different transient and ac analysis results

[nassim_el85]

Hi

I have simulated an amplifier using spectre. But when I want to calculate the low-cutoff frequency of the amplifier, it seems that the transient and ac analysis shows different responses. What is the problem? Is it about the simulation settings?

 

[FvM]

Apart from the question how you determine a cut-off frequency from the transient analysis, there are several possible reasons to get a different behaviour:
- bias point in AC analysis is not corresponding with transient analysis conditions
- large signal effects play a role in transient analysis

 

[nassim_el85]

I have find the low-cutoff frequency by 0.7 * (maximum gain) approximately.
- For the first point: The DC analysis shows desired results, also I tried to set some important nodes voltages of the amplifier, but no changes in the results.
Is there any option to specify one point of my transient analysis as the DC point for ac analysis, as can be done in some other softwares like HSPICE?
- For the second point: How can I find it out? I have some kind of non-linear elements in my structure which may cause this problem.

---------- Post added at 12:18 ---------- Previous post was at 11:33 ----------

I have also do the transient analysis using a very low amplitude input signal. The results are worse!!!!

 

[LvW]

What about the usage of initial conditions?
I recommend not to use initial conditions but to let the program compute the bias point accorcing to the dc supply voltages.
In this case both ac and tran anylyses (small amplitudes!) should operate with the same bias point (verify !) and, thus, give the same results - unless you have severe non-linearities within the circuit.

 

[nassim_el85]

I have applied a very small amplitude signal to the input of the amplifier, but the results from ac and transient for the low-cutoff frequency are different yet.
for example, from the ac analysis the low-cutoff frequency is around 10 Hz and the gain of the amplifier in this frequency is 370. But when I apply a 10 Hz input signal to the amplifier, the transient analysis shows a gain 500. The ac and transient analysis shows approximately equal gain results at high frequencies (over 300 Hz).
Something that is more strange is that when I decrease the amplitude of the input signal, the gain increases!

---------- Post added at 12:16 ---------- Previous post was at 11:44 ----------

Next, if the problem is about the nonlinearities within the circuit, how should I calculate the low-cutoff frequency? from the ac analysis or from transient? If I should calculate it from the transient, what amplitude should I applied to the input of the amplifier?

 

[FvM]

cut-off frequency is basically a small signal parameter, which should answer your question

If you want to further investigate the reasons for level dependent gain, you should show the circuit and related waveforms.

 

[nassim_el85]

How should I measure the cutoff frequencies after fabrication? Shouldn't I use a procedure similar to the transient analysis? If I do the design based on the ac analysis, would I have proper characteristics after fabrication?

 

[leo_o2]

Both transient and ac is needed if possible.

 

[dick_freebird]

Transient analyses run with pretty loose (relatively) tolerances and a
high gain circuit can be significantly perturbed by tens of uV numerical
error residue, at the wrong node. You might try jacking up the accuracy
and see if the result moves; if so then you are not accurate enough.

 

[nassim_el85]

I am accurate enough. I think the problem is about the non-linearity.

 

[LvW]

I am accurate enough. I think the problem is about the non-linearity.

Why do you believe this?
I don`t think so because you can choose an amplitude that is as small as you like..