Bulk and fine delays - orders of
magnitude and practical values.
The bulk delay is obtained by shifting the write pointer of one antenna
relatively to a center value. The read pointers sequence has a fixed
timing which is the
same for all antennas. The fractional part of the delay is performed in
the software by linear ponderation of the phase after the FT and prior
to long term integration.
One increment in address pointer has the effect of delaying the data
presented to the correlators by 16 samples, or 4ns. For +/- 2km the
geometrical delay has to vary +/-6.6 us, or +/- 1650 integer
values . The demultiplexing scheme allows to embed the bulk delay as
long as its excursion is smaller than 1/16th the sector playing
time, or 33us in our case. This is sufficient for PdB and its future
needs.
For +/- 2km the geometrical delay changes at the maximal rate of 0.5
ns/s so at
fastest the integer part will change every 8 seconds. The hardware is
such that changes in bulk delay will be effective at the 32 pps event
next to the time the delay card has received the new command word. The signal
will be correlated during the following 1/32th sec period and the
products read out one period after. The correlator can execute an
integer delay change any time. However,if a coordinated phase
compensation with the LO2 is needed, it might be a good idea to wait
for the next 1pps event to apply the change.
The fractional part of the delay is applied on a whole signal segment
of 1/32th sec. During that period the sky delay is assumed to be
constant (residual<15ps). Whenever the bulk delay value needs
to be changed, the software should first change the integer part of the
delay and later (3 to 4 periods of 1/32th sec) change the corresponding
fractional part, in order to compensate for several latency
times across the system .