open-source-silicon.dev

Screenshot from 2021-06-30 11-01-48.png

This is a simple clock buffer with 0.001p step. It's obviously using 0.5ns at points instead.

Xyce uses an adaptive timestepping algorithm. The initial dt is varied throughout the sim. There are some flags that can change that behavior I believe.

But it is supposed to be limited by the tstep as a maximum, right?

I....don't remember that clause in the reference manual. Possibly?

Users may control the maximum internal step size by
specifying the step’s ceiling value in the .TRAN command (see the Xyce Reference Guide [3])

Ah ok. The initial and ceiling vals are different parts of the invocation

Screenshot_20210630-190223.png

Example 2 is the form you want. Where the final value is the step cap

Other spice tend to be different, i think. e.g. ngspice
. tran tstep tstop &lt; tstart &lt; tmax &gt;&gt;

<https://xyce.sandia.gov/downloads/_assets/documents/Reference_Guide.pdf|https://xyce.sandia.gov/downloads/_assets/documents/Reference_Guide.pdf>

You probably have this sorted out, but the time step ceiling is the 4th argument of the .TRAN line.   The 3rd argument is the time at which it starts transient output, I believe.  So usually you’d just have that be zero.

oh, I guess other spice are like that and I've always misunderstood

I’m not sure about all the other spices, but we patterned this after what we saw in Spice3 years ago.

We also added one thing to it, which is the “schedule” option.  This allows different max time steps in different time ranges.   That is also set on the tran line, as an expression.

Glad it got sorted out!  Sorry I couldn’t reply earlier.

As I understand it, the old `.tran &lt;tstep&gt; &lt;tstop&gt; &lt;tstart&gt;` comes from the early days of SPICE when the plots were all text-based. So, `tstep` is how big the time interval will be per line printed by your old needle printer on the old green-and-white paper.

Ah, the bad old days when the printers went BRAAAAAP!... BROOOOP!