Evas Sim

EVAS Simulation

Use this skill only when visible/public EVAS simulation is allowed. Do not use hidden files, hidden checks, or reconstructed hidden logic. Use veriloga for writing the .va module itself.

Applicability

EVAS is for voltage-domain event-driven Verilog-A:

• V(node) <+
• @(cross(...)), @(above(...)), @(initial_step), @(timer(...))
• transition(...)
• if/else, case, for, arrays, real/integer parameters

Do not rely on EVAS for:

• I(...) <+
• ddt, idt, idtmod, laplace_*
• noise sources
• transistor-level devices
• KCL/KVL solving
• AC/DC analysis
• Spectre subckt hierarchy

If EVAS passes but Spectre is the target, say not Spectre-compiled.

Commands

Verify installation:

evas list

Fallback:

python -m evas list

Run a visible/public testbench:

evas simulate path/to/tb.scs -o output/run-name

If the task provides a visible smoke runner, prefer:

python3 run_visible_smoke.py

Minimal Testbench Syntax

Use a Spectre-style .scs file near the .va files:

simulator lang=spectre
global 0

Vclk (clk 0) vsource type=pulse val0=0 val1=0.9 period=10n rise=10p fall=10p width=5n

XDUT (clk out) my_module

tran tran stop=100n maxstep=100p
save clk out
ahdl_include "./my_module.va"

Keep DUT source and testbench separate:

• .va: model only
• .scs: sources, instance, tran, save, ahdl_include

Use relative ahdl_include paths when possible.

Output Files

After a successful run, inspect:

• tran.csv: waveform data, primary artifact
• strobe.txt: optional $strobe / display output
• tran.png: optional plot

Absence of strobe.txt or tran.png is not a failure by itself.

Save Syntax

• Save explicit signals to keep tran.csv small.
• Fixed point: save out:6f
• Scientific notation: save vin:10e
• Integer/digital display: save code:d
• Escape Spectre bus angle brackets: DOUT\<3\>, not DOUT<3>.

Common Failures

• evas: command not found: activate the environment or use python -m evas.
• Empty tran.csv: add explicit save statements.
• Missing compile marker: check ahdl_include, module name, and parse errors.
• All outputs zero: look for unsupported current-domain constructs or missing output contributions.
• Visible EVAS pass but hidden fail: visible test is weaker than hidden; do not infer hidden correctness from public smoke alone.