the default values for the parameters of the scrambler block are the
following:
Mask: 0x8A
Seed: 0x7F
length:7

Is this a 7-bit scrambler? If yes, why the mask corresponds to an 8-bit
number? (0x8A = ‘10001010’)
What is the polynomial that corresponds to this mask?
According to this link (which I found as reference): http://wiki.spench.net/wiki/GNU_Radio_Notes#GLFSR_Source

the mask 0x8A is related with the polynomial x^8+x^4+x^2+1

Is that correct? Do I miss something or is there an error on default
values?

On Fri, May 15, 2015 at 1:32 AM, Marcus Müller [email protected]
wrote:

So, yes, I think I agree, and length should at least be 8 here (your
polynomial has degree 7), and it’s a bit uncommon to see a LFSR being used
without the x^0 coefficient.

Agree. This was a poor implementation decision by someone who clearly
didn’t know what he was doing at the time

I’ve been wanting to fix this for a long time; looks like 3.8 would be a
good chance to break this properly.

HI Thanasis,
first things first: That is gr::digital::scrambler_bb, which has doxygen
documentation [1].
The length is really just the length of the underlying shift register –
which ought to be at least (order of generator poly)+1, ie.
floor(log2(mask))+1
I recommend having a quick glance at its source code [2], revealing that
it’s actually but a block wrapper around gr::digital::lfsr[3].
If I read lfsr’s code correctly, you’re right, the highest bit of the
mask will always be ignored, because it [4] will always be ANDed with a
0:
unsigned char newbit = (popCount( d_shift_register & d_mask
)%2)^(input & 1);
since d_shift_register never sees the 8th bit being set:
d_shift_register = ((d_shift_register>>1) |
(newbit<<d_shift_register_length));

So, yes, I think I agree, and length should at least be 8 here (your
polynomial has degree 7), and it’s a bit uncommon to see a LFSR being
used without the x^0 coefficient.