Winding Mistakes

Common Mistakes in Transformer and Coil Winding, and How to Catch Them

The winding and magnetics errors that quietly wreck yield and margin, each with its symptom, root cause, and a numeric fix.

The most expensive mistake in magnetics is estimating copper from a single mean length of turn on a thick multi-layer coil. Symptom: you run a spool dry mid-build or the copper line on the quote is 8 to 12 percent light. Root cause is that outer turns are longer than inner ones, so one MLT understates the outer bands. Fix: split the coil into layer bands or bump MLT by the layer buildup. On a 100-turn coil where the outer layer circumference is 15 percent above the inner, a flat average run through the Copper Wire Length tool underbuys roughly 7 to 8 percent of wire.

Planning to gross capacity instead of good output is the second trap. Symptom: the schedule promises 1,920 coils and the shift ships 1,676. Root cause is ignoring the two derates that always apply: uptime near 90 percent and first-pass yield near 97 percent. Multiplied together they cost about 87 percent of nameplate, so 1,920 gross becomes 1,676 good, a 244-coil gap. Fix: always commit off the good number from Turns Count or Winding Machine Output, and pull uptime and yield from the last real run, not the design intent or a hopeful 98 percent.

Overfilling the bobbin is a setup-stage error that surfaces as a winding jam or a hi-pot fail. Symptom: the last layer will not seat, or the machine throws tension faults near completion. Root cause is treating the winding window as fully usable copper. Round magnet wire realistically fills only 40 to 60 percent of the window once insulation and layer paper are counted. Fix: check the build before setup so total copper cross-section stays inside that band. A design penciled at 70 percent fill will not wind cleanly, and forcing it drives scrap wire and first-pass fallout up several points.

Confusing first-pass yield with final yield after rework overstates good output every time. Symptom: the capacity model looks fine but the line keeps missing dates. Root cause is entering a post-rework 99 percent where the machine actually delivers 97 percent first pass. On a 1,920-gross shift, that 2-point gap is about 38 coils that quietly move to a rework queue and consume winder time twice. Fix: score first-pass yield from turns-count, resistance, and layer-short checks before any rework, and model recovered units as extra cycles, not a higher yield number.

Missing the thermal side of the design fails units in the field, not on the bench. Symptom: a transformer passes hi-pot and production test, then runs hot and returns. Root cause is checking copper loss or core loss alone instead of the sum against the insulation class limit. Core Loss Estimate plus I-squared-R copper loss must feed Thermal Rise Margin, and the total rise has to sit under the class ceiling, roughly 105C for Class A and 130C for Class B. Fix: budget both losses together and hold at least a 10 to 15C margin below the class limit for ambient swing and hot-spot factor.

Unit and conversion slips corrupt otherwise correct math. Symptom: a wire length reads 2 feet when it should be 2,000, or core loss comes back 1,000 times off. Root cause is a mismatched conversion factor, mixing millimeters with inches on MLT, or reading a loss curve in watts per pound against a kilogram core mass. Fix: verify the conversion factor produces a sane order of magnitude before trusting the output. In Copper Wire Length, a 0.005 factor turns a 400 raw product into 2 feet; a stray decimal there swings the copper buy by 100x.

Letting impregnation hide as the real bottleneck starves an otherwise fast cell. Symptom: winding output is high but finished goods still queue and dates slip. Root cause is that VPI or dip-and-bake cure often runs several hours per batch, so the oven, not the winder, sets the cell rate. Fix: run Impregnation Batch Size and Hi-Pot Test Capacity with real cycle counts and find the lowest good-output number. If impregnation clears 1,676 and winding clears 2,400, the winder idles; the fix is denser fixturing or faster-cure resin, not a faster machine.

Ignoring scrap copper recovery distorts both cost and waste tracking. Symptom: the bill of materials shows full copper spend and scrap looks like a pure loss. Root cause is forgetting that start-up tails, offcuts, and reject coils carry recoverable metal. Fix: net Scrap Copper Value against material cost. Bare magnet-wire scrap commonly recovers 60 to 85 percent of the copper spot price after the processing fee and insulation strip loss, so on a shop scrapping 3 to 5 percent of copper by weight, ignoring recovery overstates true material cost by a meaningful fraction of a point.

Published 2026-07-01.