).
Measurements made in the days of the Ohio
Sky Survey showed that the peak intensities of the highest sidelobes
were about 0.5% of the height of the peak of the main beam. The value
of 0.5% = 0.005 = 1/200 is often converted into decibels and stated
as "-23 dB" or " 23 dB down" (computed as 10*log(0.005), meaning the
peak intensity of such a sidelobe is 0.005 that of the peak of the
main beam). Almost 30 years later, using the June 1994 data on the
11.53 Jy source OY372 (referred to above), I saw a somewhat different
pattern of sidelobes. The first sidelobe on each side of both the
positive horn response and the negative horn response, instead of
reaching a minor peak 23 dB down instead reached a plateau (a level
area) only about 10 dB down (an intensity of 10% or so of the main
peak). We wondered whether something had happened to the reflectors
or the horns in the intervening 30 years. We don't have an answer to
that question yet (and it now becomes a moot point as the telescope
is soon to be destroyed by the golf course developers).
In the above two paragraphs I was talking about a one-dimensional main
beam and sidelobe pattern. A similar pattern occurs in the
declination coordinate as well. How could the sidelobe pattern in
declination be relevant to the Wow! signal? Since Wow! was only seen
once (at one declination setting), we have little ability to
determine the actual declination of the source sending the signal.
Since our antenna pattern has a main beam with an HPBW of 40
arcminutes in declination plus a whole series of sidelobes both
higher and lower in declination, there is a great uncertainty of
where, in declination, the Wow! source was located. Of highest
probability would be the declination range within 20 arcminutes
either side of the declination setting of the telescope (i.e., with
the HPBW). The next highest probability would be from the half-power
level out to where the intensity of the main beam has dropped to
about 10% of the peak. An even lower probability would be assigned to
Wow! coming in the sidelobes. I deduced that the flux density of Wow!
was about 54 Jy (see the section above) based on the assumption that
the declination of Wow! was exactly the same as the setting of the
telescope. If the source generating the Wow! signal were in the main
beam but at a level where the antenna pattern was down 10 dB from the
peak (at an intensity of 0.1 of the peak), the deduced flux density
would have been 54/0.1 = 540 Jy. If the source generating the Wow!
signal were in a sidelobe at a level where the antenna pattern was
down 23 dB from the peak (at an intensity of 0.005 of the peak), the
deduced flux density would have been 54/0.005 = 10,800 Jy. [Note
however, that from WOWFIT, the half-power beam width of Wow!
corresponded very closely to the main beam width expected from a
point source. A sidelobe has a width about one half that of the main
beam. Thus, either the Wow! source was an extended source that came
in a sidelobe or else it came in the main beam; the latter of these
possibilities is the more likely.]
I have been told that some
people think there are sidelobes of the Wow! signal showing up on the
computer printout. I don't think so. The peak intensity of Wow! is
about 30.76 sigma (from WOWFIT) corresponding to the character "U" in
channel 2 on that printout. A sidelobe that is 10 dB down should then
show up as an intensity of 0.1 * 30 = 3 in channel 2. However, an
intensity less than 4 is considered to be in the noise and not
reliable as a significant signal. Similarly, a sidelobe that is 23 dB
down should then show up as an intensity of 0.005 * 30 = 0.15 (a
blank) in channel 2 (clearly in the noise). A sidelobe of a main-beam
response in channel 2 must itself also be in channel 2, unless the
frequency of the source or our receiving frequency were changing
rapidly;.we know the latter was not true and the printout provides
evidence that the former was not true either. Looking at the computer
printout there are isolated intensity values of one 5, two 6s and one
7 near or coincident in time with Wow!. None of these are in channel
2. One 6 (in channel 7) occurs at the same time as the channel-2
"Q"and the 7 (in channel 16) occurs at the same time as the channel-2
"U". Sidelobes do not generate simultaneous signals in other
channels, since sidelobes, by definition, occur both before and after
the main beam response. Having looked carefully at the computer
printout, I see no evidence of sidelobes; the printout supports the
calculations that say sidelobes should not be visible because they
should be buried in the noise. It is unfortunate that Wow!,
although strong, was not strong enough to show sidelobes. It is known
that when a horn is offset from the focus, the main beam and the
sidelobes develop asymmetries with respect to the time of the peak
(i.e., the main beam no longer looks like a symmetrical normal curve
but more like a distorted normal curve). The further a horn is offset
from the focus, the greater are the asymmetries (e.g., corresponding
sidelobes on opposite sides of the main beam are noticeably different
in amplitude). Thus, if Wow! had been strong enough to show
asymmetrical sidelobes, we could have compared those sidelobes to
ones obtained in both horns from very strong point sources, and we
would might have been able to deduce in which horn the signal was
received. The closest we came in seeing sidelobes was the
sequence of "11" for the second and third points in channel two
following the last of the six data values (viz., the "5"). The
location of these data points is about where we would expect to see
the first sidelobe, although the data points on the other side of the
peak at the same distance have intensities represented by blanks. An
intensity of 1 sigma is, by definition, noise. As you look at the
computer printout, you see many isolated values and sequences of
blanks, 1s and 2s. These all represent noise. An isolated intensity
of 3 or even a sequence of two 3s is still mostly noise because
either of those can occur randomly with a probability high enough so
that you would expect to see them several times within a few pages of
printout. It is also important to remember that the
computations for updating the baseline and rms values generate
relatively slow changing values of those two parameters for each
channel. If, something in the receiver (say, the gain) changed
rapidly, the baseline and rms values would not adapt rapidly enough
to capture all of that change. This could cause a momentary higher or
lower intensity on the printout for a given channel. So some of the
data on the printout may be off by 1 or 2 sigmas due to this effect.
However, the Wow! source could only be minimally affected by this
effect because the intensities were high enough to trigger the
cancellation of the baseline and rms updating as the source went
through the beam. Even more importantly, having a sequence of six
data points that rise and then fall in a manner that yields over a
99% correlation coefficient with the expected antenna pattern gives a
very high confidence that the data points are very little affected by
any gain fluctuations in the receiver or other similar equipmental
effects.
In conclusion on this matter, I do not see sidelobes
in the Wow! data, nor do I expect to see them.
Intermittency, Duration, and Modulation of
Signal
Several persons have commented about three
related issues: (1) the degree of intermittency (and the related
issue of the duration) of the signal; and (2) whether the Wow! signal
was modulated or unmodulated. Let me give you my thoughts.
How long was the signal present and was it "intermittent"? The
computer printout showed 6 significant data points (with intensities
ranging from 5 up to 30 sigmas). Each data point represented 10
seconds of data acquisition plus about 2 seconds of computer
analysis. Thus, the signal lasted for about 6 * 12 = 72 seconds. The
very curious thing about this signal was the fact that we should have
seen it twice within a period of about 5 minutes as our two beams
sequentially scanned the source, but we only saw one of the beam
responses. Thus, if the signal came in the negative horn (the first
one to be able to see the source), the signal could not have lasted
more than about 2 minutes - 2.5 minutes or we would have seen it also
in the second horn (positive horn). Similarly, if the signal came in
the positive horn (the second one to be able to see the source), the
signal could also not have lasted more than about 2 minutes - 2.5
minutes or we would have seen it also in the first horn (negative
horn). Thus, based on what I have just said, I would place a limit of
about 2.5 minutes on the duration of the Wow! signal. However, there
are other considerations.
The signal could actually have
been present for up to almost 24 hours earlier than the 2.5 minutes
referred to above because it takes that long for the earth's rotation
to move the beam across a source between one pass and the next pass.
[Note that we know it did not occur about 24 hours later because we
stayed at the same declination (i.e., strip of sky) for the next 30
days or so and didn't see the Wow! signal again. A few years later,
when the same strip of sky was again scanned many times, the Wow!
signal was nowhere to be found.]
However, there is still
another factor to consider. The signal could actually have been
present for years (or millennia, for that matter) prior to its
detection for the following reason. Just before the data acquisition
and analysis (i.e., the "run") began, the declination of the
telescope was changed. In the days (and years) previous to August 15,
1977 the radio telescope was not pointed at the declination where
Wow! was seen; thus, we couldn't have detected that signal. I should
note that during the Ohio Sky Survey many years earlier, we did
survey the same declination we did when the Wow! signal was
discovered. However, we were using a wideband receiver (8 MHz
bandwidth). A narrowband signal averaged over a wide bandwidth would
be reduced in intensity so much that it would have been buried in the
noise. Thus, even if Wow! were present then, we wouldn't have seen
it.
Now, let me deal with the term "intermittency". To
me, an intermittent signal is one that is present part of the time
and absent the remainder of the time. The Wow! signal certainly
qualifies. However, it would be wrong to say that the transmitter
sending this signal must have turned off abruptly. After all, if a
transmitter were sending a signal in our direction at the time we
were seeing it but then shifted direction, that transmitter could
still be transmitting but we wouldn't see it. Is the signal
"intermittent" in that case? I think the answer is yes from our
limited point of view, but no from the senders point of view.
Therefore, I need to make sure when someone says the signal was
"intermittent" that I understand what they mean by that term.
In conclusion on this first issue, it remains an open question for
me as to how long the Wow! signal was present, and I don't see any
chance that it can ever be definitively answered.
Now let me comment about the second issue of modulation. One example
of an
unmodulated radio signal is one sent at a constant frequency with a
constant peak amplitude (intensity or energy). An AM or FM radio
station, as it is just coming on the air and before you hear persons
speaking or music being played, is sending an unmodulated signal (and
you will hear a hissing sound from your radio if you turn the volume
up sufficiently). When you hear the voice or music, then you are
receiving a modulated signal. For a modulated AM (amplitude
modulated) radio signal, there is radio energy at each of many
frequencies, with the particular frequencies and the amplitudes of
the energy at those frequencies changing rapidly (many times each
second). For a modulated FM (frequency modulated) radio signal, the
frequency of the output signal keeps changing rapidly although the
amplitude is kept fairly constant. Did the Wow! signal have
modulation?
We collected one data point per channel every
12 seconds and collected a total of only 6 data points for Wow! Any
variation of signal amplitude within the 12-second interval would not
have been detected. The signal could have been varying in any of a
variety of ways and we would not have seen it. Since the pattern of
the 6 intensities followed our antenna pattern so well (with a
correlation coefficient of between 99% and 100%, i.e., almost
perfect), the signal falling on our telescope had an average value
that did not change appreciably over the 72-second observing time.
Saying that the average value didn't change does not tell you
anything about the short- term variations in the signal. The signal
could have been varying (modulated) at a frequency faster than once
every 5 seconds (or 0.2 Hz, corresponding to one half the data
collection period) and we wouldn't have sen that modulation since our
observatory was not equipped to detect such modulation. Also, any
modulation occurring at a frequency slower than once every 144
seconds (about 0.00694 Hz, corresponding to twice the duration of the
72-second Wow! signal) would not have been seen, except for the
following consideration. If we assume that the reason we saw the Wow!
signal in one horn but not in the other horn is due to a very slow
modulation of the on-off type (e.g., on for 200 seconds, then off for
200 seconds, repeating this pattern), we could then attribute what we
saw as a modulated signal (probably representing data). Would an ETI
(extraterrestrial intelligence) send data at such a slow speed if
they had discovered the same laws of physics (electronics) as we but
have a technology far beyond what we have? I don't think so.
In conclusion on this second question, if the Wow! signal was
modulated at a frequency less than 0.00694 Hz (a period longer than
144 seconds) or at a frequency greater than 0.2 Hz (a period shorter
than 5 seconds), we would not have seen that modulation, and hence we
could say that modulation is within the realm of possibility. Outside
that frequency range, I think we would have seen the modulation, if
it existed.
Speculations, Hypotheses, and
Investigations
After I showed
the computer printout of the Wow! source to John Kraus and Bob Dixon,
we immediately talked about it, speculating and making hypotheses.
Quickly, John and Bob began to investigate the various possibilities
(I wasn't heavily involved in this aspect since I was continuing to
examine the incoming data from the telescope). I'll now discuss some
of the possibilities. Some were ruled out and I will state why they
were ruled out. Note that the words "ruled out", in scientific
parlance, means "to assign a very low probability to".
The positions of all of
the planets in our solar system were looked up in an ephemeris (i.e.,
a book that provides information about a wide range of astronomical
phenomena). None of the planets were close to the Wow! source
position. Of course, one would not expect a planet to be generating a
narrowband radio emission. Normally, when a planet is observed in the
radio band, we detect the radio emission over the entire radio band
(assuming the telescope is sensitive enough). That radio emission is
"thermal emission" due to the temperature of the planet. Remember
that every body with substance (mass) generates radio waves
(including human beings). Radio telescopes have detected the thermal
emission from most of the planets plus our moon. Besides the thermal
emission, non-thermal radio emission from Jupiter in the decametric
radio band (i.e., wavelengths of 10s of meters) was first detected
from the early days of radio astronomy. This emission was moderately
narrowband and occurred from charged particles moving in the magnetic
field of Jupiter. So, not only did the Wow! source emission not fit
the pattern of this Jupiter-style emission nor the thermal-type
emission, but, in addition, none of the planets were in the proper
position in the sky.
Asteroids are essentially small planets. Hence, they have
negligible magnetic fields and hence negligible non-thermal
radiation. Since their masses and surface areas are so much smaller
than our planets, they generate much less thermal radiation. However,
the ephemeris was consulted for the locations of some of the larger
asteroids, but none were in the vicinity.
If a satellite from
the U.S. or Soviet Union or other country were broadcasting around
1420 MHz, the Big Ear would have been easily able to detect it when
it was in the beam. The frequency band around 1420 MHz (a few MHz on
either side) was declared off limits for satellite transmission or
earth-based broadcasting over the entire world. Thus, no satellite
should have been sending out any transmission in this protected band.
If a satellite were violating this agreement, it is quite possible
for the signal to be narrowband. For example, the AM (amplitude
modulated) radio stations in the frequency range of around 0.5 - 1.6
MHz (500 - 1600 kHz) transmit over a bandwidth of approximately 10
kHz, the same bandwidth as each of the 50 channels in our receiver.
[Note that the bandwidths of FM radio and television are much wider
than 10 kHz.] An investigation of the orbits of all known satellites
revealed that none were in our beam at the time of the Wow!
source.
There are two major ways to rule out airplanes and other aircraft:
(1) no
aircraft transmitters operate in the protected radio band around 1420
MHz; and (2) aircraft move with respect to the celestial background.
The Wow! source intensity pattern received matched almost perfectly
the pattern expected from a small-angular-diameter (point) radio
source on the "celestial sphere" (i.e., at such a large distance that
there is no perceptible motion relative to the background stars). An
aircraft, which would show a significant motion with respect to the
stars, would also cause the received pattern of intensities to depart
noticeably from that expected for a point source.
A check was made
for known spacecraft and none were near the direction of Wow!. In
addition, a spacecraft is not supposed to be transmitting in the
protected band.
No transmitter on earth or in space
should have been transmitting in the protected band around 1420 MHz.
I have already stated how a transmitter in space (an aircraft, a
satellite, or other nearby spacecraft) would not be able to generate
a point-source type response in our receiver. But how about a ground-
based transmitter?
A ground-based transmitter is fixed to
the ground. The Big Ear radio telescope is also fixed to the ground.
Therefore, even if a signal from such a transmitter were getting
directly into our receivers, there would be no relative motion and
hence, no way to have the signal intensity almost perfectly reproduce
the antenna pattern.
On the other hand, if a ground-based
transmitter were sending a signal out into space and it reflected off
a piece of metallic space debris, couldn't that signal come back into
the Big Ear receiver? The answer is yes! In fact, this hypothesis was
one that I kept in the back of my mind as being slightly possible.
However, now my belief is that it is much less likely than I earlier
thought. For an earth-based signal to be reflected from a piece of
space debris and give us the response that we saw in the Wow! signal,
several things would have to be true: (1) the ground-based
transmitter would have to be transmitting in the protected band
around 1420 MHz (and this is not supposed to be happening); and (2)
the piece of space debris would have to be metallic (very possible),
not tumbling (quite unlikely), and not moving significantly with
respect to the celestial sphere (not likely for nearby debris but
possible for debris not orbiting the earth).
Even though a
ground-based transmitter is not supposed to be transmitting in the
1420 MHz band, it is theoretically possible for a harmonic of a lower
frequency transmission to occur in the 1420 MHz band. For example, if
a transmitter were designed to send a narrowband signal at 710 MHz,
it unavoidably would also send a much weaker version of that signal
at twice that frequency (i.e., 1420 MHz). Similarly, if a transmitter
were designed to send a narrowband signal at 473.33 MHz, it
unavoidably would also send a much weaker version of that signal at
triple that frequency (again, 1420 MHz). In other words, weak signals
are always generated by a transmitter at integer multiples
(harmonics) of the fundamental frequency. Filters are used to lower
the intensity of these harmonics but the intensities cannot be
reduced to zero. Since the Big Ear represented a very sensitive
receiver, it could have detected such harmonics. Note that most of
these fundamental frequencies (e.g., 710 MHz, 473.33 MHz, etc.) occur
in the bands used by television and radio; TV and radio signals are
nearly always much broader in bandwidth than the 10 kHz width signal
of Wow!
In order to generate an intensity response
virtually identical to that of a celestial source of small angular
diameter (point source), a piece of space debris could not be
tumbling except at a very slow rate of one turn every hour or slower,
and it couldn't be moving with respect to the celestial sphere
(background of stars) more than about one arcminute during the 72
seconds the Wow! signal was observed. These two constraints are
uncharacteristic of most space debris. Thus, for the reasons stated
above, I now place a low probability on this alternative as the
explanation for the Wow! source.
When an electromagnetic wave
(such as light or radio waves) travels past a star or galaxy or other
condensation of matter, that wave is deflected slightly. If a radio
source (including a radio beacon from an intelligent civilization)
were located in the same line of sight but further away than this
condensation of matter, it is possible for the waves to be seen (or
imaged) as a ring or multiple points of enhanced light or radio
waves. This phenomenon is called "gravitational lensing". Many
instances of this phenomenon have been reported in recent years, both
in optical and radio images. Could this be involved with the Wow!
source? I think the short answer is "Yes, but ".
Typically, the lensing phenomenon (rings, bright spots, etc.) remain
in the images taken over a period of many days or months or even
years, depending on the motion of the source and the condensed
matter. On the other hand, the Wow! signal, which should have been
seen twice (two beams) in about 5 minutes, was seen only once. The
lensing effect probably would not have changed significantly in 5
minutes. Of course, if Wow! were a signal from an intelligent
civilization, the beings responsible for transmitting the signal
could have directed it to another direction in their sky, or could
have turned off their transmission within the 5-minute period.
When we look at the stars in our sky, we see them "twinkling". That
twinkling is due to each photon coming from the point source
experiencing a slightly different travel path on the way to our eyes
than other photons. The earth's atmosphere accounts for nearly all of
the differences imposed on these photons. We do not see the planets
twinkle because a planet has an observable angular diameter and the
effects applied to the photons from the various directions of the
planet tend to average out.
When radio and optical waves
travel through the interstellar medium (which is somewhat like our
atmosphere except much more rarefied), those waves (photons)
experience a kind of twinkling effect called "interstellar
scintillation". It is possible for there to be an enhancement of the
signal passing through this interstellar medium due to a partial
coherence effect. If this effect did occur for the Wow! source, it
still points to a signal originating many light-years away from us,
thus tending to give more support for the hypothesis of a signal of
an extraterrestrial origin.
Thus, since all of the possibilities of a terrestrial origin
have been either ruled out or seem improbable, and since the
possibility of an extraterrestrial origin has not been able to be
ruled out, I must conclude that an ETI (ExtraTerrestrial
Intelligence) might have sent the signal that we received as the Wow!
source. Of course, being a scientist, I await the reception of
additional signals like the Wow! source that are able to be received
and analyzed by many observatories. Thus, I must state that the
origin of the Wow! signal is still an open question for me. There is
simply too little data to draw many conclusions. In other words, as I
stated above, I choose not to "draw vast conclusions from 'half-vast'
data".
Having near perfect hindsight, what could we or should we have
done differently that would have allowed us to obtain more
information about the Wow! signal when it was received? The
modifications made to the N50CH data acquisition and analysis program
made in the years following Wow! should have been made sooner. Of
course, both Bob Dixon and I were fully employed elsewhere in time-
consuming jobs and the radio observatory work as volunteers was done
after normal working hours and on weekends; thus, the programming
went much slower than it would have if we had both been employed at
the Radio Observatory. Some of the especially important changes done
later that should have been done earlier included: (1) the horn
squint correction error found and corrected; (2) the overprinting of
minus signs for signals coming in the negative horn; and (3) the
search strategy algorithms applied and the detections stored in the
computer.
Another aspect of the computer programming is the
documentation. Because Bob and I were so busy with our regular jobs,
we see, in hindsight, that we were not careful enough to document all
of the changes to the computer program. We were anxious to make
changes to improve the program and to get the observing program back
on the air so that we would lose as little observing time as
possible. As a result, we did not always print out the latest version
of each subroutine or main program and organize that printout into a
chronological set of manuals. We also did not write up ongoing
summaries of the major changes to the software in a separate book.
Because of this, I had a difficult time trying to reconstruct in my
mind what attributes of the software were in existence at the time
the Wow! signal occurred. Although I had a complete listing of the
software, it was made in early 1983, about 5 1/2 years after the Wow!
occurrence, and after many significant changes to the N50CH sampling
and analysis program had been made.
The feed horn tracking
system (movable cart on which the dual feed horns were mounted),
although discussed in the early days, didn't get implemented until
just a few years ago. Being able to track the Wow! source might have
given more information about it.
Obtaining the 4-million
channel SERENDIP receiver a few years ago with its 0.6-Hz channel
widths, would have been very valuable for Wow! Unfortunately, the
technology at that time was not capable of allowing a SERENDIP
receiver to be built, although a receiver with channels much narrower
than 10 kHz was within the "state of the art" (all we needed were the
right volunteers, time and money, all of which were in short supply,
especially the money).
Also needed was greater computing
power at the observatory site. Although the IBM 1130 computer was
capable for its time, we were maxing out its capabilities. A second
computer to allow concurrent software development and offline
analysis would have been helpful (again money was a problem).
This report has been my attempt to summarize most of the key
information
about the fantastic Wow! signal. Even though, after 20 years from its
occurrence, my memory for all of the details is not complete and may
even be faulty for a few of those details, I hope that you understand
the overall picture and appreciate both the joys of the "Search for
ExtraTerrestrial Intelligence - SETI" and the challenges we faced in
conducting that original search.
Copyright © 1997-2008 Ohio State University Radio Observatory and North American
AstroPhysical Observatory.
|