Carbon Contact PCBs in Key Matrices: Identify Ghost Paths Before Changing Ink

System boundary

Polymer carbon contacts, their mating mechanics, row drivers, column sensing and the key-event policy form one interface. The printed contact is not a complete keyboard controller or a safety function.

Integration interfaces

Integration risks

Define one complete scan state

Name each row and column and state which row is currently selected. In the example used here, the selected row is driven low, unselected rows are high impedance, and both column inputs have pull-up resistors to a 3.3 V supply. Closed switches are represented by their contact resistance. This is a declared analysis configuration, not a recommendation for every controller. A driver that actively drives unselected rows high requires a different model and can introduce output-contention concerns.

Record input-low and input-high guarantees, pull-up tolerance, output-low impedance, input leakage and the delay between changing a row and sampling a column. Include any external diodes or resistors. A controller's internal pull-up is not necessarily a precise fixed resistor. The schematic and firmware configuration together define the circuit that the printed board must support.

Trace the three-key path to an unpressed position

Use a two-row, two-column corner of the keyboard. Close R1–C1, R1–C2 and R2–C1, leaving R2–C2 open. When R2 is selected low, C1 is directly connected to that row through its closed contact. C2 also reaches the selected row, but indirectly: C2 to R1 through one closed contact, R1 to C1 through another, then C1 to R2 through the third. The unselected row is part of the current path even though it is not being actively driven.

The controller observes column voltages, not the physical route followed by current. If the indirect path pulls C2 below the applicable input-low threshold, the scan can report R2–C2 as pressed. All three real contacts may satisfy their individual resistance requirements. The false fourth position is an observability limitation of this combination, rather than proof that the open contact surface has become conductive.

Include both pull-ups in the ghost-voltage calculation

For an illustrative steady-state calculation, make every closed contact 2 kilohms and both column pull-ups 100 kilohms. Neglect row-driver resistance and input leakage. The path from C2 to C1 contains two contacts in series, or 4 kilohms; C1 connects to ground through the third 2-kilohm contact. Both pull-ups feed this network. Ignoring the pull-up on C1 would give the wrong numerical voltage at C2.

Solving the two node equations gives C1 approximately 0.125 V and C2 approximately 0.247 V. If the controller's guaranteed low threshold for this hypothetical supply state were 0.8 V, both columns would be read low. That threshold is assumed for the example, not a universal logic limit. The calculation shows why good contacts can produce a ghost; it does not establish the permissible resistance of a supplied keypad.

Separate ghosting from material and timing faults

Build a result table containing the physical keys held, the selected row, raw column states and the decoded key events. A fault that appears only with a rectangular three-key combination strongly motivates the indirect-path check. It does not exclude contamination or intermittent contacts, so preserve the single-key readings and inspect the open positions independently.

Do not use contact resistance as accidental isolation

Raising contact resistance might stop one indirect path from crossing a threshold, but it also weakens a legitimate single-key low. Tolerance, wear, actuation force, humidity and pull-up variation can change which condition fails first. A design that depends on three uncertain contact resistances being large enough to reject a ghost while one remains small enough to register a key needs an explicit worst-case analysis; a nominal bench success is not a robust isolation strategy.

Lower resistance can improve legitimate detection while making the ghost path electrically stronger. This is not evidence that a better conducting ink is intrinsically worse. It means the material improvement exposed a limitation of the scan architecture. Keep the material acceptance target tied to the required contact behaviour, while the electronics owner resolves combinations that the circuit cannot distinguish.

Choose isolation or an explicit combination policy

Per-key directional isolation can interrupt unwanted paths when implemented with a topology and voltage budget suitable for the scanner. Its forward drop, leakage, orientation, assembly and fault behaviour must be reviewed rather than assumed invisible. Alternatively, a controller may identify ambiguous combinations and suppress events. That avoids certain false reports but also limits which legitimate simultaneous key presses can be accepted.

Specify the required rollover behaviour as a user-interface requirement. Emergency, access-control or other consequential commands need a system-specific treatment of ambiguous and failed inputs. Neither a printed carbon pad nor a software debounce interval by itself provides that assurance. Test the chosen rule against the complete set of required combinations and confirm what happens when keys are pressed and released in different orders.

Test the printed contact in its actual mechanical stack

A polymer carbon contact printed on an FR4 board works with an actuator, dome or conductive mating element. Contact pressure, alignment, surface state and wear belong to that assembly. Do not transfer ceramic firing temperatures or ceramic-resistor stability claims to the polymer contact. The cure and later solder-processing history remain material-process controls separate from the matrix ghost-path analysis.

Measure representative contacts with the actual scan excitation where feasible, and retain the settled voltage as well as any separate resistance measurement. Exercise required key combinations, not merely repeated single-key presses. For a suspected timing problem, retain raw waveforms and row transitions before adding a longer debounce setting; debounce addresses event stability, whereas an indirect steady-state path remains present after unlimited settling.

Release the matrix behaviour with the board drawing

The useful handoff is a pin-labelled matrix, contact construction, row-state definition, electrical threshold budget and tested combination policy. Record both raw electrical observations and the events reported to the application. This separates a board-level contact discrepancy from a controller interpretation and makes later firmware changes reviewable.

Revisit the analysis after changing pull-ups, logic supply, controller mode, contact resistance limits, isolation components or required simultaneous commands. A mechanically identical carbon-contact board can behave differently with a new scanner. Maintaining the electrical interface alongside the artwork prevents a supplier from being asked to fix an ambiguity that the board alone cannot resolve.