captab

Dear greywanderer,

greywanderer said:

1. What is "coupling" referring to? Is wire-wire parasitic, "coupling" capacitance?

I believe the "coupling" connotation refers to the fact that there is potential capacitance between wires and devices as well as to circuit ground. An example that suggests that connotation to me is in the Troubleshooting article at URL:

support.cadence.com/.../ArticleAttachmentPortal

greywanderer said:

2. I assume the variable cap reports the maximum capacitance. How does spectre know the operating range for this information? Is it assumed to be the max allowed range based on the process?

I don't get the impression it is the "maximum capacitance" greywanderer.  I think the definition provided in the spectre user manual and shown as Figurre 1 makes it clear that the capacitances reported are based on the specific environment of the capacitance and is not a "defined maximum" or at some "maximum range". Do you agree based on the definition?

Shawn

Figure 1

Hi Shawn,

For question 1 -> It seems that article is refering to "node" vs "node-node" capacitance. So, "coupling" is just non-linear "node-node" cap? Is "Cgd" of a transistor a "coupling" "variable" cap?

For question 2 -> For Cxy/Cxx how does the simulation differentiate between a fixed cap and a variable cap?

Thanks!

Dear greywanderer,

Let me provide a comment or two to my interpretation of your additional questions....other experts in the Forum may have more insight than I (well,..., definitely have!!).

greywanderer said:

It seems that article is refering to "node" vs "node-node" capacitance. So, "coupling" is just non-linear "node-node" cap? Is "Cgd" of a transistor a "coupling" "variable" cap?

and

greywanderer said:

For Cxy/Cxx how does the simulation differentiate between a fixed cap and a variable cap?

I think, perhaps, there may be a basic misunderstanding of what captab is providing. I will also admit that the misunderstanding may, perhaps be on my end!.

In my understanding, the capacitances spectre reports in captab are derived from the derivatives shown above. Recall that charge is conserved by the simulator, and hence an accurate assessment of the variation in charge at a node to ground or between nodes may be made to estimate node to ground or intranodal capacitance. Hence, it is not necessary to have any knowledge concerning the nature of the capacitances at a given node or between nodes as the computations used to generate captab values do not depend on them. The capacitances may be parasitic capacitances associated with one or more traces (i.e., independent of voltage) or device related capacitances (voltage variable), but that will not impact the computation at all. In fact, i don't believe there can be a unique mapping between the capacitances reported by captab and the type of capacitor (i.e., voltage dependent or voltage independent). Any capacitor only serves as a conduit for charge to be injected or removed from a node.

As a simple example, I've put together a couple of pages that details the computation of the capacitance spectre will report for a parallel combination of a fixed and variable capacitor with the hope that might better illustrate what I am attempting to write greywanderer. I hope this is somewhat useful or, at least, provides you with some insight.

Shawn

captab_example_sml_031322.pdf

Hi Shawn,

Thanks for sharing your write up. Perhaps that will help clarify my question. I understand how "C0+C1" can be calculated. But the mystery for me is how "C0" and "C1" are separately reported.

Well, the simulator knows whether a capacitance is fixed or voltage dependent, because it has to compute it. This is also covered in the simulator documentation - if you run <SPECTREinstDir>/bin/cdnshelp and search for captab, there's a section Printing the Node Capacitance Table which explains what it outputs. 

Andrew

Dear grewamderer,

Andrew beat me to respond, but that is not unusual!!

Intuitively, if a small change in node voltage changes the value of the capacitance, then it clearly is not a fixed capacitor. Since spectre has to   compute dq/dv, it will know if a given capacitor between two points has a voltage sensitivity and, if so, will be classified as "variable".

Shawn

Andrew/Shawn,

I was hoping to learn something more about the internals of this simulation. I presume that "captab" is a small signal analysis, so the circuit is "linearized" around the operating point. But once this happens, wouldn't the simulator lose the non-linear details of the circuit? Could you guys point me to some documentation if this is not easy to explain here?

Thanks!

Dear greywanderer,

greywanderer said:

I presume that "captab" is a small signal analysis, so the circuit is "linearized" around the operating point. But once this happens, wouldn't the simulator lose the non-linear details of the circuit?

No. If one performs a transient analysis, one can also get captab information at one or more transient timepoints. Hence, the transient waveforms will include any non-linear effects in capacitor value. However, the captab values are derived from a small-signal analysis at a given timepoint in the simulation. Details on how to do this may be found from the Cadence URL:

https://support.cadence.com/apex/techpubDocViewerPage?xmlName=anasimhelp.xml&title=Virtuoso%20Analog%20Design%20Environment%20L%20User%20Guide%20--%20Setting%20Up%20an%20Analysis%20-%20Transient%20Analysis&hash=pgfId-1013885&c_version=IC6.1.8&path=anasimhelp/anasimhelpIC6.1.8/chap5.html#pgfId-1013885&shareID=1647292204754

and included as Figure 2 below. In addition, you might be interested in Andrew Beckett's response to saving the captab information to a file at the Forum post:

https://community.cadence.com/cadence_technology_forums/f/custom-ic-design/30192/entering-infoname-and-info-captab-analysis-in-ade

Shawn

Figure 2

transient_ana_captab_infotimes_information_sml_031422v1p0.pdf

Dear greywanderer,

greywanderer said:

I presume that "captab" is a small signal analysis, so the circuit is "linearized" around the operating point. But once this happens, wouldn't the simulator lose the non-linear details of the circuit?

No. If one performs a transient analysis, one can also get captab information at one or more transient timepoints. Hence, the transient waveforms will include any non-linear effects in capacitor value. However, the captab values are derived from a small-signal analysis at a given timepoint in the simulation. Details on how to do this may be found from the Cadence URL:

https://support.cadence.com/apex/techpubDocViewerPage?xmlName=anasimhelp.xml&title=Virtuoso%20Analog%20Design%20Environment%20L%20User%20Guide%20--%20Setting%20Up%20an%20Analysis%20-%20Transient%20Analysis&hash=pgfId-1013885&c_version=IC6.1.8&path=anasimhelp/anasimhelpIC6.1.8/chap5.html#pgfId-1013885&shareID=1647292204754

and included as Figure 2 below. In addition, you might be interested in Andrew Beckett's response to saving the captab information to a file at the Forum post:

https://community.cadence.com/cadence_technology_forums/f/custom-ic-design/30192/entering-infoname-and-info-captab-analysis-in-ade

Shawn

Figure 2

transient_ana_captab_infotimes_information_sml_031422v1p0.pdf