Designer’s Guide Community

modern_analog

Junior Member

Offline

Posts: 12

Cgs, Cgg, CGd etc. in cadence
Jun 16^th, 2004, 11:37am

Do the values of caps which we get from cadence .op carry any physical significance.

ie is Cgg = Cgs + Cgd + Cgb. In this case, should I take the modulus of the values.

I feel it is not the case, but just want to re-confirm, since some of the caps are negative (due to derivative of Q being negative)

If this is not the case, how can we have cadence give us the capacitance at any node and its constituents. For instance if cadence says node n1 has 1pf of cap, and node n1 has 4 transistors connected to it, how can we figure out which transistor is contributing how much cap and what are the constituents (Cgs, Cgd etc.) of that transistor.
We used to have captab option in hspice.

thanks

Back to top

IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: Cgs, Cgg, CGd etc. in cadence-
Reply #1 - Jun 16^th, 2004, 2:26pm

I assume you're talking about bsim3v3 models here.

Here's a note that I wrote some years ago on this:

Quote:

The values you are seeing are correct because the values of these
parameters are the partial derivatives of the terminal charges with
respect to the terminal voltages, so they can be negative. They are NOT
physical two terminal capacitors. See the UCB bsim3v3 documentation,
equations 4.3.28 - 4.3.31.

An earlier spectre implementation only gave the values of three
capacitors which were closer to the intuitive value of what the
capacitance should be, but many customers asked for the complete set of
partial derivatives, so that is what spectre currently provides.

Here is a summary of the differences in the capacitance operating point
figures between 4.4 and 4.4.1 BSIM3V3 models in spectre.
---------------------------------------------------------------

In the following section Cxx refers to the parameter calculated by
the bsim evaluator inside the code. The lower case cxx refers to the
value reported by the operating point.

In 4.4:

cgs = -Csg + pModel->OverlapCgs * pInst->Width * pInst->MFactor
cgd = -Cdg + pModel->OverlapCgd * pInst->Width * pInst->MFactor
cgb = -Cbg + pModel->OverlapCgb * pInst->Length * pInst->MFactor
cbd = Cjd
cbs = Cjs

In 4.4.1:

Cgsovl = pInst->pSDModel->OverlapCgs * pInst->MFactor;
Cgdovl = pInst->pSDModel->OverlapCgd * pInst->MFactor;
Cgbovl = pInst->pSDModel->OverlapCgb * pInst->MFactor;

cgs = Cgs - Cgsovl
csg = Csg - Cgsovl
cgd = Cgd - Cgdovl
cdg = Cdg - Cgdovl
cbd = Cbd
cbs = Cbs
cjd = Cjd
cjs = Cjs

Here's a mapping between the parameter name in 4.4, and what it
corresponds to in 4.4.1:

4.4 4.4.1
------------
cgs = -csg
cgd = -cdg
cgb = -cbd
cbd = cjd
cbs = cjs

The parameter names are those reported by the operating point in the two
versions.

Using the above note, you can translate from the partial derivatives spectre outputs to be in line with the bsim3v3 specification, into the more traditional capacitance terms.

Note that spectre can also output a captab. See
"spectre -h info". You'll see that you can do what=captab
to output a capacitance table.

Andrew.

Back to top

IP Logged

superwormsz New Member Offline Posts: 2	Re: Cgs, Cgg, CGd etc. in cadence Reply #2 - Sep 2^nd, 2004, 6:53am I have a question vdsat=vgs-vth????
Back to top	IP Logged

Andrew Beckett Senior Fellow Offline Life, don't talk to me about Life... Posts: 1742 Bracknell, UK	Re: Cgs, Cgg, CGd etc. in cadence Reply #3 - Sep 2^nd, 2004, 10:55am superwormsz, That doesn't appear to be a question... Andrew.
Back to top	IP Logged

101questions Junior Member Offline Posts: 31	Re: Cgs, Cgg, CGd etc. in cadence Reply #4 - Sep 2^nd, 2004, 11:17am Andrew, Thanks for the clarifications. What about channel to bulk capacitance? Is it counted as a part of cjs, cjd?
Back to top	IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: Cgs, Cgg, CGd etc. in cadence
Reply #5 - Sep 2^nd, 2004, 9:02pm

No, that would have been (traditionally) cgb. I just noticed a typo in my note above - I think it should say that the mapping should be:

Quote:

4.4 4.4.1
------------
cgs = -csg
cgd = -cdg
cgb = -cbg
cbd = cjd
cbs = cjs

(cgb was wrongly saying it mapped to -cbd).

So if you want to find the channel-bulk capacitance, look at -cbg.

At least I think that's correct ;)

Andrew.

Back to top

IP Logged

101questions

Junior Member

Offline

Posts: 31

Re: Cgs, Cgg, CGd etc. in cadence
Reply #6 - Sep 2^nd, 2004, 11:12pm

Andrew,

I'm puzzled. Cgb cap is between gate and bulk, right?
Channel to bulk cap represents capacitance of depletion layer beneath the channel, if the channel exists, that is in strong inversion.
If we change the bulk potential of mos in strong inversion, while gate, drain and source are are constant, we charge cjs, cjd, bulk-to-channel cap, and that's it. Cgb in strong inversion is almost zero, at least in active area.
Am I missing something?

Back to top

IP Logged

Andrew Beckett Senior Fellow Offline Life, don't talk to me about Life... Posts: 1742 Bracknell, UK	Re: Cgs, Cgg, CGd etc. in cadence Reply #7 - Sep 3^rd, 2004, 9:15am You're probably right. It's years since I've really thought about this, so I'll have to spend some time checking things. Won't be for a few days, given other commitments, as well as being out of the office. Sorry... Andrew.
Back to top	IP Logged

modern_analog

Guest

Re: Cgs, Cgg, CGd etc. in cadence
Reply #8 - Sep 17^th, 2004, 9:12am

This absolutely makes no sense...

I did a simple experiment. I took a common source transistor and put a voltage source with a resistor (R1) at the gate and a resistor (R2) at the drain.

pole at the gate of the transistor is 1/(R*C) where C is Cgs (not Csg) + (gain)*Cgd (not Cdg)

At the drain, I got a zero at gm/(Cdg) and pole at 1/(R*Cdb+Cdg)

Note the transistor was in saturation.
It seems Cgd = overlap of the gate, drain ~ 0
but Cdg is very high. In fact, I read somewhere that Cdg is proportional to Cgs(or Cgg to be more exact).

So, my question is
a) Is the above correct, ie gate cap is split between a big Cgs and very small Cgd.
b) How is Cdg calculated. It seems to be a big factor
c) In all the above Cdd ~ 0 because I had a postive Cds cancel a negative Cdg. Strange? That basically means, I should not consider Cdd in my equations.

ps: In all my statements above, I could have made a mistake with the correct cap values since they are so small.
These were representative values
Cdb=-24f
Cdd=1f
Cdg=-244f
Cds=267f

Cgd=.7f
Cgg=766f
Cgs=-800f

Csg = -365f

Back to top

IP Logged

Andrew Beckett Senior Fellow Offline Life, don't talk to me about Life... Posts: 1742 Bracknell, UK	Re: Cgs, Cgg, CGd etc. in cadence Reply #9 - Sep 19^th, 2004, 3:10pm You also have captab in spectre (see spectre -h info), as well as the explanations I gave earlier. Andrew.
Back to top	IP Logged

Geoffrey_Coram

Senior Fellow

Offline

Posts: 1998
Massachusetts, USA

Re: Cgs, Cgg, CGd etc. in cadence
Reply #10 - Sep 24^th, 2004, 11:22am

It should also be noted that some of the peculiarities arise from the BSIM3 implementation itself. Philips' MOS Model 11 is supposed to do a better job, in terms of making the capacitance model reciprocal. (With a nice non-reciprocal capacitor, one could power the world simply by driving the capacitor in a cycle and extracting energy ...

)

Back to top

If at first you do succeed, STOP, raise your standards, and stop wasting your time.

IP Logged

cricket_fan Guest	Re: Cgs, Cgg, CGd etc. in cadence Reply #11 - Apr 8^th, 2005, 12:09pm In order to look at Cgg vs a dc bias voltage , I have to do a dc simulation but it looks like I need to have an ac source for simulationg capacitance. I want to look at Cgg vs Vgb....in cadence. thanks MMN
Back to top	IP Logged