Hi all:
    I have a trouble in design the single bit fourth-order cascade(2-2) sigma-delta modulator.
    the modulator output is: Out(z)=Y1(z)*H1(z)+a*Y1(z)*H1(z)*H2(z)+b*Y2(z)*H2(z),Y1(z) is the output of
the first two order,Y2(z) is the output of the second two order.the digital error cancelling logic was modeled by verilogA.the clock frequency of sampling is 26MHz.I do 4096 point of DFT,input signal frequency is 107.91KHz(m=17).
    When I use spectre to do transient analysis and then do DFT analysis,the result is unstable.
for example:
    if the DFT setup is from 1.935us to 159.467us, the SNR result is about 83dB.
    but change the DFT setup from 1.935-T to 159.467-T(T is the clock period),the result is only about 40dB,the waveform in the low frequency region is very flat.
    But if I use matlab to analysis the same output date,the result is stable in about 83dB.
    I now fall in trouble deeply,I can't make decision which result I should trust in.cadence or matlab? or there
are some mistake I made? please help me! thank you very much.
    forgive my poor English.