Aic Lab Cadence 11 Folded v02
Aic Lab Cadence 11 Folded v02
Aic Lab Cadence 11 Folded v02
Analog IC Design
Lab 11 (Mini Project 02)
Fully-Differential Folded Cascode OTA
Note: The DC loop gain and closed loop bandwidth specs may be difficult to include (precisely) in the initial
hand analysis. After following the suggested design procedure, you may tune the circuit on the simulator
(while taking trade-offs into account) to achieve these specs.
Page 2 of 9
2) The OTA current will be divided as follows: ISS = 40uA for the input pair (CS), and 40uA for the cascode
branches (CG). The NMOS current sources in the bottom needs to sink 80uA (2 x 40uA). For more details
on how to select the folded cascode current split check this paper: “Optimum Split Ratio for Folded
Cascode OTA Bias Current: A Qualitative and Quantitative Study” DOI:
10.1109/ICM48031.2019.9021755.
3) Since this is a relatively difficult design, we will directly assume values for L and gm/ID (or V*) based on
designer’s experience and folded cascode trade-offs matrix.
• For the input pair use short L and bias it in MI or WI, e.g., L = 0.2um and gm/ID = 15. This maximizes
the GBW (good efficiency) and minimizes the input capacitive loading (avoid reducing the DC LG).
You will have to tune your gm/ID to achieve the CL bandwidth spec.
• For the current source transistors use relatively long L and bias them in SI, e.g., L = 1um and gm/ID =
10. These transistors contribute significant offset and noise. A large gm will not help the gain but will
increase the noise.
• For the cascode transistors use moderate L and bias them in MI or WI, e.g., L = 0.5um and gm/ID =
15. These transistors do not contribute significant offset and noise, so they don’t need to be large. A
large gm helps the gain and doesn’t increase the noise.
• These assumptions greatly simplify the design process and can be later tuned in simulations (taking
the trade-offs matrix into consideration) to achieve exact specifications.
4) From the assumed L and gm/ID (V*), use the charts to find the sizing (W) of all transistors.
5) From the assumed V*, select suitable biasing for the cascode transistors (VCASCP and VCASCN).
Hint: Set VCASCN ≈ VGSN + V* and VCASCP ≈ VDD - |VGSP| - V*
Page 3 of 9
Create a testbench similar to the one shown below.
Page 4 of 9
• Plot diff gain (magnitude in dB and phase) vs frequency.
• Calculate circuit parameters (DC gain, BW, GBW, UGF, and PM).
➔ Cadence Hint: Use Cadence calculator expressions to calculate circuit parameters (Ao, Ao in dB, BW,
GBW, UGF). You may use Cadence calculator to create other useful expressions.
• Compare simulation results with hand calculations in a table (use SS parameters from OP simulation
in your hand analysis).
1
Note that this type of CMFB will have a limited linear range and may affect the output swing specification.
Page 5 of 9
• The CMFB limits the output swing. Max Vout is VDD – V* – |VGSP|. One reason why we selected
PMOS CD is to avoid increasing VTH by body effect (increasing VTH will limit output swing even
more).
• The output range now is from 2V* to VDD – V* – |VGSP|. Select Vocm to be around the middle of
this range to have maximum symmetric output swing (it will be different from the value selected in
the behavioral model).
• The output of the CD buffer is close to VDD; thus, we select NMOS input for the error amplifier.
• The error amplifier is a simple differential amplifier with diode connected loads. We don’t need high
gain from the error amplifier, but we need low impedance nodes to avoid deteriorating the stability
of the CMFB loop.
• The bias point output of the error amplifier is equal to VDD - |VGSP|. That’s why we use the error
amplifier output to control the PMOS current source in the folded OTA rather than the NMOS
current source.
• Note that we select the non-inverting error amplifier output to maintain -ve feedback in the CMFB
loop.
Use hierarchy editor to change the model of the CMFB circuit to use the actual circuit instead of the
behavioral one.
Report the following:
1) Schematic of the OTA and CMFB circuit with DC node voltages and transistors OP parameters (id, vgs,
vds, vdsat, vth, gm, gds, region) clearly annotated.
• Set VICM at the middle of the CMIR.
• What is the CM level at the OTA output? Why?
• What are the differential input and output voltages of the error amplifier? What is the relation
between them?
2) Diff small signal ccs:
• Use AC analysis (1Hz:10Gz, logarithmic, 10 points/decade).
• Set VIDAC = 1 and VICMAC = 0.
• Set VICM at the middle of the CMIR.
• Plot diff gain (magnitude in dB and phase) vs frequency.
• Calculate circuit parameters (DC gain, BW, GBW, UGF, and PM).
Page 6 of 9
Note that we use two baluns at the output in order to allow stability analysis. We divide the output to diff
and CM and break the diff/CM loops by 0V dc sources, then we combine them again to VOUTN and VOUTP
to close the feedback loop. We use a 2pF input capacitance and 1pF feedback capacitance to provide a
closed loop gain = 2. Note that you need to connect VOUTN to VINP and VOUTP to VINN to maintain
negative feedback. We use large resistors across the feedback capacitance to close the loop in DC. This will
set Vicm = Vocm.
➔ Cadence Hint: Instead of using two baluns at output, you can use diffstbprobe (or cmdmprobe in older
versions) to simulate LG of differential and CM loops.
Page 7 of 9
PART 6: Closed Loop Simulation (Transient Analysis)
Use the same testbench as in Part 5, but change diff and CM input sources as explained below.
Report the following:
1) Differential and CMFB loops stability (transient analysis):
• Apply a differential input pulse (initial value = 0, pulse value = 100mV, delay = 1us, period = 2us,
pulse width = 1us, rise = fall = 10ns).
• Run transient analysis for 3us with 10ns max step.
• Plot the transient signals at VINP, VINN, VOUTP, VOUTN, and VOCM overlaid in the same figure.
• Do you notice any differential/CM ringing? Are both loops stable with adequate PM?
• Set differential input to zero and apply the same previous pulse at the balun CM input.
• Run transient analysis for 3us to test the fully differential capacitive amplifier stability.
• Plot the transient signals at VINP, VINN, VOUTP, VOUTN, and VOCM overlaid in the same figure.
• Do you notice any differential/CM ringing? Are both loops stable with adequate PM?
2) Output swing:
• Apply a differential sinusoidal input with freq = 100kHz and amplitude = 150mV.
• Run transient analysis for three periods (30us) with 0.1us max time step.
• Plot the transient signals at VINP, VINN, VOUTP, VOUTN, and VOCM overlaid in the same figure.
• Plot the transient signals at VIDIFF and VODIFF overlaid in the same figure.
• Calculate the diff input and output peak-to-peak swings and the closed loop gain.
Lab Summary
• In Part 1 you learned:
o How to generate and use gm/ID design curves.
• In Part 2 you learned:
o How to design a fully-differential folded cascode OTA meeting desired specifications.
• In Part 3 you learned:
o How to set the CM output voltage of a fully-differential OTA with an ideal CMFB circuit.
o How to simulate the small-signal differential characteristics of a fully-differential folded
cascode OTA in open-loop configuration with an ideal CMFB circuit.
• In Part 4 you learned:
o How to design a CMFB circuit.
o How to set the CM output voltage of a fully-differential OTA with an actual CMFB circuit.
o How to simulate the small-signal differential characteristics of a fully-differential folded
cascode OTA in open-loop configuration with a real CMFB circuit.
• In Part 5 you learned:
o How to simulate the small-signal differential gain of a fully-differential folded cascode OTA in
closed-loop configuration with a real CMFB circuit.
o How to simulate the stability of both the main OTA loop and the CMFB loop of a fully-
differential folded cascode OTA in closed-loop configuration.
Page 8 of 9
• In Part 6 you learned:
o How to simulate the stability of both the main OTA loop and the CMFB loop of a fully-
differential folded cascode OTA using transient simulation.
o How to simulate the output swing of both the main OTA loop and the CMFB loop of a fully-
differential folded cascode OTA using transient simulation.
Acknowledgements
Thanks to all who contributed to these labs. Special thanks to Dr. Sameh A. Ibrahim for reviewing and editing
the labs. If you find any errors or have suggestions concerning these labs, contact
[email protected].
Page 9 of 9