Comparison
Replacing DCS SCADA with a third-party SCADA system
Short answer
Yes on data, no on engineering. A third-party SCADA reads and writes a PCS 7 controller through an S7 driver or an OPC server, but you rebuild faceplates, symbol links, navigation, alarms and diagnostics by hand - about 1,200 hours for a 500-object unit, which outweighs the license savings. Do it for monitoring layers and small stations; keep the integrated operator station for full process units.
On this page
- Can you replace a DCS operator station with a third-party SCADA?
- What does the integrated operator station give you for free?
- Should the link be an S7 driver or an OPC server?
- How do you design the gateway between the DCS and the SCADA?
- How big is the rebuild on a full process unit?
- How do you prove the link before building screens?
- When is the third-party SCADA the right decision?
- Questions asked
Yes, and plants do it - but only for the right scope. Any SCADA package with an S7-capable driver or an OPC client can read and write a PCS 7 controller, so the connection is never the blocker. What does not transfer is everything the integrated operator station generated from the engineering system: faceplates, symbol linking, navigation, alarms, hardware diagnostics and sequence displays. This comparison sizes that rebuild in tags and hours, then gives the decision for the three cases that come up in practice.
Can you replace a DCS operator station with a third-party SCADA? #
Yes for data, no for the engineering: any SCADA package with an S7-capable driver or an OPC client can read and write a PCS 7 controller, but nothing transfers the operator functions the integrated station generated automatically. Keep two acceptance criteria apart. A live process value on the new screen proves data access. It proves nothing about commands, alarms, faceplates, diagnostics, sequence displays, or how engineering changes are coordinated afterwards.
The controller keeps executing its control algorithms either way; what you are replacing is the operator layer and the workflow that maintained it. The same trade holds on every integrated DCS platform that exposes an OPC server: the connector is standardized, the operator station is not.
What does the integrated operator station give you for free? #
The integrated station generates the tag database, faceplates, symbol placement, navigation, alarms and diagnostics out of the engineering compile; a third-party SCADA rebuilds every one of those by hand and re-coordinates them after every change. PCS 7 ships faceplates with its technological block types, places and links symbols from the plant hierarchy, derives the screen hierarchy the same way, and turns block diagnostics into alarms without separate alarm engineering. Row by row, this is the bill the replacement inherits:
| Operator function | Integrated PCS 7 OS | Third-party SCADA rebuild |
|---|---|---|
| OS tag database | Generated from block and chart configuration at compile | Defined and maintained by hand, item by item |
| Faceplates | Supplied with the block types, editable in a faceplate tool | Built as templates with command and status handling |
| Symbols | Placed and linked from the plant hierarchy | Placed and bound per instance |
| Screen navigation | Hierarchy derived from the plant hierarchy | Built and maintained by hand |
| Alarms | Generated from technological block diagnostics | Each limit, acquisition and presentation engineered |
| Hardware diagnostics | Presented integrated with the process views | Exposed separately and displayed, if budgeted |
| Sequence displays | Integrated SFC visualization | A separate solution, if sequences still matter |
| Engineering changes | Controller and OS in one project, one compile | Two projects, every change issued and tested twice |
The last row is the one that keeps costing. In the integrated world, one project holds the controller and its screens, so a loop change compiles through to the faceplate. After the swap, each change is made twice. When you rebuild the alarm layer, include the link itself on day one - creating a SCADA-PLC communication alarm for link loss is a small, well-defined pattern worth copying.
Should the link be an S7 driver or an OPC server? #
Use the S7 driver with a gateway data block when the external system needs defined values and a few commands; use OPC when the plant already runs a suitable server, because the documented PCS 7/Open OS route for third-party connections requires at least one OPC server. The PCS 7/Open OS application guide states that requirement directly: a third-party system connects through at least one OPC server on the OS side. That makes OPC the vendor-documented path for anything beyond a driver read, and the server side is where the cost hides, because some DCS vendors license that access per tag. For moving OS data into databases or other applications, WinCC/IndustrialDataBridge is the documented transfer option.
| Decision input | S7 driver plus gateway DB | OPC server |
|---|---|---|
| What it requires | A driver that supports the AS controller, plus one interface DB | At least one OPC server on the OS side and a client connection in the SCADA |
| What it carries | The values and commands you defined in the DB | Whatever the server exposes: values and quality, alarms only if alarm services are provided |
| Licensing basis | SCADA driver and tag licensing | Possibly per tag on the DCS side, so count interface tags first |
| Best fit | Monitoring with a few defined commands | Plant-standard clients, an existing server, data collection |
One label is not a full interface. Classic OPC Data Access carries values and quality codes; alarms and events arrive only if the server provides those services, and no OPC profile delivers faceplates or sequence displays. Where the server offers OPC UA, use it: same data, none of the DCOM configuration that classic OPC carries.
How do you design the gateway between the DCS and the SCADA? #
Put every crossing value and command through one global data block with documented direction, owner, data type, range and failure behavior per field, and keep permissives and interlocks in the controller where they already run. A single gateway DB is what keeps the control charts readable; bindings scattered through the continuous function charts blur the logic and make every later change harder. Build it like this:
- List what each equipment type exchanges: process values, units, operating states, modes, setpoints, alarm bits, command requests and command results.
- Give every field one owner and one direction; controller-produced status never shares a word with SCADA-produced requests.
- Document data type, valid range, access rights and behavior on communication failure for each field.
- Handshake every command through request, accept or reject, complete, and clear, and drive the display from the returned controller state.
- Map the gateway fields to their control functions inside the engineering project, so the interface stays traceable.
- Put the interface definition under change control, and issue controller and SCADA changes against the same revision.
Permissives and protective interlocks stay in the control application - a screen button is not an interlock - and during commissioning, writes go only to an approved test point.
How big is the rebuild on a full process unit? #
About 1,200 engineering hours for a 500-object unit, against roughly 100 hours on the integrated route: the swap costs about 1,100 extra hours before it writes a single productive screen, and every later change touches two projects instead of one. Work the numbers on a concrete case - a mid-size unit with 200 analog control loops and 300 discrete equipment blocks (motors, on/off valves), 500 objects total, at illustrative mid-range planning rates.
Size the interface first. Assume 15 tags per loop, covering process value, setpoint, output, mode, alarm limits, tuning and scaling, and 6 per block, covering command, feedback, permissive, two status words and fault: 200 loops x 15 = 3,000 loop tags, 300 blocks x 6 = 1,800 block tags, 4,800 tags to license where the DCS side charges per OPC tag, and to configure and test regardless.
| Rebuild item | Basis | Hours |
|---|---|---|
| Faceplate templates | 15 templates x 12 h | 180 |
| Symbol placement and binding | 500 objects x 0.5 h | 250 |
| Alarm engineering | 500 objects x 0.75 h | 375 |
| Hierarchy and navigation | fixed scope | 40 |
| Hardware diagnostics displays | fixed scope | 60 |
| Gateway mapping in the controller | 500 objects x 0.4 h | 200 |
| Command handshake logic and test | 300 blocks x 0.4 h | 120 |
| Total | 1,225 |
That is 1,225 hours, about 31 working weeks at 40 hours a week. On the integrated route the same 500 objects cost roughly 500 x 0.2 = 100 hours of block and OS configuration, because the compile produces the tags, faceplate instances, symbols, hierarchy and alarms. The swap therefore buys about 1,125 extra hours. Planning rates move by half either way with object mix; the ratio does not flip.
How do you prove the link before building screens? #
Commission the tag database first: read a known changing value from a test client outside the graphics, check its quality indication, then trace one command back and confirm the returned state - never the echoed request. Commission the interface in layers, because a blank or frozen indication can come from the controller mapping, the server, the tag definition or the graphic binding:
- Read one known changing value, a counter or a ramping analog, from a test client with no graphic in the path, and confirm the update rate and quality code.
- Write one command to the approved commissioning point and watch accept, complete and clear arrive in the gateway.
- Trace one measured value from its control function through the gateway to the client, then one command in the reverse direction.
- If the client reaches exactly one controller and none of the others, that is routing, not licensing - troubleshooting a SCADA station that reaches only one S7-1500 covers the subnet and routing checks.
When is the third-party SCADA the right decision? #
For monitoring, trends and reporting over a healthy operator station, and for small stations under about 100 objects; for a full process unit with sequences and interlocks, keep the integrated station - the rebuild never pays back the license savings. If this swap sits inside a wider modernization, settle the platform question first: SCADA and PLC platform selection for a plant revamp walks that decision before this one. Then apply the three cases:
- Full process unit with sequences, interlocks and ongoing changes: keep the integrated station. About 1,100 extra hours plus permanent two-project change control is not recovered by licensing.
- Monitoring, trends and reports over a live integrated station: connect the third-party SCADA to the OPC server read-only. Faceplates and commands drop out of scope, and the rebuild shrinks to gateway mapping, about 200 hours on the example unit.
- Small standalone station of about 100 objects where the plant standard demands one package: the example rates give 40 loops x 15 + 60 blocks x 6 = 960 tags and roughly 290 hours - 60 for five templates, 50 for symbols, 75 for alarms, 20 for hierarchy, 20 for diagnostics, 40 for gateway mapping and 24 for command testing. A defensible one-off; cap commands to a handshake-tested list.
In every winning case the third-party station still needs its machine at the operator position, and an industrial panel PC in the cabinet door or console is the usual host for the client. Below about 100 objects, take the plant standard; above it, keep the integrated station.
Questions asked
Can the integrated operator station and the third-party SCADA run in parallel?
Yes, and it is the cheapest first step. Leave the integrated station in charge of commands, faceplates and diagnostics, and connect the third-party SCADA to the OPC server read-only for trends, reports and plant-wide screens. You get the plant-standard visualization without the thousand-hour rebuild, and you widen its scope later only where the acceptance tests pass.
Where do controller diagnostics end up if the SCADA does not show them?
They still exist: the controller keeps generating them and the engineering system still presents them. What you lose is the operator's instant view, because the integrated station turns technological block diagnostics into alarms without extra engineering. Decide explicitly whether maintenance accepts going to the engineering system for faults, or budget the diagnostics displays, about 60 hours on the example unit, as part of the rebuild.
How many tags does one control loop need on the link?
Plan 10 to 20 per analog loop, covering process value, setpoint, output, mode, alarm limits, tuning and scaling, and about 5 to 8 per motor or valve block. Multiply by the object count before licensing anything: where the DCS vendor charges per OPC tag, that fee scales with the total, 4,800 tags on the example unit.
What proves that a command from the SCADA actually moved the equipment?
Only the controller's returned state, never the echoed request. Design the handshake so the request clears only when the feedback or status word from the control function confirms the result, and display that state on the screen. If the gateway value goes stale on a link failure, the display must show invalid, not a silent last-known-good value.