A 1.5 pF capacitor in a signal path changes by more than half as its voltage moves. Find why, and replace it so the value holds within 1 % from 0.3 V to 1.5 V.
Circuit context: SKY130 poly resistors and capacitors as two-terminal blocks: a divider on a 1.8 V rail and a 1.5 pF filter capacitor. Replace a filter capacitor whose value follows the signal voltage.
An anti-alias filter uses the capacitor in the subcircuit holdcap. The designer measured 1.5 pF with 0.9 V across it and signed it off. On real signals the filter corner moves with the signal level, and the output is distorted.
The voltage across the capacitor swings between 0.3 V and 1.5 V. Measure the capacitance over that range and find what changes it.
Then replace the device inside holdcap. The capacitance must be 1.5 pF within 3 % at 0.9 V. Between 0.3 V and 1.5 V it may vary by at most 1 % of that value.
What the Grader Measures
Your subcircuit holdcap (ports top, bot) under the grader's own bench, one run per corner (tt):
| Check | Corner | Rule |
|---|---|---|
| Capacitance at 0.9 V | tt | 1.5 pF ± 3 % |
| Variation from 0.3 V to 1.5 V | tt | ≤ 1% |
Full specification
.subckt holdcap top bot
* X instances of sky130_fd_pr__cap_mim_m3_1 (w, l in um),
* sky130_fd_pr__nfet_01v8, sky130_fd_pr__nfet_01v8_lvt or sky130_fd_pr__pfet_01v8
.ends
| Check | Rule | How the grader measures it |
|---|---|---|
| Capacitance at 0.9 V | 1.5 pF ± 3 % | AC at 1 kHz with bot at 0 V and 0.9 V on top: C = −Im(i)/(2πf) |
| Variation from 0.3 V to 1.5 V | ≤ 1 % | 13 bias levels in 0.1 V steps: (largest C − smallest C) divided by C at 0.9 V |
The grader runs its bench once, with .lib sky130 tt at 27 °C. The top port is always the more positive one.
Examples
Example 1
The Starter: An NMOS Capacitor, 14.15 µm × 14.15 µm- Design
.subckt holdcap top bot * a MOS capacitor: gate on top; source, drain and body on bot XM bot top bot bot sky130_fd_pr__nfet_01v8 W=14.15 L=14.15 .ends
- Measured by the grader
Capacitance at 0.9 V: 1.497 pF (needs 1.5 pF ± 3 %) Variation from 0.3 V to 1.5 V: 57.76% (needs ≤ 1%)
The value at 0.9 V is right. At 0.3 V the gate is below the threshold and the capacitance is 0.70 pF. At 1.5 V it is 1.56 pF.
Example 2
A Larger NMOS Capacitor, 20 µm × 20 µm- Design
.subckt holdcap top bot XM bot top bot bot sky130_fd_pr__nfet_01v8 W=20 L=20 .ends
- Measured by the grader
Capacitance at 0.9 V: 2.987 pF (needs 1.5 pF ± 3 %) Variation from 0.3 V to 1.5 V: 57.74% (needs ≤ 1%)
More area scales the whole curve. The value is now twice the target, and it still moves by the same 58 %.
Example 3
A PMOS Capacitor with Its Well on Bot- Design
.subckt holdcap top bot XM bot top bot bot sky130_fd_pr__pfet_01v8 W=13.8 L=13.8 .ends
- Measured by the grader
Capacitance at 0.9 V: 1.516 pF (needs 1.5 pF ± 3 %) Variation from 0.3 V to 1.5 V: 18.52% (needs ≤ 1%)
A positive gate over an n-well gathers electrons under the oxide. This is accumulation, and it has no threshold to cross. The capacitance still moves from 1.31 pF to 1.59 pF.
Constraints and Contract
- The netlist defines .subckt holdcap top bot ... .ends exactly once, with these ports in this order.
- Inside holdcap: only X instances of sky130_fd_pr__cap_mim_m3_1, sky130_fd_pr__nfet_01v8, sky130_fd_pr__nfet_01v8_lvt or sky130_fd_pr__pfet_01v8, with comments and .param lines. Ideal capacitors are not accepted.
- The grader writes the .lib line itself. It replaces everything outside holdcap with its own bench.
Hints
Hint 1
Run the starter and read cv, the capacitance at each bias level. Where is it lowest, and what is the transistor doing there?
Hint 2
A MOS capacitor needs a channel under its gate to reach its full value. Below the threshold voltage it has none.
Hint 3
A larger MOS capacitor has the same curve, only taller. Look for a device whose two plates are both metal.
Hint 4
Measure a small capacitor of the new kind first, and work out its capacitance per area. Then size it for 1.5 pF.
What the grader checks: The grader places holdcap under its own bench. It measures the capacitance at 13 bias levels from 0.3 V to 1.5 V, at tt and 27 °C. Engine: ngspice 42 (libngspice0 42+ds-3build1) with the pinned SKY130 models, measured 2026-11-12 with tools/spice-judge.mjs. The run proves the value and its variation in this model. The SKY130 MIM model has no voltage coefficient and no stray capacitance to the substrate. A real MIM capacitor has a little of both.
Run simulates your netlist as written, with your own analyses and measurements. Submit sends it to the grader, which keeps only the subcircuit holdcap and makes 1 ngspice run. The numbers come from ngspice and the SKY130 models the site pins; they prove the measured cases, not every condition a chip meets.
See all 2 steps and the final capstone