analyses_examples

analyses_examples — ngspice analyses with Verilog-A (OSDI) devices (Enhancement-62)

A hands-on tour of every ngspice analysis working with OSDI-compiled Verilog-A devices, plus the two Enhancement-62 additions: generic .dc @inst[param] sweeps and the .disto limitation warning.

Each .cir deck below is standalone and commented — run any of them with

ngspice -b <deck>.cir

after compiling the two models (the verify/plot scripts do this automatically):

openvaf-r analyses_blocks.va -o analyses_blocks.osdi
openvaf-r analyses_dio.va -o analyses_dio.osdi

The models (analyses_blocks.va, analyses_dio.va)

module	what it is	teaches
ores	resistor, parameter real r	plain (model-card) parameters
ocap	capacitor via ddt()	reactive elements in AC/PZ
otres	resistor with $temperature dependence	temp sweeps
ires	resistor with (* type="instance" *) parameter real r	instance-kind parameters
odio	exponential diode with $limit	nonlinear device for .disto

The (* type="instance" *) attribute is the key to instance-level parameter access: it is what makes N1 a 0 mm r=2k, print @n1[r], alter @n1[r], and .dc @n1[r] possible. A plain parameter is model-kind only (set it on the .model card / altermod).

The decks

deck	analysis	expected result
tf.cir	.tf	TF = 0.75, Z_out = 750 Ω, Z_in = 2 kΩ — exact
pz.cir	.pz	single pole at −1/(RC) = −1e6 rad/s — exact, identical to built-in R/C
sens_dc.cir	.sens	dV/dr1 = −1.875e-4, dV/dr2 = +6.25e-5 — the analytic divider derivatives
sens_ac.cir	.sens … ac	dV/dacmag = 0.5 − 0.5j at the pole frequency (= H there)
temp_sweep.cir	.dc temp	I = 1V / R(T) at every point (°C→K handled)
param_sweep.cir	.dc @n1[r] (new)	I = 1V / r at every swept value
nested_sweep.cir	nested .dc @n1[r] … V1 … (new)	family of I = V/r curves, inner level resets
alter.cir	alter / altermod	instance value beats model card; readback via print @n1[r]
disto.cir	.disto	prominent warning that OSDI nonlinearities are invisible to disto (was silent zeros)

Plots

python3 plot_analyses.py regenerates plots/:

plot	shows
plots/param_sweep.png	the new .dc @n1[r] sweep vs analytic 1/R
plots/nested_sweep.png	the nested sweep's I = V/r curve family
plots/temp_sweep.png	.dc temp vs analytic 1/R(T)
plots/ac_lowpass.png	RC Bode plot with the .pz pole (−3 dB) and the .sens ac point (−45°) marked

What Enhancement-62 changed in ngspice

1. .dc @inst[param] sweeps (dctrcurv.c): the sweep code accepted only V/I sources, resistors, and temp. A new generic sweep type resolves @inst[param] through the device's own parameter tables and refreshes the device per point exactly like alter — for any device type, nestable, with the original value restored afterwards.
2. .disto warning (CKTdisto.c): distortion analysis needs higher-order derivatives the OSDI ABI doesn't carry, and ngspice used to skip OSDI devices silently — an OSDI diode reported exactly zero distortion where the identical built-in diode reported 1.8e-6. It now warns loudly, naming each affected device type.

Everything else in the table above already worked and is pinned by verify_analyses.py (19 checks).