Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Hands-on: design an inverter

In this tutorial you design a CMOS inverter from an empty project: draw the schematic, set up a testbench, simulate it, measure its switching threshold and delay, and resize it. It puts the MOSFET and CMOS logic chapters into practice. Allow about 30 minutes.

What you will build:

Value
Supply1.8 V
NMOSW = 1 µm, L = 180 nm
PMOSW = 2 µm, L = 180 nm, then 4 µm
Load20 fF (about the input of five similar inverters)
GoalMeasure VM, tpHL and tpLH; balance the delays

Step 0: a project

Start IC Studio. On the welcome page choose New project…, give it a name and a folder, and keep the generic 180 nm technology. (To follow along without saving, Open demo project works too; everything stays in memory until you press Ctrl+S.)

Step 1: create the cell

  1. Press Ctrl+N.
  2. Name the cell my_inv, keep schematic checked, and press Create.

The Schematic workspace opens on an empty canvas. The left panel has the Device Library, with the shortcut key of each device on its tile.

Step 2: place the transistors

  1. Press P and click on the canvas to place a PMOS. It is named M1. Press Esc.
  2. Press N and click below the PMOS to place an NMOS (M2), leaving a few grid points between them. Press Esc.
  3. Click the PMOS. In the inspector on the right, check Width is 2u and Inductance / L (the channel length) is 180n. Click the NMOS: Width 1u, length 180n. These are the defaults.

Each MOSFET has four pins: d (drain), g (gate), s (source) and b (body). Hover a pin to see its name.

Step 3: wire the inverter

  1. Press W for the Wire tool.
  2. Click the PMOS drain, then the NMOS drain: the two drains are now one net, the output.
  3. Click the PMOS gate, then the NMOS gate: the input net.
  4. Press S and click above the PMOS to place a Supply Net symbol (net vdd). Wire the PMOS source and the PMOS body to it.
  5. Press G and place a Ground symbol below the NMOS. Wire the NMOS source and body to it.

A wire ends by itself when you click on a pin or another wire. Unconnected pins are drawn as small squares, so a square left on a transistor means a missing wire.

Step 4: name the nets

  1. Press T (Label). In the left panel, type in in the Net: field and click on the gate wire.
  2. Type out and click on the drain wire.

Named nets make the results readable (v(in), v(out)) and, later, help LVS match the layout to the schematic.

Step 5: build the testbench

A circuit alone does nothing: it needs a power supply, an input signal and a realistic load. These parts form the testbench.

  1. Supply: press V and place a voltage source to the left. Wire its p pin to a new Supply Net symbol (S) and its n pin to a Ground symbol. In the inspector set Name to VDD and DC value to 1.8.

  2. Input: place a second voltage source. Wire p to the in net (or put an in label on its p pin) and n to ground. Set Name to VIN, DC value to 0 and Transient to:

    PULSE(0 1.8 1n 100p 100p 4n 10n)
    

    That is: from 0 to 1.8 V, starting at 1 ns, 100 ps rise and fall, 4 ns high, repeating every 10 ns.

  3. Load: press C, place a capacitor from out to ground and set its Capacitance to 20f.

Check the Problems tab in the bottom panel: it should be empty. If it reports a floating net or an unconnected pin, fix the wire it points to (click the finding to select the device). The Netlist tab now shows something like:

MP out in vdd vdd pmos w=2u l=180n
MN out in 0 0 nmos w=1u l=180n
CL out 0 20f
VDD vdd 0 DC 1.8
VIN in 0 DC 0 PULSE(0 1.8 1n 100p 100p 4n 10n)

(Your device names may differ; the connections are what matter.)

Step 6: the operating point

  1. Press Ctrl+3 for the Simulation workspace. Operating Point is enabled by default.
  2. Press F5.
  3. Go back to the schematic (Ctrl+2): every net shows its voltage. With in at 0 V, out is at 1.8 V.
  4. In the Simulation workspace, open the Operating Point tab: the NMOS is in cutoff, the PMOS in triode with almost no current. As predicted: no static current.

Step 7: the transfer curve (DC sweep)

  1. In Analyses, enable DC Sweep. Set Source to VIN, Start 0, Stop 1.8, Step 10m.
  2. Press F5 and open the DC Sweep tab. Tick v(in) and v(out) in Signals.

You see the transfer curve. To find the switching threshold exactly, open Tools > Waveform Calculator, choose the dc dataset and evaluate:

cross(v(out) - v(in), 0)

It returns the input voltage where out = in: VM ≈ 0.826 V.

Also plot i(VDD): the supply current is zero at both ends and peaks at about 118 µA near VM, where both transistors are on.

Step 8: the delay (transient)

  1. Enable Transient with Step 10p and Stop time 10n. Press F5.

  2. On the Transient tab, plot v(in) and v(out). Zoom into the edge at 1 ns by scrolling over the plot.

  3. In the Waveform Calculator, pick the tran dataset and evaluate the two delays:

    delay(v(in), 0.9, v(out), 0.9, rise, fall)     → t_pHL ≈ 36 ps
    delay(v(in), 0.9, v(out), 0.9, fall, rise)     → t_pLH ≈ 52 ps
    

    Each measures from the input crossing 0.9 V (50 % of 1.8 V) to the output crossing 0.9 V.

The rising output is slower: the PMOS is weaker than the NMOS, as the sizing table predicted.

Step 9: balance the inverter

Predict first: if the PMOS is made twice as wide, what happens to VM, tpLH and tpHL?

  1. Back in the schematic, select the PMOS and set Width to 4u.
  2. Press F5. Calculator results are kept; click each expression in Results and evaluate it again.
PMOS WVMtpHLtpLH
2 µm0.83 V36 ps52 ps
4 µm0.90 V39 ps37 ps

The threshold moved to the middle and the delays are now balanced. tpHL got slightly worse: the bigger PMOS adds drain capacitance to the output that the NMOS must also discharge. Every sizing decision trades something.

To compare the two runs on one plot, tick the earlier run under Compare runs: it is drawn dashed.

Step 10: save

Press Ctrl+S. The cell is saved as human-readable JSON in your project folder: the schematic and the simulation setup (sim.json), so the next F5 repeats the same analyses.

Exercises

  1. Load. Double the load to 40f. How do the delays change? (Delay is roughly proportional to CL.)
  2. Supply. Lower VDD to 1.2 (and the pulse to PULSE(0 1.2 ...), with the measurement levels at 0.6). The gate gets slower: Ron grows as the overdrive shrinks. The switching energy C·V2 falls by more than half.
  3. Long channel. Set both lengths to 360n. Why does the delay grow so much?
  4. Variation. Choose Simulation > Monte Carlo…, add the measurements dc: cross(v(out) - v(in), 0) and tran: delay(v(in), 0.9, v(out), 0.9, rise, fall), and run 200 samples. Manufacturing variation spreads VM by about ±17 mV (1σ) and tpHL by about ±1.5 ps. Real chips are designed to work across this spread.

From the command line. The same inverter as a netlist file runs without the GUI. Save the netlist from Step 5 with .dc VIN 0 1.8 10m and .tran 10p 10n lines added as my_inv.cir, then:

ics new myproj
ics schematic myproj my_inv my_inv.cir     # netlist → schematic and simulation setup
ics sim myproj my_inv

See Command line.

Next: learn what the layout of this inverter means in From circuit to silicon, then lay it out.