analog: bench.t27 -- ngspice wall time and a fixed-step backward-Euler reference for the M0/M1 circuits (Closes #10505)

.trinity/seals/analog_AnalogBench.json

@@ -0,0 +1,28 @@
+{
+  "built_by": "t27c-bootstrap@0.5.2+9b556b7c3",
+  "closure": [],
+  "config": "gen=c,rust,verilog,zig;tests=zig",
+  "gen_hash_c": "sha256:89195447f8fee90c8db58f4eee50988ada367390d06d6c41e171c817e16f2680",
+  "gen_hash_rust": "sha256:c26da9d6dd68f89c6d9e3d5b8b1ee978836bb638dd9130c6ad1d02d4f14b2b06",
+  "gen_hash_verilog": "sha256:1fb746ee68114b0e7506ec46cc0277e02a9ac9a7f2236a4f894032ede144f252",
+  "gen_hash_zig": "sha256:3cc22bec9ee00c4031f82679cb6acabba759bee6dd4538ffb35dbd5a025f9d14",
+  "module": "AnalogBench",
+  "ring": 12,
+  "seal_schema": 2,
+  "sealed_at": "2026-10-18T16:09:11Z",
+  "sealed_by": "t27c-bootstrap@0.5.2",
+  "spec_hash": "sha256:daa6ede3a542d8bfe2ce93995f3808f0549037a0b63538f2179ec107dc2b2f60",
+  "spec_path": "specs/analog/bench.t27",
+  "tests": {
+    "failed": 0,
+    "failing": [],
+    "forced": false,
+    "passed": 11,
+    "total": 11,
+    "vacuous": 0
+  },
+  "toolchain": {
+    "test_runner": "sha256:1982ae948e24d00c5d221ccec12dfce1eb01204b53e37b0f857f93b16d64629d",
+    "zig": "0.16.0"
+  }
+}

specs/analog/decks/lad10_adaptive.cir

@@ -0,0 +1,31 @@
+10-node RC ladder (beyond the M1 cell), ngspice default adaptive transient
+* specs/analog/decks/lad10_adaptive.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* The same circuit and run as lad10_fixed.cir (not a circuit of specs/analog/linear.t27), written as a user would write
+* it: a PWL source and the default options (trapezoidal integration, reltol 1e-3, ngspice's own step control). tmax is
+* its default, min(tstep, tstop/50) = 50 ns.
+* IN = 1 V until 5000 ns and 0 V from 5050 ns: step 99 of the run ends at 5000 ns, step 100 at 5050 ns, and the source
+* carries the value of the step's END point, as the backward Euler recurrence does. uic with IC=0 starts all nodes at 0 V.
+* Run: ngspice -b -r lad10_adaptive.raw lad10_adaptive.cir   (the raw file holds float64 values)
+V1 in 0 PWL(0 1 5000n 1 5050n 0)
+R1 in 1 1k
+R2 1 2 1k
+R3 2 3 1k
+R4 3 4 1k
+R5 4 5 1k
+R6 5 6 1k
+R7 6 7 1k
+R8 7 8 1k
+R9 8 9 1k
+R10 9 10 1k
+C1 1 0 1n IC=0
+C2 2 0 1n IC=0
+C3 3 0 1n IC=0
+C4 4 0 1n IC=0
+C5 5 0 1n IC=0
+C6 6 0 1n IC=0
+C7 7 0 1n IC=0
+C8 8 0 1n IC=0
+C9 9 0 1n IC=0
+C10 10 0 1n IC=0
+.tran 50n 10000n uic
+.end

specs/analog/decks/lad10_fixed.cir

@@ -0,0 +1,35 @@
+10-node RC ladder (beyond the M1 cell), backward Euler on a fixed 50 ns grid
+* specs/analog/decks/lad10_fixed.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* Circuit: R1..R10 = 1 kOhm in series from IN through nodes 1..10; C1..C10 = 1 nF from nodes 1..10 to ground; all
+* capacitors start at 0 V. This is NOT a circuit of specs/analog/linear.t27 (its NMAX is 5): it is here so the ngspice
+* wall time can be read against circuit size, and it has no board number.
+* Run: 200 steps of 50 ns, IN = 1 V for the first 100 steps and 0 V for the next 100.
+* The grid is made exact as in rc_fixed.cir: a B source declares no breakpoint, the circuit rests at 0 V while the source
+* is 0, the source turns on at 425 ns (half a step from the nearest grid point) and turns off at 5425 ns, and from the
+* first internal point after 425 ns every step is 50 ns. Step k (k = 0..199) ends at the k-th internal point after 425 ns.
+* tstop = 425 ns + 201 steps. gear with maxord 1 is backward Euler; trtol = 1e7 stops the truncation test.
+* Run: ngspice -b -r lad10_fixed.raw lad10_fixed.cir   (the raw file holds float64 values)
+B1 in 0 V = u(time-425n) - u(time-5425n)
+R1 in 1 1k
+R2 1 2 1k
+R3 2 3 1k
+R4 3 4 1k
+R5 4 5 1k
+R6 5 6 1k
+R7 6 7 1k
+R8 7 8 1k
+R9 8 9 1k
+R10 9 10 1k
+C1 1 0 1n IC=0
+C2 2 0 1n IC=0
+C3 3 0 1n IC=0
+C4 4 0 1n IC=0
+C5 5 0 1n IC=0
+C6 6 0 1n IC=0
+C7 7 0 1n IC=0
+C8 8 0 1n IC=0
+C9 9 0 1n IC=0
+C10 10 0 1n IC=0
+.options method=gear maxord=1 trtol=1e7
+.tran 50n 10475n 0 50n uic
+.end

specs/analog/decks/lad3_adaptive.cir

@@ -0,0 +1,16 @@
+M1 3-node RC ladder, ngspice default adaptive transient
+* specs/analog/decks/lad3_adaptive.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* The same circuit and run as lad3_fixed.cir, written as a user would write it: a PWL source and the default options
+* (trapezoidal integration, reltol 1e-3, ngspice's own step control). tmax is its default, min(tstep, tstop/50) = 50 ns.
+* IN = 1 V until 5000 ns and 0 V from 5050 ns: step 99 of the run ends at 5000 ns, step 100 at 5050 ns, and the source
+* carries the value of the step's END point, as the backward Euler recurrence does. uic with IC=0 starts all nodes at 0 V.
+* Run: ngspice -b -r lad3_adaptive.raw lad3_adaptive.cir   (the raw file holds float64 values)
+V1 in 0 PWL(0 1 5000n 1 5050n 0)
+R1 in 1 1k
+R2 1 2 1k
+R3 2 3 1k
+C1 1 0 1n IC=0
+C2 2 0 1n IC=0
+C3 3 0 1n IC=0
+.tran 50n 10000n uic
+.end

specs/analog/decks/lad3_fixed.cir

@@ -0,0 +1,20 @@
+M1 3-node RC ladder, backward Euler on a fixed 50 ns grid
+* specs/analog/decks/lad3_fixed.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* Circuit: R1, R2, R3 = 1 kOhm in series from IN through nodes 1, 2 and 3; C1, C2, C3 = 1 nF from nodes 1, 2, 3 to
+* ground; all capacitors start at 0 V (specs/analog/linear.t27, circuit 3).
+* Run: 200 steps of 50 ns, IN = 1 V for the first 100 steps and 0 V for the next 100.
+* The grid is made exact as in rc_fixed.cir: a B source declares no breakpoint, the circuit rests at 0 V while the source
+* is 0, the source turns on at 425 ns (half a step from the nearest grid point) and turns off at 5425 ns, and from the
+* first internal point after 425 ns every step is 50 ns. Step k (k = 0..199) ends at the k-th internal point after 425 ns.
+* tstop = 425 ns + 201 steps. gear with maxord 1 is backward Euler; trtol = 1e7 stops the truncation test.
+* Run: ngspice -b -r lad3_fixed.raw lad3_fixed.cir   (the raw file holds float64 values)
+B1 in 0 V = u(time-425n) - u(time-5425n)
+R1 in 1 1k
+R2 1 2 1k
+R3 2 3 1k
+C1 1 0 1n IC=0
+C2 2 0 1n IC=0
+C3 3 0 1n IC=0
+.options method=gear maxord=1 trtol=1e7
+.tran 50n 10475n 0 50n uic
+.end

specs/analog/decks/lad5_adaptive.cir

@@ -0,0 +1,20 @@
+M1 5-node RC ladder, ngspice default adaptive transient
+* specs/analog/decks/lad5_adaptive.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* The same circuit and run as lad5_fixed.cir, written as a user would write it: a PWL source and the default options
+* (trapezoidal integration, reltol 1e-3, ngspice's own step control). tmax is its default, min(tstep, tstop/50) = 50 ns.
+* IN = 1 V until 5000 ns and 0 V from 5050 ns: step 99 of the run ends at 5000 ns, step 100 at 5050 ns, and the source
+* carries the value of the step's END point, as the backward Euler recurrence does. uic with IC=0 starts all nodes at 0 V.
+* Run: ngspice -b -r lad5_adaptive.raw lad5_adaptive.cir   (the raw file holds float64 values)
+V1 in 0 PWL(0 1 5000n 1 5050n 0)
+R1 in 1 1k
+R2 1 2 1k
+R3 2 3 1k
+R4 3 4 1k
+R5 4 5 1k
+C1 1 0 1n IC=0
+C2 2 0 1n IC=0
+C3 3 0 1n IC=0
+C4 4 0 1n IC=0
+C5 5 0 1n IC=0
+.tran 50n 10000n uic
+.end

specs/analog/decks/lad5_fixed.cir

@@ -0,0 +1,24 @@
+M1 5-node RC ladder, backward Euler on a fixed 50 ns grid
+* specs/analog/decks/lad5_fixed.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* Circuit: R1..R5 = 1 kOhm in series from IN through nodes 1..5; C1..C5 = 1 nF from nodes 1..5 to ground; all
+* capacitors start at 0 V (specs/analog/linear.t27, circuit 4).
+* Run: 200 steps of 50 ns, IN = 1 V for the first 100 steps and 0 V for the next 100.
+* The grid is made exact as in rc_fixed.cir: a B source declares no breakpoint, the circuit rests at 0 V while the source
+* is 0, the source turns on at 425 ns (half a step from the nearest grid point) and turns off at 5425 ns, and from the
+* first internal point after 425 ns every step is 50 ns. Step k (k = 0..199) ends at the k-th internal point after 425 ns.
+* tstop = 425 ns + 201 steps. gear with maxord 1 is backward Euler; trtol = 1e7 stops the truncation test.
+* Run: ngspice -b -r lad5_fixed.raw lad5_fixed.cir   (the raw file holds float64 values)
+B1 in 0 V = u(time-425n) - u(time-5425n)
+R1 in 1 1k
+R2 1 2 1k
+R3 2 3 1k
+R4 3 4 1k
+R5 4 5 1k
+C1 1 0 1n IC=0
+C2 2 0 1n IC=0
+C3 3 0 1n IC=0
+C4 4 0 1n IC=0
+C5 5 0 1n IC=0
+.options method=gear maxord=1 trtol=1e7
+.tran 50n 10475n 0 50n uic
+.end

specs/analog/decks/rc_adaptive.cir

@@ -0,0 +1,12 @@
+M0 RC low-pass, ngspice default adaptive transient
+* specs/analog/decks/rc_adaptive.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* The same circuit and run as rc_fixed.cir, written as a user would write it: a PWL source and the default options
+* (trapezoidal integration, reltol 1e-3, ngspice's own step control). tmax is its default, min(tstep, tstop/50) = 50 ns.
+* IN = 1 V until 5000 ns and 0 V from 5050 ns: step 99 of the run ends at 5000 ns, step 100 at 5050 ns, and the source
+* carries the value of the step's END point, as the backward Euler recurrence does. uic with IC=0 starts OUT at 0 V.
+* Run: ngspice -b -r rc_adaptive.raw rc_adaptive.cir   (the raw file holds float64 values)
+V1 in 0 PWL(0 1 5000n 1 5050n 0)
+R1 in 1 1k
+C1 1 0 1n IC=0
+.tran 50n 10000n uic
+.end

specs/analog/decks/rc_fixed.cir

@@ -0,0 +1,19 @@
+M0 RC low-pass, backward Euler on a fixed 50 ns grid
+* specs/analog/decks/rc_fixed.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* Circuit: R1 = 1 kOhm from IN to node 1 (OUT), C1 = 1 nF from node 1 to ground, OUT starts at 0 V
+* (specs/fpga/rc_filter_ax7203.t27; node numbers as in specs/analog/linear.t27).
+* Run: 200 steps of 50 ns, IN = 1 V for the first 100 steps and 0 V for the next 100 (vin_at of linear.t27).
+* How the grid is made exact. ngspice opens every .tran with steps of tstep/100 that double up to tmax, and opens that
+* ramp again after every breakpoint, so a PWL or PULSE source puts internal points off the grid. A B source declares no
+* breakpoint. While the source is 0 V the circuit rests at 0 V and every step of the ramp leaves it at exactly 0 V, so the
+* ramp cannot touch the state. The source turns on at 425 ns, half a step from the nearest grid point; from the first
+* internal point after 425 ns every step is 50 ns. Step k of the run (k = 0..199) ends at the k-th internal point after
+* 425 ns; the source is 1 V until 5425 ns and 0 V after. tstop = 425 ns + 201 steps.
+* Integration: gear with maxord 1 is backward Euler. trtol = 1e7 stops the truncation test from shrinking a step.
+* Run: ngspice -b -r rc_fixed.raw rc_fixed.cir   (the raw file holds float64 values)
+B1 in 0 V = u(time-425n) - u(time-5425n)
+R1 in 1 1k
+C1 1 0 1n IC=0
+.options method=gear maxord=1 trtol=1e7
+.tran 50n 10475n 0 50n uic
+.end

specs/analog/decks/rc_pwl_naive.cir

@@ -0,0 +1,14 @@
+M0 RC low-pass, backward Euler with a PWL source: the negative control of rc_fixed.cir
+* specs/analog/decks/rc_pwl_naive.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* The same circuit, integration (gear, maxord 1 = backward Euler) and trtol as rc_fixed.cir, but the source is the
+* natural PWL. The corners at 5000 ns and 5050 ns are breakpoints: ngspice reaches them exactly and restarts its step
+* ramp after each (steps of a fraction of the interval that double up to tmax), and it also opens the run with such a
+* ramp. So the internal points are NOT the 50 ns grid and the result is backward Euler on a different, nonuniform
+* grid: v(1) at 5000 ns differs from the backward Euler recurrence at the 1e-5 level, not at the 1e-15 level.
+* Run: ngspice -b -r rc_pwl_naive.raw rc_pwl_naive.cir   (the raw file holds float64 values)
+V1 in 0 PWL(0 1 5000n 1 5050n 0)
+R1 in 1 1k
+C1 1 0 1n IC=0
+.options method=gear maxord=1 trtol=1e7
+.tran 50n 10000n 0 50n uic
+.end

specs/analog/decks/rlc_adaptive.cir

@@ -0,0 +1,14 @@
+M1 series RLC, ngspice default adaptive transient
+* specs/analog/decks/rlc_adaptive.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* The same circuit and run as rlc_fixed.cir, written as a user would write it: a PWL source and the default options
+* (trapezoidal integration, reltol 1e-3, ngspice's own step control). tmax is its default, min(tstep, tstop/50) = 5 ns.
+* IN = 1 V until 500 ns and 0 V from 505 ns: step 99 of the run ends at 500 ns, step 100 at 505 ns, and the source
+* carries the value of the step's END point, as the backward Euler recurrence does. uic with IC=0 starts the capacitor
+* voltage and the inductor current at 0.
+* Run: ngspice -b -r rlc_adaptive.raw rlc_adaptive.cir   (the raw file holds float64 values)
+V1 in 0 PWL(0 1 500n 1 505n 0)
+R1 in 1 100
+L1 1 2 10u IC=0
+C1 2 0 1n IC=0
+.tran 5n 1000n uic
+.end

specs/analog/decks/rlc_fixed.cir

@@ -0,0 +1,17 @@
+M1 series RLC, backward Euler on a fixed 5 ns grid
+* specs/analog/decks/rlc_fixed.cir -- test data for specs/analog/bench.t27 (#10505). A plain netlist, no .control block.
+* Circuit: R1 = 100 Ohm from IN to node 1, L1 = 10 uH from node 1 to node 2 (the capacitor voltage), C1 = 1 nF from
+* node 2 to ground; capacitor voltage and inductor current start at 0 (specs/analog/linear.t27, circuit 2).
+* Run: 200 steps of 5 ns, IN = 1 V for the first 100 steps and 0 V for the next 100.
+* The grid is made exact as in rc_fixed.cir: a B source declares no breakpoint, the circuit rests at 0 while the source
+* is 0, the source turns on at 42.5 ns (half a step from the nearest grid point) and turns off at 542.5 ns, and from the
+* first internal point after 42.5 ns every step is 5 ns. Step k (k = 0..199) ends at the k-th internal point after 42.5 ns.
+* tstop = 42.5 ns + 201 steps. gear with maxord 1 is backward Euler; trtol = 1e7 stops the truncation test.
+* Run: ngspice -b -r rlc_fixed.raw rlc_fixed.cir   (the raw file holds float64 values)
+B1 in 0 V = u(time-42.5n) - u(time-542.5n)
+R1 in 1 100
+L1 1 2 10u IC=0
+C1 2 0 1n IC=0
+.options method=gear maxord=1 trtol=1e7
+.tran 5n 1047.5n 0 5n uic
+.end