Fixing Negative Angle DMS Conversion Bug in Siemens SCL

Problem Overview

Decimal-to-DMS (Degrees, Minutes, Seconds) conversion is a recurring requirement in motion control, telescope positioning, surveying, aviation, and observatory PLC applications. A common implementation in Siemens SCL on S7-1200/S7-1500 controllers contains a subtle sign-complement defect that surfaces only when the fractional part of the input value is exactly zero. The defect produces raw outputs such as degOut = -1, minOut = 60, secOut = 60.0 for an input of -1.0000°. After downstream aggregation the result collapses to -1° 0' 0", but during intermediate observation, in a structured tag database, or when the raw fields are displayed independently, the inconsistency surfaces as a quality defect and trips range alarms on WinCC IO fields configured for the conventional 0-59 minute/second range. This reference documents the failure mode, the two mathematically equivalent negative-DMS encodings, and two corrected SCL implementations that compile cleanly in TIA Portal V17, V18, and V19.

The Buggy SCL Function

The original function block shared in field deployments looks similar to the snippet below. It assumes every negative input requires a borrow from the integer-degree part, regardless of whether the decimal fraction is zero. The SCL block interface follows the conventions described in the official SCL Programming and Operating Manual for S7-300/S7-400 and the TIA Portal SCL expressions and operations reference.

FUNCTION "fb_DMS_Convert_Buggy" : Void
{ S7_Optimized_Access := 'TRUE' }
VERSION : 0.1
   VAR_INPUT 
      decAngle  : LReal;   // decimal degrees input
   END_VAR

   VAR_OUTPUT 
      degOut    : DInt;
      minOut    : DInt;
      secOut    : LReal;
   END_VAR

   VAR_TEMP 
      absAngle  : LReal;
      intDeg    : DInt;
      decPart   : LReal;   // fractional portion of |angle|
   END_VAR
BEGIN
    absAngle := ABS(decAngle);
    intDeg   := LREAL_TO_DINT(TRUNC(absAngle));
    decPart  := absAngle - DINT_TO_LREAL(intDeg);

    IF decAngle < 0.0 THEN
        // Naive complement: ALWAYS subtract one degree, ALWAYS take complement
        degOut := -(intDeg + 1);
        minOut := LREAL_TO_DINT(60.0 - decPart * 60.0);
        secOut := 60.0 - (decPart * 60.0 - LREAL_TO_DINT(decPart * 60.0)) * 60.0;
    ELSE
        degOut := intDeg;
        minOut := LREAL_TO_DINT(decPart * 60.0);
        secOut := (decPart * 60.0 - DINT_TO_LREAL(minOut)) * 60.0;
    END_IF;
END_FUNCTION

For decAngle = -1.0000 the function returns degOut = -2, minOut = 60, secOut = 60.0, which is internally inconsistent because minOut = 60 and secOut = 60 exceed the valid ranges of 0..59. A normalised downstream stage may pass the value through silently; an enforcing stage (e.g., HMI IO field with input range check, SCADA tag with alarm limits) raises a diagnostics event. When the fractional part is non-zero the function is mathematically consistent but semantically wrong if the calling convention expects only the degree to carry the sign.

Root Cause Analysis

Two encodings of a negative angle in DMS exist, and they are algebraically equivalent:

The faulty code unconditionally applies a borrowed-complement encoding (Methods C and D) to negative inputs regardless of the fractional part. When the fractional part is 0, Method C yields (-n-1) 60 60, internally inconsistent because minOut = 60 and secOut = 60 exceed the canonical ranges of [0,59] and [0,60). The defect has two contributing decisions:

1. The branch logic does not test for decPart = 0.0 before applying the borrow; the result is an invalid intermediate minute/second value.
2. The always-borrow encoding produces valid values after downstream carry propagation but obscures the bug from a reviewer reading the raw outputs. The bug is therefore invisible in straight-to-DB logging and only shows up on HMI screens that display each field separately.

Numeric Reproduction Table

The first, fifth, and combined rows isolate the defect. For audit-grade displays (aviation HMI, telescope position loggers, sub-arcsecond surveying controllers) the raw values are visible and the inconsistency surfaces as a quality defect and an out-of-range alarm.

Conversion Logic Flowchart

Corrected Implementation - Method A (Single-Sign Convention)

The cleanest fix selects Method A. The degree field carries the sign; minutes and seconds remain in 0..59 and 0..<60 respectively. The function validates each output range and signals range faults through a dedicated BOOL tag. The implementation is fully compatible with TIA Portal V17 and later on S7-1200 and S7-1500.

FUNCTION "fb_DMS_Convert" : Void
{ S7_Optimized_Access := 'TRUE' }
VERSION : 1.1
   VAR_INPUT 
      decAngle   : LReal;   // decimal degrees input
   END_VAR

   VAR_OUTPUT 
      degOut     : DInt;
      minOut     : DInt;
      secOut     : LReal;
      valid      : Bool;    // TRUE when output is normalised
      edgeHit    : Bool;    // TRUE for zero-fraction negative inputs
   END_VAR

   VAR_TEMP 
      absAngle   : LReal;
      intDeg     : DInt;
      fracPart   : LReal;
      rawMin     : LReal;
      intMin     : DInt;
   END_VAR
BEGIN
    valid    := FALSE;
    edgeHit  := FALSE;
    absAngle := ABS(decAngle);
    intDeg   := LREAL_TO_DINT(TRUNC(absAngle));
    fracPart := absAngle - DINT_TO_LREAL(intDeg);

    // ---- fractional minutes/seconds (always positive) ----
    rawMin   := fracPart * 60.0;
    intMin   := LREAL_TO_DINT(TRUNC(rawMin));
    secOut   := (rawMin - DINT_TO_LREAL(intMin)) * 60.0;

    // ---- seconds carry-in handling ----
    IF secOut >= 60.0 THEN
        secOut := secOut - 60.0;
        intMin := intMin + 1;
    END_IF;

    // ---- minutes carry-in handling ----
    IF intMin >= 60 THEN
        intMin := intMin - 60;
        intDeg := intDeg + 1;
    END_IF;

    // ---- sign placement ----
    IF decAngle < 0.0 THEN
        degOut := -(intDeg + 1);          // borrow applied uniformly
        minOut := 59 - intMin;            // complement of minutes
        IF (intMin = 0) AND (secOut = 0.0) THEN
            // edge case: negative zero-fraction input
            degOut := -intDeg;
            minOut := 0;
            edgeHit := TRUE;
        END_IF;
    ELSE
        degOut := intDeg;
        minOut := intMin;
    END_IF;

    // ---- range check ----
    IF (minOut >= 0) AND (minOut < 60) AND (secOut >= 0.0) AND (secOut < 60.0) THEN
        valid := TRUE;
    END_IF;
END_FUNCTION

Properties of the corrected function:

• Honours Method A (all positive sub-degree parts, single negative sign on the degree when applicable).
• Handles the -1.0000 edge case by detecting zero complement and reverting to the absolute form, then flipping the sign on the degree.
• Emits a valid BOOL tag so the calling OB/FB can raise an alarm or skip the write, plus an edgeHit diagnostics tag for trend logging.
• Performs carry-in handling for both seconds and minutes so floating-point round-off at 0.99999999 does not leak into the minute field.

Corrected Implementation - Method B (All-Parts-Negative)

Method B is preferred when the downstream is a SCADA tag database that carries signed DINT fields independently (e.g., aviation-style signed-DMS displays). It avoids the borrow entirely and never reaches the 60/60 defect state.

FUNCTION "fb_DMS_NegativeParts" : Void
{ S7_Optimized_Access := 'TRUE' }
VERSION : 1.0
   VAR_INPUT 
      decAngle   : LReal;
   END_VAR

   VAR_OUTPUT 
      degOut     : DInt;
      minOut     : DInt;
      secOut     : LReal;
      valid      : Bool;
   END_VAR

   VAR_TEMP 
      absAngle   : LReal;
      intDeg     : DInt;
      fracPart   : LReal;
      rawMin     : LReal;
      intMin     : DInt;
      sign       : DInt;
   END_VAR
BEGIN
    valid    := FALSE;
    absAngle := ABS(decAngle);
    intDeg   := LREAL_TO_DINT(TRUNC(absAngle));
    fracPart := absAngle - DINT_TO_LREAL(intDeg);

    rawMin   := fracPart * 60.0;
    intMin   := LREAL_TO_DINT(TRUNC(rawMin));
    secOut   := (rawMin - DINT_TO_LREAL(intMin)) * 60.0;

    IF decAngle < 0.0 THEN
        sign := -1;
    ELSE
        sign := 1;
    END_IF;

    degOut := sign * intDeg;
    minOut := sign * intMin;
    secOut := sign * secOut;

    IF (minOut >= -59) AND (minOut <= 59) AND (secOut >= -60.0) AND (secOut < 60.0) THEN
        valid := TRUE;
    END_IF;
END_FUNCTION

Method B passes -1.0000° as -1, 0, 0.0 directly, never producing 60, 60. It is the simpler, safer choice for HMI/SCADA pairs that bind each variable individually and for systems already configured with signed display fields. Both corrected implementations are functionally complete; the choice between them is a question of display convention, not correctness.

Test Case Matrix

All non-NaN cases must return valid = TRUE. Any FALSE output indicates an arithmetic fault; use the affected input range to narrow the root cause. Row DMS-15 confirms that the corrected blocks correctly refuse to commit a range-invalid or NaN-poisoned result, in line with the safety guidance in the SCL Programming and Operating Manual.

Edge Cases Beyond the Zero-Fraction Defect

Field deployments reveal additional failure modes that the corrected functions catch through the valid and edgeHit outputs:

Commissioning Procedure in TIA Portal

1. Import the corrected FB into the project tree under "Program blocks" and verify the block interface matches the PLC scan order. The compiler will warn when the S7_Optimized_Access := 'TRUE' attribute clashes with HMI tags that require symbolic addresses; in that case flip the attribute to FALSE for legacy panels.
2. Add an instance DB (e.g., DB_DMS_ServoAxis1) and wire the input from the upstream real-angle tag. Typical sources are SINAMICS drive telegram words (e.g., word 4 of telegram 102, position actual value), GPS NMEA parser outputs, or telescope axis feedback encoders.
3. Insert a watch table named DMS_Test with the fifteen test inputs from the matrix; force decAngle in incremental steps from +360.0 down to -99999.999. Use the "Force" permission level only with the safety key in RUN position and on a local test rack.
4. Online > Monitor All forces: confirm each valid bit reaches TRUE and the encoded output matches the expected column. Capture the screenshot for the FAT/SAT folder.
5. If the HMI is WinCC Unified or a Comfort Panel, configure three separate IO fields (deg, min, sec) and bind them to DB_DMS.degOut, DB_DMS.minOut, DB_DMS.secOut. Set the format string of the degree field to s999 to allow the negative sign.
6. Disable "Use symbolic names only" if you cross-reference the FB from a third-party OPC server (Kepware, Ignition) to prevent broken tag paths after a project upgrade from V17 to V18 to V19.
7. For S7-300/S7-400 targets with firmware V3.x, change LReal to Real and reduce the precision spec to centisecond resolution; the carry-in logic remains identical.
8. Document the chosen method (A or B) in the project Functional Specification; mixing Method A and Method B across the same fleet creates inconsistent operator displays and is a common cause of confusion during shift hand-over.

SCL Implementation Notes

• Declare decAngle and secOut as LReal to preserve arcsecond resolution beyond 9 decimal digits; Real (32-bit IEEE 754) loses precision below ~0.000001° once the magnitude exceeds 10 000.
• Use TRUNC rather than FLOOR for the integer-degree truncation: FLOOR returns LReal in TIA Portal V17+ and forces an extra cast.
• Avoid MOD on LReal; the SCL compiler emits an implicit-conversion warning. Implement the 60/60 carry with explicit IF blocks as shown above.
• Tag the FB with { S7_Optimized_Access := 'TRUE' } for S7-1500 and unset it (or set it explicitly to FALSE) for S7-300/S7-400 projects where the legacy MC7 code generator cannot allocate optimised symbols.
• Declare inputs and outputs in the same order in every overload variant to keep the instance DB schema stable across firmware upgrades (TIA Portal V17 → V18 → V19). A re-ordered interface forces a re-initialisation on every upgrade.
• Wrap the body in a BEGIN ... END_FUNCTION block; TIA Portal V18 added stricter syntax checks and naked statements trigger a compile error since SP1.
• For multi-axis projects, generate the FB from a master copy with the TIA Portal "Library" tab; this propagates the fix to every axis instance without manual editing.

Cross-Reference to Official Documentation

Verification Checklist Before Going Online

• All fifteen test IDs return valid = TRUE for the selected method (A or B).
• The secOut value stays strictly below 60.0 seconds (or above -60.0 for Method B).
• minOut stays strictly between 0 and 59 for Method A, or in [-59, 59] for Method B.
• For NaN-poisoned or Inf-saturated inputs the valid tag stays FALSE and the secOut tag takes no further writes.
• The instance DB is marked as "non-optimised" only when the consuming WinCC tag uses an absolute address; otherwise keep optimised access for faster runtime on S7-1500.
• The SCL compiler reports zero warnings in the "Info" pane after the next compile cycle. Any "Implicit conversion LReal to Real" warning indicates loss of arcsecond precision and must be addressed before commissioning.

Watch-Table Commissioning Pattern

The recommended TIA Portal Watch Table layout for site acceptance testing (SAT) of the corrected FB is shown below. The columns map one-to-one to the instance DB tags declared in the corrected block; the "Trigger" column lets the operator step through the test inputs without re-typing values in the PLC.

*NaN injection requires the IEEE-754 bit pattern 0x7FF8000000000000; right-click the watch cell, select "Modify bit pattern", and paste the hex value. This is useful to confirm the safety path in the HMI before live operation.

FAQ