I have designed a basic 3 part blend for differential length for SPI, now installed on HAM2 X in the supersensor 6 path. This is not meant to be a highly optimized blend filter, I'm still puzzling through how to try to do the modeling for this and I know Sina is working on something that will probably be better. It should be good enough to try turning on and seeing if we can get some improvement in, ferinstance, the IMC. First attached image compares the installed version ("new") to the 250mhz CPS/GS13 blends that we use most of the time for X. 

Second image shows the complementary form of the filter magnitudes and the sum, showing the total for the super sensor is 1 at all the cross-overs. 

This is mostly meant to be a starting point, I expect we will need to do some more development on this.