E9.2.1 Keep the Leak Under a Budget

MediumImplementngspice 42 + SKY130 · serverReview the topic

Size an NMOS hold switch for 125 °C. It must conduct with at most 500 Ω at the slow corner. It must leak at most 150 pA at the fast corner while it blocks 1.8 V.

Circuit context: An NMOS hold switch on a sensor node: 2 pF held for 10 ms at up to 1.8 V, on SKY130 at 125 °C. Resize a minimum-length hold switch whose hot leak is four times its budget.

A sensor node holds its voltage on 2 pF for 10 ms while an ADC converts it. The switch that isolates the node may leak at most 150 pA. More leak moves the held voltage by more than 0.75 mV.

When the switch is on, it must settle the node quickly, so its on-resistance may be at most 500 Ω.

The first design used the minimum length, 2.6/0.15 µm. Its on-resistance is 448 Ω at the slow corner. Its leak at the fast corner and 125 °C is 644 pA, more than four times the budget.

Resize the subcircuit sw so that it meets both limits. You may change W and L, and you may use more than one transistor.

What the Grader Measures

Your subcircuit sw (ports a, b, en) under the grader's own bench, one run per corner (ss, ff):

CheckCornerRule
On-resistance at 125 °Css50 Ω to 500 Ω
Off current at 125 °C, 1.8 V across the switchff≤ 150 pA
Full specification
The subcircuit the grader tests (keep the name and the ports)
.subckt sw a b en
* X instances of sky130_fd_pr__nfet_01v8 only; the body of each device goes to node 0
.ends
CheckCornerRuleHow the grader measures it
On-resistancess, 125 °C50 Ω to 500 Ωen = 1.8 V, a = 50 mV, b = 0 V: 50 mV divided by the current
Off currentff, 125 °C≤ 150 pAen = 0 V, a = 1.8 V, b = 0 V: the current into a

The grader runs its bench twice, with .lib sky130 ss and ff. Both runs are at 125 °C with gmin lowered to 1e-18, so the current it reads is the device's own.

The enable signal is 0 V or 1.8 V. It cannot go below ground.

Examples

Example 1

The Starter: 2.6/0.15
Design
.subckt sw a b en
* the first design: minimum length, sized for the on-resistance only
XM1 a en b 0 sky130_fd_pr__nfet_01v8 W=2.6 L=0.15
.ends
Measured by the grader
On-resistance at 125 °C: 448.2 Ω (needs 50 Ω to 500 Ω)
Off current at 125 °C, 1.8 V across the switch at ff: 643.6 pA (needs ≤ 150 pA)

The on-resistance is fine. With 1.8 V on the drain the short channel loses threshold to DIBL. At the fast corner and 125 °C the leak is 4.3 times the budget.

Example 2

One Step Above Minimum Length: 3/0.18
Design
.subckt sw a b en
XM1 a en b 0 sky130_fd_pr__nfet_01v8 W=3 L=0.18
.ends
Measured by the grader
On-resistance at 125 °C: 455.5 Ω (needs 50 Ω to 500 Ω)
Off current at 125 °C, 1.8 V across the switch at ff: 189.1 pA (needs ≤ 150 pA)

A slightly longer channel already cuts the leak 3.4 times at the same on-resistance. It is still 26 % over the budget.

Example 3

As Long as Possible: 12/1
Design
.subckt sw a b en
XM1 a en b 0 sky130_fd_pr__nfet_01v8 W=12 L=1
.ends
Measured by the grader
On-resistance at 125 °C: 482.6 Ω (needs 50 Ω to 500 Ω)
Off current at 125 °C, 1.8 V across the switch at ff: 210.7 pA (needs ≤ 150 pA)

A 1 µm channel has little DIBL. But it needs 12 µm of width for the on-resistance, and leak grows with width.

Constraints and Contract

  • The netlist defines .subckt sw a b en ... .ends exactly once, with these ports in this order.
  • Inside sw: X instances of sky130_fd_pr__nfet_01v8, comments and .param lines. No resistors, no sources and no other device types.
  • The grader writes the .lib line for each corner itself. Everything outside sw is replaced by its bench.

Hints

Hint 1

Measure the leak of the starter with 0.1 V across the off switch, then with 1.8 V. Lesson E9.2 names the effect that makes the difference, and shows how it changes with length.

Hint 2

A longer channel needs more width for the same on-resistance, and width adds leak. Hold the on-resistance near its limit, step the length, and watch the leak.

Hint 3

Your own bench must copy the grader's conditions. Add .options gmin=1e-18 temp=125, and change the .lib line to ss for the on-resistance and to ff for the leak. With the default gmin the simulator adds 1.8 pA of its own.

What the grader checks: The grader places sw under its own bench at 125 °C with gmin = 1e-18. It reads the on-resistance at the ss corner, with 1.8 V on en and 50 mV across the switch. It reads the off current at the ff corner, with 0 V on en and 1.8 V across the switch. Engine: ngspice 42 (libngspice0 42+ds-3build1) with the lean SKY130 library, measured 2026-11-12 with tools/spice-judge.mjs. It proves those two numbers. It does not check the switch near the top of the signal range, its charge injection or its speed. The SKY130 model of this device has no GIDL and no gate current, so the leak is subthreshold and junction current only.

Run simulates your netlist as written, with your own analyses and measurements. Submit sends it to the grader, which keeps only the subcircuit sw and makes 2 ngspice runs. 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 1 steps and the final capstone

Report a problem with this page

Your netlist SPICESaved locally
Ln 1, Col 1Tab indents · Ctrl/⌘ + Enter runs
Not run yet
Run your netlist to see its measurements and the ngspice log here.
ngspice 42 + SKY130 · server