How to minimize the noise floor of amplifier

[flamingo]

Hi, guys:
When I simulate amplifier in Hspice using .fft to get linearity performance, I find the output frequency spectrum's noise floor is two high, about -70dB. How can I minimize this value, my ckt is somewhat huge.

 

[flatulent]

This is caused by a basic characteristic of SPICE simulations. You are measuring the convergence errors at each time point of the simulator and not the performance of the amplifier. SPICE is not intended to do noise calculations in the time domain.

 

[tlihu]

SPICE is not intended to do noise calculations in the time domain.

So, which tool is better?

 

[flamingo]

the convergence errors at each time point of the simulator and not the performance of the amplifier. SPICE is not intended to do noise calculations in the time domain.

Are the convergence errors same for all nodes in a ckt when simulated in Hspice? if so, it doesnot agree with the simulation results. It is obvious that input source and the output of pre-stages' .fft results show lower noise floor than final outputof ckt.

 

[flatulent]

for the above two posts, use the ac analysis with the noise option.

In the time domain the convergence closeness is set by two user changeable opttions, relative tolerance and absolute tolerance. Making these smaller will reduce the error but increase the simulation time. The other problem is the accuracy of the FFT caused by limited numerical resolution.