On this page
A ratio-trimmed array cannot be released from the phrase “trim to ratio” alone. The numerator, denominator, node access, excitation, loading, temperature and time after the last cut all affect the observed endpoint. An instrument may report a precise ratio that is not the ratio used by the assembled circuit, or a mathematically correct endpoint may be unreachable because the remaining trim directions only move it away from target. This guide turns the functional requirement into a measurable quantity, checks observability and adjustment direction, and defines evidence for stopping without inventing a universal trim resolution.
Key design decisions
- Write the exact functional equation and identify which accessible nodes reproduce it.
- Allocate endpoint error among trim increment, measurement uncertainty, loading, settling and recovery.
- Prove that the selected trim sequence approaches the acceptance region without consuming required adjustment direction.
Start from the circuit function, not resistor labels
Identify the output that matters: divider gain, bridge balance, weighted sum, differential gain or another transfer quantity. Write it using the actual network connections. Two arrays with the same list of resistor values can require different endpoints because shared nodes or loads alter sensitivity. If the final circuit adds parallel paths, decide whether trimming occurs before or after those paths are present.
Assign names to every measured terminal and retain the polarity convention. State whether the array is measured as bare ceramic, with jumpers, or after selected assembly operations. A resistance ratio between isolated elements is not automatically equivalent to a loaded voltage ratio. The endpoint definition must be reproducible by production and independently checkable by incoming or final test.
Confirm that the endpoint is observable at available nodes
Map force and sense leads through pads, conductor traces, switches and fixtures. Determine whether the instrument measures each resistor separately, measures a voltage ratio under one excitation, or solves several readings. Contact resistance may cancel in a high-impedance voltage observation but contaminate a low-resistance element reading. Leakage and finite input resistance can change a high-impedance divider.
Prefer an endpoint that minimizes reconnection and exposes the same electrical boundary each time. If switching is necessary, validate channel isolation, settling and thermoelectric offsets. Include a known check network or repeated reference reading. The best mathematical endpoint is unsuitable if it cannot be measured with uncertainty smaller than the remaining acceptance window.
Calculate how each adjustable element moves the endpoint
For a simple unloaded divider with output across R2, y = R2/(R1+R2). Increasing R2 raises y, while increasing R1 lowers it. The partial derivatives quantify how much a possible trim increment changes the endpoint. For a larger array, calculate a sensitivity vector at the measured state because the effect changes as resistance changes.
Use sign as well as magnitude. A laser cut normally increases an element's resistance; it cannot undo an overshoot by decreasing that same element. Select coarse and fine elements or an ordered sequence that retains both required endpoint direction and adequate final resolution. Shared resistors require simultaneous evaluation of all outputs they influence.
dy ≈ (∂y/∂R1)ΔR1 + (∂y/∂R2)ΔR2 + ... + (∂y/∂Rn)ΔRn
- y: defined measurable functional endpoint
- Ri: resistance of adjustable or fixed element i
- ΔRi: proposed resistance change from the next trim action
- ∂y/∂Ri: local endpoint sensitivity to element i
Small changes around the current state; nonlinear behavior and instrument loading are included in the full endpoint model when significant.
Build an endpoint budget before selecting the stop band
Separate target tolerance from the capability of observing and stopping near it. Budget instrument accuracy, repeatability, switching, fixture contacts, test temperature, self-heating, digitization, controller delay and minimum effective cut. Include recovery between the last cut and the release reading. A narrow displayed band is not useful if these terms occupy most of it.
Avoid simply adding every specification at its maximum when dependencies are known, but do not use root-sum-square without a justified uncertainty model. Define the internal control band and the final acceptance band separately if appropriate. The control band must leave margin for expected post-trim movement; that margin is established from project evidence, not copied from another array.
Model endpoint loading and excitation
For voltage-ratio measurement, include the instrument input resistance, guards and any connected circuit branches. For resistance measurements, record test current and compliance. Self-heating can move elements unequally, especially when values or geometries differ. Use excitation representative enough to expose the relevant behavior but below any project-specific perturbation limit.
Measure at a second lower excitation to check whether the endpoint shifts with power. Reverse polarity when it helps reveal offsets. If the trim system and verification station use different instruments, perform a correlation study using the same arrays and connection boundary. Do not declare one station correct solely because it has more digits.
Test endpoint reachability with a numerical example
Suppose an unloaded divider uses R1 = 10.00 kΩ and R2 = 9.92 kΩ, giving y = 0.497992. The target is 0.500000 and only resistance-increasing cuts are available. Increasing R2 moves y upward, while increasing R1 moves it downward. The first directional choice is therefore R2. This example illustrates sign; it does not state a trim capability.
The target condition R2 = R1 would require R2 to increase by about 80 Ω if R1 remained fixed. If the remaining safe geometry cannot supply that change, the endpoint is unreachable and should be held before trimming. If the estimated final effective increment corresponds to 150 ppm of y while only 100 ppm remains in the internal band, a finer element, different starting values or a wider approved tolerance is needed.
Verify stop, recovery and independent reproduction
Record the endpoint immediately before each final trim action, after the controller settles and after the defined recovery interval. Return the array to the reference temperature and repeat the measurement without further cutting. Then reproduce the result with the independent station or connection method specified by the control plan.
Inspect the trim geometry for termination approach, excessive cut interaction, debris or damage to adjacent conductors and glaze. Correlate unusual electrical steps with images and timing. A successful displayed endpoint followed by an out-of-band recovered result is a process signal, not permission to redefine the official measurement time after the event.
Release one endpoint definition with explicit failure signatures
The drawing or test record should state the endpoint equation, terminal map, excitation, load, switching sequence, temperature, stabilization, stop band, final band, recovery interval, allowed adjustable elements and measurement uncertainty. Store individual resistor readings when available so ratio behavior can be diagnosed later.
Failure signatures include endpoint movement opposite to calculated sensitivity, repeatable overshoot, dependence on connection order, power-dependent ratio, recovered drift, disagreement between trim and verification stations, or an unreachable target. Stop and investigate rather than adding opportunistic cuts. The approved endpoint applies only to the named network, layout and test boundary.
| Item | Required evidence | Failure signature |
|---|---|---|
| Functional equation | Node-level circuit model | Measured ratio does not predict circuit output |
| Sensitivity direction | Calculated or controlled element perturbation | Cut moves endpoint the wrong way |
| Observation boundary | Force/sense and loading map | Connection order changes result |
| Stop budget | Increment, delay and uncertainty allocation | Overshoot consumes final band |
| Recovery | Timed return measurement | Endpoint exits band without more trimming |
Send the ratio-array endpoint definition
Provide the circuit and measurement data required to review a reachable, reproducible trim endpoint.
- Complete network equation, target outputs, resistor values, tolerances and shared-element relationships.
- Terminal map, fixture connections, excitation, instrument input, switching and temperature conditions.
- As-printed distributions, allowable trim elements, trim geometry and directional adjustment limits.
- Stop band, final acceptance band, recovery interval, uncertainty and independent verification method.
The drawing-upload form loads as you reach this section.