photo etching process control

The EMI shield photo etching process is only as good as the control behind it. Photo etching, chemical etching, photo chemical machining and PCM are the same route under four names: resist defines the shape, chemistry removes the metal, and nothing touches the work. What the process has is chemistry, time and registration rather than a tool to wear out or a force to measure, and all three drift. A shield program is therefore a control program, because the same artwork can produce a part that meets its drawing on Monday and misses it on Friday if the bath has loaded up or the resist has aged. This page is about that control layer: the variables that move, the checks that catch them, the signatures that identify a deviation, and the evidence a buyer can reasonably ask for.

1. A Chemical Process Has No Stop Button

Etching removes metal at a rate, not to a stop. There is no depth stop, no tool offset and no way to measure the cut while it is happening, so the only control is knowing the rate and stopping the clock at the right moment.

That single fact changes what capability means. A machined feature is limited by the tool and the machine; an etched feature is limited by how well the rate is known and held over the panel, the lot and the season. Two panels etched from the same artwork in the same week will differ slightly, and a supplier who claims otherwise is describing a different process.

The practical consequence is that quality is built from monitoring rather than from inspection at the end. Inspection proves the lot; the rate check is what keeps the next lot from failing.

A control plan for an etched part therefore looks like a list of measurements rather than a list of tools: bath condition, resist lot, exposure, rate, registration, final dimension. Each has a frequency, and the frequency is what makes the plan real rather than aspirational.

2. The Variables That Actually Move

Most of the variables in the process are stable. A few move constantly, and those are the ones a control plan has to name.

Etchant concentration and temperature drift as metal dissolves into the bath and as the etcher adds chemistry to compensate. Metal loading changes with the mix of parts on the panel. Filtration and agitation decide whether fresh etchant reaches a feature or whether a boundary layer slows the cut in a corner.

On the imaging side, resist thickness and adhesion decide whether an edge is clean or ragged. Exposure dose and development time decide the feature’s starting width. Age and storage of the resist, and the cleanliness of the laminator, decide whether pinholes appear in a field that should be solid.

The etch time itself is the last variable, and it is the one operators can see. The dangerous ones are upstream: a slightly warmer bath shortens every etch in the shift, and the parts still look right until they are measured.

Bath chemistry is measured by titration and by the etcher’s own dosing records, and the two should agree. When they diverge, the dosing is compensating for something the titration can already see, usually metal loading, and the rate will follow within a shift. Reading those two numbers together is the earliest warning the process gives.

3. Etch Rate, Depth and Half-Etch

Rate is measured, not assumed. A rate coupon etched with the lot gives the depth removed per unit time, and the etch time for the production panels follows from it. Where the coupon and the panel differ, the difference is usually agitation or loading rather than chemistry.

Half-etch is the same calculation with a smaller target. Depth is controlled by time and is less precise than a through-cut, because the same rate error applied to a shallow depth is a larger fraction of the result. A drawing that treats a half-etch depth like a machined step will be disappointed; a drawing that accepts a band will not.

Rate also sets the artwork compensation. When the rate shifts, the aperture changes while the feature position does not, which is why an over-etched lot fails on width and a misregistered lot fails on position. Telling those two failures apart quickly is most of what process diagnosis is.

4. Registration and Two-Sided Alignment

A through-cut is etched from both faces, so the two images have to line up. Registration is set by the phototool and by how the panel is located during exposure, and it is checked on the first piece of a lot rather than assumed.

The signature of poor registration is asymmetry: one wall steeper than the other, an aperture offset toward one face, or a slot that drifts along its length. Because the effect is directional, it is easy to separate from a rate problem, which is symmetric.

Alignment marks belong in the artwork for this reason. Without them the operator has no reference to measure against, and a small registration error becomes a scrap panel discovered at final inspection.

5. Monitoring, SPC and Re-qualification

A control plan for this process is short and specific. Aperture and outline are measured at a defined frequency per lot, rate is checked with every batch, and the results are plotted rather than filed. Plotting is what turns a series of passing lots into a trend that can be acted on before a failure.

Control limits are set from the process’s own history, not from the drawing. A process running comfortably inside a wide band is in control even when the parts sit near the middle of that band; a process running at the edge of a tight band is not, however good the last measurement looked.

Re-qualification follows the chemistry: after a bath is regenerated or dumped, after a change of resist supplier or thickness, and after any maintenance that touches agitation or filtration. Each of those can shift the rate, and a rate shift invalidates the etch time rather than the drawing.

Reacting to an out-of-control point matters more than detecting it. The usual response is to re-check the coupon, then the bath, then the resist lot, in that order, because the cheapest explanation is also the most common. A lot that fails on aperture is quarantined before anything is reworked, since a second etch moves width and position together and turns a recoverable lot into scrap.

6. Traceability and Release Evidence

A lot should be traceable back to the material and forward to the parts. In practice that means a lot number that links the material certificate, the bath log and the rate coupon to the dimensional report for that run.

The dimensional report should carry sample size and the measured range rather than a single good part, because a range is what the drawing’s tolerance is actually applied to. Where plating is specified, the plating certificate and the sequence belong in the same package.

Change notification is the part buyers underestimate. A change of resist, of etchant supplier or of a filtration setup can move the rate without touching the drawing, and a program that requires notification should say so at the start rather than discover it at a containment.

A release package for a shield lot is short: material certificate with the heat number, bath and rate record for the run, dimensional report with sample size and range, plating certificate where relevant, and the first article produced on the same line. Buyers who ask for more usually receive a binder that contains the same five documents.

7. Deviation Signatures

Most process problems announce themselves in a small number of ways. The useful skill is reading the symptom back to the cause before the lot is scrapped.

The table is not exhaustive, but it covers the deviations that reach a customer. Each one is cheaper to catch at the coupon than at final inspection.

8. Specifying So the Process Stays Controllable

A drawing can make a process easier or harder to control, and the difference shows up in yield rather than in argument.

Tolerance should follow thickness and be stated only where function lives. Critical dimensions should be few enough to measure at the agreed frequency. Aperture should be accepted as a range with a stated sample size, because that is what the process produces and what inspection can defend.

Where a half-etch depth matters, it should be called out as a band and checked on a coupon. Where registration matters, the datum should be a feature that exists on both faces.

Finally, the drawing should ask for what the buyer will actually use: a first article, a range on the dimensions that matter, and notification when the process changes. A supplier who can show the rate trend is a better partner than one who can only show a passing report. A tolerance band tied to thickness, a sample size and a named datum say more about intent than a table of two-decimal limits ever does.