stanford ee272b project - bandgap voltage reference in SKY130 technology

IEEE Solid State Circuits Society Code-a-Chip Jupyter Notebook is located here, at bandgapReferenceCircuit/jupyter/notebook.ipynb

2020-2021 Spring

table of contents

1. single sample silicon measurements

2. table: design goals vs. post-layout, extracted circuit performance

3. ngspice simulation: temperature coefficient vs. corners and output voltage monte carlo

4. layout pictures

5. transient on/off switch power

6. what's a bandgap circuit?

7. how's the bandgap circuit work?

8. prerequisites for this repo

9. steps to set up the repository

10. design files <- look here to get the lvs-passing final schematic file

11. running all tests

12. test descriptions

13. viewing results

14. viewing unscripted results

single sample silicon measurements

measurement setup

close up of the thermoelectric cooler

wiring diagram for the test setup

Output reference voltage vs. load and temperature

Output reference voltage vs. load for fixed temperature

Output reference voltage vs. VDD

Power supply rejection ratio of output reference voltage at one frequency

⚠️Important disclaimer: As you can see, there is a large discrepancy between the measured temperature coefficient and the one obtained from simulation. My guess as to why they are different is a difference between the BJT models and what was fabricated. This is only a guess now because I haven't done any investigation and I could only measure one sample. Either I or someone else needs to tweak the BJT size parameters in the schematic and investigate in simulation if the change of BJT parameters can induce a temperature coefficient that is on the same order as what was measured. In other words, there is more work to be done for this project and repository :), but read on with this disclaimer in mind.

table: design goals vs. post-layout, extracted circuit performance

VDD = 1.8 V, TT corner, no mismatch

ngspice simulations: temperature coefficient vs. corners and output voltage monte carlo

temperature coefficient

ppm.py

output voltage monte carlo

variation.py

0 <= Temperature <= 70

layout pictures

transient on/off switch power

what's a bandgap circuit?

a circuit which provides a voltage that's (ideally) independent of temperature. the final circuit in this repository is shown below.

how's the bandgap circuit work?

A bandgap circuit consists of the following blocks

a bipolar junction transistor (BJT) given some current, will produce a voltage between its base and emitter. This voltage is dependent on temperature and that dependence is typically around -2mV/degC. Notice that the dependence is negative (the negative dependence is usually called "complementary" to absolute temperature, or CTAT). the Vt generator is typically produced by taking the difference in base-emitter voltages between a pair of BJTs with different current densities, which is achieved by either supplying the BJTs with different currents, or by making the area of one BJT larger than the other. The temperature dependence of this difference, i.e. the output of the Vt generator, is typically +0.085mV/degC, much smaller than the CTAT voltage's dependence. Because this Vt dependence is positive, it's often reffered to as a proportional to absolute temperature (PTAT) voltage. The PTAT voltage is scaled (by M in the figure) such that when the CTAT and PTAT voltages are added, the resulting temperature dependence is zero, since the positive and negative temperature dependencies cancel out.

this circuit has its nuances

the bandgap reference core generates both the CTAT and PTAT voltages. assuming Va=Vb, the currents in R1 and R2 will be the same and they will correspond to the CTAT voltage Veb (since V=IR). The generated Vt appears across R3, so the current through R3 corresponds to the PTAT voltage. These PTAT and CTAT currents are summmed at node Vb and mirrored to the output, where that current generates the reference voltage across R4.
the amplifier works to keep Va=Vb so that the above process can happen (we assumed Va=Vb). the self-bias circuit supplies the amplifier with a bias current.
the startup circuit helps the circuit find its stable operating point after the circuit is powerd on.
the current mirror mirrors current!

prerequisites for this repo

ngspice, skywater-pdk, Xschem_sky130, and Xschem. if you're missing one of these, please follow the steps in this video. the written documentation for Xschem is available here.

note: Xschem_sky130 is different from the skywater-pdk but is entirely necessary. make sure you have it installed.

steps to set up the repository

1. enter bash and source caddy.bashrc
2. in project-paths.json fill in the absolute paths for each of the entries. my paths are left in that file as an example.
3. python edit-xschemrc-paths.py
4. xschem &. see if xschem opens without error. important: please launch xschem from this top directory. xschemrc must be in the current working directory, i.e. only launch xschem from the root of this repository.
5. File -> Open -> /path/to/repo/schematics/tsmc_bandgap_real.sch
   hopefully, you can now see the full bandgap circuit schematic. try generating a netlist from this schematic; click netlist in the top right corner. to verify this worked, click the simulation tab in the top toolbar and then click edit netlist. a nano editor window should pop up, showing the various circuit components in the netlist.

design files

the bandgap circuit is compact and the entire design schematic is contained in the following file
schematics/tsmc_bandgap_real.sch
note: you will need xscheme to open the above! if you just want to run the tests for this design, see the next section.

running all tests

first, each .spice file must be generated through xschem. all tests are listed in tests.json.

1. xschem &
2. one by one, open the schematics in table below
3. click Netlist in the top right corner of the screen
4. repeat steps 2,3 for all tests
5. in sims/ you should see the different tests in their .spice format
6. python run-tests.py
7. all of the tests will run. to view the relevant results, follow the steps of "viewing results"

test descriptions

test	description	path to relevant schematic
operating point	this simulation solves for a dc operating point for the circuit at 27 degrees celcius. the operating point represents the dc behavior in steady state	schematics/tsmc_bandgap_real_op.sch
transient	this simulation is a time based simulation; the supply, VDD, is ramped up over 5 microseconds. a power on reset pulse is then applied. the circuit then settles into a desired operating point. this simulation performs the test at three temperatures: 0, 27, 70 degrees celcius.	schematics/tsmc_bandgap_real_tran.sch
transient, with VDD and some process variation	this test varies the following quantities: VDD, threshold voltage (Vth), gate oxide thickness (tox), offset voltage in the subthreshold region, subthreshold swing, ideal max forward beta, transport saturation current, and transport saturation curent's temperature effect exponent. each variation is according to a normally distributed Gaussian distribution.	schematics/tsmc_bandgap_real_tran_gauss.sch
temperature sweep	this simulation solves for an initial operating point at temp = 0 degC and then does an incremental Newton Raphson to solve for the opearting points at successive temperatures.	schematics/tsmc_bandgap_real.sch

running an individual test

as an example, the following commands simulates only one SPICE netlist:
ngspice -b -r ./sim/test-name-here.raw -o ./sim/test-name-here.out ./sim/test-name-here.spice

viewing results

There is a script to interpret the ngspice data generated from each test, i.e. to get the most salient information you can run these scripts

script	description	path to relevant script
vdsat.py	this script will read from the operating point simulation all of the vdsats and the vdsat margins. all margins should be positive to ensure the transistors can be in saturation.	simulation_post_processing/vdsats.py
ppm.py	this script will read from the transient simulation (no variation) to plot the start up of the circuit at 0, 27, and 70 degrees celsius. the script will also print out the ppm of the circuit.	simulation_post_processing/ppm.py
variations.py	this script will read from the transient simulation with variations to plot many startups of the circuit at 0, 27, and 70 degrees celsius. the script will also print out the mean and standard deviation of the resulting reference voltage	simulation_post_processing/variations.py
tempsweep.py	this script will read from the DC temperature sweep simulation to plot the voltage reference value over temperature. the script will also print out a ppm value according to the DC sweep.	simulation_post_processing/tempsweep.py

viewing unscripted results

a printout of relevant operating point voltages is available at sims/tsmc_bandgap_real_op.out.

running your own simulations

if you want to play with simulations yourself, use schematics/tsmc_bandgap_real.sch as your sandbox. the difference between this schematic and the schematics in the test descriptions table is that the schematics in the table have specific simulation commands attached to them. schematics/tsmc_bandgap_real.sch has many ngspice commands in the schematic, which are enabled and disabled by the "spice_ignore" variable in the command listing. if you want a set of simulation commands to be ignore, set spice_ignore = True, but if you want to use a set of simulation commands, set spice_ignore = False.