in now days i design FIXED POINT FIR that wil be implemented on FPGA
the filtering unit using cic decimation followed by 2 fir LPF.

the input to the unit is 32 bits - 30 fractions and 2 for real number.
the end of the unit in 57 bits and i take just the fractions -51 downto 22.
when i analyze the FILTER FREQUENCY RESPONSE ( for the whole unit) by FFT to the impulse response - i get this difference between matlab analyze and modelsim

http://images.elektroda.net/84_1336288760.jpg --FPGA RESPONSE
http://images.elektroda.net/3_1336288760.jpg -- MATLAB RESPONE

my question is -
while cut any one bit from the output we probably attenuate the value by 6 db.
i cut the bits 51 downto 21 from the 57 output.
i miss 150 db attenuation from the matlab respone .
does somebody know the meaning of cutting bits from the middle of the output vector?
i mean if i will cut the lower 27 lower bits i will attenuate the frequency respone by 162 db , but i dont know what will be the attenuation when i cutting the middle bits..

i will be very happy for help