archive: Re: SETI Search Schemes

Re: SETI Search Schemes

David Woolley ( david@djwhome.demon.co.uk )
Sun, 30 Aug 1998 18:30:23 +0100 (BST)

[ Requoted to internet conventions ]
Don Adams wrote:
>
> [ Unidentified ] wrote:
> >Chip Cohen wrote:
> > >No Paul; I was referring to the fact that the receive effort is not using
> > >Doppler Invariant Pattern Recognition for the polychromatic FHSS modality.
>
> >What H___ is he talkin' about! Wow what "Techno-babble"........

Often "techno-babble" is necessary to explain something compactly,
when, if one takes the typical software support case, the alternatives
are either "it won't work" or a long a detailed explanation, which may
still be too much to digest.++

> My guess is it's something like a "Cepstrum" which is a

The way I read it is that it is a statement of a particular problem,
given certain facts and assumptions, not a statement about the solution
to that problem, although I think people could reasonably be assumed to
infer that there is one from what was written.

As Chip is declining to interpret it, I'll have a go at my interpretation,
(note that I think Don probably has a similar reading, but commented
on the implications, not the literal meaining; also, much of the
interpretation may be obvious to those with an average science
background).

I think that FHSS stands for frequency hopping spread spectrum (more
below) and that "modality" is a noise word (the dictionary doesn't
help here) - maybe replace "modality" by "method of operation".

"polychromatic" literally means of many colours. As light and radio
are really the same thing but with differing wavelengths (for light
it is around the thickness of an oil film on water, for radio it is
of a similar size to that of the components of a (non-dish) antenna),
polychromatic would seem to indicate listening at multiple wavelengths.
(For radio communications, frequency is often more useful than wavelength,
but, as both travel at the speed of light, the frequency at which waves
pass is simply the wavelength divided by the speed of light.)

"pattern recognition" refers to the fact that to detect an ETI signal we
have to recognize something about it that makes it look different from
all the other signals and noise we are receiving. If you tune an AM,
broadcast radio, you use the strength of the signal (a simple pattern)
and that fact that it sounds like speech or music (easy for the brain,
but rather more difficult to automate) for pattern recognition.

Knowing what pattern to look for in SETI is much more difficult.
The simple approach looks for a single wavelength/frequency which
is stronger than the surrounding ones and where the frequency/wavelength
is fixed or varying smoothly with time.

"doppler invariant" means that you must be able to recognize this
pattern even though the frequency is varying with time, in a way that
we don't know in advance. This variation happens because the receiver
is accelerating with respect to the sender as a result of the motion of
their planets etc. If you are on the earth and the source is due east
of you, the rotation of the earth will be taking you towards the source,
so you will see a shorter wavelength (shorter by the distance you have
travelled between the waves). This is called the Doppler effect, and the
change in frequency/wavelength, the doppler shift. As the earth rotates
so the source is at its highest in the sky, you are moving across the
direction of the signal, so the waves are not shortened or lengthened.
During this six hour period the frequency will have changed. (Note that
the rate of change is itself changing, although I've largely ignored
this below.)

A similar effect can happen at the other end. We know the effect from
the earth, but we don't know that from the other end.

The approach used by the current generation project Argus software is
to try to recognize the signal over a time period short enough, and
over a range of frequencies wide enough that the Doppler effect has
not changed the frequency too much in that period; this results in a
table which is typically shown with frequency ranges horizontally and
time slots vertically. Only once that is done is there any Doppler
independence attempted, in that we then look for any signal whose
frequency is smoothly varying with time (I'm not sure how good the
current software is at doing this).

This doesn't detect signals showing more random variation in doppler
shift. On the other hand, the demodulation (conversion to audio
frequency) of an AM broadcast signal is quite doppler invariant, in
that it uses one of the frequencies in the signal (carrier frequency)
as a reference for the others, so the whole transmission can vary in
frequency without compromising the ability to demodulate it (this is
an oversimplification - the doppler invariant property is actually the
symmetry of the signal around the carrier frequency). The tuning process
in an AM radio is not doppler invariant. I'm not sure whether Chip's
concern is with smooth variations or random ones, but will stick with
smooth ones for the moment.

>From the point of view of trying to detect a single frequency
extra-terrestial signal (if it is truly single frequency, the only
information it really conveys is that there is something tranmitting
in a particular direction), there are a lot of things that can vary.
Obviously the direction of the source can vary. So can the frequency,
and so can the rate at which the frequency changes. The frequency can
vary because of different choices in sending frequency and because of
the doppler effect.

In another article, I think Chip used the term "phase space". I think
this just means the characterstics of the signal that can vary
independently.

If we knew direction, frequency and the way the frequency changed with
time, we could detect extremely weak signals, because we could measure
the signal for a very long time and demonstrate that it was stronger
than the surrounding noise. However, we don't know any of these, and
have to try and search all of them at once. For frequency, we guess
likely centre values, then use software that tries to efficiently work
out what is happening on a lot of frequencies at once. Current software
has to divide up the range into fixed slots, and moreover, those slots
can't be too narrow because of the doppler effect.

As indicated above, the Argus approach only very partially deals with the
problem of the Doppler effect. The seti@home project make a much more
intensive approach, by running the software multiple times with different
values for the rate of change of frequency. However, this is achieving
doppler independence by brute force; what one would ideally like is a
mechanism that can detect a doppler modified signal in noise without
knowing or guessing the way that it is doppler shifted. I don't know
if Chip has an efficient way of doing this, or is just suggesting that
it is necessary to look for such things.

The above has been based on searching for a single frequency signal.
It has been demonstrated that, for a signal corrupted only by an
idealised form of noise, you get the most accurate transmission for
a given level of noise if you send a signal that is almost like noise
itself, and spread over a wide range of frequencies. It is also the
case that spreading over a wide range of frequencies tends to result in
a signal that is less damaged by delayed copies of itself.

If you have a single frequency signal, and you only receive two
reflections where one has travelled exactly half a wavelength more, the
two cancel each other out. At cellular telephone frequencies, this can
happen because of reflections off buildings; on short wave radio links
this happens because the signals are being reflected by moving areas in
the upper atmosphere. Chip also argues that variations of the material
on the path through space will cause similar effects.

For these reasons, and because it is possible to construct such signals
which are difficult to jam in a military context, "spread spectrum"
techniques are used for some modern radio systems. Spread spectrum means
that they use a wider range of frequencies (the light analogy again,
than would be implied by the information being sent). There are two
fairly common systems:

- direct sequence, which I won't explain, but is used by some cellular
telephones, the GPS satellite navigation system, etc., and is allowed
for use by amateur radio operators, in the USA;

- frequency hopping, where the signal is sent for a short time on one
frequency then moved to an other frequency, in a way that is apparently
irregular, but but is known to to both sender and receiver.

My impression is that frequency hopping is mainly used by the military,
but it may well also be more tolerant to variations in the path length
and multiple reception paths, than, say, direct sequence. I think Chip
is suggesting that any established interstellar communication link
would use this method, because it is the most ideal method given our
current state of knowledge; I have reservations, because both parties
have to understand and agree on the parameters of the method, and each
exchange between them could take hundreds of years, so you wouldn't know
whether the other side had understood for a long time. Also, a temporary
collapse in the receiving civilisation (a dark age) might mean they lost
the knowledge needed to find the signal and the sender might take a few
hundred years to discover that the connection had failed - as a result,
it might be prudent to transmit an easy to detect signal containing the
parameters for the more complex one, at regular intervals.

My guess is that project Argus is more likely to find someone who wants
to be found, and I would expect them to transmit a very simple signal.

To complicate this, I did an AltaVista search for '+polychromatic +SETI'.
I found one false positive (on the film Independence Day), two hits
referring to Chip's articles in back issues of SetiQuest magazine and
one to a review of a lecture by Chip. The magazine references require
the magazine to find any details. The lecture reference contains
one paragraph saying that the concern is actually to do with earth
transmission reflected back from the same material in space that could
distort incoming signals, and, I think, that trying to find evidence
on multiple frequencies is a way of eliminating these. That seems to
conflict with its use in connection with ways of transmitting a signal.
Either way, the evidence points to the term as having been coined by Chip.

The URL for the lecture summary is:

<http://www.seds.org/spaceviews/9510.html#lectsum>

There is another article on SETI issues on the same page. A URL for
the magazine back issue list is:

<http://www.setiquest.com/nback.htm>

(The polychromatic reference from Independence day was to a scene alluding
to "polychromatic lights" in Close Encounters of the Third Kind.)

(Searches for '"aperture engine"' produce 30 false positives, all
apparently relating to a web/database graphics software package (I'm
using those terms to hint to those who know, not to explain to those who
don't), whose home site is www.aperture.com. There were no good hits,
so this looks like another coined term - but note that today's coined
terms can become tomorrow's everyday words.)

Trying to put this together, the phrase seems to be saying that project
Argus is not good at handling signals with unknown Doppler shifts,
cannot detect the sort of transmission likely to be used on an established
link and is adversely affected by only looking on one frequency.

I'm having trouble working out whether the polychromatic and FHSS are
correlated, i.e. whether the wide frequency range of the signal doubles to
allow monitoring of multiple frequencies to suppress local interference,
or whether these are two independent aspects of the search strategy.

Nowhere does it actually say how you would perform the search, even if
the articles do.

++ Unfortunately salesmen know that customers don't want to know
technical details so tend to give them a false expectation that it is
possible to explain things simply. E.g., although Windows is sold as
very easy to use, it is actually very complex internally, and when
things go wrong or you stray outside the normal useage, that complexity
can be revealed to someone expecting something that is simple to use to
break in simple ways or handle unusual useage in simple ways.