EN
Quick Links
Two years ago a customer's engineering supervisor called me about a PCB manufacture disaster. They would certainly created a power supply board with 2mil trace width and 2mil spacing for the control area, and specified 2oz copper throughout the entire stack-up so the power courses might bring the present. The bare boards came back with approximately a 40% yield. Connection screening flagged dozens of dead internet per panel. Under the microscopic lense, the traces in the control section really did not look thin-- they looked like they 'd been sandblasted. Some were gone completely.
Here's the awkward component: that board was doomed from the moment the design rules were composed. The producer didn't slip up. The legislations of damp chemistry simply overtook the artwork.
Let's put the spec in devices that make the trouble obvious.
A 2mil trace is 50.8 microns large. 2oz copper is 70 microns thick. Read that again: the copper aluminum foil is 38% thicker than the trace is vast. You are asking the etchant to sculpt a structure that's taller than it is wide-- basically a 70-micron wall standing on a 50-micron impact. Then you ask it to do the very same to the area right next door, 2mil away.
Wet etching is isotropic. The chemistry doesn't recognize which instructions is "down." It strikes the copper sideways at almost the very same price it eats through vertically. The market evaluates this with the etch variable-- the proportion of upright etch depth to side undercut per side. In production, etch variables run between 2.5 and 4.0 depending upon chemistry, devices, and process control. Acidic cupric chloride on inner layers normally holds 3.0 to 3.5. Alkaline systems on outer layers sit reduced, around 2.5 to 3.0.
Do the arithmetic with an etch factor of 3.0 on 2oz copper:
Upright etch deepness needed: 70 microns
Side undercut per side: 70 ÷ 3.0 ≈ 23 microns
Total width loss throughout both sides: ~ 46 microns
Out of a 50.8-micron trace, the etchant takes 46 microns. What remains is a 4.8-micron sliver-- if it makes it through whatsoever. At an etch factor of 3.5, the trace comes out around 10 microns broad. Regardless, it's a blade edge, not a conductor. It can't bring current, it fails resistance targets, and it barely holds up against handling before breaking.
And the 2mil spacing? The very same 23-micron undercut on each trace edge diminishes the space from 50 microns to about 4 microns. The threat of copper bridging, micro-shorts, and dendrite growth in time skyrockets.
That's not a yield issue. That's a physics issue putting on a return's clothing.
Right here's the component that obtains shed in the blame game. Engrave payment-- the CAM process of widening art work to represent undercut-- is standard practice in every serious fab. A great producer will certainly adjust a 5mil style to a 5.5 or 6mil artwork size so the completed trace lands at 5mil. This works perfectly when the compensation is a small fraction of the attribute dimension.
It falls short entirely when the attribute is currently at the edge of what's possible.
To obtain a finished 2mil trace after ~ 46 microns of overall undercut, the art work would need to start at approximately 97 microns-- nearly 4mil large. Currently the spacing mathematics collapses: 2 4mil-wide artwork traces positioned 2mil apart in the design means the withstand attributes themselves are just ~ 51 microns apart on the panel. The resist can not survive that. The etchant undermines and raises it, and you obtain unpredictable, ragged etching throughout the entire layer. Fine-line etching on thick copper does not degrade gracefully-- it falls short chaotically.
This is why the sector's minimum trace size tables review the means they do. For 1oz copper, a 3mil trace is the ragged edge. For 2oz, the functional minimum leaps to 5mil. At 3oz it's 9mil. At 4oz, 12mil. These aren't traditional assumptions-- they're the sizes where engrave settlement quits being a gamble. Style a 2mil attribute on a 2oz layer and you're requesting a process that doesn't exist outside of research study laboratories.
The yield collapse isn't subtle. In my experience throughout about 30 heavy copper PCB projects over the last 3 years, the pattern corresponds:
Trace size listed below ~ 3x the copper density: returns in the 35-- 55% range, with failures dominated by opens up and near-opens
Trace size around 4x the copper density: yields in the 70-- 85% array, with periodic insusceptibility drift
Trace width at 5x the copper density or above: returns recuperate to 95%+.
The scariest failures aren't the dead shorts. They're the intermittent opens-- traces that pass connection testing at the factory since they're a 5-micron filament eventually, then crack under thermal cycling or resonance and fall short in the field six months later on. I've traced field failings back to boards that passed electric examination at the fab with flying colors. The trace looked penalty in the AOI images. Nobody X-rayed every 2mil internet, and the ones that had engraved down to hair-width were unnoticeable up until they broke.
There's also the resistance tale. Also when a slim trace makes it through, its cross-section isn't rectangular any longer-- it's trapezoidal, broad near the bottom and slim at the top. Regulated insusceptibility is calculated on rectangle-shaped geometry. A trapezoidal trace moves the real insusceptibility off the target in an instructions that relies on the stack-up and the undercut account. For differential sets, the mistake substances: the traces obtain narrower and the space in between them obtains bigger, so differential insusceptibility drifts greater than single-ended. Boards that "need to" have hit 100 ohms gauged 96. Boards at 90 ohms gauged 87. Minimal layouts headed out of specification entirely.
I want to be fair to the designers who request 2mil/2oz. Nobody defines that combination to be tough. The reasoning typically seems like this: "I need 3 amps on this power rail, the calculator states a 1oz trace needs to be X wide for that, so I'll go 2oz and keep the trace narrow." Or: "The room is tight, and 2oz lets me reduce the power traces.".
The blunder is treating copper weight as a knob that just influences present capability. It's not-- it's a handle that drags the whole layer's minimum trace size and spacing along with it. Thicker copper buys you current handling on large traces and concurrently ruins your capacity to path anything great on the exact same layer. You can not have both on one copper weight, since the etching procedure does not negotiate.
I've seen teams escape this catch, and the remedies comply with a consistent pattern:
Divide the copper weights throughout layers. Place the 2oz and even 3oz copper on a devoted power layer with broad traces and charitable polygons. Place the fine-pitch control logic on a 0.5 oz or 1oz layer. 2 copper weights in one stack-up is completely standard-- most multilayer PCB fabs manage mixed-ounce stack-ups without dramatization. The board costs a bit more. The yield difference spends for it often times over.
Maintain 2mil transmitting on thin copper, period. If the design truly requires 2mil trace and room-- for BGA getaway directing, as an example-- that section belongs on a layer slim enough to etch it dependably. 0.5 oz copper with an excellent alkaline line and LDI imaging can hold 2mil attributes. 2oz copper can not, in any type of fab, in any chemistry, with any kind of CAM compensation. It's not a capability question; it's a geometry concern.
Widen prior to you switch copper weight. A 3-amp rail at 1oz needs approximately a 50-60mil trace for a moderate temperature level surge. At 2oz, the exact same rail requires about fifty percent that-- still far above 2mil. If the routing area really won't suit it, the response isn't 2oz on fine traces. The solution is a specialized power plane or a much heavier copper layer with no fine functions on it.
Ask for a DFM testimonial before dedicating to the stack-up, not after. Every serious fab offers cost-free layout for manufacturability look at Gerber upload. The ones worth dealing with will flag a 2mil/2oz mix prior to the order is put and tell you what the return will certainly be. If your maker quotes your 2mil/2oz board without a single remark, that's a warning sign-- either they're going to make it with terrible return and pass the expense on, or they're mosting likely to make it and allow you discover the trouble at test.
Heavy copper and penalty lines are not a design choice-- they're a contradiction in terms. The etchant that makes 2oz copper beneficial coincides etchant that eliminates 2mil traces, and no quantity of process wizardry alters the proportion. The etch aspect math is public, it's steady, and it does not care about target dates.
The groups that deliver high-current boards on schedule figured this out by hand: put the copper where the existing is, put the resolution where the signals are, and never ask one copper layer to do both. Every 2mil/2oz board I have actually ever before seen created has actualy been either a low-yield prototype, a field-failure investigation waiting to occur, or an extremely pricey lesson in etching chemistry. None shipped twice.
Article written by Zeon
Hi, I'm Zeon — 20 years in PCB and electronics manufacturing. Front-end design and R&D, components sourcing, precision SMT, DIP through-hole assembly, and complete unit assembly. That's the full path from concept to finished product, and it's the path I've walked for two decades.
Hot News2026-09-17
2026-09-09
2026-09-07
2026-09-04
2026-09-03
2026-09-01
2026-08-31
2026-08-27
