what difference between couple and decouple when doing RCX ext.

I'm designing a ADC now,when doing assura RCX extract,select couple
option in assura RCX option and run post-simulation will lost about
20db SFDR conparing to use decouple option of assura rcx,why?

 

I'm designing a ADC now,when doing assura RCX extract,select couple

I guess you are referring to the coupled/decoupled capacitance extraction.

This is very well explained in the manual : in coupled mode, net-to-net
parasitic capacitances are extracted as-is, whereas in decoupled mode, they
are lumped to ground.

As an example, consider the parasitic cap between net A and net B. In coupled, you'll get a cap
between net A and net B, whereas in decoupled you get two caps, one between A and ground and another
between B and ground.

In decoupled, capacitive coupling effects are lost, so it's less accurate, but it can speed up
simulation time a lot.

Stéphane

 

thank you.After reading assura refernece book and do some
experiments,i get to know the difference between coupled and
decoupled.
Still one more question,i'm designing a switch capacitor ADC.Using
coupled RC rxtracion,Posting-simulation shows a 11dBc SFDR lost
coomparing to using decoupled extration,and the highest harmonic
appear in high frequency,could you please tell me what's wrong?

 

When you do a coupled extraction, you get charge injection between
signal lines (and the loading that this charge injection causes). For
example, if I put a voltage spike on netA, it will couple to netB--
netB will get a charge injection due to the voltage changing on netA.
At the same time, netA sees the loading--some current flows out of
netA (into netB).

When you do a decoupled extraction, you do not get charge injection
between signal lines, but you do get the loading. Assura does not
extract capacitance from netA to netB; it simply pretends this
capacitance is to ground. So, if I put a voltage spike on netA, netB
doesn't care. However, netA will inject some charge into the ground
reference.

It is likely that your SFDR reduction is caused by coupling between
two nets. I can't (and likely no one on this list can) tell you which
nets without seeing a schematic and knowing the operation of your ADC.
However, your coupled results are likely more accurate than the
uncoupled results.

 

Still one more question,i'm designing a switch capacitor ADC.Using

I'm no ADC expert, and I do not know the details of your implementation.

However, if you get this 11dB loss between decoupled and coupled, then it's likely that some
parasitic coupling is causing this. If you can figure out where this coupling occurs, you'll be able
(maybe) to improve your layout to reduce it.

I may suggest you try to understand which nets are critical with respect to coupling in your design,
then produce a "refined extracted view" with only those caps that you expect to be the guilty ones
and check with simulation if this is really the case.

To produce a refined extracted view, go to Tools->Parasitics, then Parasitics->Setup, then
Parasitics->Refine Extracted View. You will need to specify those parasitics that you want to keep
in the refined view by adding special symbols to your schematic - namely, spcapacitor2 for
net-to-net capacitances - these symbols are found in the sbaLib library
($CDS_INST_DIR/tools/dfII/etc/cdslib/artist/sbaLib).

See "Virtuoso Parasitic Simulation User Guide" for details.

Good luck,

Stéphane