Siemens FB41 CONT_C PID Block: GAIN, TI Reset, and TD Effects

1. Overview

The Siemens FB41 "CONT_C" (continuous controller) is the workhorse PID block shipped with the SIMATIC Standard PID Control library for STEP 7 / S7-300 / S7-400. It implements a parallel-form (non-interacting) PID algorithm with separate proportional, integral, and derivative paths, integral-action anti-windup, bumpless transfer between manual and automatic modes, configurable process-variable / setpoint / manipulated-variable scaling, a tracking input, and a disturbance feed-forward input. Understanding how the three primary tuning constants — GAIN, TI (the "reset" or integral time), and TD (the derivative time) — actually shape the controller output is essential for field commissioning, especially because Siemens documents TI in minutes per repeat rather than the repeats per minute used in many DCS systems.

2. FB41 in the Standard PID Control Library

FB41 is part of the "Standard PID Control" library that ships with STEP 7 V5.x and remains importable into TIA Portal for S7-300/400 targets. The canonical Siemens documentation is the manual Standard PID Control (function blocks FB41–FB43), entry ID 1137085 in the Siemens Industry Online Support:

• Siemens Standard PID Control manual (entry 1137085)
• PID Control with S7-300/400 application example (entry 60546408)
• S7-1200/1500 PID_Compact documentation (entry 109769928)

Each FB41 instance requires an instance DB (created automatically when the FB is called) plus a static data area. The instance DB is typically renamed to a project-specific name, e.g. DB201 "PID_FURNACE", and houses all the tuning parameters in a structure that can be monitored and forced online.

3. I/O Pinout of FB41 CONT_C

The block exposes a comprehensive set of inputs and outputs. The most important pins for tuning and observation are listed below; the full set is in the Siemens manual referenced above.

4. The Continuous PID Algorithm in FB41

FB41 evaluates a parallel-form PID structure. The ideal continuous-time equation is:

The output of FB41 is the manipulated variable LMN in percent of final control element range. The error e(t) is the difference between the scaled setpoint and the scaled process variable, after the dead-band is applied. The three terms are independent in the parallel structure, so the contribution of each term to the final output can be inspected individually via LMN_P, LMN_I, and LMN_D. This separation is one of the most useful commissioning features of the block: you can trend each term and see exactly which knob to turn.

For the discrete implementation, FB41 uses the backward-rectangle (backward-Euler) approximation for the integral action and a first-order-lag filtered derivative for the rate term. With sample time TS (set by the calling OB, typically OB35 at 1000 ms) and a fixed internal derivative gain of N = 10 used to bound the high-frequency derivative gain, the discrete update is equivalent to:

The integral action is held (frozen) when INT_HOLD is TRUE. The integral state is re-initialised to I_ITLVAL on a rising edge of I_ITL. The output is clamped to [LMN_LLM, LMN_HLM], and the integral action uses an anti-windup back-calculation so that the I-term cannot grow beyond what the output clamp allows. DISV is added directly to the raw I-term, making it a feed-forward action that influences the I-state without going through the P or D paths.

5. GAIN: The Proportional Action

GAIN sets the controller's proportional gain Kp. The relationship between the gain, the proportional band (PB), and the error in percent of the PV span is:

If the setpoint is in percent (0–100) of the process range, a gain of 1.0 produces an output change equal to the error in percent. A gain of 2.0 doubles the reaction, and a gain of 0.5 halves it. The P-only step response to a step error of magnitude E is a constant value in time:

This is the classic P-controller offset: the output never integrates away the steady-state error because the I-action is not present. Without I-action, a constant disturbance of magnitude d in PV units will always leave a residual error of:

The effective proportional band (in % of span) for a given gain is:

For example, a gain of 2.0 corresponds to a proportional band of 50 %, meaning a 50 % step in error drives the output from 0 to 100 %. A gain of 0.5 corresponds to a 200 % PB, which is a sluggish loop that never reaches 100 % output on a 100 % error.

6. TI: The Integral (Reset) Action — Siemens Convention

TI is the "reset time" or "integral action time" in MINUTES. This is the single most frequently misunderstood parameter on FB41. To be unambiguous about Siemens convention:

To see this, consider a constant step error of magnitude E applied at t=0 with no D-action and no disturbance. The controller output for t ≥ 0 is:

The proportional contribution is Kp*E. The integral contribution at time t is Kp*E*t/TI. They are equal when t = TI. So TI is the time, in minutes, for the integral to "repeat" the proportional correction. Equivalently:

The valid Siemens range for TI is typically 0.0 to 9999.0 minutes; the input is a REAL and accepts fractional values such as 0.25 min (= 15 s).

For comparison with a DCS that expresses the integral as "repeats per minute" (rpm):

For example, a DCS tuning of "0.5 repeats per minute" maps to FB41 TI = 2.0 minutes. A DCS tuning of "5 repeats per minute" maps to FB41 TI = 0.2 minutes. The "minutes per repeat" naming maps directly to TI on FB41.

For a step input of magnitude E applied at t=0, the integral contribution as a function of time is:

The slope of the integral ramp is therefore Kp * E / TI. Doubling Kp doubles the I-ramp slope. Doubling TI halves the I-ramp slope (slower integration).

7. TD: The Derivative Action

TD is the derivative time in MINUTES. It sets how much rate action contributes to the output per unit of error rate. The continuous-time contribution of the D term is:

In Siemens FB41, the D-action is filtered through a first-order lag to attenuate high-frequency measurement noise. The user-tunable time constant of this lag is TM_LAG (in seconds). The discrete filtered D-action, expressed with TI, TD in minutes and TS in minutes, is approximately:

Setting TD = 0.0 disables derivative action. The valid range is typically 0.0 to 9999.0 minutes. Practical TD values for industrial processes:

8. Step Response of a PI Controller in FB41

For a constant step error E applied at t=0, with the D-action disabled (TD = 0.0) and no disturbance, the FB41 output evolves as:

The P-action (blue) jumps instantly to Kp * E and stays there. The I-action (orange) starts at zero and ramps linearly with slope Kp * E / TI. The total PI output (green) is the sum. At t = TI, the I-action has caught up to the P-action; at t = 2*TI, the I-action is twice the P-action; and so on. This is the geometric meaning of "minutes per repeat" that Siemens uses for TI.

For a PI controller with GAIN = 1.5, TI = 2.0 min, and a step error of E = 10 (e.g., 10 % deviation of PV below SP), the output at t=0 is 15 %, and the I-ramp climbs at 7.5 % per minute, so it adds 15 % more after 2 minutes, 30 % more after 4 minutes, etc. In practice, the ramp stops when the process reaches setpoint and the error drops back to zero.

9. Anti-Windup, Output Limiting, and Bumpless Transfer

Integral windup is the phenomenon where the I-term keeps integrating while the manipulated variable is saturated at 0 or 100 %, so the controller cannot immediately back off the output when the PV returns to setpoint. FB41 implements anti-windup by tracking the I-term separately and clamping it so that LMN = LMN_P + LMN_I + LMN_D stays inside [LMN_LLM, LMN_HLM]. The integral action is held (frozen) when INT_HOLD = TRUE.

On switching from manual to automatic (MAN_ON goes from TRUE to FALSE), the I-term is loaded with the current MAN value minus the current P and D contributions. This guarantees a bumpless transition: the output does not jump at the moment of the switch. The reverse transition (automatic to manual) is the responsibility of the user — you must drive MAN with the current LMN before flipping MAN_ON to TRUE.

The blue trace is the MAN input; the green trace is the actual LMN output. In manual mode (MAN_ON = TRUE), LMN tracks MAN exactly. The instant MAN_ON drops to FALSE, the I-term is loaded with the current LMN minus the P and D contributions, so the output continues smoothly. After the switch, the controller reacts to the error and adjusts LMN.

10. Sample Time and Where to Call FB41

FB41 must be called in a cyclic OB. The call period TS is the OB period. Common practice:

• OB35 (cyclic interrupt, default 1000 ms): typical choice for process control loops at 1-second sample time. The OB35 period can be reconfigured in HW Config from 1 ms up to 60 s.
• OB1 (main cyclic): the call period is the OB1 cycle time, which depends on the program. Often 10–100 ms. Faster than OB35, but OB1 cycle time can vary with program load and may cause derivative noise issues.
• OB32, OB33, OB34: additional cyclic interrupt OBs at default 500 ms / 200 ms / 100 ms. Useful for medium-speed loops.

Rule of thumb: choose TS at most one tenth of the dominant process time constant. For a temperature loop with time constant of 5 min, a TS of 5–10 s is appropriate. For a flow loop with time constant of 0.5 s, a TS of 50–100 ms is appropriate.

11. Setpoint and Process-Variable Scaling

FB41 supports independent scaling on the setpoint and PV paths:

Typical use cases:

• PV_FAC / PV_OFF: linearise a 4-20 mA input that has already been converted to engineering units. For example, a 0–100 °C RTD scaled to 0.0–100.0 °C: PV_FAC = 1.0 and PV_OFF = 0.0. If the input is in 0.1 °C units, use PV_FAC = 0.1.
• SP_FAC / SP_OFF: scale the setpoint to match the PV range; useful when SP comes from another controller in a cascade. Often left at 1.0 / 0.0.
• DEADB_W: suppress small noise-induced controller action. A DEADB_W of 0.5 means errors below 0.5 % of span are treated as zero. Useful on level loops with a noisy dP cell.

12. Output Scaling

The final manipulated variable that leaves FB41 is:

Defaults are LMN_FAC = 1.0 and LMN_OFF = 0.0, giving LMN directly in percent. To convert LMN (%) to a 4-20 mA output for a valve, the scaling is normally done outside FB41 in the analog-output driver block (e.g., FB79 or a CP/IM analog output with scale block). When the controller drives a motor-driven valve through FB42 (CONT_S) or FB43 (PULSEGEN), LMN in % is interpreted as the desired valve position.

If the valve has a split-range (e.g., 4-12 mA heats, 12-20 mA cools), use LMN_FAC = 2.0 and LMN_OFF = -100.0 to map 0–100 % controller output to −100 % to +100 % valve signal, then write the scaled value to two analog output channels (or use FB43 PULSEGEN with a single bipolar output).

13. Cascade, Tracking, and Feed-Forward

FB41 is a single-input controller but can be cascaded to build multi-loop control. Cascade is typically implemented by wiring the primary controller's LMN into the secondary controller's SP_INT:

The primary controller's LMN becomes the secondary controller's SP_INT. The secondary controller must be tuned 3–5× faster than the primary for cascade stability. In a temperature/flow cascade, the primary is slow (TI in minutes) and the secondary is fast (TI in seconds).

For feed-forward, connect a measured disturbance to the DISV input of the primary. The block adds DISV directly to the integral contribution, allowing the controller to compensate for known disturbances without waiting for the PV to deviate. A typical example is a heat-exchanger outlet temperature controller receiving a feed-forward of the inlet flow rate; the larger the flow, the more steam is needed at the same outlet temperature.

14. Common Commissioning Errors and How to Avoid Them

15. Verification Procedure After Configuration

1. Open the FB41 instance DB online and confirm all input values match the intended tuning constants. Force the controller to manual (MAN_ON = TRUE), set MAN to 50 %, and verify the controlled variable (e.g., valve position) follows.
2. Apply a small setpoint step (5–10 % of span) in automatic mode, with the loop in a stable operating point. Trend LMN_P, LMN_I, LMN_D, PV, and ER. Confirm the P-term jumps, the I-term ramps, and the D-term spikes on the rate of change.
3. Verify the time scale. The I-term should ramp at the predicted slope Kp*E/TI (in LMN units per minute). For example, with GAIN = 2.0, TI = 1.0 min, and ER = 5 %, the I-ramp slope should be 10 % per minute.
4. Check the steady-state offset. With a PI controller, the offset should decay to zero within 4–5 * TI minutes. If a permanent offset remains, the I-term is being held (INT_HOLD) or windup is still active.
5. Perform a load step (disturbance) and confirm the controller recovers. The closed-loop time constant should be roughly Tcl ≈ TI * (Kp_loop / Kp_total); this is a coarse check, not a substitute for proper tuning.
6. Switch MAN↔AUTO several times and confirm there is no step change in LMN. The transition is bumpless if and only if MAN = LMN at the moment of the switch.
7. For D-action loops, double the TD and confirm the output reacts to PV ramps more strongly; reset to original value when finished.
8. For feed-forward loops, drive the disturbance input and confirm LMN shifts immediately, before the PV has had time to deviate.

16. Open-Loop Step Test Procedure for Tuning

To tune a loop from scratch with model-based methods, perform an open-loop step test:

1. Place the controller in manual (MAN_ON = TRUE) and bring the process to a stable operating point. Record the steady-state PV and LMN.
2. Apply a step change in MAN of about 5–10 % (small enough to stay linear, large enough to be measurable above noise).
3. Trend the PV. Identify the dead time θ (time from the step to the first detectable PV change) and the time constant τ (time from the start of the PV change to 63 % of the total PV change). Compute the process gain Kp = ΔPV / ΔMAN in PV units per % LMN.
4. Use Cohen-Coon (good for processes with significant dead time) or Lambda tuning (good for setpoint tracking) to compute GAIN, TI, TD.
5. Apply the new tuning in manual first (verify SP matches PV and the controller is ready), then switch to automatic and perform a small setpoint step to verify behaviour.

Lambda parameter λ is the desired closed-loop time constant; a typical choice is λ = 3*θ for aggressive setpoint tracking or λ = 5*θ for smoother response.

17. Worked Tuning Example: Steam-Heated Heat Exchanger

A heat exchanger heats water from 30 °C to 70 °C. The PV is the outlet water temperature in °C (0–100 °C range). The control valve is steam. An open-loop step test with a 5 % step in MAN from 40 % to 45 % produced:

• Process gain Kp = 0.8 °C per % LMN (i.e., 1 % more steam gives 0.8 °C higher outlet)
• Time constant τ = 90 s = 1.5 min
• Dead time θ = 15 s = 0.25 min

Cohen-Coon tuning:

FB41 settings to apply:

Call FB41 in OB35 with TS = 1000 ms. For a 5 °C setpoint step (e.g., SP from 65 °C to 70 °C), the predicted response is roughly 5% / 5 = first-order with a closed-loop time constant of about 0.4 min, reaching 63 % of the step in 24 s. Verify by trending the actual response and fine-tuning by inspection.

18. FB41 vs. CONT_S (FB42) vs. PID_Compact

For S7-300/400 systems that already use FB41, there is no functional reason to migrate. For S7-1200/1500 systems, use PID_Compact (FB 1130) instead; FB41 can be imported into TIA Portal for S7-1200/1500 targets but its derivative gain filter constant is fixed and the sample-time model is calibrated for S7-300/400 cyclic OBs.

19. FB41 and TIA Portal Migration Notes

When migrating an S7-300/400 project containing FB41 to TIA Portal on the same S7-300/400 hardware, the FB41 instance is preserved with its tuning values intact. When migrating to S7-1500, Siemens recommends replacing FB41 with PID_Compact (FB 1130), which uses different pin names and a different default cycle (typically the PID cycle OB used by the new PID technology object). Key translation points:

For new S7-1500 designs, prefer PID_Compact for its built-in pre-tuning and fine-tuning routines; the tuning constants from a working FB41 loop can be transferred as starting values.

20. FAQ