[Magic-dev] Query about 'units microns' statement in tech file extraction

[Graham Petley]

Hallo,

   Can someone explain the use of the 'units microns' statement in the
extraction section of the technology file? It doesn't work as I expect.
   My layout has a Magic database unit of 1nm which means that the extraction
lambda value is 0.1 (there are 0.1 centimicrons in one Magic db unit). The top
of the extraction secton looks like

extract
style	pharosc013-55nm
    cscale   1
    lambda   0.1
    rscale   1
    step     500
    sidehalo 510
    units    microns

   For poly coupling or sidewall capacitance, the process has a value of 50aF
for each 1um length of poly for wires which are spaced 0.2um apart.

   I calculate the sidewall capacitance value without using the 'units microns'
statement as below.

1um poly length, .2um poly sep, 50aF cap
1um            , 1um          , 10aF
1nm            , 1um          , 0.01aF
1nm            , 1nm          , 10aF cap

   Then, according to the maint2 manual, p43, "Cap is half the capacitance in
attofarads per lambda when the edges are 1 lambda apart", so 5 is entered into
the tech file

    sidewall (*poly)/a ~(*poly)/a ~(*poly)/a (*poly)/a 5.0

   A 10um (10000nm) parallel poly run with 0.2um (200nm) separation will have a
sidewall cap 

Cc = 10/200 * 10000 aF = 500aF = 0.5fF and this is the result Magic gives me.

   If I use 'units microns' then I expect that "Cap is half the capacitance in
attofarads per micron when the edges are 1 micron apart". From the calculation
above, this means that the same value can be used (1um length, 1um sep has same
cap as 1nm length, 1nm sep).

Cc = 10/0.2 * 10 aF = 500aF = 0.5fF as before.

   But I find that the result is 1000X too small. It looks like "Cap is half
the capacitance in attofarads per micron when the edges are 1 lambda apart",
which means entering a sidewall value of

    sidewall (*poly)/a ~(*poly)/a ~(*poly)/a (*poly)/a 5000

   Is this the desired behaviour? Personally I find mixing lambda and microns
like this is confusing and it's easier to choose a Magic db unit like 1nm for
extraction which allows fairly easy conversion of foundry numbers.
   All of this is using version 7.5.95. I'm not sure if earlier versions can
handle fractional values of lambda like 0.1.

Best regards, Graham Petley