EM-based models leverage the tight integration of the EM simulator (EMSight) and linear simulator in the AWR Microwave Office software to interactively construct discontinuity models based on full-wave electromagnetic simulations of the parameterized structure. The model automatically constructs and simulates the discontinuity using EMSight. A database is constructed from the EM simulations and then interpolated for the specific data points. When all points are filled in the database and saved in a file, the simulation is extremely fast with EM-based accuracy. The models require the EMSight option in AWR Microwave Office software to be functional. An addition advantage of X-models is they can predict when the model's accuracy will start to decay, something that closed-form models cannot do. See “Upper Frequency Limitations” for more details.
When using X-models, you should understand three issues.
Which models are X-models
How to properly set up substrate parameters for X-models
How to fill a new database
X-models have model names that end in either X or X$. Currently they fall into one of five substrate types: Microstrip, Stripline, CPW, Suspended Microstrip, or Suspended Stripline.
The simplest way to view all of the X-models for a given substrate type is to open the standard example shipped with the software to help generate new X-models (example specifics are discussed in a further section):
In the Open Example Project dialog box that opens, type "autofill" to filter for only the X-model autofill examples.
Open the specific type you are interested in and view the only schematic in those projects. This schematic contains all of the X-models for that given type.
There are many X-model database files shipped with the AWR Design Environment software installation, see “X-model Tables Shipped with the AWR Design Environment” for the exact list. If your substrate values match any of these, there will be no issues using X-models in your first simulation. However, if your substrate parameters do not match a shipping substrate, simulation cannot occur and a "The X-model Database has not been filled" error message displays. There are two options to correct this error.
The simplest way to correct this error is to understand the relationship between the X-model fixed parameters and statistical parameters. For details on these parameter definitions, see “Interpolation Database Concept”. For microstrip, the fixed parameter is ErNom and its statistical parameter is Er. By default, the ErNom parameter has the same value as the Er value, as shown in the following example by using the syntax "Er@".
This syntax indicates "use the value of the Er parameter". However, you can use any value for the ErNom parameter. The fixed parameter (ErNom for microstrip) chooses the proper X-model database to use and the Er parameter is used in the simulation. These values can be 10% different and still be accurate. You can also set your fixed parameter to one of those shipped with the software as long as the Er you are using is less than 10% different than this value. For example, if you want to simulate with Er=10 and see that there is a 10.2 and 9.8 available in the shipping substrates, you can set Er=10 and ErNom = 10.2 and the X-models will work correctly.
The second way to correct this problem is to fill new X-model database tables. Since this is a lengthy process (but only has to be done once per substrate values), Cadence recommends using the first approach before filling a new database. The following section discusses how to fill a new database.
When you decide you want to fill a new X-model database, Cadence strongly suggests using the autofill examples shipped with the AWR Design Environment software to do so. These examples come with design notes to help you know exactly what to do, and scripts to help you give the X-model files meaningful names after they are filled. These examples automatically fill the entire database for a particular set of fixed parameters. In this way, all of the required EM simulations can be completed and stored without user interaction. Typically, you should "autofill" all of the discontinuity models for one substrate at one time. To find these examples:
In the Open Example Project dialog box, type "autofill" to filter for only the X-model autofill examples.
Open the specific type you are interested in filling and read the design notes.
After the autofill example is done, the new X-model database files are ready to use with the project. See “X-model Filenames and Locations” for details on the X-model names and file locations.
The software decides to fill a new X-model database based on the substrate settings and the Autofill parameter on each X-model. If the proper X-model database files do not exist for a model and its substrate values, and Autofill is set to 1, then the files for that model are created. A common mistake is to use one X-model in a design, (for example, a MBEND90X) when the database has not been filled. Based on the error, you change the Autofill parameter to 1 and wait for a long time while that specific table fills. If you then decide to use a different X-model such as the MTEEX, for example, then that database needs to fill. It is much better to use the autofill examples and run them overnight so you know all tables for a given substrate are filled and ready.
Because there are fixed and statistical parameters for the substrates that X-models use, you can vary the statistical parameters using Tuning, Yield, and Optimization. The models issue a warning when the statistical value varies 10% from the fixed value. The default configuration for the fixed parameters using the "@Er" type syntax does not work with Tuning, Yield, and Optimization. You must change these values (ending in Nom) to the fixed value number. If you had to change the fixed parameters to find the right X-model database files, then you don't need to do anything different. If you did not have to do this, simply set the Nom parameters to be identical to the non-Nom parameter names (Er and ErNom, for example) and then vary the non-Nom parameter. If you don't make this change one of two things will happen. If the Autofill on the individual X-models is set to 1 (not the default) then a new database file must be filled at each iteration. If Autofill is set to 0, simulation errors about database files not being filled display.
The following sections list the X-models and Er values shipped with the AWR Design Environment software.
The following substrates are filled for the MSTEPX, MTEEX, MBENDRWX, MGAPX, MCROSSX, MBEND90X, MLEFX and MOPENX models:
Er = 2.2 Er = 2.33 Er = 2.45 Er = 2.5 Er = 2.94 Er = 3 Er = 3.2 Er = 3.25 Er = 3.38 Er = 3.4 Er = 3.48 Er = 3.66 Er = 4.5 Er = 4.99 Er = 6 Er = 6.15 Er = 9.2 Er = 9.7 Er = 9.8 Er = 10.2 Er = 10.8 Er = 11.2 Er = 12.4 Er = 12.9
The following substrates are filled for the CPWABRGX, CPWBENDX, and CPWTEEX models:
Er = 3.38, H=813um, Hab=35.6um, No Cover, No Gnd Er = 3.38, H=1000um, Hab=50um, NoCover, No Gnd Er = 9.8, H=508um, Hab=25.4um, HCover=508um, Gnd Er = 12.4, H=280um, Hab=1.5um, No Cover, Gnd Er = 12.9, H=100um, Hab=3um, No Cover, No Gnd
The following substrates are filled for the CPWLINX model:
Er = 3.38, H=813um, Hab=35.6um, No Cover, No Gnd, T=35.6um, Rho=1, Tand=0, Acc_10 Er = 3.38, H=1000um, Hab=50um, No Cover, No Gnd, T=0um, Rho=1, Tand=0, Acc_10 Er = 9.8, H=508um, Hab=25.4um, HCover=508um, Gnd, T=0um, Rho=0, Tand=0, Acc_10 Er = 12.4, H=280um, Hab=1.5um, No Cover, Gnd, T=0um, Rho=1, Tand=0, Acc_10 Er = 12.9, H=100um, Hab=3um, No Cover, No Gnd, T=2um, Rho=1, Tand=0.001, Acc_10
The following substrates are filled for the MSBND90X, MSCROSSX, MSOPENX, MSSTEPX, and MSTEEX, models:
Er = 3.38, Tand=0, H1=500um, H2=1000um
The following substrates are filled for the MS1LINX, MS2CLINX and MSBCPLX models:
Er = 3.38, Tand=0, H1=500um, H2=1000um, T=25um, Rho=1, Acc=1
The following substrates are filled for the SSBND90X, SSCROSSX, SSOPENX, SSSTEPX, and SSTEEX models:
Er = 3.38, Tand=0, H1=1000um, H2=500um, H3=1000um
The following substrates are filled for the SS1LINX, SS2CLINX and SSBCPLX models:
Er = 3.38, Tand=0, H1=1000um, H2=500um, H3=1000um, T=25um, Rho=1, Acc=1
The following sections discuss X-model specifications.
The parasitic reactance associated with discontinuities is the circuit realization of stored energy in evanescent higher-order modes of the structure. As the frequency approaches the cutoff frequency of the first higher-order mode of the structure, the stored energy increases dramatically, making this structure unsuitable for typical circuit design. The upper frequency limit of this model is set at 80% of an estimate of the cutoff frequency of the first higher-order mode encountered in the discontinuity. For example, a microstrip line with a substrate thickness which is small compared to wavelength encounters a first higher mode when the effective width of the widest line of the structure is one-half of a guided wavelength. An approximation of this condition can be obtained with the following equation:
where Zo=Characteristic Impedance of TxLine with W/H = 4.0 and c = speed of light in the chosen units. The upper frequency limit is defined by the 80% of this calculated cutoff frequency:
For previous frequencies, this maximum frequency, an extrapolation based upon results within the limits of the parameterized model, is used. A warning is issued indicating that results produced are not based on surrounding EM simulations. Typically, this warning is not seen during linear circuit design, but rather during harmonic balance-based nonlinear simulations where the number of harmonics used causes this limit to be violated.
In the program installation directory there is a folder named
EM_Models that contains several defined X-models that Cadence provides.
These models are stored as binary (non-ASCII) files that have an
extension, and are name coded to their function. The naming convention is
<substrate>_<parameters>_<model>, where <substrate> is the type of
substrate, <parameters> are the parameters used in filling the model, and <model>
is the model name.
The location of newly user-filled X-models is the
directory. To locate this directory choose .
You can set up AWR Design Environment configurations so the
EMModelUser directory is a
different folder (such as a network path). This is useful if a group of designers wants to
share a common set of X-models. See the
Installation Guide for details on
how to redirect folder names.
Although most substrate dielectric values are covered by the X-model files included with the AWR Design Environment software, there may be times when you need to create a new X-model file. When you do so, the file is assigned a name based on the discontinuity type and an incrementing 3-digit number to ensure that it is unique. For example, if a Microstrip 90-Degree Angle Bend model named "MS90B000.EMX" exists; if you create another model it is named "MS90B001.EMX". This naming method is useful for avoiding naming conflicts, however you should rename new X-model files as follows.
After creating a new X-model file, Cadence recommends that you rename it to prevent future confusion about what the file may represent, especially when the files are shared within an organization, transferred between computers, or if you must save the models when upgrading the software. Cadence recommends creating unique file names to distinguish them from other X-model files. For example, you can append your company name to the beginning of the file name and use the Cadence convention for defining the discontinuity and substrate.
Note that the
.EMX extension and directory location of the file are
the only significant items in loading the files. The rest of the information is stored in the
binary file and is used to load the correct file. Therefore, you can rename the file once
created to apply user significance. For example, you could rename the
MS90B9_80.emx file (the EM data for a 90-degree bend on alumina) to
AluminaBend.emx, and the program still functions correctly.
The autofill examples that ship with the AWR Design Environment software contain scripts to help rename the files generated by the autofill process. See the design notes for these examples for information on running the scripts.
The following table lists all current Cadence X-models and the name given to them automatically by the software when a new model is created.
The model operates via the interpolation of EM simulations of surrounding evaluation points of the model parameters. Importantly, the EM simulation at the surrounding points are stored to disk, eliminating the need to perform repetitive EM simulations at these points. The EM simulations are performed on an N-dimensional grid of the parameters. To limit the size of the database required for storage in RAM, the model parameters are divided into four groups consisting of the following:
|Independent||Parameters that typically vary during the design process on one particular substrate.|
|Scalable||Parameters that relate to the independent parameters via a scaling of the structure to a different size.|
|Fixed||Parameters that do not vary in the design process.|
|Statistical||Parameters that do not vary in the design process but vary in manufacturing and require small changes for statistical and sensitivity analysis.|
The database is constructed by varying the independent parameters and keeping the other three parameter types constant. A finite range and number of values for each independent variable is selected and EM simulations are required at each of these evaluation points. All EM simulation points the model requires are checked to ensure they have been previously simulated. Any required simulations that are missing from the database are automatically completed and stored. After the EM simulations are complete, an interpolation of the data are performed to predict the performance of the discontinuity at the specified evaluation point.
From the previous discussion, the data storage requirements suffer the problem of dimensionality. For every additional independent parameter added, the number of EM simulation points increases by a factor equal to the number of points simulated in that dimension. For example, if there are three independent parameters: W1, W2, and W3, with eight values for each parameter, and 10 frequency values, the database would require 8*8*8*10 = 5120 EM simulations to be completely filled. As a result, the number of independent variables is kept to a minimum.
Scalable parameters can be changed without needing to create a new database. This type of parameter scales the EM structure so that the solution for the model can be related to a scaled frequency solution. As a result, when a data table has been created with the scaled solution already contained, the model does not have to create those particular solutions. For the MTEEX model, simulations at one particular height of a substrate can lead to electrical parameters of a second height by scaling the frequency by the ratio of the two H dimensions. Scalable parameters do not add a dimension to the database and can be evaluated for any substrate height. Therefore, when observing the EM simulations, the simulation frequencies may be quite different than the project simulation frequency range due to frequency scaling.
Fixed parameters are fixed for the given database. These parameters do not vary for a typical design so variation of these parameters is not evaluated in one given database. Importantly, you should not attempt to tune on a fixed parameter as this results in the generation of a new database for each value of the fixed parameter evaluated.
Statistical parameters are not varied during the design process but do vary due to manufacturing tolerances. To perform statistical analysis, these parameters require a small variance to quantify the final design or to perform design centering. During acquisition of the database this type of parameter is considered fixed. A first-order model of the dependence upon this parameter is used to vary the solution. This method is valid for small variations from a nominal value. For microstrip models, the only statistical parameter is the relative dielectric constant of the substrate. Small variations in Er <10% recommended and <20% required is more than adequate to perform statistical evaluation. Another possible use involves the frequency- or temperature-dependent nature of the dielectric constant. This model supports these types of variations in Er up to the previously mentioned range.