| Author |
Message |
dragonwell
Joined: 31 Mar 2004
Posts: 28
|
 26 Aug 2004 8:56 extract problem in C(at)dence |
|
|
|
|
When I perform parastic RC extract, i must set "skip_soft_connect_check" switch, otherwise it will generat lots of problems of "Figures causing multiple stamped connections".
What is "soft_connect"?
What is "stamped connection"?
And, how to solve this problem?
thx!
|
|
| Back to top |
|
 |
pantic
Joined: 06 May 2003
Posts: 93
Helped: 2
|
| 26 Aug 2004 9:13 Re: extract problem in C(at)dence |
|
|
|
|
| Which software are you using to extract? Diva, Assura ..?
|
|
| Back to top |
|
|
dragonwell
Joined: 31 Mar 2004
Posts: 28
|
| 26 Aug 2004 11:01 Re: extract problem in C(at)dence |
|
|
|
|
| diva
|
|
| Back to top |
|
|
Hughes
Joined: 10 Jun 2003
Posts: 715
Helped: 85
|
| 26 Aug 2004 14:41 extract problem in C(at)dence |
|
|
|
|
| A soft connection usually means the connection is formed by a relatively high-resistive layer, namely, nwell or pwell. If you add two well taps in a well, then multiple stamped connection error will arise if these two well taps are not connected by metal.
|
|
| Back to top |
|
|
rajesh13
Joined: 07 Jul 2004
Posts: 126
Location: Milky Way
|
| 26 Aug 2004 14:57 Re: extract problem in C(at)dence |
|
|
|
|
Soft Connect is basically a virtual connction to fool the the tool for getting extracted netlist.
like you have several VDD & GND rails in case of I/Os So these rails are connected to each other when the ring is complete in the supply cells. But for one cell level they may not be connected. So by using soft connect option, this problem can be overcome.
|
|
| Back to top |
|
|
Hughes
Joined: 10 Jun 2003
Posts: 715
Helped: 85
|
| 27 Aug 2004 3:27 Re: extract problem in C(at)dence |
|
|
|
|
| rajesh13 wrote: |
Soft Connect is basically a virtual connction to fool the the tool for getting extracted netlist.
like you have several VDD & GND rails in case of I/Os So these rails are connected to each other when the ring is complete in the supply cells. But for one cell level they may not be connected. So by using soft connect option, this problem can be overcome. |
That is another type "soft connect" -- connected by software. It is not the problem dragonwell encountered.
"Connection by software" can be achieved in diva by add a property "joinableNet" to the cellview.
|
|
| Back to top |
|
|
Question
Joined: 11 Aug 2004
Posts: 76
|
| 27 Aug 2004 9:38 extract problem in C(at)dence |
|
|
|
|
I have the same probled as you. This is the answer from Cadence.
'>If the customer does *not* get this same message when running *without*
>parasitics, then we have a problem in the Assura deck. If they get this
>message in both parasitic and non-parasitic LVS, then the issue is as I
>describe below. Please let me know if the customer does not see this
>message when running without parasitics.
>
>
>The tie-downs are probably not connected to each other in metal. This check
>is intended to warn the user that the substrate of the chip *may* be
>connected to different voltages. Since the tie-downs are not connected to
>each other, they may go to different I/O pads and these I/O pads may have
>different voltages applied to them. This check is done on NWELLS too.
>
>
>These conditions are flagged by the soft connect checks:
>
>
>1) Substrate or NWELL with two or more tie-downs that are not connected to
>each other in metal. This is the condition described above.
>
>
>2) Substrate or NWELL with no tie-downs connected to metal. This is
>reported as a floating well or floating substrate. The substrate can never
>be floating and NWELLs are also normally not supposed to be floating (no
>power supply connected). Of course, NWELL resistors are an exception.
>
>
>It is common to have these two conditions during the development of a chip.
>Therefore, the Assura decks provide the 'skip_soft-connect-checks' switch
>to mask these conditions. Eventually, the chip designer will need to
>correct these conditions to ensure proper biasing of P/N junctions and
>proper operation of devices. '
But I can not understand yet.
|
|
| Back to top |
|
|
nxing
Joined: 10 May 2004
Posts: 446
Helped: 13
Location: China
|
| 30 Aug 2004 6:22 extract problem in C(at)dence |
|
|
|
|
Connect all of your VDD and GND together, you can solve this problem.
Good Luck!
|
|
| Back to top |
|
|
dragonwell
Joined: 31 Mar 2004
Posts: 28
|
| 06 Sep 2004 14:03 Re: extract problem in C(at)dence |
|
|
|
|
| Has connected all the nwell ties by metal & connected all vdd/gnd, but still the problem
|
|
| Back to top |
|
|
nxing
Joined: 10 May 2004
Posts: 446
Helped: 13
Location: China
|
| 07 Sep 2004 6:40 extract problem in C(at)dence |
|
|
|
|
| Did you connect all the nwell and psub together TO vdd and gnd? or just connect them together? You need take the first action.
|
|
| Back to top |
|
|
zx2051
Joined: 23 Oct 2003
Posts: 20
|
| 07 Sep 2004 10:56 extract problem in C(at)dence |
|
|
|
|
| use dracula, the file is called lpe.com,you can do it just as run LVS
|
|
| Back to top |
|
|
dragonwell
Joined: 31 Mar 2004
Posts: 28
|
| 07 Sep 2004 13:55 Re: extract problem in C(at)dence |
|
|
|
|
hi nxing,
yes, i have connected all nwell to vdd & psub to gnd.
|
|
| Back to top |
|
|
Hughes
Joined: 10 Jun 2003
Posts: 715
Helped: 85
|
| 07 Sep 2004 14:52 extract problem in C(at)dence |
|
|
|
|
Hi dragonwell,
If you still connot fix the problem, I suggest you a debug procedure. First of all, I suppose you use diva extract tool.
1. Modify the divaEXT.rul file: Remove the geomStamp statement; add the nwell and psub into the geomConnect statement. The statements may look like:
ntapcont=geomAnd(cont ntap)
ptapcont=geomAnd(cont ptap)
geomConnect(
via(ntapcont ntap m1)
via(ptapcont ptap m1)
; other connections including label statements.
)
; other statements
saveDerived(
geomAndNot(ntap geomGetNet("vdd!" ntap))
"ntap not connected to vdd! explicitly"
)
saveDerived(
geomAndNot(ptap geomGetNet("gnd!" ptap))
"ptap not connected to gnd! explicitly"
)
; comments out all device extract rules using nwell or pwell as one of their terminals.
2. Re-extract the layout. Check the markers.
|
|
| Back to top |
|
|
