Spectre S-parameter analysis port definition

Hi Radu,

You don't have to specify the ports on the analysis UI; this is only needed if you've either not given a port number on the instances of the port component, or you want to have a subset of the ports on the schematic, or the ports in a different order than the port numbers on the schematic.

So, if you have lots of ports on the schematic, and the port numbers are sequential on those ports, then you can just leave the field blank. That avoids the need to use wildcards, I'd expect.

Andrew.

Hi,

Thanks for the quick answer. I know this and have tried it, but unfortunately I hit a snag, there is a block in the schematic which is from an IP lib and has a port like instance in it (strangely it is an analogLib psin instance when I hit Q in schematic, but appears as port in the nertlist) with no num property, and spectre gives me an error, it seems it expects all ports to have num property. I had hoped if I leave the ports list empty it would just do the ports which have num, but no luck. I would need to somehow identify all such instances in the schematic and somehow handle them.

I am not adding ports / running sims, I am in fact developing a skill tool that does some simulations automatically on a schematic. I do not add my ports physically in the schematic, I generate a netlist format file, generate the port strings automatically (e.g. PORT1 (net018 0) port r=50 dc=0 type=dc etc), which I include in the main netlist. The tool does not alter the schematic in any way. It is very easy for me to add unique consecutive num properties to the ports, which I did at first, until I hit the error. If wildcard was a solution, I can give my ports appropriate names and would not have to deal with the issue. Otherwise I would have to somehow warn users that ports exist in their schematic and that should change it or somehow alter the main netlist wherever I find a port to add an unique number.

I do not run the sim directly from ADE-L, but generate an ocean file, so I could also modify the ocean file manually to avoid adding ports in the analysis field, but would rather not do this.

Radu

Hi Radu,

Rather odd that there would be a port (or psin, which is just an older component which netlisted to port) within the IP - that strikes me as being a bad thing to do in IP! You shouldn't have simulation test bench components within the design - rather bad practice.

A couple of ways you could resolve this:

1. In the hierarchy editor, use "Bind to open" for the psin instance within the IP - this will mean that it doesn't get netlisted. That may or may not be OK - as maybe it's dependent upon having the impedance present (I'd be worrying a lot though if that was the case)
2. Add an alter statement in your included netlist (change the dev to the hierarchical instance name of the psin, and the value to the unused contiguous port number you want to use):
   fixPort alter dev=port2 param=num value=2

The second approach should then allow you to run the sp analysis (I just tried this, and it worked a treat).

Andrew.

Hi Radu,

Rather odd that there would be a port (or psin, which is just an older component which netlisted to port) within the IP - that strikes me as being a bad thing to do in IP! You shouldn't have simulation test bench components within the design - rather bad practice.

A couple of ways you could resolve this:

1. In the hierarchy editor, use "Bind to open" for the psin instance within the IP - this will mean that it doesn't get netlisted. That may or may not be OK - as maybe it's dependent upon having the impedance present (I'd be worrying a lot though if that was the case)
2. Add an alter statement in your included netlist (change the dev to the hierarchical instance name of the psin, and the value to the unused contiguous port number you want to use):
   fixPort alter dev=port2 param=num value=2

The second approach should then allow you to run the sp analysis (I just tried this, and it worked a treat).

Andrew.

Thank you for the suggestion, I will see if I can get rid of that block and if not, I will try the second option. It seems a bad idea not to make use of the num property feature.

Radu