[sdiy] Fun with LFSRs and such

[Tom Wiltshire]

On 27 Mar 2012, at 22:14, ASSI wrote:

> On Tuesday 27 March 2012, 21:40:16, Tom Wiltshire wrote:
>>> That'd be a non-maximum length LFSR. <snip>
>> 
>> It need not be.
> 
> Yes it need be.  If you want to have more long runs vs. short runs, you must 
> chose a non-maximum length LFSR and the "right" cycle within it

Ok, I think I see what you mean now. The longest run in the maximal cycle is n, which isn't that long. 16 ones or zeros for a 16-bit LFSR. And the distribution is clearly in favour of short runs.

Can non-maximum length cycles produce chains of ones and zeros that are longer than n? Could I get 27 zeros in a row out of a 16-bit LFSR? Or does it only alter the distribution?

Thanks,
Tom