In the ranging experiment, because the signal waveform sampling process is not synchronized to the trigger pulse frequency, there will be a variable time offset when processing signal data from each range scan.
With reference to Greg's Matlab code "read_data_RTI.m",
if ii is the number of the signal data point at which start(ii)>0 then the offset is given by
offset = -s(ii-1)/[s(ii)-s(ii-1)]
The value of the offset should lie in the range 0 < offset < 1
A value of, say, 0.5 means that the trigger waveform had a zero crossing point which is exactly halfway between the occurrence of signal data points ii-1 and i.
Typical variation of the offset is shown in the first picture for a data set measured using my tin can radar.
A simple way of compensating for the offset is to average the signal data points in a range scan using the value of offset as a weight
s_new(ii)=(1. - offset)*s(ii-1) + offset*s(i)
It is also possible to offset the signal data by applying a linear phase shift in the Fourier Transformed domain but the above approach is simpler to understand.
Having looked at both approaches, the effect in practice doesn't seem to matter too much, i.e. the sampling frequency of 44.1 or 48kHz, as used by most people to record the signal data seems adequate.
As Greg has already pointed out, the data processing step which IS effective, however, is 2 pulse cancellation. This tries to remove noise due to stationary radar returns (normally the ranging experiment is only used to "see" moving targets) by subtracting, on a point by point basis, the data from one range scan from that in the next. The effectiveness of this approach is shown in the other two pictures.