[Discuss-gnuradio] Improving Mode-S/ADS-B demodulation using parallel detectors and blocks of samples

[Eric A. Cottrell]

Thinking over the design of the Mode-S/ADS-B receiver I have a possible 
improvement.  It changes the serial method of demodulation to a parallel 
block method.

Mode-S/ADS-B is a smaller bandwidth signal in a wider bandwidth channel 
of 8 MHz.  There is a frequency tolerance of +/- 1 MHz.  Desired signals 
are from different sources and can overlap in time and frequency.  There 
are also undesired signals (ATCRBS) that share the same channel and have 
a frequency tolerance of +/- 3 MHz.  The incoming signal is oversampled 
at eight or ten MSPS.

The recommended designs use serial processing and will only process one 
data transmission at a time.  If two data transmissions overlap then the 
stronger one is selected and the weaker one is considered interfering.  
In one recommended design there is mention of using megabytes of ROM 
data to minimize the effect of the interfering signal on the selected 
signal.  The interfering signal is likely at a different frequency so 
the ROM data is a lookup table based on the the effects of the 
interfering signal on the sample values.  This process is done after the 
AM Detector.  I assume a real ADS-B receiver has a single ADC that 
digitizes the output of a hardware AM Detector which would limit the 
possible methods.

The USRP output is a complex signal.  If I use the recommended designs 
then I am throwing information away by taking the magnitude of the 
complex signal and recreating the information later using the ROM lookup 
tables.

I want to experiment with implementing multiple detectors across the 
channel.  The detectors feed blocks of candidate samples to the 
Mode-S/ADS-B decoder. It seems to me that SNR improves by narrowing the 
detector bandwidth.  The interference effect of multiple signals will be 
reduced as well.  This may require a bit of processing power to 
implement. :)

73 Eric