Explainer
What a SCADA Operator Interface Must Do: Core Functions
Short answer
A SCADA operator interface carries six functions: acquisition and display of binary and analog data, alarm and event recording, a history database of process values, measurand processing for derived values, presentation through graphics, lists and reports, and command handling from binary commands to recipes. Keep alarm and event records separate from the history database: one records occurrences, the other retains values.
On this page
- The six functions at a glance
- Acquisition and display
- Alarms and events are not the history database
- Working the numbers for one analog tag at a 1 s log rate
- Measurand processing: derived values before limits and trends
- Presentation: graphics, lists and reports
- Command handling and the systems around it
- The specification, one line per function
- Questions asked
A SCADA operator interface is not one program but six functions you specify separately: acquisition and display of binary and analog data, alarm and event recording, a history database of process values, measurand processing for derived values, presentation as graphics, lists and reports, and command handling. The rule that settles most arguments: alarms and events are recorded occurrences; the history database retains values. They are different stores with different write triggers, and the major products keep them apart for that reason. Write the specification along these six lines and every later decision — driver, database, screen, command — has one home.
The six functions at a glance #
| Function | What it handles | What it must deliver |
|---|---|---|
| Data acquisition and display | Binary and analog data | Presented values and a stored copy in the process database |
| Alarms and events | Important process changes and operator actions | Timestamped occurrence records |
| History database | Process values | A retained record of values over time |
| Measurand processing | Raw measurements | Derived values for limit supervision and trending |
| Presentation | Current and stored data | Graphical object states, lists, reports, logging |
| Command handling | Operator intent | Binary commands, setpoints, recipes, batches, scripts, command procedures |
Each function has its own inputs, its own storage and its own failure mode, so each gets its own line in the specification. A panel that displays beautifully and logs nothing has failed the history function just as surely as a blank screen has failed presentation.
Acquisition and display #
Acquisition brings binary and analog values from the controllers into the process database and puts them on screen. Scale comes first: a documented plant system connects over 2700 devices across production lines, utilities and support systems on Ethernet and RS485, and every one of those devices becomes at least one acquired tag. The specification therefore names the tag list, the protocol and driver for each network, the scan rate per tag class, and where the timestamp comes from — the controller or the interface. Where values must reach a reporting database, logging PLC data to SQL through a gateway keeps that traffic off the operator panel.
Alarms and events are not the history database #
The alarm and event function records occurrences: important process changes and operator actions, each with a timestamp. The history database retains values on a schedule. The write triggers differ — an event is written when something happens, a historian value when the log interval or the deadband says so — and so do the volumes. Documented products such as AVEVA Historian and Fernhill SCADA keep process values and event records as distinct data, queried separately. When you choose the product that holds the values, compare SCADA historian functions with a fixed test method across candidates rather than trusting datasheets.
One failure mode belongs in this section: an interface that derives events by watching tags misses transitions that happen while the link is down, which is the same reason why HMI operation records miss PLC start stop transitions. Events that must survive a comms loss are timestamped in the controller, not on the panel.
Working the numbers for one analog tag at a 1 s log rate #
Inputs, with sources: one analog loop tag, chosen here as the design input for a typical fast loop; a history log rate of 1 value per second, the rate such a loop is normally given; one limit on that tag, crossing into alarm once per day and back once, so 2 event records per day; and the day itself, 60 x 60 x 24 = 86,400 seconds.
The rule in words: the historian writes log rate times seconds per day; the event log writes one record per recorded transition; the ratio between the two is the case for two separate stores.
The arithmetic:
- 1 value/s x 86,400 s/day = 86,400 value records per day in the history database
- 2 event records per day in the event log
- 86,400 / 2 = 43,200 historian records for every event record
The decision: at a 1 s log rate and two alarms a day, one tag outwrites its own event log by a factor of 43,200. Any time a tag logs faster than it alarms — past a few hundred values per event — the two streams belong in separate destinations with separate retention, and a 1 s trace is never rebuilt from event records, because they hold the steps and none of the path between them.
Measurand processing: derived values before limits and trends #
Measurand processing calculates derived values — totals, rates, averages — so that limit supervision and trending act on the computed result rather than the raw measurement. A limit watched on a raw counter is not a limit on the total. Each derived value is specified as a formula with named input tags, an execution rate and a storage destination. In tools that run scripts per screen, a value computed only while its screen is open is a different tag from one computed every scan, and the specification has to say which one you are getting.
Presentation: graphics, lists and reports #
Presentation delivers current and stored data as graphical object state presentation, lists and reports. Object state means appearance driven by tag state — running, stopped, faulted — not decoration, and the alarm list and event list are first-class screens with their own filters. The reporting side covers logging and scheduled reports built off the history database. Screen design decides whether the operator can act at 2 a.m., which is the discipline behind designing HMI screens for clear operator decisions. When the station is a PC-based panel running the SCADA client or a visualization runtime, industrial panel PCs are the class of hardware that mounts at the machine.
Command handling and the systems around it #
Command handling covers binary commands, setpoints, recipes, batches, scripts and command procedures. Each command in the specification names its write target, the value or sequence written, the permission level required, whether it demands confirmation, and the event record it leaves behind — who, what, when. One rule is not negotiable: interlocks and safety functions belong in the controller, and the interface sends and logs commands; it is never the guard. The same layer interfaces outward to planning and analysis systems such as a CMMS, handing over events and history rather than duplicating those stores on the panel.
The specification, one line per function #
- Acquisition and display: tag list per device, protocol and driver for each network, scan rate per tag class, timestamp source.
- Alarms and events: trigger list with limits and deadbands, event classes for process changes and operator actions, a destination separate from trends, retention, acknowledgement behavior.
- History database: tags logged, log rate, compression or deadband, retention period, and a restore test before anyone relies on it.
- Measurand processing: each derived value as a formula with its input tags, execution rate and storage destination.
- Presentation: screen list, the states each object can show, alarm and event lists as dedicated screens, report schedule.
- Command handling: command list with write targets, permission levels, confirmation for writes, the event record each write produces, and the named owner of interlocks — the controller, never the screen.
Questions asked
Can I rebuild a 1 second trend from alarm and event records?
No. Event records are written on transitions, not on a schedule, so a trend rebuilt from them shows step changes with nothing in between. Use the history database for the trace and the event log for what happened and who acted: one tag at a 1 s log rate writes 86,400 value records a day against a handful of event records.
Where should the event log live on a single-panel installation?
In a destination separate from trend data: the panel's internal memory, an external database, or a scheduled file export all work. The point of the separation is that clearing or rotating trends never touches the audit record, and a scheduled export off the panel keeps events survivable if the panel itself fails.
What happens to event records when the panel loses communication with the controller?
An interface that derives events by watching tags misses transitions that occur while the link is down, and the values it sees on recovery can imply a change that was never recorded. Where that gap matters, timestamp the events at the source in the controller or buffer them there.
Are recipes and batches part of command handling?
Yes. A recipe is a set of setpoint writes and a batch is a sequence of command procedures, so both live in the command layer, with every write logged as an action with its origin. The controller still owns the interlocks and the sequence logic.
Does the history database need settings of its own?
Yes. Log rate, deadband or compression, and retention are history-function settings, separate from the acquisition scan rate. A tag scanned every second does not have to be stored every second; the compression rule decides what the historian keeps, and it needs a restore test before anyone relies on it.