show low_freq cutoff in hspice, help please!!!!

[sunjiao3]

Hi, friends, I am designing a cascoded amplifier. And I wish to evaluate its bandwidth and gain of every stage(differential). I use commands as follows:

.probe AC VOUTDB=PAR('DB(V(out1)+V(out2))-DB(V(in1)+V(in2))')

And then, I read the gain showen in avanwaves and the 3db bandwidth by curser.

Every thing gose well until I want to evaluate the low-frequency cutoff of it. The line seems like 2 steps downwards rather than the proper one of "first rist, then drop". And I have found the circuit is right in other simulators. So, what's wrong with my command?

Here are the avanwaves pictures I got.

 

[sunjiao3]

Dear friends, I changed some options in my netlist and found quite different results.
First, I changed the .option accurate=1
Then, I simulate AC like .AC DEC 1000 1e-6 1G.
The, the first the step of the graph I posted disappeared!!!! It seems to be a flat band from 1e-6 to the -3db frequency, about 137MHz. So, the problem seems come from the simulator and its settings.
So, which one can I trust, and how to set the simulator to get the trustable resutls.
Thank you in advance.

 

[sunjiao3]

I feel disappointed to find that there is no answer at all. By the way, I find that while simulating using smartspice, there is no such "downstair stages" at all. The graph seems right. So, which one I can trust? The result of hspice or smartspice?