Intro

• Here are a couple of python scripts to process MOS operating point simulation data and present it in a nice table format.

• The scripts MOS_OP2 and MOS_OP in this main folder are to process simulation data from a cadence Spectre simulation.

• In the LTSpice folder you find a script that works with LTspice (and likely other spice3 based simulators like ngspice).

• The scripts have been tested with a limited set of models. You might have to make some modifcations if you use different MOS models.

• Any issues, file an issue through Github.

MOS_OP2

• Browse through MOS_OP below first.
• MOS_OP2.py has a simpler interface but it has to be run on a machine with access to the cadence psf command.
• psf_utils has to be installed in your python3 installation (see below under 'Tools Needed').

Usage

• invoke with python3 ./MOS_OP2.py config_device_names.json or make MOS_OP2.py executable
• device_names.json now also needs to include the path to the simulation result dcOpInfo.info and element.info files.
• Example config_device_names.json for a circuit called tb_diff_amp.

{"simulation_dir": "/workdir/pk171/simulation",
 "design_name": "presized_OTA_tb/spectre/schematic/psf",
 "transistor_names": 
{ "M1b": "I0.M1",
 "M2b": "I0.M2",
 "M3b": "I0.M3",
 "M4b": "I0.M4",
 "M5b": "I0.M5",
 "M6b": "I0.M6",
 "M7b": "I0.M7",
 "M8b": "I0.M8"}
}

• The syntax is largely self-explanatory and based on MOS_OP (see below);
  • simulation_dir needs to point to the simulation directory that cadence ADE is using;
  • design_name is the path in simulation_dir to the folder that contains the .info files;
    • the script will look for the .info files in the folder that is the join of simulation_dir and design_name
  • transistor_names is a dictionary of transistor names and their full schematic names.
• if the .info files are not in ascii, the script will invoke psf; make sure it is in your PATH
• a file operating_point.csv is written in the working directory

MOS_OP

Quick Demo

• clone the repository
• start a terminal on a computer with python3, numpy and pandas
• install psf_utils with pip3 install psf_utils (consider using a virtual environment)
• go into the example files of the repository: cd example_files_presized_OTA_tb
• execute the script with an example device dictionary file python3 ../MOS_OP.py dcOpInfo.info.ascii element.info.ascii device_names_I0.json producing the following output on your terminal and creating a operating_point.csv.

                  M1b         M2b         M3b         M4b         M5b         M6b         M7b         M8b
w                8.0u        8.0u       24.0u       24.0u       96.0u       32.0u       16.0u        8.0u
l              500.0n      500.0n      500.0n      500.0n      500.0n      500.0n      500.0n      500.0n
m                 1.0         1.0         1.0         1.0         1.0         1.0         1.0         1.0
as               4.0p        4.0p        8.0p        8.0p       26.0p       10.0p        6.0p        4.0p
ad               2.0p        2.0p        6.0p        6.0p       24.0p        8.0p        4.0p        2.0p
ps               9.0u        9.0u       27.0u       27.0u      108.0u       36.0u       18.0u        9.0u
pd               8.0p        8.0p       16.0p       16.0p       52.0p       20.0p       12.0p        8.0p
ids             99.5u       99.6u      -99.5u      -99.6u     -453.9u      453.9u      199.2u      100.0u
vgs            736.2m      736.5m     -722.3m     -722.3m     -730.5m      644.9m      644.9m      644.9m
vds               1.3         1.3     -722.3m     -730.5m        -1.3         1.2      513.5m      644.9m
vdsat          218.6m      218.7m     -258.5m     -258.5m     -267.6m      228.2m      218.7m      218.5m
region     saturation  saturation  saturation  saturation  saturation  saturation  saturation  saturation
vbs           -513.5m     -513.5m         0.0         0.0         0.0         0.0         0.0         0.0
vth            512.9m      513.0m     -479.6m     -479.6m     -479.3m      398.1m      412.0m      412.8m
vgsteff        223.2m      223.3m      243.7m      243.7m      252.0m      246.0m      232.1m      231.4m
gm             771.2u      771.5u      688.4u      688.9u        3.0m        3.3m        1.5m      755.2u
gds             14.2u       14.2u       11.1u       11.0u       36.6u       61.1u       45.9u       18.8u
gmb            140.7u      140.7u      226.0u      226.2u      997.0u      748.9u      343.7u      173.5u
gmoverid          7.7         7.7         6.9         6.9         6.7         7.2         7.5         7.6
self_gain        54.2        54.1        61.9        62.4        82.6        53.3        32.5        40.1
cgs             20.5f       20.5f       57.5f       57.5f      229.9f       82.8f       41.3f       20.6f
cgsovl           3.8f        3.8f       13.7f       13.7f       54.8f       15.4f        7.7f        3.8f
cgb              1.1f        1.1f        4.7f        4.7f       18.9f        4.7f        2.4f        1.2f
cgbovl         450.6z      450.6z      417.8z      417.8z      417.8z      450.6z      450.6z      450.6z
cgd              3.5f        3.5f       13.7f       13.7f       54.8f       14.2f        7.1f        3.6f
cgdovl           3.8f        3.8f       13.7f       13.7f       54.8f       15.4f        7.7f        3.8f
cbd              2.2f        2.2f        8.8f        8.8f       31.0f        9.9f        6.0f        2.9f
cjd              2.2f        2.2f        8.8f        8.8f       30.9f        9.7f        5.8f        2.8f
cbs              9.6f        9.6f       10.4f       10.4f       42.4f        7.4f        3.7f        1.9f
cjs              8.4f        8.4f         0.0         0.0         0.0         0.0         0.0         0.0
csd            122.0a      122.0a      -17.6a      -16.9a       -8.5a      496.5a      210.5a      116.0a
cm               7.7f        7.7f       20.6f       20.6f       82.4f       30.8f       15.4f        7.7f
cmb              1.4f        1.4f        6.6f        6.6f       26.4f        7.1f        3.5f        1.8f
cmx              1.6f        1.6f        2.7f        2.7f       10.4f        7.0f        3.4f        1.7f
fug              4.9G        4.9G        1.4G        1.4G        1.6G        5.1G        4.7G        4.7G

• there is another example with more devices: python3 ../MOS_OP.py dcOpInfo.info.ascii element.info.ascii device_names_all.json
• operating_point.csv is overwritten during each execution.

Usage

The python script MOS_OP.py takes in simulation results from a spectre operating point simulation and tabulates them nicely on the terminal and saves them in a csv file.

Usage:

python3 MOS_OP.py <name dcOPInfo.info ascii file> <name element.info ascii file> <json file with device name dictionary>

The operating point table is printed to stdout and is saved in operating_point.csv; if there is an existing operating_point.csv, it will be overwritten.

After a spectre DC Analysis with Save Operating Point, spectre saves the operating point information in the dcOPInfo.info file and the transistor sizing information in the element.info file in the folder with the simulation results; that folder is typically located in <simulation results folder>/<name of the cell>/spectre/schematic/psf/. So if you are simulating presized_OTA_tb and you ask ADE to save your simulation results in ~/simulation, you should look in ~/simulation/presized_OTA_tb/spectre/schematic/psf/ after completing the simulation.

These files are typically in binary format. You can convert them to ascii with the psf command from cadence; on a workstation where you can execute virtuoso you should be able to run:

psf dcOPInfo.info dcOPInfo.info.ascii
psf element.info element.info.ascii

Those ascii files can be read by MOS_OP.py. There are example files in the example_files_presized_OTA_tb folder.

Next, you need to make a json dictionary of the (subset of the) devices you want the script to display the operating point for. The format is:

{"device1_label_used_by_script": "spectre_reference_of_device1",
....
 "devicen_label_used_by_script": "spectre_reference_of_devicen"}

So, for example, the following dictionary names I0.M1, i.e. the M1 device in subcircuit I0, as "M1b" in the print out and csv file:

{"M1b": "I0.M1",
 "M2b": "I0.M2",
 "M3b": "I0.M3",
 "M4b": "I0.M4",
 "M5b": "I0.M5"}

There are dictionary examples in the example_files_presized_OTA_tb folder: e.g., device_names_I0.json

You do not need to include all devices, you can select devices of interest only.

Output

The output parameters have their 'usual' meanings and most are taken directly from the cadence results.

Only for the capacitances, the script modifies the cadence output. The signs of transcapacitances are converted to the convention in Tsividis' MOS book, namely, $C_{xx} = \partial Q_x/\partial V_{x}$ and $C_{xy} = -\partial Q_x/\partial V_{y}$ when $x\neq y$; so with this definition, if $C_{xy}$ was a classical two-terminal capacitor then the sign of the capacitance would indeed be positive. Note that in spectre, $C_{xy} = \partial Q_x/\partial V_{y}$ for all $x$ and $y$.

The following transcapacitances are calculated corresponding to small-signal model in Fig. 8.5 in Tsividis` MOS book: $C_m = C_{dg}-C_{gd}, C_{mb} = C_{db}-C_{bd}, C_{mx} = C_{bg}-C_{gb}$.

Small-signal equivalent model for the intrinsic part of the MOS transistor, i.e. not including cjs, cjd, cgsovl, cgdovl, cgbovl

Note: The current $g_m v_{gs}$ is going to be bigger than $j\omega C_m v_{gs}$, except at very high frequencies; since $C_{m} < C_{gs}$, those frequencies are beyond the $f_T$ of the transistor. Similarly when $v_{bs} \neq 0$, the current due to $g_{mb}$ is typically larger than due to $C_{mb}$. And, $C_{mx} \approx 0$ in strong inversion.

For more details see Y. Tsividis and C. McAndrew, Operation and modeling of the MOS transistor, 3rd ed. New York: Oxford University Press, 2011.

Tools Needed

• python 3
• numpy
• pandas
• psf_utils is a library for python that enables reading the ascii psf files available from https://pypi.org/project/psf-utils/ and can be installed with: pip install psf-utils or pip3 install psf-utils. You need version 1.7 or later.

Example Circuit

The simulation example files provided in example_files_presized_OTA_tb are for this single-ended two-stage Miller OTA:

Two OTAs have been placed in a testbench:

The OTA instances are I0 and I7. Both have the same DC operating point. The operating point for the transistors in the I0 instance can be obtained with this device names json file with the following output and csv output. The output contains the transistor sizes which you can also verify from the netlist.

The 0.25um CMOS transistor models used for the simulation are available at Nagendra Krishnapura's CAD Tools page.

Other Jupyter Notebook Examples

There are a couple of jupyter notebooks included; they are not polished and will likely not work out of the box on your system.

• MOS_Operating_Point_v0.1d.ipynb uses psf_utils to extract, explore and process the operating point information in a notebook.

• Miller_OTA_Performance_from_Operating_Point.ipynb reads in the operating_point.csv and uses approximate analytical expressions to compute the performance parameters of the OTA. These are just drafts, use at your own risk, the formulas can be wrong.

Thanks

Thanks to @KenKundert for updating psf_utils so it can read the element.info files (Dec. 2023).