Resolving S7-1200 1215C Analog Input Instability at ADC Code Edges

This technical reference addresses a recurring field complaint on the SIMATIC S7-1215C DC/DC/DC integrated analog input where the scaled process value oscillates at specific code-transition voltages (most commonly 1.00 V, 3.00 V, 5.00 V, 7.00 V, 9.00 V) while remaining stable immediately above and below. The instability is fundamental to the integrated 10-bit successive-approximation ADC and cannot be resolved with conventional software averaging, S7-1200 channel smoothing, or shunt capacitors. The article quantifies the problem, documents the diagnostic procedure, and provides a tested migration path to a SIWAREX weighing module appropriate for load-cell and force-measurement applications.

1. S7-1200 1215C DC/DC/DC Integrated Analog Input Specifications

The CPU 1215C DC/DC/DC ships with two integrated analog inputs (AI0, AI1) and two integrated analog outputs (AQ0, AQ1). The relevant published specifications, per the SIMATIC S7-1200 Programmable Controller System Manual, are summarised below.

The published resolution confirms the field observation: 10 bits yields 1024 codes across a 10 V span, which is insufficient for load-cell / mass-weighing applications that typically require 16 to 24 bits of useful resolution and accuracies better than 0.05 %.

Reference: SIMATIC S7-1200 Programmable Controller System Manual (Siemens Online Support, entry ID 109751826).

2. Root Cause: ADC Code-Transition Quantization

The 1215C integrated analog front-end uses a successive-approximation register (SAR) ADC. The input voltage is compared against an internal reference ladder and the closest of 2^N discrete output codes is returned. Each code is exactly FSR / 2^N wide. For a 10 V full-scale range:

LSB_voltage = 10 V / 1024 = 9.7656 mV
LSB_counts  = 27648 / 1024 = 27.0 counts

When the applied voltage sits inside the ±0.5 LSB window of a code boundary, the SAR comparator flips between two adjacent output codes at every conversion. This is not a malfunction, calibration drift, or wiring problem - it is the deterministic behaviour of every N-bit quantiser. The S7-1200 1215C simply exposes 27-count jumps (≈9.8 mV) at every code boundary, and the user program is fast enough to capture the alternation.

2.1 Why 1.00 V, 3.00 V, 5.00 V, 7.00 V, 9.00 V Are the Hot Spots

Code boundaries occur at multiples of one LSB:

V_boundary(k) = (k + 0.5) * 9.7656 mV   for k = 0, 1, 2, ... 1023

Nearest "round" values to a code boundary in the 0–10 V span:
  0.99 V ≈ code 101    | 1.00 V ≈ code 102 boundary
  2.99 V ≈ code 306    | 3.00 V ≈ code 307 boundary
  4.99 V ≈ code 511    | 5.00 V ≈ code 512 boundary
  6.99 V ≈ code 716    | 7.00 V ≈ code 717 boundary
  8.99 V ≈ code 920    | 9.00 V ≈ code 921 boundary

A stabilised power supply set to exactly 1.000 V has a real output of (for example) 1.0001 V to 0.9999 V with sub-millivolt ripple. This voltage traverses code 101/102, 306/307, 511/512, 716/717, and 920/921 boundaries - which is precisely the unstable trace observed in the original report. Setting the supply to 0.99 V or 1.01 V places the input well inside a single code, so the digital value is rock-solid.

2.2 Why Software Filters Cannot Fix It

A moving average, exponential filter, or median filter operates on the post-quantisation digital value. If the underlying ADC alternates between 102 and 103 counts at the sample rate, an exponential filter with α = 0.1 reaches a steady state of approximately 50 % duty cycle (≈ 102.5), which the S7-1200 truncates to 102. The filter does not remove the LSB flicker; it merely attenuates the rate at which the digital value moves. The LSB still toggles ±27 counts on a 0–27648 scale, which is ±0.098 % of span - a value 10 to 100 times larger than a typical load-cell application's noise budget.

2.3 Why Shunt Capacitors Cannot Fix It

A capacitor across the input (typically 100 nF to 10 µF with a 100 kΩ source) forms a single-pole low-pass filter with a time constant of τ = R × C. With a 100 kΩ source and 100 nF, τ = 10 ms, which is slow enough to attenuate the flicker slightly but only because the average voltage moves toward a fractional value - the ADC still quantises to the nearest code, and the S7-1200 still exposes the LSB. To meaningfully integrate out the LSB, τ must be much larger than the 0.625 ms conversion time, which makes the channel response time unsuitable for dynamic mass measurement.

3. Diagnostic Procedure to Confirm Code-Edge Behaviour

Use this procedure to determine whether the observed oscillation is a code-edge quantisation artefact (correct diagnosis) versus a real source-side problem such as ground loops, thermocouples, or floating references (wrong diagnosis, see NI Knowledge Base: Troubleshooting Unexpected Analog Measurements).

1. Disconnect the field wiring from the integrated AI0 and AI1 terminals of the 1215C.
2. Connect a calibrated, low-noise DC reference (Fluke 5520A, Burster 1421, or equivalent) to the channel. Verify the reference output with a 6.5-digit DMM.
3. Disable all software filters in the user program. In TIA Portal, navigate to Project tree → Devices → [CPU 1215C] → Properties → Analog inputs and set Smoothing to None.
4. Sweep the input from 0.05 V to 9.95 V in 10 mV steps and record the raw AIW0 value (or NORM_X / SCALE_X result).
5. Plot the data. Stable plateaus (27 counts wide on the 0–27648 scale) with sharp 27-count transitions are the fingerprint of ADC code-edge noise. Any single-LSB oscillation inside a plateau indicates either source noise or ground loop, not the integrated AI's quantisation.
6. Repeat the sweep with the field source (load-cell amplifier). If the field source shows the same plateaus/edges, the integrated AI is the limiter. If the field source shows a different noise signature, the source is the limiter.

4. TIA Portal Software Filter Options (Interim Mitigation Only)

For non-critical applications, TIA Portal provides two filter mechanisms on the integrated AI. Neither eliminates the LSB flicker; they only damp the rate of change.

4.1 Channel-Level Smoothing

Open the CPU's device properties, expand Analog inputs, select the input channel, and set Smoothing to one of: None, Weak, Medium, Strong. Smoothing is implemented as an exponential filter in the firmware:

y[n] = α · x[n] + (1 − α) · y[n−1]

α values (TIA Portal V18 S7-1200):
  None   : α = 1.0   (no filter)
  Weak   : α ≈ 0.5
  Medium : α ≈ 0.1
  Strong : α ≈ 0.02

4.2 Application-Level Moving Average (Ladder Logic Example)

// 16-sample moving average, integer math, TIA Portal V18
// AIW0 = raw 0–27648 scaled value from CPU 1215C integrated AI

VAR
    RingBuf  : ARRAY[0..15] OF INT;
    Idx      : INT;
    Sum      : DINT;
    AvgVal   : INT;
END_VAR

BEGIN
    // Subtract oldest sample from running sum
    Sum := Sum - RingBuf[Idx];
    // Insert new sample
    RingBuf[Idx] := "AIW0"_I;       // 0–27648
    Sum := Sum + RingBuf[Idx];
    // Compute average
    AvgVal := DINT_TO_INT(Sum / 16);
    // Advance ring index
    Idx := Idx + 1;
    IF Idx > 15 THEN Idx := 0; END_IF;
END_FUNCTION_BLOCK

4.3 Oversampling (Effective Resolution Enhancement)

For a stationary signal, oversampling at a factor N provides an effective resolution gain of log2(√N) ≈ 0.5 · log2(N) bits. To gain 4 extra bits (10 → 14 bits) requires 256× oversampling, i.e. 256 × 0.625 ms = 160 ms per averaged output. This technique is valid only if the input is stationary; for dynamic mass measurement it is too slow.

None of these filter paths are suitable for laboratory-grade mass measurement. The only viable fix is a higher-resolution acquisition module.

5. Hardware Filtering and Wiring Best Practices (Required but Not Sufficient)

Even after migration to a high-resolution module, the source-side signal path must be clean. Reference: NI - Troubleshooting Unexpected Analog Measurements for the general taxonomy of analog-input faults.

6. Migration Path: SIWAREX Weighing Modules

For load-cell, force, and mass applications Siemens provides the SIWAREX product line. The integrated AI on the 1215C is not designed for this use case; the correct replacement is a dedicated weighing module.

The SIWAREX WP231 is the most common direct replacement for load-cell applications requiring both high accuracy and high speed. Its 24-bit ADC corresponds to:

LSB_voltage_WP231 = 10 V / 16 777 216 = 596 nV
LSB_count_WP231  = 8 388 608 / 16 777 216 = 0.5 count (effectively 0.5 LSB)

This is a 16 384× improvement over the integrated AI in voltage resolution, which is why code-edge noise is no longer observable.

6.1 Recommended Selection Logic

• Static mass / tank weighing, simple SCADA integration: SIWAREX CS (lower cost, 50 Hz, RS-232).
• Dynamic mass flow / belt scale / checkweigher: SIWAREX WP231 (24-bit, 1 ms).
• Force-displacement / tensile testing: SIWAREX WP241 (PROFINET for tight HMI integration).
• Two independent load cells in one platform: SIWAREX WP251.

7. Step-by-Step Migration: SIWAREX CS to S7-1215C

The CS module is the lowest-cost entry point and the simplest migration from the integrated AI. The wiring and configuration steps are documented in the SIWAREX CS manual (entry ID 109751826 family).

7.1 Prerequisites

• CPU 1215C DC/DC/DC, firmware V4.2 or later (TIA Portal V16+ for current GSD).
• CM 1241 RS-232 communication module (6ES7241-1AH30-0XB0) installed in the left-of-CPU signal stack.
• SIWAREX CS (7MH4910-0AA01) with the dedicated 7MH4900-0AKxx cable (length 1.5 m, 3 m, 5 m, or 10 m).
• Load cell with 4-wire or 6-wire connection (recommend 6-wire for long cable runs).

7.2 Wiring

CM 1241 (RS-232, X11)
  Pin 1 (TD)  →  SIWAREX CS X2.3 (RxD)
  Pin 2 (RD)  →  SIWAREX CS X2.2 (TxD)
  Pin 5 (GND) →  SIWAREX CS X2.5 (GND)
  Shield      →  CS X2.4 / earth (one side only)

SIWAREX CS X1 (load cell, 6-wire)
  X1.1  EXC+   X1.2  SENSE+   X1.3  SIG+
  X1.4  EXC−   X1.5  SENSE−   X1.6  SIG−

7.3 TIA Portal Configuration

1. Open the TIA Portal project, navigate to Devices > [CPU 1215C], and add the CM 1241 to the configuration. Set the RS-232 protocol to Modbus Master (RTU) at 115 200 Bd, 8E1.
2. Install the SIWAREX CS HSP or GSD and add the CS as a Modbus device. The Modbus register map is documented in the SIWAREX CS manual, key registers:

3. Use the MB_MASTER instruction (Siemens S7-1200 Modbus library) in the user program to poll registers 0x0001 / 0x0003 / 0x0007 at 10 Hz.
4. After zero / span calibration via the SIWATOOL V7 configuration software, the integer gross weight reading is scaled to engineering units (e.g. kilograms, grams, newtons) and made available to the HMI or control logic.

8. Verification Procedure

After commissioning the SIWAREX module, run the following verification to confirm the instability has been eliminated.

1. Apply a known calibration weight (typically 50 % of full scale and 100 % of full scale) and confirm the reading is within the SIWAREX accuracy class specification (SIWAREX CS Class III OIML R76: 0.01 % FS; WP231: 0.005 % FS).
2. Log the gross weight at 1 Hz for 60 minutes with a stable mass on the platform. Calculate the standard deviation σ.
3. Expected σ for a 50 kg / 5 000 d platform with SIWAREX CS: ≈ 0.5 d (0.5 count of 5 000). With the integrated AI in the same mechanical system: ≈ 27 counts of 27 648 = 0.098 % of span - approximately 50× worse.
4. Sweep the calibration voltage from 0 % to 100 % of full scale in 5 % increments. Confirm a monotonic, linear response with no plateaus, no missing codes, and no 27-count transitions.

9. When the Integrated AI Is the Right Tool

The integrated 1215C analog input is appropriate for non-precision applications:

• Slow-changing signals (process times > 1 s).
• Tolerances of ≥ 1 % of full scale.
• Threshold / setpoint comparisons where absolute accuracy is irrelevant (e.g. "tank empty" at 0–10 V from a capacitive probe).
• Cost-sensitive applications where a SIWAREX module (≈ €350 for the CS, ≈ €600 for the WP231) cannot be justified.

For these cases, a combination of TIA Portal channel smoothing (Medium) and an application-level moving average (16 samples) provides adequate performance and avoids the migration cost.

10. Summary of Key Constraints

11. Frequently Asked Questions