Measuring and Extracting

This section provides guidelines as well as procedures for performing measurements and extractions of MOSFET devices.

Measurement and Extraction Guidelines

The following guidelines are provided to help you achieve more successful model measurements and extractions.

Setting Instrument Options

Before starting a measurement, you can quickly verify instrument options settings. Save the current instrument option settings by saving the model file to <file_name>.mdl from the model window. Some of the Instrument Options specify instrument calibration. For the most accurate results, calibrate the instruments before taking IC-CAP measurements.

Typical DC and cv instrument options are:

Measuring Instruments

Ensure that the measuring instruments (specified by unit names in the inputs and outputs) are correctly connected to the DUT. Refer to Table 78 for a list of nodes and corresponding measurement units. The quality of the measuring equipment (instruments, cables, test fixture, transistor sockets, and probes) can influence the noise level in the measurements and extracted parameter values.

For some measurements the instruments or test hardware must be calibrated to remove non-device parasitics from the DUT. For MOS devices, stray capacitance due to probe systems, bond pads, and so on should be calibrated out prior to each measurement.

Extracting Model Parameters

For a given setup, you can find the extraction transforms in the Extract/Optimize folder. IC-CAP's extraction algorithms exist as functions; choose Browse to list the functions available for a setup.

When the Extract command is selected from the setup, all extractions in the setup are performed in the order listed in the setup. This order is usually critical to proper extraction performance. Extractions are typically completed instantly and the newly extracted model parameter values are placed in Model Parameters.

Simulating Device Response

Simulation uses model parameter values currently in Model Parameters. A SPICE deck is created and the simulation performed. The output of the SPICE simulation is then read into IC-CAP as simulated data.

Select a simulator from Tools > Hardware Setup or define a SIMULATOR variable. DC simulations generally run much faster than cv simulations. CV simulations can be done in a much shorter time by executing the calc_mos_cbd_model transform instead of running the simulator.

If simulated results are not as expected, use the Simulation Debugger (Tools menu) to examine the input and output simulation files. The output of manual simulations is not available for further processing by IC-CAP functions (such as transforms and plots). For more information refer to "Using the Simulation Debugger" in the IC-CAP User's Guide.

Displaying Plots

The Display Plot function displays all graphical plots defined in a setup. The currently active graphs are listed in the Plots folder in each setup.

Measured data is displayed as a solid line; simulated data is displayed as a dashed or dotted line. After an extraction and subsequent simulation, view the plots for agreement between measured and simulated data. Plots are automatically updated each time a measurement or simulation is performed.

Optimizing Model Parameters

Optimization of model parameters improves the agreement between measured and simulated data. An optimize transform whose Extract Flag is set to Yes is automatically called after any extraction that precedes it in the transform list.

Extracting Parameters

This section describes the general procedure for extracting model parameter data from the UCB MOSFET transistor. The general procedure applies to all types of parameters; differences between extracting one type and another are primarily in the types of instruments, setups, and transforms used. Also included in this section is information specific to DC and capacitance measurements and extractions.

Parameters are extracted from measured data taken directly from instruments connected to the inputs and outputs of the DUT. Using the extracted parameters simulated data can be generated by the simulator. Once measured and simulated data have been obtained, each data set can be plotted and the resulting Plots visually compared in the Plot window.

IC-CAP also extracts model parameters from simulated data. This capability is useful for creating a set of model parameters from the parameters of another model (parameter conversion) or for testing the accuracy of the extraction.

The general extraction procedure is summarized next, starting with the measurement process.

1	Install the device to test in a test fixture and connect the test instruments.

2	Ensure the test fixture, signal source and measuring instruments, and workstation are physically and logically configured for the IC-CAP system.

3	Load the model.

4	Select the DUT. In the DUT Parameters folder, enter the W and L device parameters for the selected DUT.

5	In the Macros folder, select the appropriate macro to enter the process parameters.

6	Select the setup.

7	Issue the Measure command.

8	Issue the Extract command.

9	Issue the Simulate command.

10	Display the results.

11	Fine tune the extracted parameters if needed by optimizing.

DC Measurement and Extraction

In DC parameter extraction, the extracted parameters are directly related to the geometries of the devices being tested. For a DUT to accurately extract DC model parameters, it must have the correct L (drawn or mask channel length) and W (drawn or mask channel width) device parameters. Before executing an extraction or simulation, edit each DUT to ensure the L and W parameters are correct.

Before starting the extraction, enter several process parameters. The most important of these is TOX. Determine TOX by reading the process information for the device, or by measuring the oxide capacitance; TOX is measured in meters. Enter its value directly in Model Parameters, or run the init_parameters macro. Also use the init_parameters macro to enter initial values for XJ, LD, and RS. These initial values can contribute to the accuracy of the extracted parameters. They are overwritten by new values when the XJ, LD, and RS are extracted during the extraction process.

Accurate results depend on the sequence of the extraction. Follow this DC extraction sequence.

Do all of the measurements, followed by all of the extractions, and finally, the simulations. Extraction usually provides a reasonable fit to the measured data, but you can optimize data to attain an increased level of accuracy. Execute the optimization after extracting the DC parameters for each setup.

To perform DC parameter measurements:

1	Choose File > Open > Examples. Select <filename>.mdl and choose OK. Open the model window.

When the model window appears you are ready to begin measurement and extraction operations.

2	Select the DUT large. Enter the values for L and W. To include the effect of WD, enter the following expression for W:

    < value > - 2 · WD

where value is the drawn width and WD is defined as a model variable.

3	In Macros, select init_parameters. Enter the values for TOX, XJ, LD, and RS. Default values can be used by simply choosing OK in each dialog box.

4	Select the idvg setup and issue the Measure command.

5	Repeat these steps for narrow/idvg, short/idvg, and short/idvd.

To perform DC parameter extractions:

1	Select large/idvg. Select the transform extract and execute the selection to extract the LEVEL 2 classical parameters.

2	Repeat this procedure for narrow/idvg, short/idvg, and short/idvd.

All DC model parameters have now been extracted and their values are listed in Model Parameters.

Notes on DC Parameter Extraction

These procedures assume that the large device is large enough to make small geometry effects irrelevant. This condition exists when the device geometries are much larger than LD and WD. For a typical process, 50*50 microns should be sufficient. To improve accuracy, enter the approximate values of LD and WD in Model Parameters so they can be taken into consideration in the first extraction step. A more accurate value for each is produced by the second and third extractions.

When a very large device is not available and you cannot enter LD and WD, try the following:

1	Use the largest available device for the large setup and execute all four steps of the DC extraction.

2	Repeat the extraction sequence starting with the first step (you do not need to re-enter the parameters). The previously extracted parameters (particularly LD and WD) are used as the initial values.

To extract DC parameters when only one size of device is available, extract model parameters using the following sequence. This sequence does not extract geometry-dependent parameters but does extract a subset of parameters to fully model that size device.

1	Perform the large/idvg extraction to obtain the classical parameters.

2	Perform the short/idvd extraction to obtain the saturation parameters.

Enter the same L and W device parameters for both DUTs. The model can be reconfigured so that it has only one DUT with two setups, one similar to idvg and one similar to idvd. Copy the setup from large/idvg and the setup from short/idvd (copy complete setups so the appropriate extraction and optimization functions are included).

If you cannot determine the L and W for a single geometry device (as might be the case with a packaged transistor), set estimated values. The actual values are less important than the ratio between them. An incorrect ratio of W/L results in extraction of an unreasonable value for UO. In general, the mobility parameter UO should be set between 200 and 800. Start the extraction after setting the ratio of L and W to 1, then change the ratio of L to W to scale back the extracted value of UO.

Capacitance Measurement and Extraction

Capacitance parameters can be extracted before or after the DC parameters. The extraction requires that two different DUTs be measured; model parameters are extracted from the second DUT.

The extraction in the cbd1/cjdarea setup requires a single geometry to be measured and produces the parameters CJ, MJ, and PB. The extraction uses a transform set_CJ to find the initial zero bias value of CJ then uses optimization to obtain all three parameter values.

The extraction in the cbd2/cjdperimeter setup requires two geometries to be measured (one in the cbd1/cjdarea setup and one in the cbd2/cjdperimeter setup) that produces the parameters CJ, MJ, PB, CJSW, and MJSW—and therefore a more complete capacitance model.

The extract transform uses the MOSCV_total_cap function to simultaneously solve for the bottom area and sidewall capacitance parameters. To extract the capacitance contributions from the bottom area and the sidewall periphery the geometries must have different area-to-perimeter ratios. The device measured with the cbd1/cjdarea setup should have a high bottom area to perimeter area ratio and the device measured with the cbd2/cjdperimeter should have a low bottom area to perimeter area ratio.

Place the device to be measured into the test fixture. Ensure that the CMs (Capacitance Meters Units) connected to the device correspond to the same CMs in Table 78 for each of the next two measurements. Calibrate the capacitance meter before taking each measurement.

The extractions of the sidewall capacitance parameter sets use the measured data from both setups—measure both setups before performing the extraction.

1	In Macros, select init_cap_parameters and Execute.

2	Enter the drain area and perimeter information.

3	Connect the low terminal of the capacitance meter (CM low) to drain and connect the high terminal (CM high) to bulk.

4	Select cbd1/cjdarea and Measure to measure the first drain-bulk junction capacitance.

5	Repeat steps 3 and 4 for cbd2/cjdperimeter if both geometry sizes are being measured.

Perform the model parameter extractions.

1	If a single geometry was measured, select cbd1/cjdarea. If two different geometries were measured, select cbd2/cjdperimeter.

2	Choose Extract.

Optimization is usually not required for capacitance data.

Notes on Capacitance Parameter Extraction

The drain-to-substrate and source-to-substrate junction capacitances are modeled as a combination of the sidewall and bottom (area) capacitances. To extract the parameters for these capacitances, first measure capacitance against voltage on two different size capacitors. Then execute the extraction command using two setups: cjdarea and cjdperimeter. Execute cjdarea on a square-shaped capacitance with a small sidewall to bottom ratio, and cjdperimeter on a long, narrow junction with a large sidewall to bottom ratio.

Each p-n junction should be reverse-biased when measured. Extraction is performed by the MOSCV_total_cap function. The parameters CJ, MJ, CJSW, MJSW, and PB are calculated from a combination of the two measurements.

Before running the extraction, specify the area and perimeter of the capacitance. Enter these numbers by executing the init_cap_parameters macro. This sets the variables defined at the model level for the area and perimeter of the two DUTs. The parameters AD or AS (area) and PD or PS (perimeter) in the cbd1 and cbd2 DUTs are set by these variables.

Simulating

To simulate any individual setup choose Simulate with an active setup. Simulations can be performed in any order once all of the model parameters have been extracted.

IC-CAP provides a special function, MOSCVmodCBD, to speed up capacitance simulation in the cbd1 and cbd2 DUTs. This function models the simple pn junction capacitance and provides a fast simulation of the CBD capacitance. Use this function to execute a simulation by specifying the transform calc_mos_cbd_model in the setups for the two DUTs and Execute the Transform rather than issuing the Simulate command. For more information, refer to Chapter 9, "Using Transforms and Functions," in the IC-CAP User's Guide.

For more information on simulation, refer to Chapter 6, "Simulating," in the IC-CAP User's Guide.

Displaying Plots

Plots can be displayed from the Plots folder for the setup. To display plots issue the Plot Display command from a DUT to display the plots for all setups in that DUT. The plots use the most recent set of measured and simulated data. Viewing plots is an ideal way to compare measured and simulated data to determine if further optimization would be useful. For more information on plots, refer to Chapter 10, "Printing and Plotting," in the IC-CAP User's Guide.

Optimizing

The optimization operation uses a numerical approach to minimize errors between measured and simulated data. As with the other IC-CAP commands, optimization can be performed at either the DUT or setup level.

Optimization is typically interactive in nature, with the best results obtained when you specify the characteristics of the optimization function.

For more information, refer to Chapter 7, "Optimizing," in the IC-CAP User's Guide.