Netlist Reduction

[Sylvio Triebel]

Hallo Markus,

die v3_ct_Filter Schaltung scheint eine ziemlich niedrige Sensitivitaet auf
parasitaere Wiederstaende zu haben... (bei power supply rejection sieht es mit Sicherheit
anders aus!)

Die Uebertragungsfunktion ist fuer C-only extrahierte Netzlisten
nahezu identisch mit der RC-extrahierten Netzliste. Dadurch konnte ich auch keinen
nennenswerten Einfluss der Optionen bei der Netzlisten-Reduktion finden.
-->plot
Das Ergebnis weicht nur unbedeutend von der Schematic Simulation ab... ich glaube nicht,
dass hier der Fehler liegt, den du suchst. (Die Ueberhoehungen sind bei der extrahierten etwas
staerker ausgepraegt)

Interessant ist, dass mit den neuen Algorithmen die Simulation der RC-Extraktion
nur noch um Faktor 4-10 langsamer als bei einer C-Extraktion und immerhin ~4x schneller
als vorher (oder ca. 20x schneller als mit av_extracted). Das Ergebnis sieht nahezu identisch aus.
Ohne Reduktion ist diese Schaltung RC-extrahiert schon kaum noch simulierbar (einige Tage...).

Viele Gruesse,
Sylvio

 

[Svenn Are Bjerkem]

Sylvio Triebel wrote:

Hallo Markus,

I think this is a typical example what happens when you use mozilla
suite to read mail and usenet news. I have *almost* done this a couple
of times myself, but luckily I do not have an automatic signature
attached on usenet articles so I could catch myself before pressing the
send button.

So, people, don't get upset about a little German on the group.

--
Svenn

 

[Bernd Fischer]

No realy I'm not upset I'm just curious about what's up with the
'new algorithms' are they for reducing a rc extracted netlist
or for extraction?

Howerver ;-)

Bernd

Svenn Are Bjerkem wrote:

Sylvio Triebel wrote:

Hallo Markus,


I think this is a typical example what happens when you use mozilla
suite to read mail and usenet news. I have *almost* done this a couple
of times myself, but luckily I do not have an automatic signature
attached on usenet articles so I could catch myself before pressing the
send button.

So, people, don't get upset about a little German on the group.

 

[Sylvio Triebel]

Hi Bernd,

I'm not sure if the algorithms are realy new ;-) I just implemented
some new ideas for our inhouse tool... to reduce rc-extracted netlists or
av_extracted views.
Mathematically 'exact' algorithms like TICER are good to eliminate
most parasitics reliable, but with some additional heuristic algorithms
and interactive features to play with the netlists the tool becomes
even for big circuits usable...

Are you not satisfied with your resimulation flow? ;-)
Do you know perhaps Jivaro netlist reduction? I'm not sure if it is better
(except perhaps for substrate networks) and they don't tell too much
how exactly it works...

Sylvio.


Bernd Fischer > wrote:

No realy I'm not upset I'm just curious about what's up with the
'new algorithms' are they for reducing a rc extracted netlist
or for extraction?

Howerver ;-)

Bernd

Svenn Are Bjerkem wrote:

Sylvio Triebel wrote:

Hallo Markus,



I think this is a typical example what happens when you use mozilla
suite to read mail and usenet news. I have *almost* done this a couple
of times myself, but luckily I do not have an automatic signature
attached on usenet articles so I could catch myself before pressing
the send button.

So, people, don't get upset about a little German on the group.