Cross-section transition

[alpasla]

The linear transition between cross-sections does not work well: the offset sections still taper with a sinusoidal profile. This is evident in the transition method defintion in phidl.path. However forcing the offset sections to be follow linear profiles looks bad too. The transition with the sine profile does work well.

Here is a minimum working example:

import phidl
import phidl.path as pp

readout_layer = (1,0)

readout_CPW_strip_w = 2
readout_CPW_slot_w  = 3
readout_CPW_w       = readout_CPW_strip_w + 2*readout_CPW_slot_w

bondpad_CPW_strip_w = 20
bondpad_CPW_slot_w  = 30
bondpad_CPW_w         = bondpad_CPW_strip_w + 2*bondpad_CPW_slot_w

readout_xsection      = phidl.CrossSection()
readout_xsection.add(width=readout_CPW_strip_w, offset=0, layer=readout_layer, name='strip')
readout_xsection.add(width=readout_CPW_slot_w, offset=+(readout_CPW_strip_w/2+readout_CPW_slot_w/2), layer=readout_layer, name='slot1')
readout_xsection.add(width=readout_CPW_slot_w, offset=-(readout_CPW_strip_w/2+readout_CPW_slot_w/2), layer=readout_layer, name='slot2')


bondpad_xsection      = phidl.CrossSection()
bondpad_xsection.add(width=bondpad_CPW_strip_w, offset=0, layer=readout_layer, name='strip')
bondpad_xsection.add(width=bondpad_CPW_slot_w, offset=+(bondpad_CPW_strip_w/2+bondpad_CPW_slot_w/2), layer=readout_layer, name='slot1')
bondpad_xsection.add(width=bondpad_CPW_slot_w, offset=-(bondpad_CPW_strip_w/2+bondpad_CPW_slot_w/2), layer=readout_layer, name='slot2')


bondpad_taper_xsection = pp.transition(cross_section1 = bondpad_xsection,
                                       cross_section2 = readout_xsection,
                                       width_type = 'linear')    

bondpad_taper_l = 100
bondpad_taper_path = pp.straight(length = bondpad_taper_l)
bondpad_taper          = bondpad_taper_path.extrude(bondpad_taper_xsection)

phidl.quickplot(bondpad_taper)

[MechanicalPenguin225]

Just stumbled on this too. Looks like it might happen when a section is offset from the crossection center ?

EDIT : looking in the code, looks like it's the case. Segment width does taper linearly, but its centerpoint is still mappsed out using a cosine transition