Common Mistakes

PCB and SMT Manufacturing Mistakes That Wreck Your Numbers

The eight errors that make SMT throughput, panel utilization, and paste estimates lie to you, each with the symptom that gives it away and a fix tied to a real number.

The most common SMT throughput error is quoting nameplate CPH as if it were real output. A machine rated at 40,000 CPH runs a mixed board at 12,000 to 18,000 effective placements per hour once you account for large parts, fine-pitch nozzle changes, and vision cycles. Symptom: quoted cycle time is half of measured. Root cause: nameplate assumes optimal chip-shooter conditions on 0402 parts only. Fix: pull the actual placement count per board from the program, divide by measured cycle time, and feed that real rate into the SMT Placement Rate calculator instead of the datasheet number.

Feeder setup time gets buried, then blows up small-run cost. Symptom: a 50-board job takes a full shift while the machine only ran 40 minutes. Root cause: nobody counted 30 to 60 unique feeders at 3 to 5 minutes each to load, plus first-article verification. That is 90 to 300 minutes of changeover against 40 minutes of placement. Fix: run the numbers through the Feeder Setup Time and Feeder Setup Cost calculators before quoting, and set a minimum run quantity where changeover falls below 30 percent of total job time, usually around 200 to 500 boards.

Panel utilization math fails when people forget the machine border and V-score kerf. Symptom: the fab quotes fewer boards per panel than your spreadsheet predicted, and cost per board jumps 15 to 25 percent. Root cause: you used the raw 457 by 610 mm panel area instead of the usable area after a 5 to 10 mm routing rail on each edge and 2 mm inter-board spacing. Fix: subtract the rails and spacing first, then run Boards Per Panel and PCB Panel Utilization. A 100 by 80 mm board on an 18 by 24 inch panel typically nets 40 to 55 percent utilization, not the 70 percent a naive area ratio suggests.

Solder paste estimates go wrong on unit basis, not just quantity. Symptom: you order paste by the pad but the line burns through three times the volume. Root cause: ignoring stencil aperture volume, print transfer efficiency around 70 to 90 percent, and squeegee waste of 10 to 20 percent per print. A 500 g jar covers fewer boards than pad-area math implies. Fix: base the Solder Paste Usage estimate on deposited volume per board times boards per hour, then add 15 percent for waste and skips, and track actual grams per 1,000 joints as a sanity check, typically 0.3 to 0.6 g.

Utilization and throughput get confused, so line balancing decisions go backwards. Symptom: the slowest machine sits at 95 percent while a faster machine idles at 60 percent, yet line output is flat. Root cause: treating Pick-and-Place Utilization as the same thing as SMT Line Throughput. A line runs at the speed of its bottleneck, so an unbalanced placement split caps output regardless of average utilization. Fix: measure per-station cycle time, move components to level the load within 10 percent station to station, and confirm the gain with the SMT Line Throughput calculator before buying more capacity.

Yield loss gets applied once instead of compounding across steps. Symptom: your good-board cost is understated and margin evaporates at final test. Root cause: multiplying by a single 98 percent yield when print, reflow, AOI, and ICT each shave 0.5 to 2 percent. Five steps at 98.5 percent compound to about 92.7 percent, not 98.5. Fix: multiply stage yields in series, then inflate material and machine cost per board by the reciprocal. At 92.7 percent rolled throughput yield you must start 1,079 boards to ship 1,000, and your PCB Cost Per Board input must reflect that.

Mixed units quietly corrupt paste, aperture, and spacing math. Symptom: apertures look right in the CAD file but deposits are 30 percent low. Root cause: mixing mils and millimeters, or treating a 5 mil stencil as 0.005 inch when the aperture was drawn in microns. One mil is 0.0254 mm, so a 4 mil versus 5 mil stencil is a 25 percent volume swing on a fixed aperture. Fix: lock every input to one unit system before it touches a calculator, and verify stencil thickness in microns against the paste maker's area ratio guidance of 0.66 or higher.

Stale program data makes every downstream estimate confidently wrong. Symptom: throughput and paste numbers were accurate last quarter and now miss by 20 percent. Root cause: the BOM changed, a 0603 became an 0402, a feeder slot moved, but the placement count and nozzle map fed to the calculators were never refreshed. Fix: version the machine program with the BOM, re-extract placement count and unique-part count on every revision, and re-run SMT Cycle Time and Feeder Setup Time whenever component count shifts by more than 5 percent.

Published 2026-07-01.