Forum Replies Created
-
Posts
-
A quick google of 'API-MS-WIN-CORE-COM-L1-1-0.DLL' would indicate that you are missing some Windows DLLs. You might want to run Windows Update to check you have all the latest files.
Hi Paco,
I have implemented the two parasitic extraction commands:
1) extractParasitic(metLayerLayer, areaCap, perimCap, gndNetName)as mentioned previously.
2) extractParasitic2(met1Layer, met2Layer, areaCap, perimeterCap). This finds all intersections (logical AND) of the two metal layers, and computes the capacitance between the nets of these two layers according to the intersection area and perimiter multiplied by the areaCap and preimCap terms, respectively.
Both functions screen out same-net capacitance e.g. in the first case if the metLayer net is the same as the gndNetName, no parasitic will be reported. In the second case if the two layers are part of the same net no capacitance will be reported.
I have not done extensive testing so bug reports/feedback appreciated. Windows versions of 4.3.39 are currently available, I expect to have Linux ones built later.
regards
Keith
Hi Paco,
I have finally got round to extracting parasitics of conductor layers to ground. It's still in its early stages, but I would be grateful of any feedback. The extraction command in the extraction rule file is:
extractParasitic( conductorLayer, areaCap, perimCap, gndNetName)
where conductorLayer is the layer to be extracted, and must be a layer defined in the geomConnect() statement i.e. have connectivity. areaCap and perimCap are the capacitances of the layer area in F/um^2 and F/um and gndNetName is the name of the ground net to be used in spice simulation e.g. "VSS", "GND" or even spice global ground "0".
This will be rolled out in 4.3.39 after a bit more testing. Obviously there are limitations:
– it is a simple 2D model so no good for small geometry tracks which need a 3D model
– it does not consider coupling to other layers e.g. a metal2 track running over a metal1, or two metal1 tracks running in parallel.Let me know if this is useful or if there are things that would make it more useful!
regards
Keith
Just a reminder: if you try and register for the forum with an email address that is potentially fake (e.g gmail, yahoo etc) *and* a google search of that address shows no sensible postings from that address about IC design, your registration will be rejected.
If you really want to register with some fake email address, you'll need to email the glade support email address with a request. This is just to deter spammers, so please don't take it personally.
One other thing about the Linux installations:
You should set LD_LIBRARY_PATH – so that the $GLADE_HOME/lib directory is found first.
If you don't, you may find that Glade picks up a different version of Qt libs.
In 4.3.37 for example, Glade was switch to Qt 4.8.4 from 4.8.2; it *may* work with the old libs but to be on the safe side check with
Code:ldd gladewhich should print a list of tha actual shared libs the application can find.
Glade was featured in a paper/poster at DATE 2013: https://www.peardrop.co.uk/glade/abstract_poster.pdf
I have just rebuilt the Linux builds of 4.3.36 and re-loaded them onto the website. Alas the previous build was linked to python 2.7 by mistake, and this causes several problems as people have discovered recently.
Please go and re-download 4.3.36 for Linux if you are seeing messages like 'ImportError: No module named _weakref' or 'ImportError: No module named site'.
This does not affect the Mac or Windows platforms though.
Hi Jofre,
The regression tests include the last example you submitted using the 'p_simple1' pcell, which has parameters replaced using dbReplaceProp and pass OK.
Do you have a new testcase you can share?
Keith
It takes 8 arguments as above. The python wrapper always adds another but that is invisible to the user.
– Keith
The new version fixes some problems with e.g. move/copy not working correctly with shortcut keys, a bug which was introduced in 4.3.22 and missed being detected by the regression tests.
The version will be uploaded to the website today (15th).
– and uploading is complete for all platforms except Solaris which is an on-demand only build.
OK, I have found a nasty bug introduced in version 4.3.22 when the underlying window handling was changed to QMdiArea / QMdiSubWindow. Basically the event filter on the canvas did not seem to catch events 100% of the time, sometimes not at all. Although this might be a Qt bug, I've changed the even filter to work at the application level to catch shortcut keystrokes.
Fix will be in v 4.3.34, to be released later today after running the regression tests.
Just a preliminary observation: Stretch does not consider infix mode. The use model is to select e.g. an edge then use the stretch command to move it by some distance. So there's no need to specify an initial coordinate.
The other problem with infix – I still need to look at…
There is something weird going on here when infix is turned off AND the option 'Always pop up dialog options' is set. Needs debugging, will take a look as soon as I can…
In the meantime, use infix mode if possible, or use the menu command rather than the bindkey if you don't want infix…
One important point – if you have problems getting glade to run on Windows, check if you have python27.dll installed in either of
C:windowssystem32
C:windowssysWOW64If you're running a 64 bit version of Windows, the one in system32 should be the 64 bit one (this may not seem intuitive but then this is is Microsoft…) and the one in sysWOW64 should be the 32 bit dll.
Better still just use the ones in the default installations and start Glade from that directory.
If you're not sure, try using Dependency Walker (http://www.dependencywalker.com/) which will show you just which DLL's your executable is using and whether there are any conflicts.
