Hi,

I want to perform the THD and the DFT analysis of a Source Follower with Spectre simulator, but I don't understand the correct setup of the transient analysys for good accuracy (i.e. reltol, abstol, max timestep, simulation time, etc.). Is the strobing technique necessary to gain in terms of accuracy. The follower has to be used as input buffer in a ADC.

Thanks, Matteo

Thanks a lot for all the clarifications.

Assuming the setup described above; the THD value and DFT show a significant variation with respect the simulation time. How can I choose the end time of the transient analysis? The transient analysis has to be stopped according to the number of samples and the strobing period (slightly beyond the last sample)?

Matteo,

  A source follower, even one driving an ADC is a linear circuit
[just like an op-amp]. There is no need to worrying about sampling
effects, aliasing, etc. Chose the input frequency based on the ADC
requirements, for example, the Nyquist Frequency for a 20MHz ADC
is 10MHz. So use a 10MHz sine wave to test the follower.

The transient stop time for a 10 MHz input frequency is 1/10MHz or
100ns.  

You should use the Spectre error presets to control the simulation,
in general, setting reltol, etc. is an inefficient way to improve
accuracy.

Strobing eliminates the interpolation when the output data points
are selected for the dft. You also can use the Spectre Fourier
analysis. The Fourier integral uses the points that the simulator
selects, see Ken's book for a more detailed discussion).

For a source follower, windowing should not be required because
source follower is not a sampled circuit.

One other thing you might want to try is to perform the DFT across
several cycles. For example, if the transient stop time is 400ns for
the DFT, then there will be three noise bins between each harmonic
bin. This will give you insight into the noise floor of the DFT. This
allows allows you to run transient noise analysis and see the effect
of noise on the buffer.

In the attached plot, I simulated a source follower and plotted the
DFT for simulation setups:
1) 10MHz input, 64 point DFT, tstop=100ns
2) 10MHz input, 64 point DFT, tstop=100ns
   The results are equivalent to the results with strobing, in
   this case I just made maxstep small because the simulation
   is fast, maxstep=100ps (~1/2 * ( 100ns/64))
   This works okay for linear circuits, but makes the simulation
    time for sampled circuits very long, so use strobing for the S/H
3) #1 with the tstop=400ns and 256 point DFT
   This inserts three noise bins between each harmonic
4) #2 with the tstop=400ns and 256 point DFT

                                                         Best Regards,

                                                            Sheldon

Hi,

At the moment the follower drives  a capacitor that will be the SH holding capacitor. The next step on my analysis will be the simulation of the whole front-end (follower+switch+holding capacitor).

Thanks to all, Matteo