SETI public: Just a simple idea...

[Marko Cebokli <s57uuu@hamradio.si>]

Hello!

When studying the ARGUS technical manual on the web, a simple idea came
accross my mind, how to double the frequency coverage and therefore
probability of detection with an extremely simple trick. Please excuse
me if this has been discussed before, but the archives are down, so I
couldn't check.

Looks that a typical ARGUS station is like this: the KK7B type phasing
SSB receiver, modified for 20kHz bandwidth, feeding one channel on a
sound card. The samples are then fed to the real part of the FFT input
array, and zeros to the imaginary part. A complex FFT is done, and so
on. In this setup, we need a delicate analog 90 degrees audio phase
shifter, that requires precison components, etc. Also, half of the FFT's
processing power is wasted. (* note 1) 
I have seen people use DSP to replace the phase shifter and do some
filtering, but not in the following way:

Since virtually all sound cards are two-channel (stereo), my
modification would be like this:
Leave away the 90 deg phase shifter and feed the two (I and Q) channels
into the two inputs of the sound card. Then put the samples from one
channel into the real part of the FFT input array, and the samples from
the other channel into the imaginary part. Do a complex FFT like before.
The FFT itself will provide the required 90 deg shift perfectly, with
mathematical precision! And more: both sidebands will be analysed, well
separated in the output array, representing  40kHz of radio spectrum. (*
note 2)

Let me summarise:

using:
   -less hardware
   -the same amount of software processing
we get:
   -better sideband separation (* note 3)
   -double the frequency coverage.

Now is this a free lunch or what?


73, Marko S57UUU

* note 1: if zeros are put into the imaginary input, only the bottom
half of the FFT's output samples will contain useful data. The upper
half will just be a mirror image of the lower half.

* note 2: The 20kHz filters in the receiver do not need to be modified,
and neither does the sample rate to be increased, since a signal 20kHz
below th LO will give the same 20kHz tone as a signal 20kHz above the LO
frequency. But the FFT magic will put them into different output bins!
Depending on which channel is connected to real/imag, the output of a
1024 point FFT might look like this: output bin #0 coresponds to the LO
frequency. Output bin #1 to the frequency just above it, and so on until
bin #511, that represents the frequency 20kHz (actually Fsampling/2)
above LO. Output bin #1023 will represent the frequency just below the
LO, and so on until bin #512, which represents a frequency 20kHz BELOW
the LO.

* note 3: The 90 deg shift in the FFT is mathematically perfect, so the
sideband separation depend only on the channel phase quadrature and
amplitude balance. It is possible to improve even on that, with an
(possibly automatic/adaptive) software routine that compensates for the
imperfect quadrature/balance of the analog circuitry, pushing separation
to 60 and more dB. Although, for a SETI application, this amount of
separation is probably not needed.