Between “no data” and “full historian” sits a wide, useful middle: the CPU logging its own story to the memory card.
The toolkit
DataLog instructions (DataLogCreate/Write and family) append records — your chosen tags plus timestamp — into CSV files on the card, downloadable via the web server. Production counts per shift, temperature profiles per batch, energy snapshots, fault context records: a few dozen lines of SCL and the machine keeps its own diary. Traces remain the high-speed forensic tool (cycle-resolution, triggered, short); DataLogs are the long-duration, low-rate diary. Choose by question: “what led to the trip?” → trace; “how did this drift over six weeks?” → DataLog.
The budgets that bite
Cards wear: log rates in the seconds-to-minutes class are fine for years; logging at cycle time to flash is a card-murder plan — that need means trace or external historian. Files grow: rotate by the built-in record limits and create-new-file logic, and actually download/clear on maintenance visits (a checklist line, or the files silently plateau). Timestamps mean something only with NTP — the same clock discipline as everything else diagnostic.
The pattern we ship: one “flight recorder” DataLog per machine (state, counters, key analogs at 10–60 s), one “event context” log written on faults (a burst of pre-fault values held in a ring buffer), both harvested on every service visit into the machine dossier.
FAQ
Can the customer’s Excel people use the output? That is the point — CSV over the web server, no toolchain. For anything bigger, the same interface DB feeds SCADA later without rework.
Does logging survive CPU swaps? Files live on the card; carry the card (and its logs) to the spare CPU — then update the dossier, because serials changed.
Zone Otomasyon leaves every retrofit machine keeping a diary — the first year of data usually pays for the habit. Monitoring options up to Zsense.