A cluster probe cell shares handling equipment across several test modules. Extra modules help only when handling, alignment, scheduling and recovery can keep them supplied. Begin with a capacity bound, then use event logs to discover where the real cell spends its time.
Define one completed wafer
Specify the wafer size, test recipe, compatible modules, lot mix and completion rule. Count a unique wafer once when its required tests and result records are complete. Keep test visits, retests and accepted wafers as separate measures.
Draw the route from incoming carrier through identification, alignment, loading, probing, unloading and return. Mark which operations use the shared handler and which can run concurrently. If a tester, aligner or buffer is also shared, give it its own resource line. This simple route map prevents hidden serial work from disappearing into a module headline rate.
Calculate a bound before predicting output
On narrow screens, scroll the table horizontally. Keyboard users can focus it and use the arrow keys.
| Resource | Record | Why it matters |
|---|---|---|
| Test modules | Number eligible for this recipe; occupancy seconds per wafer | A nominally installed module may be unavailable for the actual test. |
| Shared handler | Total exclusive handler seconds per completed route | Include required loaded and empty moves, handshakes and occupied waiting. |
| Alignment and buffers | Service time, capacity and permitted routes | Blocking and starvation can reduce the achievable schedule. |
| Recovery and retest | Event counts, duration, repeated visits and wafer disposition | These consume capacity and can change the product mix. |
Define occupancy consistently, including any transfer time that blocks the module. Count handler demand separately even when a transfer occupies both resources. Overlap assumptions need a feasible schedule; adding resource rates together overstates throughput.
Test whether the schedule can achieve it
Research on dual-arm cluster-tool scheduling shows how wafer mix, robot activity and residency constraints affect feasible cycles. Its fabrication-tool model is useful scheduling context, not a performance specification for a probe cell.
For the real cell, log wafer ID, module, event start/end, queue entry and fault state on a common time base. Compare pure-recipe runs with a representative mixed lot. Distinguish a module waiting for a wafer from one holding a finished wafer while waiting for unloading; they suggest different constraints.
Include recovery in the acceptance trial
Agree on permitted trials for a module taken out of service, an unreadable ID, a retest and an interrupted route. Verify that the controller preserves wafer identity, resumes the correct recipe and prevents duplicate completion records. Follow the equipment maker's recovery procedure.
Report elapsed run time, unique completed wafers, retest visits, downtime and queue distributions together. Select an improvement against the observed constraint, then rerun the same mix. Use wafer identification and traceability to plan identity checks across the handoffs.
Use the working record
Download the editable working record. It includes the task-specific fields, instructions, fictional example and source references needed to preserve the evidence behind a decision. Complete it with the people responsible for the actual process.