For maximum value, you need maximum possible mantissa and maximum possible exponent.
Since the exponent's maximum value that can be stored is x, but here we are using a bias of q
So, actually exponent value will be b^(x-q)
(You can relate it with the {2^(8) -1} - 127 in case of IEEE, as 2^(8) -1 maximum value (8 bits available for exponent) and 127 is the bias)
Now it is given the significant can contain P digits,
So in normalised form
Expression will look like
(1 + 1/b + 1/b^2 + 1/b^3.........)b^(x-q)
It will evaluate to
b^(x-q)(1-b^(-p))
