# Alarm analytics for the current session Open **View → Alarm Analytics…** to inspect frequent offenders, chattering alarms, standing durations, and acknowledgement response times. Collection starts when this runtime opens, even if the analytics window is closed or notifications are paused. Analytics use processed alarms, after QtALH alarm filtering. Analytics are held in memory for this runtime only. Closing QtALH loses the session. There is no shared collector, database, or import of old ALH logs. The configuration editor does not collect analytics. ## Read the dashboard Choose **Last 15 minutes**, **Last hour** (the default), **Last 8 hours**, **Last 24 hours**, or **Since Session Start**. Select a group/channel scope or search for part of a channel name or full path. The suppression filter selects all observed activity, activity while unsuppressed, or activity while suppressed. Historical counts and durations use the suppression state at the time of each observation; suppression labels in rows describe the channel's current state. Tables support sorting and selection. **Channel details** or a double-click opens the selected channel's measurements, coverage, and incomplete-sample information. The four tabs are: - **Frequent offenders:** Observed normal-to-alarm activations, each channel's share of activations, peak severity, last activation, and coverage gaps. Changes between alarm severities do not create another activation. Charts show the ten largest activation counts and an activation trend over retained history. - **Chattering alarms:** The default is five activations within 60 seconds. Adjust the count and window to recalculate chatter from retained observations. The table shows the current detection result, recent activation count, and maximum count in a detection window within the report's detail coverage. Select a channel to see its activation timeline. - **Standing alarms:** Currently active process alarms, ordered by observed standing duration. Acknowledgement does not end standing time; recovery does. The table also shows accumulated active time and observed time in the selected range. The chart shows the ten longest currently observed standing durations. - **Acknowledgement times:** Completed sample count, mean and maximum confirmed response time, outstanding age, and incomplete samples. The histogram bins are 0–1 seconds, >1–5 seconds, >5–15 seconds, >15–60 seconds, >1–5 minutes, >5–15 minutes, >15–60 minutes, >1–4 hours, and >4 hours. A channel already active when first observed has an unknown true onset. Its standing age is labeled **At least**: this means continuously observed duration, not an estimate of how long the alarm existed before observation. Outstanding acknowledgements with unknown onset use the same lower-bound label. ## Acknowledgements and coverage gaps A response sample starts at an observed transition into an unacknowledged state. Local acknowledgement completes the sample after the local state changes. Global acknowledgement requires IOC confirmation; making a request or attempting a failed write does not complete a sample. The engine distinguishes locally requested confirmation from externally observed acknowledgement without claiming an operator identity. A recovered alarm that remains latched continues waiting for acknowledgement. Automatic clears, masking, unknown onset, and interrupted measurements do not enter response-time statistics. A clear coincident with an ambiguous nonlatched IOC recovery is excluded rather than assumed to be an operator acknowledgement. Completed samples belong to the range containing their confirmation time; their observed onset may precede that range. Shelved, disabled, and NoAck channels continue contributing observed process activity, with suppression labels and filters. Mask changes and unshelving do not create activations. Silence, display filters, and NoLog do not affect collection. Analytics never acknowledge alarms, modify masks or shelves, send notifications, or write output PVs. Connection/access failures and Cancel create coverage gaps. They do not count as process-alarm activations or chatter. Recovery establishes a fresh baseline; QtALH does not invent the events or continuous duration that might have occurred while observation was unavailable. Synthetic initial ERROR placeholders are not process alarms. The dashboard reports diagnostic gaps and channel availability. Elapsed measurements use a monotonic clock. Detected clock discontinuities, suspend, or long event-loop stalls interrupt measurements conservatively. Fresh observations are needed before continuous measurement resumes. UTC timestamps identify observations and exports; changing the wall clock does not directly change measured response durations. ## History limits and reloads The detail buffer retains at most 24 hours, 250,000 compact records, or 64 MiB, whichever limit is reached first. It stores state transitions, not PV values. Session activation totals, active/observed durations, acknowledgement totals, means, maxima, and histogram counts survive detail eviction. A **PARTIAL DETAIL** notice identifies the start of available history. Rolling reports, timelines, and recalculated chatter use only that coverage. Missing history is never represented as a measured zero. Since-session summaries remain available, while their activation timelines can cover a shorter retained period. Rows with no process activity can still show unavailable or ambiguous channels. Reloading the same saved configuration preserves summaries for uniquely matched full hierarchy identities. It interrupts open measurements and waits for fresh observations. Removed channels retain their history with a Removed label; ambiguous identities are not merged. A failed configuration reload leaves the session unchanged. Opening another configuration starts a new session. **Reset Analytics** clears only analytics and establishes new observation baselines. Alarm, shelving, acknowledgement, and notification state remain intact. The range, search, and chatter controls are temporary dashboard settings. ## Export Use **Export table CSV…** for the current sorted table, or **Export chart CSV…** for plotted series. Exports use UTF-8, quote text safely, and protect text fields that spreadsheet software could interpret as formulas. They include the configuration, session identifier, scope/filter choices, requested and available coverage, units, and incomplete-measurement flags. The offender chart export includes both top-offender and trend series. The chatter chart export contains the selected channel's retained activation times. The standing chart export contains its top ten durations and lower-bound flags; the acknowledgement chart export contains the response histogram. CSV is an explicit report export, not persistent collection or an importable ALH file.