Cadence Tips and Tricks

Expected Min/Max for Normal Distribution Number of Samples Mean Standard Deviation Jarque-Bera Statistic (Optional) Calculate Expected Ranges Expected min/max (normal): — Expected min/max (with JB): —

That's when you use the (thanks Amit S.) Reference Histories approach: Do a run with the setup you want - default Monte-Carlo and kill it Name the run - for your convenience Now, click the Gear icon (1) that's on the same row as the Play and Stop buttons (to the right of this will be "History") In the resulting dialog check the two checkboxes on the top-left (Use reference results, use reference netlist) Now, to the right of the History drop-down, you have an icon that resembles a lock (2). Click it (which unlocks it and make is resemble the familar pencil that represents an "Edit" button) and now you can change the run name from MonteCarlo (3). Put in a meaningful name Hit the (notice how it's turned blue) run button How I feel, using Cadence

BLUF : Ensure you turn OFF (uncheck the checkbox) saving of final OP point (if your MC involves transient analysis) and ensure the Simulation > Options > Analog > Check does not have "both" or "file" for destination (disk usage) What's bad (or good) about the default setup? It saves the final Operating Point (OP) in a transient sim It writes assert information to the log file (note, you need it in the sqldb for the Device Checks to be useful) Now, given that the Monte Carlo setup is so nice as to NOT save all (but only explicitly selected) waveforms (even though the setup in ADE says to Save All) unless told to, why can't it do the same with the mostly (in MC) information that's generated from the default setup (see above two bullets)

You've just done a simulation using the Assembler Debug Environment (the good old ADE-L) and now you want to capture the numbers as text so you can put them in a spreadsheet, or whatever.. In ADE, Outputs > Export.. dump as a CSV file Then, in a terminal: perl -n -e 'if( /^([^,]+),expr,([^,]+),t,.+,t,t?$/ ){ $op = $1; $opp = $1; $ex = $2; $op =~ s/\s/__/g; print "$op = $ex\n"; print q{printf("} . $opp . q{,%s\n",aelSuffixNotation(}.$op.q{))},"\n"; }' /tmp/desired_outs.csv > /tmp/outputs.il After that, you just (in your CIW) loadi "/tmp/outputs.il" Let's break all of that down: The regex in perl is looking for expressions (you don't want signals, etc) : "expr", but it's also looking for those that are set to "Yes, plot".

I can't wait to see what's new and what's been broken.

You need checklimitdest set to "both" or "sqldb" Yes, it would be nice if the tool, sensing user intention (the user enabled Checks/Asserts and then the user went into the Checks/Asserts view in Results - two actions) could give a better message than just "Disabled". BTW, this is a handy one to SUPPRESS logging of warnings to the spectre.out when you have hundreds of sims running for Monte-Carlo (especially trim-and-sim runs where you want as many samples as you can get and each needs 5+ sims)