CZY (czy0126) errors

14 recurring problems engineers report in CZY (czy0126), ranked by impact.

#26 by errors trackedECAD

Recurring CZY (czy0126) errors, by impact

  1. 1A pre-fab review of a 48V to 2V/100A converter: in the SPICE deck the two secondary half-windings meet at the center tap with opposite dot polarity, while the winding process card says the dotted ends join at the center tap, so if the winder follows the card the secondary gets built reversed. The author also notes tools/board_netlist.py only checks matching net names, so it cannot catch this, and asks for one dotted wiring drawing shared by sim, board netlist and card, plus a first-article polarity measurement.ECAD · Agent failure · 1 report82
  2. 2A 100 A power board netlist check returned a misleading ok=True. It now says readiness PARTIAL_PASS and manufacturing_ready false, and refuses to close the gate on a new magnetics refdes whose physical pins and dot are not verified. The same PR fixes an output-path loss calc that treated the worst local 1.188 mOhm branch as the whole 100 A path, giving 13.33 W instead of about 1.35 to 2.8 W across four 25 A groups. About 39 MPNs/values are still missing.ECAD · Agent failure · 1 report74
  3. 3A 48 V to 2 V, 100 A converter first article has three documents with three answers: the capacitor selection table picks 15 x 470 uF (7.05 mF) while the V3 simulation deck uses 4 mF, the deck has Lr at 0.92 uH against a 1.0 uH design target, and simulated Vout is 1.8867 V against a 2 V plus or minus 2 percent spec. The board is still marked NOT_FOR_FAB.ECAD · Agent failure · 1 report74
  4. 4A homemade BOM gate raised an error whenever an operator swapped an erroneous OEM part, so the real EasyEDA GUI repair process cannot be reflected in an auditable report at all. It also kept counting R8 as a wrong purchase after it was set to DNP.ECAD · Agent failure · 1 report74
  5. 5Building a native EasyEDA Pro schematic for a 61-part power board from a golden netlist, the golden netlist names a FET's pins G/S/D, but the real SON/TDSON packages have 8 to 9 pads. A schematic that wires the S pin leaves pads 1 and 2 floating and still exports a netlist that matches the golden one, so the defect only appears in copper. The authors pulled pin-to-pad maps from a measured pin dump of the real devices and wrote a pad-level diff, plus a test that breaks a correct export eight ways so the ok: True is not self-certifying. Issue #9 says plainly that a PDF or screenshot cannot stand in for the CAD file.ECAD · Agent failure · 1 report74
  6. 6EasyEDA BOM values matched the schematic while the native manufacturer codes pointed at different parts, with 39 native MPN vs design-target conflicts across 51 targets and 37 cases marked ALREADY_CONFORMS by a text-only ECO; companion PR #62 adds the release check.ECAD · Agent failure · 1 report74
  7. 7Expanding combined designators in bom_draft.csv gave 84 parts against 66 in the golden SPICE netlist, with 17 control and decoupling parts never in the netlist, so even a full pass could not guarantee the real periphery is complete.ECAD · Agent failure · 1 report74
  8. 8On a 100 A power board in EasyEDA Pro, the native-inventory BOM gate only read the Nov 6 main-board snapshot while the current delivery board is the G1 candidate, so the gate described another project. The new export reads the active G1 project (83 parts, sha256 pinned before and after) and found the export renumbers excluded parts, so the gate now falls back to matching on library part plus both DNP flags.ECAD · Agent failure · 1 report74
  9. 9On the 48 V to 2 V, 100 A converter, the author reopened the native EasyEDA board (77 components, 299 pads, 98 tracks) instead of writing another conceptual report. Four 1.0 mm VOUT feeder branches and eight 1.0 mm PGND feeders alone come to 3,766 µΩ at 35 µm copper, about 37.7 W at 100 A, and even 210 µm copper leaves the necks at 628 µΩ, more than 4x the 150 µΩ budget for the whole VOUT plus PGND path. The board stays NOT_FOR_FAB.ECAD · Agent failure · 1 report74
  10. 10The fab gate never vetoed 39 real part-number conflicts because it only read schematic values: a separate authenticated native EasyEDA inventory found 39 real manufacturer MPN versus target conflicts, such as a Samsung 100nF part visually relabeled as 2.2uF. The fix hard-vetoes on native MPN mismatch and refuses a stale inventory snapshot.ECAD · Agent failure · 1 report74
  11. 11This corrects PR #77: the magnetics overlap report used the original g1_pads.json from before an already-merged driver ECO. Against the live saved EasyEDA board (77 parts, 299 pads, 65 nets), the inductor body overlaps 22 pads on 11 parts, not 30 on 15. The physical risk is real, and the fix pins the check to the exact Git blob SHA of the readback and keeps the old 30/15 as a regression test.ECAD · Agent failure · 1 report74
  12. 12Two checks fought the engineers: the netlist comparison flagged 17 real BOM parts as unexpected refs, so completing the BOM turned the check red, and the fabrication gate trusted an old BOM snapshot and text-only approval records. Now it recomputes from the live BOM and EDA-native netlist each time, while 22 refdes are still not drawn in the actual JLC EDA project and 39 values are placeholders.ECAD · Agent failure · 1 report74
  13. 13An older audit still said the 50 V case had not run, but it had. The fault sim collapsed at 7.57 ms, so only the prefix is valid, and the resonant inductor at two turns hits about 0.33 T against a 0.279 T limit. The PR adds a risk register built from the committed results that keeps six P0 items blocked, and refuses to call a sim done unless the log shows elapsed time and the measured points.ECAD · Agent failure · 1 report66
  14. 14The published 92.33% efficiency uses 100 A losses extrapolated from a 95 A point by current squared. The port difference is about 11.0 W, and the summed model losses come to about 12.35 W, so the books do not close, and the reviewer says plainly the number is an estimate, not a hardware or closed simulation result.ECAD · Agent failure · 1 report66

How it fails

  1. Agent failure14

Sourced from public community reports. Costs and delays on individual entries are modeled estimates, not measurements.