You are right on! Indeed, all the problems can be solved by telling MODELZEN how to combine those expanded instances into a group. You can do so by assigning a 'group' attribute to the device mapping entries listed in the 'devo_devicemap' dictionary of the technology configuration file.

Let’s say you'd like MODELZEN to treat two instances with the following patterns as a single transistor. That is, you'd like MODELZEN to take a pair of 'nch_svt_mac' and 'nch_svt_mac_pode' instances connected to the nets 'A', 'B', 'C', and 'D':

M1 ( A_dmy  B_dmy  C_dmy  D ) nch_svt_mac l=0.008u w=0.308u …
M1__macro ( A  B  C  D  A_dmy  B_dmy  C_dmy  D ) nch_svt_mac_pode l=72n …

and convert them into a single 'nmosfet' instance with its drain, gate, source, and bulk terminals connected to the nets 'A', 'B', 'C', and 'D', respectively.

nmosfet  M1 (.d(A), .g(B), .s(C), .b(D));

Here is how you can describe the corresponding grouping rules in the technology configuration file. For your convenience, first define a dictionary (e.g. named 'group_pode') that defines the naming patterns of the instances to be grouped and their terminal lists. Here is an example for our case:

# group property to merge transistor instances with their PODE instances
group_pode = {
    "@(INSTNAME)"           : dict(terms=['*', '*', '*', 'b']),
    "@(INSTNAME)__macro"    : dict(terms=['d', 'g', 's', 'b', '*', '*', '*']),
}

It says, if a transistor instance is found with another instance with a '__macro' suffix in its name, MODELZEN shall group them using the nets listed in the positions of 'd', 'g', 's', and 'b' as the drain, gain, source, and bulk terminals of the combined transistor instance, respectively. Here, the "@(INSTNAME)" is an embedded Python expression where 'INSTNAME' is a variable containing the instance name under consideration. With this embedded Python expression, you can describe various naming patterns. For example, "X@(INSTNAME[1:])" is the instance name of which prefix is replaced with an 'X' and "@(INSTNAME.upper())" is the instance name all of which letters are converted to uppercases. On the other hand, the asterisks '*' in the terminal lists denote the internal nets.

Next, assign this dictionary as the 'group' attribute of the devices listed in the 'devo_devicemap' dictionary of the technology configuration file. The example shown below specifies that the instances of 'nmos_svt_mac', 'pmos_svt_mac', 'nmos_lvt_mac', and 'pmos_lvt_mac' shall be grouped with their PODE instances if the instances with the said naming patterns are found in the netlist.

# global device mapping
devo_devicemap = {
    "nmos_svt_mac"  : dict(type="nmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
    "pmos_svt_mac"  : dict(type="pmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
    "nmos_lvt_mac"  : dict(type="nmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
    "pmos_lvt_mac"  : dict(type="pmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
}

Lastly, make sure the 'dev_model' entry in the technology configuration file contains a line that includes the SPICE model library defining the PODE devices (e.g. the subcircuit definition for 'nch_svt_mac_pode' and so on), since MODELZEN will now characterize the combined set of these instances as a single transistor. Otherwise, MODELZEN may fail during the SPICE/Spectre characterization phase with an error message saying that the definition for the said model or subcircuit is not found.

You are right on! Indeed, all the problems can be solved by telling MODELZEN how to combine those expanded instances into a group. You can do so by assigning a 'group' attribute to the device mapping entries listed in the 'devo_devicemap' dictionary of the technology configuration file.

Let’s say you'd like MODELZEN to treat two instances with the following patterns as a single transistor. That is, you'd like MODELZEN to take a pair of 'nch_svt_mac' and 'nch_svt_mac_pode' instances connected to the nets 'A', 'B', 'C', and 'D':

M1 ( A_dmy  B_dmy  C_dmy  D ) nch_svt_mac l=0.008u w=0.308u …
M1__macro ( A  B  C  D  A_dmy  B_dmy  C_dmy  D ) nch_svt_mac_pode l=72n …

and convert them into a single 'nmosfet' instance with its drain, gate, source, and bulk terminals connected to the nets 'A', 'B', 'C', and 'D', respectively.

nmosfet  M1 (.d(A), .g(B), .s(C), .b(D));

Here is how you can describe the corresponding grouping rules in the technology configuration file. For your convenience, first define a dictionary (e.g. named 'group_pode') that defines the naming patterns of the instances to be grouped and their terminal lists. Here is an example for our case:

# group property to merge transistor instances with their PODE instances
group_pode = {
    "@(INSTNAME)"           : dict(terms=['*', '*', '*', 'b']),
    "@(INSTNAME)__macro"    : dict(terms=['d', 'g', 's', 'b', '*', '*', '*']),
}

It says, if a transistor instance is found with another instance with a '__macro' suffix in its name, MODELZEN shall group them using the nets listed in the positions of 'd', 'g', 's', and 'b' as the drain, gain, source, and bulk terminals of the combined transistor instance, respectively. Here, the "@(INSTNAME)" is an embedded Python expression where 'INSTNAME' is a variable containing the instance name under consideration. With this embedded Python expression, you can describe various naming patterns. For example, "X@(INSTNAME[1:])" is the instance name of which prefix is replaced with an 'X' and "@(INSTNAME.upper())" is the instance name all of which letters are converted to uppercases. On the other hand, the asterisks '*' in the terminal lists denote the internal nets.

Next, assign this dictionary as the 'group' attribute of the devices listed in the 'devo_devicemap' dictionary of the technology configuration file. The example shown below specifies that the instances of 'nmos_svt_mac', 'pmos_svt_mac', 'nmos_lvt_mac', and 'pmos_lvt_mac' shall be grouped with their PODE instances if the instances with the said naming patterns are found in the netlist.

# global device mapping
devo_devicemap = {
    "nmos_svt_mac"  : dict(type="nmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
    "pmos_svt_mac"  : dict(type="pmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
    "nmos_lvt_mac"  : dict(type="nmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
    "pmos_lvt_mac"  : dict(type="pmosfet", vmax=1.2, max_leveldc=3, group=group_pode),
}

Lastly, make sure the 'dev_model' entry in the technology configuration file contains a line that includes the SPICE model library defining the PODE devices (e.g. the subcircuit definition for 'nch_svt_mac_pode' and so on), since MODELZEN will now characterize the combined set of these instances as a single transistor. Otherwise, MODELZEN may fail during the SPICE/Spectre characterization phase with an error message saying that the definition for the said model or subcircuit is not found.