setupEM 0.14.0__tar.gz → 0.14.2__tar.gz

setupem-0.14.2/PKG-INFO

@@ -0,0 +1,75 @@
+Metadata-Version: 2.4
+Name: setupEM
+Version: 0.14.2
+Summary: Python tool for configuration of gds2palace workflow with GUI.
+Author-email: Volker Muehlhaus <volker@muehlhaus.com>
+License-Expression: GPL-3.0-or-later
+Project-URL: Homepage, https://github.com/VolkerMuehlhaus/setupEM
+Project-URL: Repository, https://github.com/VolkerMuehlhaus/setupEM
+Project-URL: Changelog, https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: gds2palace>=0.9.0
+Requires-Dist: gds_prepare_for_EM>=1.6.0
+Requires-Dist: PySide6
+Requires-Dist: shiboken6
+Requires-Dist: scipy
+Requires-Dist: requests
+Requires-Dist: scikit-rf
+Requires-Dist: matplotlib
+Requires-Dist: numpy
+Requires-Dist: gdspy
+Requires-Dist: meshio
+Requires-Dist: pyvista
+Requires-Dist: pyvistaqt
+Dynamic: license-file
+
+# Python GUI for gds2palace 
+
+![Intro](https://raw.githubusercontent.com/VolkerMuehlhaus/setupEM/main/doc/png/setupEM_banner.png)
+
+## Install
+
+    pip install setupEM
+
+**Full documentation, installation guide, and usage walkthrough:**
+https://github.com/VolkerMuehlhaus/setupEM
+
+## Recent changes
+
+# What's New - October 21, 2026
+
+**Tools > Simplify GDS:** the simplification code moved to gds_prepare_for_EM (module `gds_simplify_for_em`), where the gds2palace MCP server uses it too. setupEM now requires gds_prepare_for_EM 1.6.0 or later; `pip install --upgrade setupEM` updates it.
+
+Fix: with **Merge via arrays** (and no fill removal or merging per layer), the simplified file still contained the original subcells as extra top-level cells. Without a cell name, gds2palace reads the first top-level cell, which could be one of them, e.g. a pad cell instead of the design. These left-over cells are now removed.
+
+# What's New - October 19, 2026
+
+**Stackup Editor, ADS Momentum (*.ltd) import:** metals with an EXPAND thickness operation now make the dielectric next to them thicker, as in ADS, instead of reaching into the dielectric below. Before, a thick EXPAND DOWN metal (e.g. MET1/MET2 in a GaAs stackup) overlapped the layers underneath it, and a via could end up with zero height. Files where long lines are wrapped or END_... markers follow the last entry on the same line now import too; before, they failed with "No usable dielectric slabs".
+
+# What's New - October 18, 2026
+
+**Touchstone files with different port impedances:** the option line of the combined `.sNp` file now has one Z0 per port, in port order. Before, ports with the same Z0 as port 1 were left out.
+
+Full history: [CHANGES.md](https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md)
+
+---
+
+**Note:** the `setupEM` PyPI package itself requires:
+
+- gds2palace>=0.9.0
+- gds_prepare_for_EM>=1.6.0
+- PySide6
+- shiboken6
+- scipy
+- requests
+- scikit-rf
+- matplotlib
+- numpy
+- gdspy
+- meshio
+- pyvista
+- pyvistaqt
+
+(Other Python modules mentioned above are only needed to run standalone helper scripts in this repository, not to use the installed package.)

setupem-0.14.2/README_pypi.md

@@ -0,0 +1,48 @@
+# Python GUI for gds2palace 
+
+![Intro](https://raw.githubusercontent.com/VolkerMuehlhaus/setupEM/main/doc/png/setupEM_banner.png)
+
+## Install
+
+    pip install setupEM
+
+**Full documentation, installation guide, and usage walkthrough:**
+https://github.com/VolkerMuehlhaus/setupEM
+
+## Recent changes
+
+# What's New - October 21, 2026
+
+**Tools > Simplify GDS:** the simplification code moved to gds_prepare_for_EM (module `gds_simplify_for_em`), where the gds2palace MCP server uses it too. setupEM now requires gds_prepare_for_EM 1.6.0 or later; `pip install --upgrade setupEM` updates it.
+
+Fix: with **Merge via arrays** (and no fill removal or merging per layer), the simplified file still contained the original subcells as extra top-level cells. Without a cell name, gds2palace reads the first top-level cell, which could be one of them, e.g. a pad cell instead of the design. These left-over cells are now removed.
+
+# What's New - October 19, 2026
+
+**Stackup Editor, ADS Momentum (*.ltd) import:** metals with an EXPAND thickness operation now make the dielectric next to them thicker, as in ADS, instead of reaching into the dielectric below. Before, a thick EXPAND DOWN metal (e.g. MET1/MET2 in a GaAs stackup) overlapped the layers underneath it, and a via could end up with zero height. Files where long lines are wrapped or END_... markers follow the last entry on the same line now import too; before, they failed with "No usable dielectric slabs".
+
+# What's New - October 18, 2026
+
+**Touchstone files with different port impedances:** the option line of the combined `.sNp` file now has one Z0 per port, in port order. Before, ports with the same Z0 as port 1 were left out.
+
+Full history: [CHANGES.md](https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md)
+
+---
+
+**Note:** the `setupEM` PyPI package itself requires:
+
+- gds2palace>=0.9.0
+- gds_prepare_for_EM>=1.6.0
+- PySide6
+- shiboken6
+- scipy
+- requests
+- scikit-rf
+- matplotlib
+- numpy
+- gdspy
+- meshio
+- pyvista
+- pyvistaqt
+
+(Other Python modules mentioned above are only needed to run standalone helper scripts in this repository, not to use the installed package.)

{setupem-0.14.0 → setupem-0.14.2}/pyproject.toml

@@ -13,7 +13,7 @@ authors = [
 ]
 dependencies = [
     "gds2palace>=0.9.0",
-    "gds_prepare_for_EM>=1.3.0",
+    "gds_prepare_for_EM>=1.6.0",
     "PySide6",
     "shiboken6",
     "scipy",

{setupem-0.14.0 → setupem-0.14.2}/src/setupEM/__init__.py

@@ -24,4 +24,4 @@ A Python tool for EM setup using gds2palace.
 from .setupEM import main
 
 __all__ = ["main"]
-__version__ = "0.14.0"
+__version__ = "0.14.2"

{setupem-0.14.0 → setupem-0.14.2}/src/setupEM/gds_hierarchy_scan.py

@@ -30,69 +30,16 @@ cell/reference tree directly instead, recursing into referenced cells
 without expanding them, so cost scales with the (small) number of distinct
 cells rather than the (potentially huge) number of instantiated copies.
 
+open_gds_cell() and estimate_polygon_count() live in gds_prepare_for_EM's
+gds_simplify_for_em module (shared with the gds2palace MCP server) and are
+re-exported here, so callers keep using gds_hierarchy_scan.<name>.
+
 No PySide6/setup_common/layout_preview dependency here - keeps this a leaf
 module importable from both without any circular-import concern (setup_common.py
 and layout_preview.py already need a deferred import between each other).
 """
 
-import gdspy
-
-
-def open_gds_cell(gds_path, cellname):
-    """Open gds_path and resolve the target cell: cellname if given and
-    found, else the file's first top-level cell. Shared first step for
-    every walker below. Raises on a bad file or a library with no top-level
-    cell, same as gdspy itself would - callers already wrap their own call
-    sites in a broad try/except.
-    """
-    library = gdspy.GdsLibrary(infile=gds_path)
-    top_level = library.top_level()
-    if not top_level:
-        raise ValueError(f"no top-level cell in {gds_path}")
-    cell = library.cells.get(cellname, top_level[0])
-    return library, cell
-
-
-def estimate_polygon_count(gds_path, cellname, layernumbers, purposelist):
-    """Fast pre-check of how many polygons a full read_gds()-based load
-    would end up processing, without flattening the GDS hierarchy. Walks
-    the cell/reference tree directly: a plain CellReference contributes its
-    referenced cell's own (memoized) count once, a CellArray contributes it
-    columns*rows times - the same arithmetic flatten() would otherwise do
-    by actually generating that many polygon objects. Returns None if the
-    file can't even be opened here (the real read_gds() call is expected to
-    surface that error properly instead).
-    """
-    try:
-        library, cell = open_gds_cell(gds_path, cellname)
-    except Exception:
-        return None
-
-    layernumbers = set(int(n) for n in layernumbers)
-    purposeset = set(int(p) for p in purposelist) if purposelist else None
-
-    memo = {}
-
-    def count(c):
-        if c.name in memo:
-            return memo[c.name]
-        total = 0
-        for polygonset in c.polygons:
-            for layer, datatype in zip(polygonset.layers, polygonset.datatypes):
-                if int(layer) in layernumbers and (purposeset is None or int(datatype) in purposeset):
-                    total += 1
-        for ref in c.references:
-            ref_cell = ref.ref_cell
-            if isinstance(ref_cell, str):
-                ref_cell = library.cells.get(ref_cell)
-            if ref_cell is None:
-                continue
-            sub = count(ref_cell)
-            total += sub * ref.columns * ref.rows if isinstance(ref, gdspy.CellArray) else sub
-        memo[c.name] = total
-        return total
-
-    return count(cell)
+from gds_simplify_for_em import open_gds_cell, estimate_polygon_count  # noqa: F401 (re-export)
 
 
 def layers_present_in_range(gds_path, cellname, layer_min, layer_max, purposelist):

{setupem-0.14.0 → setupem-0.14.2}/src/setupEM/momentum_import.py

@@ -294,12 +294,15 @@ def _build_tree(slabs, layers, materials, warnings, derived_layers=None):
   # (both source formats define it that way) rather than that metal's TOP, where it
   # actually needs to start - shift any via whose zmin coincides with a metal's zmin
   # up by that metal's own thickness.
+  # A metal that only shares its zmin with the via by coincidence (e.g. a thick *.ltd
+  # EXPAND DOWN metal reaching the via's bottom interface) must not shrink the via to
+  # nothing - only nudge while the via keeps a positive height.
   metal_spans = [(l.zmin, l.zmax - l.zmin) for l in layers if l.kind == "metal"]
   for l in layers:
     if l.kind != "via":
       continue
     for metal_zmin, metal_thickness in metal_spans:
-      if abs(l.zmin - metal_zmin) < _EPSILON_UM:
+      if abs(l.zmin - metal_zmin) < _EPSILON_UM and l.zmin + metal_thickness < l.zmax - _EPSILON_UM:
         l.zmin += metal_thickness
         break
 
@@ -584,7 +587,23 @@ def _ltd_kv(line):
   return {key.upper(): value.strip('"') for key, value in _KV_RE.findall(line)}
 
 
+# keywords that start a new statement in each section - any other line in these sections
+# is a continuation of the previous statement (a long MATERIAL/MASK/LAYER line wrapped
+# onto the next line, e.g. "IMAG_PERMEABILITY=0", "OPERATION=...", or a MASK's own
+# "MASK_PROPERTIES = {...}" line) and is joined back onto it
+_LTD_STATEMENT_KEYWORDS = {
+    "MATERIAL": {"MATERIAL"},
+    "OPERATION": {"OPERATION"},
+    "MASK": {"MASK"},
+    "DERIVEDMASK": {"DERIVEDMASK"},
+    "STACK": {"TOP", "INTERFACE", "LAYER", "BOTTOM"},
+}
+
+
 def _split_ltd_sections(text):
+  # END_<section> is not always on a line of its own - seen trailing the last statement
+  # ("OPERATION OperationTSV_Virtual_VIA1 DRILL END_OPERATION")
+  text = re.sub(r'\s+(END_\w+)\b', r'\n\1', text)
   sections = {}
   current = None
   buf = []
@@ -603,7 +622,11 @@ def _split_ltd_sections(text):
       current = None
       continue
     if current:
-      buf.append(line)
+      keywords = _LTD_STATEMENT_KEYWORDS.get(current)
+      if keywords and buf and line.split()[0] not in keywords:
+        buf[-1] = buf[-1] + " " + line
+      else:
+        buf.append(line)
   return sections
 
 
@@ -679,10 +702,13 @@ def _parse_ltd(ltd_path, air_thickness_um, warnings):
     elif "INTRUDE" in kv or "EXPAND" in kv:
       # both give a thickness in meters; DOWN (vs. the default UP) means this metal
       # grows downward from its interface into the slab below - same concept *.subst
-      # expresses via a negative thick value, here via an explicit direction token
+      # expresses via a negative thick value, here via an explicit direction token.
+      # INTRUDE places the metal inside the existing slab; EXPAND makes that slab
+      # thicker by the metal thickness (same as *.subst's expand="1"), handled when
+      # the stack is walked below.
       magnitude_um = float(kv.get("INTRUDE") or kv.get("EXPAND")) * 1e6
       thickness_um = -magnitude_um if "DOWN" in tokens else magnitude_um
-      operations[name] = ("metal", thickness_um)
+      operations[name] = ("metal", thickness_um, "EXPAND" in kv)
     elif "SHEET" in tokens:
       operations[name] = ("sheet", None)
     elif "WALL" in tokens:
@@ -773,7 +799,33 @@ def _parse_ltd(ltd_path, air_thickness_um, warnings):
   groundplane_detected = False
   bottom_state = None
 
-  for line in reversed(stack_lines):
+  # EXPAND metals make the dielectric LAYER next to their INTERFACE thicker (DOWN: the
+  # LAYER below, UP: the one above) - by the thickest such metal on that interface, not
+  # the sum (ADS shows TFR 0.05um + MET1 1um EXPAND DOWN on one interface as +1um). A
+  # LAYER can grow from both sides (metal EXPAND UP below it, another EXPAND DOWN above
+  # it), which adds up. Must be known before the bottom-up walk reaches that LAYER, so
+  # collected up front: (line index in file order, top-to-bottom; +1 DOWN / -1 UP) ->
+  # extra thickness in um
+  expand_um = {}
+  for idx, line in enumerate(stack_lines):
+    if line.split()[0] != "INTERFACE":
+      continue
+    mask_match = _MASK_BRACE_RE.search(line)
+    for token in mask_match.group(1).split() if mask_match else []:
+      mask = _resolve_mask_token(token, masks, masks_by_layer)
+      op = operations.get(mask["operation"]) if mask else None
+      if op is None or op[0] != "metal" or not op[2]:
+        continue
+      side = 1 if op[1] < 0 else -1
+      target = idx + side
+      if 0 <= target < len(stack_lines) and stack_lines[target].split()[0] == "LAYER":
+        expand_um[(target, side)] = max(expand_um.get((target, side), 0.0), abs(op[1]))
+      else:
+        warnings.append(f"Mask '{mask['name']}' uses EXPAND but has no dielectric LAYER on that "
+                         f"side of its INTERFACE - placed without growing any layer")
+
+  for idx in reversed(range(len(stack_lines))):
+    line = stack_lines[idx]
     tokens = line.split()
     if not tokens:
       continue
@@ -809,7 +861,8 @@ def _parse_ltd(ltd_path, air_thickness_um, warnings):
       if material_name not in materials:
         warnings.append(f"Dielectric slab material '{material_name}' not found, skipped")
         continue
-      thickness_um = float(height_m) * 1e6 if height_m else 0.0
+      thickness_um = ((float(height_m) * 1e6 if height_m else 0.0)
+                      + expand_um.get((idx, 1), 0.0) + expand_um.get((idx, -1), 0.0))
       zmin = z
       z += thickness_um
       slabs.append(DielectricSlab(material_name, thickness_um))

{setupem-0.14.0 → setupem-0.14.2}/src/setupEM/setup_common.py

@@ -4912,8 +4912,8 @@ class MainWindowBase(QMainWindow):
         self.layout_preview_window.show()
 
     def open_simplify_gds(self):
-        # local import: simplify_gds.py imports gds_prepare_for_EM lazily
-        # inside run_simplify() too, but import the dialog module itself here
+        # local import: gds_simplify_for_em.run_simplify() imports gds_prepare_for_EM
+        # lazily, but import the dialog module itself here
         # (not at module load time) for the same circular-import reason as
         # open_stackup_editor()/open_layout_preview() above
         if __package__ in (None, ""):

{setupem-0.14.0 → setupem-0.14.2}/src/setupEM/simplify_gds.py

@@ -21,19 +21,18 @@ simplify_gds.py
 
 Tools > Simplify GDS... : removes floating (unconnected) metal fill, fills
 in small cutouts and/or merges via arrays on the currently loaded GDS file,
-writing the result to a new GDS file. All operations come from
-gds_prepare_for_EM (a sibling package, already an editable install in this
-venv) and are called in-process - no subprocess/CLI involved.
+writing the result to a new GDS file. The pipeline itself is
+gds_prepare_for_EM's gds_simplify_for_em.run_simplify() (shared with the
+gds2palace MCP server), called in-process - no subprocess/CLI involved.
 
 No technology-specific layer numbers are hardcoded here: the metal layers to
 process, the via layers and the metal layers above/below each via are derived
 entirely from the XML stackup already parsed into MainWindow.metals_list
-(gds2palace's stackup_reader), same as every other setupEM feature that needs
-to know "which GDS layers are metal".
+(gds2palace's stackup_reader), with gds_simplify_for_em.layers_from_stackup().
 """
 
-import os, io, contextlib, traceback, tempfile
-import gdspy
+import os, io, contextlib, traceback
+from gds_simplify_for_em import run_simplify, layers_from_stackup
 from PySide6.QtWidgets import (
     QDialog, QVBoxLayout, QHBoxLayout, QGroupBox, QLabel, QLineEdit,
     QPushButton, QPlainTextEdit, QFileDialog, QMessageBox, QCheckBox,
@@ -112,138 +111,6 @@ def _parse_layer_list(text):
                           f"comma-separated list of layer numbers: {text!r}")
 
 
-def _flatten_cell(lib, cellname):
-    """Return (lib, top_cell) with top_cell flattened in place. Round-trips
-    through a temporary GDS file first: gdspy's flatten()/get_polygonsets()
-    hits an internal bug on some in-memory-only reference structures ('tuple'
-    object does not support item assignment) that only a GDS write+read
-    cycle avoids - same workaround gds_prepare_for_EM.py's own CLI pipeline
-    uses before every flatten() call in its main()."""
-    with tempfile.TemporaryDirectory(prefix="setupEM_simplify_") as tmp_dir:
-        tmp_path = os.path.join(tmp_dir, "pre_flatten.gds")
-        lib.write_gds(tmp_path)
-        lib = gdspy.GdsLibrary(infile=tmp_path)
-    top_cell = lib.cells.get(cellname, lib.top_level()[0])
-    top_cell.flatten()
-    return lib, top_cell
-
-
-def run_simplify(gds_path, metal_layers, output_path, cellname="",
-                  do_cutouts=False, max_hole_area=None,
-                  do_floating=False, fill_minsize=1.0, fill_maxsize=None, fill_mincount=20,
-                  do_merge=False, via_layers=None, via_merge_size=0):
-    """Load gds_path, optionally fill small cutouts, remove floating metal
-    fill, and/or merge (boolean OR) polygons per layer as a final pass, on
-    the given metal_layers, write the result to output_path. Mirrors the
-    step order gds_prepare_for_EM.py's own main() uses: cutout removal first
-    (works on the hierarchical design), floating-fill removal after that
-    (needs a flattened, per-layer-merged cell first so that touching fill
-    tiles aren't mistaken for isolated ones - see
-    find_isolated_same_size_polygons_by_layer()'s own docstring), per-layer
-    merge, then via array merging.
-
-    via_layers: list of (via layer, metal layers above, metal layers below),
-    GDS layer numbers. Vias closer than via_merge_size (microns) are merged
-    if they connect the same metal shapes above and below, see
-    gds_prepare_for_EM.merge_via_array_by_metal_overlap(). Vias without metal
-    above or below are kept unchanged. None or via_merge_size <= 0: no via
-    merging.
-    """
-    from gds_prepare_for_EM import (
-        remove_cutout_keep_hierarchy,
-        merge_polygons_by_layer,
-        find_isolated_same_size_polygons_by_layer,
-        merge_polygons,
-        merge_via_array_by_metal_overlap,
-    )
-    from gds_geometry_utils import validate_and_repair_polygons
-
-    # merge_polygons_by_layer()/find_isolated_same_size_polygons_by_layer()
-    # create cells via the bare gdspy.Cell(...) constructor, which
-    # auto-registers each one into gdspy's persistent global
-    # gdspy.current_library - harmless for the CLI (a fresh process every
-    # run) but in this long-running GUI, a second Run on the same file
-    # would reuse the same cell names and gdspy raises "already present in
-    # library". Starting from a fresh current_library each call clears that
-    # accumulated state without touching gds_prepare_for_EM itself.
-    gdspy.current_library = gdspy.GdsLibrary()
-
-    lib = gdspy.GdsLibrary()
-    lib.read_gds(gds_path)
-    top_cell = lib.cells.get(cellname, lib.top_level()[0])
-
-    if do_cutouts:
-        design_bbox = top_cell.get_bounding_box()
-        lib = remove_cutout_keep_hierarchy(lib, metal_layers, design_bbox=design_bbox,
-                                            max_hole_area=max_hole_area)
-        top_cell = lib.cells.get(cellname, lib.top_level()[0])
-
-    # floating-fill removal already ends with a flattened, per-layer-merged
-    # cell (it needs that itself, to tell a floating fill cluster apart from
-    # touching real metal) - so a separate final merge pass is only extra
-    # work when floating-fill removal did NOT already run
-    already_merged_flat = False
-
-    if do_floating:
-        lib, top_cell = _flatten_cell(lib, cellname)
-        merged_lib = merge_polygons_by_layer(top_cell, layers_list=metal_layers)
-        merged_top = merged_lib.top_level()[0]
-        lib = find_isolated_same_size_polygons_by_layer(
-            merged_top, metal_layers,
-            minsize=fill_minsize, maxsize=fill_maxsize, mincount=fill_mincount)
-        already_merged_flat = True
-
-    if do_merge and not already_merged_flat:
-        lib, top_cell = _flatten_cell(lib, cellname)
-        lib = merge_polygons_by_layer(top_cell, layers_list=metal_layers)
-
-    if via_layers and via_merge_size > 0:
-        lib, top_cell = _flatten_cell(lib, cellname)
-        lpp = top_cell.get_polygons(by_spec=True)
-        print(f'Via array merging (distance {via_merge_size} um):')
-
-        def metal_polygons(layers):
-            polygons = [p for (layer, purpose), polys in lpp.items() if layer in layers for p in polys]
-            # touching metal pieces are one metal shape for via grouping
-            return merge_polygons(polygons) if polygons else []
-
-        for via_layer, above_layers, below_layers in via_layers:
-            purposes = [purpose for (layer, purpose) in lpp if layer == via_layer]
-            if not purposes:
-                continue
-            above, below = metal_polygons(above_layers), metal_polygons(below_layers)
-            for purpose in purposes:
-                vias = lpp[(via_layer, purpose)]
-                merged = merge_via_array_by_metal_overlap(
-                    vias, above, below, via_merge_size, keep_unmatched=True)
-                top_cell.remove_polygons(
-                    lambda pts, layer, datatype: layer == via_layer and datatype == purpose)
-                for pts in merged:
-                    top_cell.add(gdspy.Polygon(pts, layer=via_layer, datatype=purpose))
-                print(f'  layer {via_layer}:{purpose}: {len(vias)} -> {len(merged)} polygons')
-
-    # Final polygon validity check/repair - always runs, regardless of which
-    # operations above were selected. Catches any polygon still geometrically
-    # invalid (self-intersecting) no matter the reason - e.g. a cutout
-    # deliberately left alone by remove_cutout_keep_hierarchy() for being
-    # larger than max_hole_area, or a merge step re-introducing an invalid
-    # "bridge" hole encoding while fusing polygons back together. A
-    # solid-body mesher like gds2palace cannot handle this even though
-    # gdspy/GDSII themselves tolerate it.
-    repaired_count, unresolved = validate_and_repair_polygons(lib, metal_layers)
-    if repaired_count:
-        print(f'Final validity check: repaired {repaired_count} invalid polygon(s)')
-    if unresolved:
-        print(f'Final validity check: WARNING - {len(unresolved)} polygon(s) could not be '
-              f'auto-repaired, still geometrically invalid in the output file:')
-        for cell_name, n, layer, reason in unresolved:
-            print(f'  {cell_name} polygon #{n} layer {layer}: {reason}')
-    if not repaired_count and not unresolved:
-        print('Final validity check: no invalid polygons found')
-
-    lib.write_gds(output_path)
-
-
 class SimplifyGdsDialog(QDialog):
     """Own top-level window (WA_DeleteOnClose, non-modal), same lifecycle as
     LayoutPreviewWindow/StackupEditorWindow."""
@@ -266,18 +133,10 @@ class SimplifyGdsDialog(QDialog):
         app_name = self.MainWindow.APP_NAME
         port_layer_min, port_layer_max = _port_layer_range(app_name)
         metals_list = self.MainWindow.metals_list
-        self._metal_layers = sorted(
-            int(m.layernum) for m in metals_list.getallplanarmetals()
-            if not (port_layer_min <= int(m.layernum) <= port_layer_max)
-        ) if metals_list is not None else []
-        # via layers with the metal layers above/below them, for via array merging
-        self._via_layers = [
-            (int(m.layernum),
-             [int(n.layernum) for n in m.above if n.is_metal],
-             [int(n.layernum) for n in m.below if n.is_metal])
-            for m in metals_list.metals
-            if m.is_via and not (port_layer_min <= int(m.layernum) <= port_layer_max)
-        ] if metals_list is not None else []
+        # metal layers to process, and via layers with the metal layers above/below them
+        # for via array merging
+        self._metal_layers, self._via_layers = layers_from_stackup(
+            metals_list, (port_layer_min, port_layer_max)) if metals_list is not None else ([], [])
 
         layout = QVBoxLayout(self)
 

setupem-0.14.2/src/setupEM.egg-info/PKG-INFO

@@ -0,0 +1,75 @@
+Metadata-Version: 2.4
+Name: setupEM
+Version: 0.14.2
+Summary: Python tool for configuration of gds2palace workflow with GUI.
+Author-email: Volker Muehlhaus <volker@muehlhaus.com>
+License-Expression: GPL-3.0-or-later
+Project-URL: Homepage, https://github.com/VolkerMuehlhaus/setupEM
+Project-URL: Repository, https://github.com/VolkerMuehlhaus/setupEM
+Project-URL: Changelog, https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: gds2palace>=0.9.0
+Requires-Dist: gds_prepare_for_EM>=1.6.0
+Requires-Dist: PySide6
+Requires-Dist: shiboken6
+Requires-Dist: scipy
+Requires-Dist: requests
+Requires-Dist: scikit-rf
+Requires-Dist: matplotlib
+Requires-Dist: numpy
+Requires-Dist: gdspy
+Requires-Dist: meshio
+Requires-Dist: pyvista
+Requires-Dist: pyvistaqt
+Dynamic: license-file
+
+# Python GUI for gds2palace 
+
+![Intro](https://raw.githubusercontent.com/VolkerMuehlhaus/setupEM/main/doc/png/setupEM_banner.png)
+
+## Install
+
+    pip install setupEM
+
+**Full documentation, installation guide, and usage walkthrough:**
+https://github.com/VolkerMuehlhaus/setupEM
+
+## Recent changes
+
+# What's New - October 21, 2026
+
+**Tools > Simplify GDS:** the simplification code moved to gds_prepare_for_EM (module `gds_simplify_for_em`), where the gds2palace MCP server uses it too. setupEM now requires gds_prepare_for_EM 1.6.0 or later; `pip install --upgrade setupEM` updates it.
+
+Fix: with **Merge via arrays** (and no fill removal or merging per layer), the simplified file still contained the original subcells as extra top-level cells. Without a cell name, gds2palace reads the first top-level cell, which could be one of them, e.g. a pad cell instead of the design. These left-over cells are now removed.
+
+# What's New - October 19, 2026
+
+**Stackup Editor, ADS Momentum (*.ltd) import:** metals with an EXPAND thickness operation now make the dielectric next to them thicker, as in ADS, instead of reaching into the dielectric below. Before, a thick EXPAND DOWN metal (e.g. MET1/MET2 in a GaAs stackup) overlapped the layers underneath it, and a via could end up with zero height. Files where long lines are wrapped or END_... markers follow the last entry on the same line now import too; before, they failed with "No usable dielectric slabs".
+
+# What's New - October 18, 2026
+
+**Touchstone files with different port impedances:** the option line of the combined `.sNp` file now has one Z0 per port, in port order. Before, ports with the same Z0 as port 1 were left out.
+
+Full history: [CHANGES.md](https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md)
+
+---
+
+**Note:** the `setupEM` PyPI package itself requires:
+
+- gds2palace>=0.9.0
+- gds_prepare_for_EM>=1.6.0
+- PySide6
+- shiboken6
+- scipy
+- requests
+- scikit-rf
+- matplotlib
+- numpy
+- gdspy
+- meshio
+- pyvista
+- pyvistaqt
+
+(Other Python modules mentioned above are only needed to run standalone helper scripts in this repository, not to use the installed package.)

{setupem-0.14.0 → setupem-0.14.2}/src/setupEM.egg-info/requires.txt

@@ -1,5 +1,5 @@
 gds2palace>=0.9.0
-gds_prepare_for_EM>=1.3.0
+gds_prepare_for_EM>=1.6.0
 PySide6
 shiboken6
 scipy

setupem-0.14.0/PKG-INFO

@@ -1,85 +0,0 @@
-Metadata-Version: 2.4
-Name: setupEM
-Version: 0.14.0
-Summary: Python tool for configuration of gds2palace workflow with GUI.
-Author-email: Volker Muehlhaus <volker@muehlhaus.com>
-License-Expression: GPL-3.0-or-later
-Project-URL: Homepage, https://github.com/VolkerMuehlhaus/setupEM
-Project-URL: Repository, https://github.com/VolkerMuehlhaus/setupEM
-Project-URL: Changelog, https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md
-Requires-Python: >=3.9
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: gds2palace>=0.9.0
-Requires-Dist: gds_prepare_for_EM>=1.3.0
-Requires-Dist: PySide6
-Requires-Dist: shiboken6
-Requires-Dist: scipy
-Requires-Dist: requests
-Requires-Dist: scikit-rf
-Requires-Dist: matplotlib
-Requires-Dist: numpy
-Requires-Dist: gdspy
-Requires-Dist: meshio
-Requires-Dist: pyvista
-Requires-Dist: pyvistaqt
-Dynamic: license-file
-
-# Python GUI for gds2palace 
-
-![Intro](https://raw.githubusercontent.com/VolkerMuehlhaus/setupEM/main/doc/png/setupEM_banner.png)
-
-## Install
-
-    pip install setupEM
-
-**Full documentation, installation guide, and usage walkthrough:**
-https://github.com/VolkerMuehlhaus/setupEM
-
-## Recent changes
-
-# What's New - October 17, 2026
-
-**Conductor model** replaces **Conductor meshing** on the Mesh and Boundaries tab (Palace only), with a third choice: **Surface impedance** (default, as before), **Solve inside (volume mesh)** and the new **Solve inside + skin layers**, which adds thin layers along the conductor faces for accurate Q and loss at high frequency. Generated model scripts now use `settings['conductor_model']` instead of `settings['filled_metals']`. Older projects and model scripts with `filled_metals` still load. setupEM now requires gds2palace 0.9.0 or later; `pip install --upgrade setupEM` updates it.
-
-# What's New - October 16, 2026
-
-**Boundary conditions per side:** **Advanced...** next to **Boundary conditions** on the Mesh and Boundaries tab opens a table with the six sides of the simulation box. Each side can have its own boundary condition (**Absorbing**, **PEC** or **PMC**) and air layer thickness; a blank entry uses the value from the tab. For example, **PEC** at zmin with 0 µm air models a backside ground plane at the box boundary. This replaces the "same on all sides / different per side" choice for the air layer, which now lives in the same table.
-
-Loading a model with different boundary conditions per side no longer changes it to one condition for all sides: before, a single **PEC** side turned all six sides into PEC.
-
-**Stackup Editor:** while the stackup has errors, the cross-section preview can't be redrawn and keeps the last valid drawing. It is now marked as not up to date: dimmed, crossed by a thin red line, with a banner above it. **Show errors** in the banner opens the editor's XML Preview tab, which lists the errors. The status line now points there too, instead of to Save.
-
-# What's New - October 12, 2026
-
-**3D Field Viewer: value markers.** Check **Probe** in the View box (or press **P**), then click the model to place a numbered marker with the field value at that exact point. Clicking works from any angle, on the full model, the clipped solid, its cut face or the **2D plane only** section; dragging still rotates. Markers are drawn on top of the model, so a marker on a buried body stays visible. A **Markers** list opens beside the view with the marker number, value and x / y / z in µm, plus **Remove**, **Clear** and **Copy** (tab-separated, for spreadsheets). For Elmer thermal temperatures, a dropdown switches the list and labels between **K**, **°C** and **ΔT** (rise above the lowest temperature in the model, in K). Right-click the model for **Jump to Maximum in Cut Plane** (with the clip plane on), **Jump to Maximum on Surface of Body N** (the outer surface of the body you clicked, e.g. the chip; results without body IDs use the whole shown surface), **Add Marker Here**, **Remove Marker N** and **Clear Markers**. Markers are updated when you change the field, result file or cycle; markers outside the new model are removed. Ported from EMStudio's field viewer.
-
-**3D Field Viewer: keys.** With the 3D view focused: **R / F / Home** reset camera, **I** isometric, **X / Y / Z** axis views (**Shift**: negative side), **+ / −** zoom, arrow keys pan, **O** parallel projection, **M** find max., **A** arrows on/off, **P** probe, **PgUp / PgDn** move the clip plane. VTK's own letter keys (e.g. **w** wireframe, **3** stereo, **q** / **e** which stopped the interactor) no longer do anything. The right-click menu now opens only on a click: a right-drag zooms without the menu popping up afterwards.
-
-**3D Field Viewer:** the **Cycle** picker of a Palace field dump showed Palace's extra AMR error-indicator cycle as a frequency, e.g. "999 GHz (cycle 2)". It is now listed as "error indicator (cycle 2)". This happened when that cycle stores its Indicator/Rank values as point data. Dumps that store them as cell data were already listed as "geometry".
-
-**3D Field Viewer:** running `field_viewer.py` / `fieldViewer` standalone with a window failed in Python environments that also have PyQt5 installed. Fixed.
-
-**Port de-embedding in `scripts/combine_extend_snp.py` fixed** (the `_deembedded.sNp` and `_dc_deembedded.sNp` files): the negative inductance of all ports was put in series at port 1, and the other ports were not de-embedded. For a 3-port, port 1 got the sum of all three values. Now each port gets the value for its own port number; a port without geometry data in `port_information.json` is not de-embedded. The raw `.sNp` and the `_dc.sNp` files were not affected. Please run `combine_snp` again to create the de-embedded files again. The same bug was fixed in gds2palace, where Gianluca Simone and David Lurz (FAU) found it.
-
-Full history: [CHANGES.md](https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md)
-
----
-
-**Note:** the `setupEM` PyPI package itself requires:
-
-- gds2palace>=0.9.0
-- gds_prepare_for_EM>=1.3.0
-- PySide6
-- shiboken6
-- scipy
-- requests
-- scikit-rf
-- matplotlib
-- numpy
-- gdspy
-- meshio
-- pyvista
-- pyvistaqt
-
-(Other Python modules mentioned above are only needed to run standalone helper scripts in this repository, not to use the installed package.)

setupem-0.14.0/README_pypi.md

@@ -1,58 +0,0 @@
-# Python GUI for gds2palace 
-
-![Intro](https://raw.githubusercontent.com/VolkerMuehlhaus/setupEM/main/doc/png/setupEM_banner.png)
-
-## Install
-
-    pip install setupEM
-
-**Full documentation, installation guide, and usage walkthrough:**
-https://github.com/VolkerMuehlhaus/setupEM
-
-## Recent changes
-
-# What's New - October 17, 2026
-
-**Conductor model** replaces **Conductor meshing** on the Mesh and Boundaries tab (Palace only), with a third choice: **Surface impedance** (default, as before), **Solve inside (volume mesh)** and the new **Solve inside + skin layers**, which adds thin layers along the conductor faces for accurate Q and loss at high frequency. Generated model scripts now use `settings['conductor_model']` instead of `settings['filled_metals']`. Older projects and model scripts with `filled_metals` still load. setupEM now requires gds2palace 0.9.0 or later; `pip install --upgrade setupEM` updates it.
-
-# What's New - October 16, 2026
-
-**Boundary conditions per side:** **Advanced...** next to **Boundary conditions** on the Mesh and Boundaries tab opens a table with the six sides of the simulation box. Each side can have its own boundary condition (**Absorbing**, **PEC** or **PMC**) and air layer thickness; a blank entry uses the value from the tab. For example, **PEC** at zmin with 0 µm air models a backside ground plane at the box boundary. This replaces the "same on all sides / different per side" choice for the air layer, which now lives in the same table.
-
-Loading a model with different boundary conditions per side no longer changes it to one condition for all sides: before, a single **PEC** side turned all six sides into PEC.
-
-**Stackup Editor:** while the stackup has errors, the cross-section preview can't be redrawn and keeps the last valid drawing. It is now marked as not up to date: dimmed, crossed by a thin red line, with a banner above it. **Show errors** in the banner opens the editor's XML Preview tab, which lists the errors. The status line now points there too, instead of to Save.
-
-# What's New - October 12, 2026
-
-**3D Field Viewer: value markers.** Check **Probe** in the View box (or press **P**), then click the model to place a numbered marker with the field value at that exact point. Clicking works from any angle, on the full model, the clipped solid, its cut face or the **2D plane only** section; dragging still rotates. Markers are drawn on top of the model, so a marker on a buried body stays visible. A **Markers** list opens beside the view with the marker number, value and x / y / z in µm, plus **Remove**, **Clear** and **Copy** (tab-separated, for spreadsheets). For Elmer thermal temperatures, a dropdown switches the list and labels between **K**, **°C** and **ΔT** (rise above the lowest temperature in the model, in K). Right-click the model for **Jump to Maximum in Cut Plane** (with the clip plane on), **Jump to Maximum on Surface of Body N** (the outer surface of the body you clicked, e.g. the chip; results without body IDs use the whole shown surface), **Add Marker Here**, **Remove Marker N** and **Clear Markers**. Markers are updated when you change the field, result file or cycle; markers outside the new model are removed. Ported from EMStudio's field viewer.
-
-**3D Field Viewer: keys.** With the 3D view focused: **R / F / Home** reset camera, **I** isometric, **X / Y / Z** axis views (**Shift**: negative side), **+ / −** zoom, arrow keys pan, **O** parallel projection, **M** find max., **A** arrows on/off, **P** probe, **PgUp / PgDn** move the clip plane. VTK's own letter keys (e.g. **w** wireframe, **3** stereo, **q** / **e** which stopped the interactor) no longer do anything. The right-click menu now opens only on a click: a right-drag zooms without the menu popping up afterwards.
-
-**3D Field Viewer:** the **Cycle** picker of a Palace field dump showed Palace's extra AMR error-indicator cycle as a frequency, e.g. "999 GHz (cycle 2)". It is now listed as "error indicator (cycle 2)". This happened when that cycle stores its Indicator/Rank values as point data. Dumps that store them as cell data were already listed as "geometry".
-
-**3D Field Viewer:** running `field_viewer.py` / `fieldViewer` standalone with a window failed in Python environments that also have PyQt5 installed. Fixed.
-
-**Port de-embedding in `scripts/combine_extend_snp.py` fixed** (the `_deembedded.sNp` and `_dc_deembedded.sNp` files): the negative inductance of all ports was put in series at port 1, and the other ports were not de-embedded. For a 3-port, port 1 got the sum of all three values. Now each port gets the value for its own port number; a port without geometry data in `port_information.json` is not de-embedded. The raw `.sNp` and the `_dc.sNp` files were not affected. Please run `combine_snp` again to create the de-embedded files again. The same bug was fixed in gds2palace, where Gianluca Simone and David Lurz (FAU) found it.
-
-Full history: [CHANGES.md](https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md)
-
----
-
-**Note:** the `setupEM` PyPI package itself requires:
-
-- gds2palace>=0.9.0
-- gds_prepare_for_EM>=1.3.0
-- PySide6
-- shiboken6
-- scipy
-- requests
-- scikit-rf
-- matplotlib
-- numpy
-- gdspy
-- meshio
-- pyvista
-- pyvistaqt
-
-(Other Python modules mentioned above are only needed to run standalone helper scripts in this repository, not to use the installed package.)

setupem-0.14.0/src/setupEM.egg-info/PKG-INFO

@@ -1,85 +0,0 @@
-Metadata-Version: 2.4
-Name: setupEM
-Version: 0.14.0
-Summary: Python tool for configuration of gds2palace workflow with GUI.
-Author-email: Volker Muehlhaus <volker@muehlhaus.com>
-License-Expression: GPL-3.0-or-later
-Project-URL: Homepage, https://github.com/VolkerMuehlhaus/setupEM
-Project-URL: Repository, https://github.com/VolkerMuehlhaus/setupEM
-Project-URL: Changelog, https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md
-Requires-Python: >=3.9
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: gds2palace>=0.9.0
-Requires-Dist: gds_prepare_for_EM>=1.3.0
-Requires-Dist: PySide6
-Requires-Dist: shiboken6
-Requires-Dist: scipy
-Requires-Dist: requests
-Requires-Dist: scikit-rf
-Requires-Dist: matplotlib
-Requires-Dist: numpy
-Requires-Dist: gdspy
-Requires-Dist: meshio
-Requires-Dist: pyvista
-Requires-Dist: pyvistaqt
-Dynamic: license-file
-
-# Python GUI for gds2palace 
-
-![Intro](https://raw.githubusercontent.com/VolkerMuehlhaus/setupEM/main/doc/png/setupEM_banner.png)
-
-## Install
-
-    pip install setupEM
-
-**Full documentation, installation guide, and usage walkthrough:**
-https://github.com/VolkerMuehlhaus/setupEM
-
-## Recent changes
-
-# What's New - October 17, 2026
-
-**Conductor model** replaces **Conductor meshing** on the Mesh and Boundaries tab (Palace only), with a third choice: **Surface impedance** (default, as before), **Solve inside (volume mesh)** and the new **Solve inside + skin layers**, which adds thin layers along the conductor faces for accurate Q and loss at high frequency. Generated model scripts now use `settings['conductor_model']` instead of `settings['filled_metals']`. Older projects and model scripts with `filled_metals` still load. setupEM now requires gds2palace 0.9.0 or later; `pip install --upgrade setupEM` updates it.
-
-# What's New - October 16, 2026
-
-**Boundary conditions per side:** **Advanced...** next to **Boundary conditions** on the Mesh and Boundaries tab opens a table with the six sides of the simulation box. Each side can have its own boundary condition (**Absorbing**, **PEC** or **PMC**) and air layer thickness; a blank entry uses the value from the tab. For example, **PEC** at zmin with 0 µm air models a backside ground plane at the box boundary. This replaces the "same on all sides / different per side" choice for the air layer, which now lives in the same table.
-
-Loading a model with different boundary conditions per side no longer changes it to one condition for all sides: before, a single **PEC** side turned all six sides into PEC.
-
-**Stackup Editor:** while the stackup has errors, the cross-section preview can't be redrawn and keeps the last valid drawing. It is now marked as not up to date: dimmed, crossed by a thin red line, with a banner above it. **Show errors** in the banner opens the editor's XML Preview tab, which lists the errors. The status line now points there too, instead of to Save.
-
-# What's New - October 12, 2026
-
-**3D Field Viewer: value markers.** Check **Probe** in the View box (or press **P**), then click the model to place a numbered marker with the field value at that exact point. Clicking works from any angle, on the full model, the clipped solid, its cut face or the **2D plane only** section; dragging still rotates. Markers are drawn on top of the model, so a marker on a buried body stays visible. A **Markers** list opens beside the view with the marker number, value and x / y / z in µm, plus **Remove**, **Clear** and **Copy** (tab-separated, for spreadsheets). For Elmer thermal temperatures, a dropdown switches the list and labels between **K**, **°C** and **ΔT** (rise above the lowest temperature in the model, in K). Right-click the model for **Jump to Maximum in Cut Plane** (with the clip plane on), **Jump to Maximum on Surface of Body N** (the outer surface of the body you clicked, e.g. the chip; results without body IDs use the whole shown surface), **Add Marker Here**, **Remove Marker N** and **Clear Markers**. Markers are updated when you change the field, result file or cycle; markers outside the new model are removed. Ported from EMStudio's field viewer.
-
-**3D Field Viewer: keys.** With the 3D view focused: **R / F / Home** reset camera, **I** isometric, **X / Y / Z** axis views (**Shift**: negative side), **+ / −** zoom, arrow keys pan, **O** parallel projection, **M** find max., **A** arrows on/off, **P** probe, **PgUp / PgDn** move the clip plane. VTK's own letter keys (e.g. **w** wireframe, **3** stereo, **q** / **e** which stopped the interactor) no longer do anything. The right-click menu now opens only on a click: a right-drag zooms without the menu popping up afterwards.
-
-**3D Field Viewer:** the **Cycle** picker of a Palace field dump showed Palace's extra AMR error-indicator cycle as a frequency, e.g. "999 GHz (cycle 2)". It is now listed as "error indicator (cycle 2)". This happened when that cycle stores its Indicator/Rank values as point data. Dumps that store them as cell data were already listed as "geometry".
-
-**3D Field Viewer:** running `field_viewer.py` / `fieldViewer` standalone with a window failed in Python environments that also have PyQt5 installed. Fixed.
-
-**Port de-embedding in `scripts/combine_extend_snp.py` fixed** (the `_deembedded.sNp` and `_dc_deembedded.sNp` files): the negative inductance of all ports was put in series at port 1, and the other ports were not de-embedded. For a 3-port, port 1 got the sum of all three values. Now each port gets the value for its own port number; a port without geometry data in `port_information.json` is not de-embedded. The raw `.sNp` and the `_dc.sNp` files were not affected. Please run `combine_snp` again to create the de-embedded files again. The same bug was fixed in gds2palace, where Gianluca Simone and David Lurz (FAU) found it.
-
-Full history: [CHANGES.md](https://github.com/VolkerMuehlhaus/setupEM/blob/main/doc/CHANGES.md)
-
----
-
-**Note:** the `setupEM` PyPI package itself requires:
-
-- gds2palace>=0.9.0
-- gds_prepare_for_EM>=1.3.0
-- PySide6
-- shiboken6
-- scipy
-- requests
-- scikit-rf
-- matplotlib
-- numpy
-- gdspy
-- meshio
-- pyvista
-- pyvistaqt
-
-(Other Python modules mentioned above are only needed to run standalone helper scripts in this repository, not to use the installed package.)