A bit later than intended, I'm now in a position to receive GPS locked phase
coded transmissions without the need for any lock up or clock recovery. So I
am looking for anyone who can transmit BPSK or higher level coding, locked to
UTC seconds pulses, with a symbol period that can be any submultiple of two
minutes, eg 5, 10, 12, 15, 20, 24, 30, 40, 60 or 120 seconds. Ideally,
frequency needs to be known, and stable, to a fraction of the symbol period. My
receiver uses a DDS clocked at 10MHz accurate to a few parts in 10^-10, so
carrier frequency resolution is to .00233Hz (429 minutes per cycle). We'll
need to agree a protocol, and start time, and the phase at the start of the
transmission.
The demodulation equipment is as follows :
The 1kHz output from the receiver is passed through a single opamp 100Hz
bandpass filter to a pair of op-amp demodulators driven from a quadrature 1kHz
squarewave. The I and Q outputs are passed through a pair of 10Hz cutoff
Butterworth filters to two 8 bit A/D converters inside a PIC16C71 for
digitisation.
Each second, timed by the GPS pulse, the PIC sums 240 readings of the I/Q
outputs, the effect of the summing being to increase the resolution above the 8
bits inherent in the A/D converter. The summed result, a pair of twos
complement 16 bit numbers, is output along a serial port to a PC. For
completeness, the PIC also reads the NMEA output from the GPS module and uses
this to initialise a modulo 120 seconds counter clocked by the 1 PPS signal.
The value of this counter is also sent with the I/Q data. Summing multiple
asynchronous 8 bit readings gives an increase in resolution approximately equal
to the square root of the number of averaged readings used, so here the
resolution is approximately equal to 12 bit coding - good for over 60dB
dynamic range, particularly when receiving signals buried in noise.
The PC takes in the received I/Q data and seconds count from the COM port, then
depending on the integration time requested, averages the I and Q readings over
the period. It uses the modulo 120 second count transmitted with the data to
synchronise timing. Phase is calcluated by taking the ATAN2 function of the
two integrated channels, and plotted.
So, any signalling period is possible, provided it is a submultiple of two
minutes which I considered to be the longest symbol period likely to be of any
use. It could be increased to 4 minutes with a very minor change to the PIC
software. The integration of separate one second readings is not quite ideal
as an integration method over long symbol durations, but is a reasonable
compromise to keep the PIC code simple and allow symbol times to be made
arbitrary lengths set only in the PC. The next generation of PIC software
will truly integrate over a number of selectable periods.
Andy G4JNT / G8IMR
|
