Resolving TIA Portal DataLog Format Issues on S7-1200/S7-1500

Problem Overview: S7-1200 / S7-1500 DataLog CSV Output Defects

The DataLog instruction family in TIA Portal (DataLogCreate, DataLogOpen, DataLogWrite, DataLogClose, DataLogDelete, DataLogNewFile) is intended to produce a portable, comma-separated log file on the load memory / SD card of an S7-1200 or S7-1500 CPU. In practice, four recurring defects break downstream consumption in Microsoft Excel, Python pandas, MES historians, and any tool that expects a clean RFC 4180-style CSV:

1. Forced "RECORDS" index column – A monotonically incrementing counter column is prepended to every record when timestamp stamping is active. It cannot be disabled through the block interface.
2. UTC offset on timestamp – When the wizard preconfigures timestamp = "1" (System time), the CPU writes the values from the realtime clock but applies no localization. Operators in CET / CEST observe a −1h or −2h drift relative to the wall clock.
3. Quoted string fields – String tags passed through the DATA parameter are wrapped in double-quote characters ("..."). The DataLog engine treats this as a hard-coded escape, even when the content contains no comma, semicolon, or quote.
4. Separator mismatch – The default separator is the comma ,. Changing it to ; in the wizard does not retroactively affect the header generation logic, and the "RECORDS" column re-appears whenever a timestamp is requested.

Root Cause: How the DataLog Engine Builds the CSV

The DataLogCreate instruction (FB block number varies by firmware; documented in the S7-1200 / S7-1500 system manual) writes a fixed header row, then a fixed record row, into a binary record file. When the CSV is later extracted from the S7 load memory (or read directly from the SD card on a CPU that supports it), the engine concatenates fields with the configured separator.

Three behaviors drive the defects:

These are not bugs in the strict sense; they are documented edge cases of the DataLog firmware implementation, but the TIA Portal wizard exposes very little of this state, which is why field engineers hit the wall.

Affected Products, Firmware, and Software Versions

The defects map to specific firmware generations. Confirm the firmware build on the CPU first — the available TIMESTAMP constants differ across versions.

For the full firmware / TIA Portal compatibility matrix, refer to the Siemens SIMATIC S7-1500/ET 200MP system manual and the SIMATIC S7-1200 system manual.

TIMESTAMP Parameter Reference

The full TIMESTAMP enum is documented in the TIA Portal online help under "DataLogCreate: Create data log" (F1 on the block). Switch the project "Editing language" to English to see the authoritative list, as localized help bundles are occasionally out of date.

For European plants, the recommended constant is TIMESTAMP = 2 on S7-1500 FW 2.6+ when local time in classic layout is needed, or TIMESTAMP = 4 when ISO-8601 is required by the downstream MES / historian. Neither constant is available on S7-1200 at any firmware; S7-1200 users must manually localize in the application code (see workarounds below).

DataLogCreate Block Interface – Key Inputs

The full block signature is published in the TIA Portal help. The parameters that drive the defects above are summarized here for quick reference:

Verified Workaround: Disable TimeStamp, Build the Header / Data in Application Code

The only field-proven path that yields a clean CSV — no RECORDS column, no double-quoted string fields, with local time and a configurable separator — is to set TIMESTAMP = 0 and emit the timestamp yourself as part of the HEADER and the record STRUCT.

Step-by-step

1. Define a UDT for the data record (example for a barcode / weight / recipe logger):

   TYPE "UDT_LogRecord" :
   STRUCT
       sDate   : STRING[10];   // "YYYY-MM-DD" or "DD/MM/YYYY"
       sTime   : STRING[8];    // "HH:MM:SS"
       sBarcode: STRING[32];
       rWeight : REAL;
       sRecipe : STRING[32];
   END_STRUCT;
   END_TYPE

2. Define a UDT for the entire log line (only the data fields you want to log):

   TYPE "UDT_LogLine" :
   STRUCT
       sBarcode : STRING[32];
       rWeight  : REAL;
       sRecipe  : STRING[32];
   END_STRUCT;
   END_TYPE

3. Build the HEADER string once, in static / init code:

   // "date,heure,componant1,componant2,componant3"
   "sHeader" := 'date,heure,componant1,componant2,componant3';

4. In the cyclic OB, populate the record with current local time and the data:

   "udt_struct".sDate   := MID(IN := DTL_TO_STRING("dtNow"), LEN := 10);
   "udt_struct".sTime   := MID(IN := DTL_TO_STRING("dtNow"), LEN := 8);
   "udt_struct".sBarcode := "iBarcode";
   "udt_struct".rWeight  := "rWeight";
   "udt_struct".sRecipe  := "sRecipe";
   
   "DataLogWrite"(ID     := "dwLogID",
                   RECORD := "udt_struct");

5. Create the log with TIMESTAMP = 0:

   "DataLogCreate"(REQ    := TRUE,
                     NAME   := 'ProcessLog',
                     HEADER := "sHeader",
                     TIMESTAMP := 0,
                     FORMAT    := 0,         // 0 = comma
                     DATA   := "UDT_LogLine",
                     ID     => "dwLogID",
                     STATUS => "wStatus");

Resulting CSV row example:

2024-06-12,14:32:07,BC-12345,12.34,Recipe_A

No RECORDS column, no surrounding quotes, comma is the only separator. Excel "From Text/CSV" imports this directly with the comma as delimiter.

Workaround: Localized Timestamp on S7-1500 FW 2.6+

If you can constrain the project to S7-1500 FW 2.6+ (and the matching TIA Portal V15.1+), the firmware now renders the local time correctly:

"DataLogCreate"(REQ       := TRUE,
                  NAME      := 'ProcessLog',
                  HEADER    := "sHeader",
                  TIMESTAMP := 2,        // local time, classic layout
                  FORMAT    := 0,        // comma
                  DATA      := "UDT_LogLine",
                  ID        => "dwLogID",
                  STATUS    => "wStatus");

For ISO-8601 output (recommended for MES systems that store the timestamp as text):

TIMESTAMP := 4;   // local time, yyyy-mm-dd hh:mm:ss[.999]

Workaround: Cleaning the Output in Excel / pandas

If the log files have already been generated with the legacy format and the customer is waiting, the following pipeline is the least painful.

1. Open the .csv in a plain text editor (Notepad++, VS Code). Find / replace , with . only on lines that contain numeric data, to convert European decimal notation.
2. Open the file in Excel. Use Data → From Text/CSV. Choose ; as the delimiter if the wizard used semicolon, , otherwise.
3. Delete the first column (RECORDS) and the trailing empty column if present.

For Python / pandas consumers, the same logic is two lines:

import pandas as pd
df = pd.read_csv('ProcessLog.csv',
                 sep=',',
                 encoding='utf-8-sig',
                 skiprows=0,
                 usecols=lambda c: c != 'RECORDS')   # drop the counter column

The official Siemens FAQ 87138437 documents the legacy post-processing flow for S7-1200 data logs.

Alternative: RecipeExport / Data Log on Recipe DB

When the application can accept a different export mechanism, the RecipeExport instruction family and the Data log on recipe DB feature introduced in TIA Portal V15.1 give finer control over the resulting file. RecipeExport is intended for production recipes, not high-frequency logging, and uses more load memory per write, but it does not impose the RECORDS column and does not quote STRING fields.

Reserve the technique for batch logs (recipe changes, parameter sets, audit trails) where the write rate is low and the file size overhead is acceptable.

Verification Procedure

1. After DataLogCreate returns STATUS = 0, open the Web server of the CPU (or read the SD card) and download the log file. Confirm the file name matches NAME and the file size grows after each DataLogWrite.
2. Inspect the first two lines (header + first record). Verify:
   • The number of comma-separated columns equals the number of HEADER tokens.
   • STRING fields are not surrounded by double quotes.
   • The first column is the one named in HEADER (not "RECORDS").
   • The timestamp column matches the chosen TIMESTAMP format and the wall-clock time of the operator's HMI.
3. Import into Excel via Data → From Text/CSV. Each column should align with its header without manual cleanup.
4. Import into Python via pandas.read_csv. Inspect df.dtypes: STRING fields are object, REAL fields are float64, no extra leading column.

Error Code Reference

Full list is in the TIA Portal online help under the DataLog instruction group.

Troubleshooting Matrix

Field-Engineering Notes

• The TIA Portal help is partially translated. Switch the project "Editing language" to English to see the full TIMESTAMP enum and the Format / separator constants. Half-translated help is the most common reason a field engineer sees only constants 0 / 1 and concludes the rest do not exist.
• The HEADER parameter is documented as a single STRING[256] at the FB interface. A STRING array is rejected at runtime with W#16#80B0 (file format error). Build the header by concatenating tokens in the application code, with the configured separator between them.
• STRING quoting in the DATA record is a hard-coded behavior of the DataLog engine. There is no "do not quote" parameter. The only mitigation is to pre-format everything inside a STRUCT and not to pass a single concatenated STRING as the DATA element.
• Daylight-saving handling on S7-1500 FW 2.6+ with TIMESTAMP = 2 / 4 is correct only if the CPU's time zone and DST flag are set in the CPU properties under "Time of day". Without that configuration, the firmware still produces local time, but it is the local time of the unsynchronized RTC.
• For high-frequency logs (sub-second), prefer a third-party data logger (PC side, MQTT / OPC UA) and disable the S7 DataLog. The DataLog is designed for sporadic process events, not 1 kHz trace.