You need to keep the windows aligned (reset by the sync pulse). The window valid simply labels one window as valid or not, it does not demarcate the window beginning/end.
If you're using the windowed X engine, you should be getting 16*17/2=528 baselines, each of which has four complex crosspol terms (for a total of 4224 numbers out the engine). I suspect the reason you're getting incorrect outputs is because your windowing scheme is not aligned. Jason On 10 Aug 2011, at 22:40, Jon Losh wrote: > Hello, > > I have a question about feeding the casper xengine block input. My impression > is that the window_valid input line is there so that if we have windows of > data (acc_len * n_ants points of data for a given frequency) that have some > gap in between (because of random delay of 10gbe, let's say), we can tell the > x-engine to ignore the nonsense in between the windows. When I run things in > Simulink, however, I get different answers out, depending on the delay > between windows. The initial correlations are correct, but the ones at the > end vary. > > Also, why am I getting 544 baselines for 16 dual pol antennas? If there are > no redundant multiplies, there should be 32*33/2 = 528. The most likely > answer seems to be that 16 of the baselines are repeated, the AxAy and > AyAx-type ones from what I'm seeing.
