Resistors, Capacitors and Varactors on a Chip
Measure the resistors and capacitors a chip offers: their size, their drift with temperature and voltage, and why ratios hold.
You Will Be Able To
- Size a resistor from its sheet resistance and its two ends.
- Measure temperature and voltage coefficients with a
TEMPsweep and a DC sweep. - Choose between a MIM and a MOS capacitor from a measured C-V curve.
- Build a ratio from unit devices so that it holds over temperature.
- Lesson
- E9.3
- Time
- 35 min
- Runs on
- ngspice
- Check
- 5 questions
A low-pass filter used a MOS capacitor to save area. Its corner was 2.28 MHz with the input at 0.3 V and 1.02 MHz at 1.5 V.
- gain, input at 0.3 V
- gain, input at 1.5 V
Read the plot as a table
| Frequency | gain, input at 0.3 V | gain, input at 1.5 V |
|---|---|---|
| 10 kHz | -0.00008335 dB | -0.0004183 dB |
| 100 kHz | -0.008327 dB | -0.04164 dB |
| 1 MHz | -0.7625 dB | -2.93 dB |
| 10 MHz | -13.05 dB | -19.88 dB |
| 100 MHz | -32.83 dB | -39.84 dB |
Only the DC level of the input differs.
The capacitor changed with its own voltage. Every resistor and capacitor on a chip has limits like this one.
Sheet Resistance and Squares
- Neutral
- Data
- Structure
- Squares
The strip holds L/W squares. Each square has the same resistance, the sheet resistance Rs.
- Ends
Each end adds a fixed resistance that does not grow with L.
- Stray capacitance
The strip and the substrate form a capacitor through the oxide. It grows with the area W·L.
Read the diagram as text
- Contact end: metal meets poly
- Poly strip: L long, W wide
- Contact end: metal meets poly
- Field oxide: thick insulator
- Substrate: at ground
- Contact end → Poly strip: current
- Poly strip → Contact end: current
- Poly strip → Field oxide: stray C
- Field oxide → Substrate
R = Rs·L/W plus the two ends.
The lab measures a 1 µm wide p+ poly strip. It reads 3628 Ω at 10 squares and 6876 Ω at 20 squares. The difference gives Rs = 324.8 Ω per square. The two ends add 380 Ω.
| Film | Device | Ω per square | At −40 °C | At 125 °C |
|---|---|---|---|---|
| p+ poly | res_high_po | 324.8 | −2.7 % | +5.9 % |
| Gate poly | res_generic_po | 51.1 | −5.6 % | +9.5 % |
| n-well | rnwell | 1700 | −6.4 % | +22.0 % |
Each value comes from two lengths, so the ends cancel. Changes are against 27 °C.
Temperature and Voltage Coefficients
- p+ poly
- gate poly
- n-well
- Temperature -40 °Cp+ poly = -2.727% · gate poly = -5.595% · n-well = -6.424%
Cold, every film reads low. The n-well is 6.4 % under its 27 °C value.
- Temperature 125 °Cp+ poly = 5.879% · gate poly = 9.549% · n-well = 22.05%
Hot, the n-well is 22.0 % high and the p+ poly only 5.9 %.
Read the plot as a table
| Temperature | p+ poly | gate poly | n-well |
|---|---|---|---|
| -40 °C | -2.727% | -5.595% | -6.424% |
| -20 °C | -2.023% | -4.004% | -5.242% |
| 0 °C | -1.225% | -2.346% | -3.434% |
| 20 °C | -0.334% | -0.6199% | -0.9998% |
| 40 °C | 0.6507% | 1.173% | 2.06% |
| 60 °C | 1.729% | 3.034% | 5.746% |
| 80 °C | 2.901% | 4.963% | 10.06% |
| 100 °C | 4.166% | 6.959% | 15% |
| 120 °C | 5.525% | 9.022% | 20.56% |
| 125 °C | 5.879% | 9.549% | 22.05% |
One dc TEMP sweep.
A temperature coefficient is the fractional change of a value per degree. A lightly doped film such as the well changes most.
Poly lies on thick oxide, isolated from the substrate. A well resistor lies inside the silicon behind a junction. A real one leaks a little and changes with its voltage.
A voltage coefficient is the change with the voltage across the device. The lab puts 1.8 V across two res_xhigh_po strips of 2 kΩ per square. The 2 µm strip drops 4.1 %. The 20 µm strip drops 0.09 %.
The field along the strip sets this change, so a longer resistor is more linear.
MIM, MOM and MOS Capacitors
- Neutral
- Structure
- MIM
A MIM capacitor is metal, insulator, metal. Both plates are metal, so the gap never changes.
- MOS
A MOS capacitor is a transistor with source, drain and body tied together. Its bottom plate is charge in the silicon, which moves with the voltage.
Read the diagram as text
- MIM top plate: thin metal, to metal 4
- Thin insulator: fixed thickness
- MIM bottom plate: metal 3
- MOS top plate: the poly gate
- Gate oxide: far thinner
- MOS bottom plate: the silicon surface
- MIM top plate → Thin insulator
- Thin insulator → MIM bottom plate
- MOS top plate → Gate oxide
- Gate oxide → MOS bottom plate
The thin gate oxide gives the MOS capacitor more capacitance per area.
| 10 µm × 10 µm | C at 0.3 V | C at 1.5 V | Per µm² at 1.5 V |
|---|---|---|---|
MIM, cap_mim_m3_1 | 206.0 fF | 206.0 fF | 2.06 fF |
MOS, nfet_01v8 | 348.6 fF | 781.0 fF | 7.81 fF |
From the imaginary part of the AC current: C = Im(i)/(2πf).
The MOS capacitor holds 3.8 times more per area at 1.5 V. At 0.3 V it has lost more than half of that.
A MOM capacitor uses the side walls of interleaved metal fingers. This library has no MOM model, so no run here shows one.
Every plate also has stray capacitance to the substrate, and the bottom plate has the most. The SKY130 MIM model here leaves it out.
The Varactor's C-V Curve
cgg of the 10 µm × 10 µm device. Source, drain and body are at 0 V.- Gate voltage -1.8 VC = 776.4 fF
Accumulation: the negative gate pulls holes to the surface. C is 776 fF, close to the oxide capacitance.
- Gate voltage 0 VC = 215.8 fF
Depletion: the surface is empty of carriers. The empty layer acts as a second capacitor in series, and C falls to 216 fF.
- Gate voltage 900 mVC = 743.4 fF
Past the threshold, source and drain supply electrons. A channel forms under the gate and C is back at 743 fF.
- Gate voltage 1.8 VC = 782.8 fF
Inversion: the channel is the bottom plate. C is 783 fF.
Read the plot as a table
| Gate voltage | C |
|---|---|
| -1.8 V | 776.4 fF |
| -1.5 V | 748.1 fF |
| -1 V | 662.8 fF |
| -500 mV | 323.5 fF |
| 0 V | 215.8 fF |
| 500 mV | 567.6 fF |
| 900 mV | 743.4 fF |
| 1 V | 753.8 fF |
| 1.5 V | 776.1 fF |
| 1.8 V | 782.8 fF |
The AC run in the lab reads about 5 fF more than cgg.
A varactor is a capacitor that a control voltage tunes, for example in an oscillator. This device is one: from 0 V to 0.9 V its capacitance rises 3.4 times.
SKY130 also offers dedicated varactor devices. This library has no model for them.
Ratios Match, Absolute Values Drift
Film thickness and doping differ from die to die. Every device of one film on a die changes by the same factor. Temperature acts the same way.
| Divider for a ratio of 1/3 | Tap ratio at −40 °C | At 27 °C | At 125 °C |
|---|---|---|---|
| Three 10 µm units | 0.3333 | 0.3333 | 0.3333 |
| One 20 µm piece over one unit | 0.3456 | 0.3454 | 0.3449 |
All p+ poly, 1 µm wide. Every unit rises by the same factor at 125 °C, so the first ratio does not move.
The long piece has one pair of ends where two units have two. Its ratio is 3.6 % high.
Build every ratio from identical units of one film: same width, same length, same ends. The same rule holds for capacitors.
Asked in interviews
Why do analog circuits rely on ratios of resistors or capacitors?
Absolute values move with process and temperature. Here a p+ poly resistor rises 5.9 % at 125 °C. Devices of one film on one die move together, so a ratio of identical units stays fixed.
When would you use a MOS capacitor and when a MIM?
A MOS capacitor gives the most capacitance per area: 7.8 against 2.1 fF per µm² here. Its value depends on its voltage. Use it for supply and bias decoupling. Use a MIM where the value must stay fixed, as in a filter or a sampling network.
Why is a 2R resistor drawn as two unit resistors in series?
A resistor is its body plus two ends. A piece twice as long doubles the body only. Two units in series double both, so the ratio to one unit is exactly 2.
Try It
Part 1 sweeps TEMP with 1 V across each resistor. Two lengths give the sheet resistance, and what is left of the short one is the ends.
Part 2 reads both dividers over the same sweep. The three-unit divider prints the same ratio cold and hot.
Part 3 sweeps the voltage across two strips of the 2 kΩ per square film. Part 4 reads each capacitance from the AC current at two bias levels.
Predict before each change. Make XL1 30 µm long: does k_long move toward 1/3 or away? Set both capacitor bias levels to 0.9 V: which capacitance changes?
Run the netlist to draw its plots here.
r10 = 3.627978e+03 r20 = 6.876250e+03 r40 = 1.337279e+04 rs_po = 3.248272e+02 r_ends = 3.797051e+02 rnw = 1.700000e+04 po_hot = 5.878870e+00 nw_hot = 2.204757e+01 k_units_cold = 3.333333e-01 k_units_hot = 3.333333e-01 k_long_cold = 3.456410e-01 k_long = 3.453826e-01 k_long_hot = 3.448704e-01 short_pct = -4.111187e+00 long_pct = -8.733373e-02 cmim_03 = 2.060450e-13 cmos_03 = 3.486139e-13 cmim_15 = 2.060450e-13 cmos_15 = 7.810333e-13Check Yourself
Common Mistakes
Wrong Draw a 2R resistor as one piece twice as long.
The body doubles but the two ends do not, so the piece is less than twice one unit.
Fix Put two unit resistors in series. The lab reads 0.3454 for the long piece and 0.3333 for units.
XL1 top tap 0 sky130_fd_pr__res_high_po w=1 l=20
XL2 tap 0 0 sky130_fd_pr__res_high_po w=1 l=10XU1 top m 0 sky130_fd_pr__res_high_po w=1 l=10
XU2 m tap 0 sky130_fd_pr__res_high_po w=1 l=10
XU3 tap 0 0 sky130_fd_pr__res_high_po w=1 l=10Wrong Use a MOS capacitor in a filter because it is the smallest.
Its value follows its own voltage: 349 fF at 0.3 V and 781 fF at 1.5 V here.
Fix Use a MIM capacitor where the value matters. Keep the MOS capacitor for decoupling, biased well above the threshold.
Wrong Trust the flat temperature curve that ngspice 42 gives for res_xhigh_po.
The model file gives the body a coefficient near −0.25 % per °C, but ngspice does not apply it to this resistor.
Fix Read the coefficient from the model file or the foundry tables, and check the design at both temperature ends by hand.
Remember
- R = Rs·L/W plus two ends. Here the p+ poly has 324.8 Ω per square and 380 Ω of ends at 1 µm width.
- Films drift differently: at 125 °C the p+ poly is 5.9 % high and the n-well 22.0 %.
- A MIM capacitor is fixed at 2.06 fF per µm². A MOS capacitor holds 3.8 times more and changes with voltage.
- The MOS C-V curve is a varactor: 216 fF in depletion, 783 fF in inversion.
- Ratios of identical units of one film hold over temperature. Absolute values do not.