Hi All,
Thanks for the great package!
I am running time reversal using kspaceFirstOrder3d using a sensor mask with 24000 sensors and experimental time series. The grid size is 384x384x180 with 1mm spacing, so max supported f=750kHz. I am using the default PML but changing it makes little difference. I run it on the GeForce GTX 1080 card and an 8 core PC.
The strange thing is that when I run the full recon with all 24000 sensors, it produces distorted images but when I split it up into 100 sets of 240 and then sum/stitch the sub-images, the results are clear, similar to results from standard back-projection. In fact, I determined that using a single sensor which gets a non-zero time series, along with many (>1000) sensors whose time series are all zeros, produces distorted results compared to using the same sensor/time-series but without other sensors present.
So it seems that even though the other sensors have all zeros as input data, they significantly affect the pressure calculations. I thought that the sensors are supposed to be passive....is that an approximation? i.e. Does a large amount >1000 of sensors affect the calculation even tough they have zero time-series amplitude as input?
Thanks,
Ivan Kosik