[sdiy] Shift register sequence period

[Dave Manley]

Tim Stinchcombe wrote:

> First, having done some calcs, the proof that it *is* non-linear is
> virtually immediate. If you start with a '1' in stage 1, and 0s for the
> rest, at the first shift the feedback is easily seen to be 0, and we then
> shift the only 1 out of the register and a 0 in, i.e. it is now filled with
> *18* zeroes - thus it *cannot* be linear as it would stick in this state for
> ever. 

Hi Tim,

Maybe I'm misunderstanding your scenario, but when the register is all 
zero, the feedback will be ~(0^0^0) = 1.  Here's the sequence going thru 
zero:

                   262137, 000000000000001111
                   262138, 000000000000000111
                   262139, 000000000000000011
                   262140, 000000000000000001
                   262141, 000000000000000000
                   262142, 100000000000000000
                   262143, 110000000000000000
                   262144, 111000000000000000
                   262145, 111100000000000000
                   262146, 111110000000000000


-Dave