50g: an interesting RAND anomaly

[DavidM]

(03-17-2018 05:43 PM)J-F Garnier Wrote:  The same happens using the HP-71B ancestor.

Thanks for that confirmation! I suspect all of the HP calculators that use the same RAND algorithm will show these results.

Admittedly, MOD 2/4/8 is simply using the last 1-3 bits of the seed, but I would have still thought the period would have been bigger.

I'm wondering if performing a right-shift on the seed of 1-2 nibbles prior to the MOD would give better results. The range of input values for my application would leave plenty of overhead for that. But I have no idea if that would simply change the parameters with no real benefit.

Does anyone know if there's a documented issue with simply using a few of the least significant bits of the RAND seed? Or have I simply stumbled upon a special case that resonates with the RAND seed generator?