Explainer
Do you need a historian or just an OPC tag server?
Short answer
No. An OPC tag server (OPC DA) supplies the current value, quality and timestamp of each item and stores nothing, so monthly figures need a historian or a SQL logger collecting from that same server. Keep a reporting client like ActiveFactory only if you also run the historian it expects.
On this page
- Does an OPC tag server store any history?
- What is the difference between OPC DA and OPC HDA?
- Why does the HMI show live values while the report shows nothing?
- Should the historian collect from the HMI or from the source server?
- What do exception and compression deadbands do to monthly figures?
- How much storage does a small collection job need?
- Can you keep a reporting client like ActiveFactory?
- Is logging to SQL Server a valid historian substitute?
- So which one do you need?
- Questions asked
No — a tag server alone will not give you monthly figures. An OPC tag server (an OPC DA server) hands a client the current value, quality flag and timestamp of each item and keeps no history at all; anything that looks backward in time — trends, monthly totals, shift reports — needs a component that has been storing samples, which means a historian or a database logger collecting from that same server. This page separates the three jobs people confuse: acquisition, storage and reporting. It covers what OPC DA and OPC HDA each answer, where the historian should collect from, which deadband settings decide what actually lands in the archive, how much disk a small collection job needs, and the decision for keeping a reporting client such as ActiveFactory against an HMI such as Genesis 32.
Does an OPC tag server store any history? #
No. An OPC DA tag server hands a client the current value, the quality flag and the source timestamp of each item, and it keeps zero history: disconnect the client for an hour and that hour is gone for good.
A tag server does one of three jobs. The OPC server acquires: it talks to the controller and exposes items. The historian stores: it samples those items, stamps them and keeps them for a retention period. The trend or report client presents: it asks a historian for an interval and renders it. Empty trends under a working screen is the signature of the middle job missing, and no client-side tool substitutes for it.
What is the difference between OPC DA and OPC HDA? #
OPC DA is the live interface: subscribe and you get changes as they happen. OPC HDA is the history interface: a client asks what an item held between two timestamps, and only a component that has been storing samples can answer.
A DA server cannot answer an HDA question, and a reporting client that speaks only DA cannot read history from anything. HDA servers are normally distributed alongside a DA server covering the same tags, but the two use separate interfaces, and on several products the historical ItemID differs from the live ItemID for the same measurement. Browse the HDA namespace and map tags separately rather than pasting DA item names into it.
Why does the HMI show live values while the report shows nothing? #
Because a changing number on the screen only proves the live chain — controller, OPC server, HMI tag — and says nothing about storage: the reporting client is asking a different question of a component that does not exist yet.
Genesis 32 sits at the client end of that chain, reading from whatever OPC server is tied to your controller or I/O system. Read the symptom for what it proves:
| What you see | What it proves | What is missing |
|---|---|---|
| Genesis 32 updates current values | Controller, OPC server and HMI tag all work | Nothing — acquisition is fine |
| A trend shows no history | No component has been storing samples | A historian collecting from that server |
| Monthly figures are unavailable | No stored series exists to aggregate | Collection plus a report aimed at it |
Confirm the live binding before spending anything: pick one tag that visibly changes, record its server, its item, its quality and its source timestamp, and check that quality is good and the timestamp advances. If it does, the gap in your reports is storage, not communications.
Should the historian collect from the HMI or from the source server? #
From the source server, whenever it exposes the items you need: that keeps the historian independent of the HMI runtime, so one restart does not blind both your operators and your reports at the same time.
Two collection routes exist. Route one: the historian subscribes to the same OPC DA server the HMI already uses. Route two: the HMI itself exposes an OPC server and the historian reads that; some HMI packages include this, some license it separately. Choose route two only for values that exist solely inside the HMI, such as calculated or internal tags the source server never exposes. Prove independence once: stop the collector, confirm Genesis 32 still runs, restart the collector, and confirm the same item, value, quality and timestamp return.
If the collector is a new box rather than an existing server, a fanless mini PC in the control cabinet is the usual host for an unattended job. What that host has to survive — warm cabinets, dust, power cuts, unattended restarts — is covered in running a compact industrial PC as an always-on server.
What do exception and compression deadbands do to monthly figures? #
They decide which values are stored: with an exception deadband of 0.5% of span on a 0–100 bar signal, only changes larger than 0.5 bar from the last reported value reach the archive.
In PI, exception reporting is documented as a per-point deadband that decides whether each new value is sent on to the data archive, and many historians add a second, similar filter called compression on the archive side. Both are per-tag settings you choose, and both mean the stored series is not every value the controller produced.
Illustrative example — a 0–100 bar transmitter with an exception deadband of 0.5% of span. The deadband is 0.5% × 100 bar = 0.5 bar. Four values arrive in sequence: 42.0, 42.3, 42.6, 41.9.
- 42.0 is stored; it becomes the reference.
- 42.3: |42.3 − 42.0| = 0.3 bar, under the 0.5 bar deadband — not stored.
- 42.6: |42.6 − 42.0| = 0.6 bar, over the deadband — stored, and the reference becomes 42.6.
- 41.9: |41.9 − 42.6| = 0.7 bar, over the deadband — stored.
Two of the four values reach the archive. A monthly average from those points depends on whether your tool interpolates between stored samples or averages only raw samples — find out which before the first report goes out.
For monthly totals, do not sum instantaneous samples. If the PLC keeps a totalizer register, collect the totalizer and report month-end minus month-start; if the figure feeds billing or environmental reporting, reconcile it against the primary totalizer each month, because deadband filtering discards values by design.
How much storage does a small collection job need? #
About 3.8 GB a year for 300 tags stored once a minute at 24 bytes per stored sample — at these counts, disk size is never the reason to skip a historian.
Illustrative estimate, before any historian compression: 300 tags, one stored sample per tag per minute, 24 bytes per stored sample (an 8-byte timestamp, an 8-byte value and roughly 8 bytes of quality and row overhead).
- Samples per day: 300 tags × 1,440 minutes = 432,000 samples.
- Bytes per day: 432,000 × 24 bytes = 10,368,000 bytes, about 10.4 MB.
- Bytes over 365 days: 10,368,000 × 365 = 3,784,320,000 bytes, about 3.8 GB.
Deadband filtering keeps slow signals well below one stored sample a minute, and historians compress on top, so treat 3.8 GB a year as the ceiling for this tag count, not the floor. The constraint that actually bites is collector uptime: every hour the collector is down is an hour no reporting tool can recover later.
Can you keep a reporting client like ActiveFactory? #
Yes, but only if you also run the historian that client was built for: ActiveFactory reads Wonderware InSQL, so the real decision is add that historian or change the reporting client — not change the tag server.
A reporting client binds to a historian's query interface, never to a live DA server; telling it your visualization is Genesis 32 simply gives it nothing to connect to. To keep it, install InSQL, feed it from the OPC DA server your Genesis 32 project already uses through a compatible connector, and point ActiveFactory at InSQL. To move on, choose a historian that ships with its own trend and reporting tools, or one that writes to SQL Server you can query directly. Clients are cheap to swap; the storage layer is the part you live with, so weigh that choice across products before committing — plant historian selection by scale and support needs walks that comparison.
Is logging to SQL Server a valid historian substitute? #
Yes, for straightforward collection: an OPC client writing timestamped rows into SQL Server covers acquisition, storage and reporting, and you inherit the buffering, restart recovery, quality handling, retention and aggregation work yourself.
For a few hundred tags and monthly reports this is a legitimate architecture, not a compromise. The work you take over from the historian vendor is the unglamorous half: what happens when the database is down, how gaps are flagged, how millions of rows aggregate. If that is acceptable, the pattern is worked through in logging PLC values from an OPC server into SQL Server; if continuous collection through maintenance windows matters more than simplicity, buy the historian instead of writing one.
So which one do you need? #
A historian or a SQL logger, never the tag server alone: if any report looks backward in time, something must have been storing samples while time passed, and an OPC DA server never does.
| Your situation | What you need |
|---|---|
| Any report that looks backward in time | A historian or SQL logger collecting from your existing OPC DA server |
| You want to keep ActiveFactory | InSQL, fed from the OPC server Genesis 32 already uses |
| Free to choose every tool | A historian or SQL logger with a query interface your reporting tool speaks |
| Only live monitoring, no reports | The tag server you already have — nothing more |
| Values exist only inside the HMI | Collect from the HMI's OPC server, after checking its licensing |
For the case that prompted the question: a tag server behind Genesis 32 produces zero monthly figures on its own. Add a collection layer — a desktop-scale historian is enough at a few hundred tags — pointed at that same OPC DA server, connect the reporting client to it, and the monthly numbers exist.
Questions asked
Is MatrikonOPC a historian or a tag server?
The name covers both jobs: the family includes plain OPC servers that only move live values between programs, and a desktop historian that collects tags and serves history to clients. The brand on the box tells you nothing about storage; check which component is actually installed and whether anything is writing samples before you count on a report.
Do OPC HDA items use the same ItemID as the live OPC DA items?
Not reliably. HDA servers usually ship alongside a DA server covering the same tags, and in some products the identifiers match, but in others the historical ItemID differs from the live one — PI is one documented case. Browse the HDA namespace and map tags separately instead of assuming the DA ItemID will resolve.
What happens to a monthly total when the collector is offline for a day?
It depends on what the signal is. If the PLC keeps a totalizer, the collector misses samples but the counter keeps running, so reading the totalizer after the gap recovers the missed delta exactly. If the report averages instantaneous values, the missing day is gone: substitute an interpolated or manual estimate and label it as such in the report.
Do we need OPC HDA at all for monthly reports?
No. OPC HDA is one standard way for a client to read history from a historian, useful when the reporting tool and the historian come from different vendors. If your historian ships with its own trend and query tools, or you log to SQL Server and report from there, you never open an HDA connection.