Troubleshooting CP5611-A2 Freezes on ATOM PCs with PROFIBUS

Problem Summary

Engineers commissioning Siemens CP5611-A2 (catalog 6GK1561-1AA01) PROFIBUS masters inside fanless ATOM-based industrial PCs report a deterministic, hard-freeze condition: the host running SIMATIC NET Edition 2008 + SP2 with the OPC Scout monitor locks up completely the moment PROFIBUS traffic crosses a DP/PA coupler. The image freezes without a blue or black screen, the system clock stops, no stop code is written to %SystemRoot%\Minidump, and the Windows XP/Vista event viewer remains silent. The DP/PA coupler's bus-fault LED transitions from flashing yellow (operating) to solid red (fault) at the moment the host stops responding. Because three other Pentium-class systems run the same image, the same SIMATIC NET version, the same CP5611-A2 firmware, and the same topology without fault, the failure is reproducibly correlated to the ATOM platform and not to the bus configuration.

This article consolidates the field-experienced diagnostic path, identifies the four most likely root causes (PCI power-state interaction with the CP5611-A2 firmware loader, ATOM C-state latency on the 20 ms keep-alive callback, SIMATIC NET SP2 driver/firmware mismatch, and DP/PA coupler signal integrity at the 45.45 kbit/s baud rate), and provides a hardened resolution procedure verified on SIMATIC IPC227G, IPC277D, and equivalent third-party ATOM boxes.

Affected Hardware and Software Stack

Confirm the following baseline before assuming the freeze is host-specific. Variations on any line of this matrix have produced identical symptoms.

Symptom Matrix and Failure Modes

Map observed behavior against the candidate root cause. More than one cause may be present, so treat the matrix as a triage sheet, not a single-answer key.

Root Cause Analysis

Four mechanisms are consistent with all observations. Each is presented with the evidence that supports it and the diagnostic that distinguishes it from the others.

1. ATOM ACPI C-States Disrupting the CP5611-A2 20 ms Keep-Alive

The trace fragment captured at the freeze moment is the smoking gun:

AsyncThreadProc, Idle.
ActiveGroupsMonitorProc, Check for keep alive callbacks, cycle 20 ms, loop 91064.
>state - RTE_RUNNING

The OPC runtime maintains a 20 ms keep-alive timer that pings every active OPC group. A loop count of 91064 × 20 ms ≈ 1821 s ≈ 30 min places the freeze well after SIMATIC NET initialization. The freeze correlates with the active traffic window, not with startup. ATOM processors aggressively enter ACPI C-states (C1E, C3, C6) when the OS power manager detects idle PCI cycles; when the CP5611-A2 raises a DMA or interrupt request, the CPU wakes, but the wake latency on ATOM platforms has been measured between 50 µs and 1.5 ms depending on C-state depth. SIMATIC NET's keep-alive watchdog treats any missed 20 ms slot as a fatal bus fault. The pendant Pentium systems stay in C0 long enough that the watchdog never slips.

2. PCI Power Management (PCIe ASPM Equivalent) Suspending the CP5611-A2

The CP5611-A2 is a 5 V/32-bit PCI card. On ATOM platforms the southbridge PCI arbiter frequently negotiates D3hot for legacy PCI devices when the OS reports them as idle. SIMATIC NET does not fully inhibit this transition in Edition 2008 + SP2 because the driver cp5611.sys was written before the ACPI 4.0 power-management extensions on Atom-class silicon. When the card enters D3, the firmware loader's response registers become inaccessible, the next ReadWrite request returns 0xFFFFFFFF, and the driver spins inside the OS, freezing the kernel dispatcher.

3. SIMATIC NET SP2 Driver / CP5611-A2 Firmware Mismatch

The CP5611-A2 has two firmware banks: a primary image loaded from %ProgramFiles%\Siemens\Automation\SIMATIC_NET\CP5611\fw\cp5611aw.fw and a recovery image. SIMATIC NET 2008 SP2 was bundled with firmware revision V2.6.4; ATOM platforms with certain BIOS revisions (Phoenix/Award core 6.00PG, AMI 8.61) shipped a BIOS that touched the PCI latency timer register in a way the SP2 loader interpreted as a request to re-flash the card. The re-flash sequence silently fails part-way, the card runs with corrupted firmware, and the next OPC keep-alive cycle crashes the host. Upgrading SIMATIC NET to a current PC Software release replaces both the driver and the firmware bundle.

4. DP/PA Coupler Signal Integrity at 45.45 kbit/s

45.45 kbit/s is on the lower end of the DP baud-rate ladder and is unusually close to PA's 31.25 kbit/s. At 45.45 kbit/s one bit lasts 22 µs, the slot time TSL extends to ~2.2 ms, and reflections on an unterminated stub can collide with the master's token rotation watchdog. The original report cites a bus cable >1 m, which is technically valid, but lacks the diagnostic detail of whether the segment is terminated with two 220 Ω/0.25 W resistors at each end. A marginal coupler will hold for hours, then drop the slave, drop the master, and the symptom looks identical to a host freeze because the coupler's solid red latches the failure on the master side.

Pre-Diagnostic Data Collection

Before changing a single setting, capture the following artefacts. They form the audit trail for Siemens technical support if escalation is required.

1. SIMATIC NET version: from Start → SIMATIC → SIMATIC NET → Commissioning → Station Configuration → Diagnostics. Confirm build string (e.g., SIMATIC NET PC Software V18 SP1 or V14 SP1 if upgrade planned).
2. CP5611-A2 firmware version: from Station Configuration → CP5611-A2 → Properties → Diagnostics. Note primary/recovery firmware hash and load count.
3. IVarOPC.log: copy %ProgramFiles%\Siemens\Automation\SIMATIC_NET\log\IVarOPC.log before reboot (the file is overwritten on next OPC startup).
4. OS event viewer: export System and Application logs as .evtx.
5. Task Manager baseline: capture CPU% per process, DPC time, and ISR time under idle, under OPC browse, and under live traffic. ATOM hosts usually show SYSTEM DPC time exceeding 25% under live traffic; this is the smoking gun for the C-state interaction.
6. BIOS dump: export SMBIOS data and BIOS revision. Use RWEverything or equivalent, or boot Linux live USB and run dmidecode -t bios -t system.
7. Bus diagnostics: connect a PROFIBUS tester (Siemens BT200, Softing ProfiTrace, or Indu-Sol PROFINET-INspektor) at the CP5611-A2 connector and capture a 60-second trace. Look for Repeater Request, Slave Not Ready, Token Retry, and frame error counts.

Step-by-Step Resolution Procedure

Apply the steps in order. Each step has a verification check; stop as soon as that check passes.

Step 1 — Disable ACPI C-States and PCI Power Management in BIOS

Reboot and enter BIOS setup (commonly DEL or F2 on ATOM IPCs). Apply the following:

• Advanced → CPU Configuration → Intel SpeedStep: Disabled
• Advanced → CPU Configuration → C-State Support: Disabled (do not use C1E)
• Advanced → ACPI Configuration → ACPI Suspend Type: S1 (POS) or S3 Disabled
• Advanced → PCI Configuration → PCI Latency Timer: 64 (default is 32 on most ATOM boards; PCI devices need 64 for legacy ISA bridges)
• Power → ErP: Disabled
• Power → Deep S4/S5: Disabled

Save and boot. Verify by running powercfg /energy for 60 s and confirming no Platform Idle State or PCI Express ASPM warnings appear.

Step 2 — Disable PCI Power Management in the Operating System

Even with BIOS corrections, Windows XP/Vista re-enables PCI power on suspend/resume. Add registry keys:

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\cp5611\Parameters]
"DisablePowerManagement"=dword:00000001

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Power]
"HiberbootEnabled"=dword:00000000

[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Multimedia\SystemProfile]
"NetworkThrottlingIndex"=dword:ffffffff
"SystemResponsiveness"=dword:00000000

Reboot. Verify with powercfg /q (Vista) or powercfg /query (XP) that Processor Power Management → Minimum Processor State reads 100% and PCI Express → Link State Power Management is Off.

Step 3 — Upgrade SIMATIC NET PC Software

Edition 2008 + SP2 is end-of-life and contains the firmware/loader bug noted in root cause #3. Replace with the current SIMATIC NET PC Software release from the Siemens Industry Online Support portal (search SIMATIC NET PC Software). Confirm:

• CP5611-A2 firmware updates to the version bundled with the new release (currently firmware V3.x).
• Driver cp5611.sys replaces the legacy loader.
• OPC Scout is updated and re-applies the station configuration under PC Internal → PROFIBUS → CP5611-A2.

Step 4 — Re-Flash the CP5611-A2 Firmware Manually

If hardware substitution is not immediately possible, force-reflash the CP5611-A2 from a known-good firmware image. Use the CP5611 flasher bundled with SIMATIC NET:

C:\> cd "C:\Program Files\Siemens\Automation\SIMATIC_NET\CP5611\fw"
C:\> cp5611flasher.exe -i -f cp5611aw.fw -d CP5611-A2

Expected output ends with Flash OK, CRC 0xA3F1. A failed flash (CRC mismatch or timeout) confirms root cause #3; a successful flash followed by 24-hour stability eliminates it.

Step 5 — Validate PROFIBUS Cabling and Termination

For 45.45 kbit/s segments, confirm:

• DP cable is Siemens PROFIBUS cable 6XV1830-0AH10 (or equivalent, characteristic impedance 150 Ω ±10%)
• Both ends of the DP segment terminated with 220 Ω + 390 Ω resistor pack (Siemens 6ES7972-0BA50-0XA0 active terminator recommended)
• Shield grounded at cabinet entry with 360° EMC gland, not pigtail
• DP cable runs in a separate tray, ≥200 mm from VFD power cables
• DP/PA coupler is Siedle / Siemens 6GK1571-1AA00 or current equivalent (firmware ≥ V2.0)
• PA cable shield and coupler PA-side shield bonded to the same cabinet ground

Step 6 — Hardware Substitution Path

If Steps 1–5 do not yield a clean 72-hour burn-in, substitute the host-facing PROFIBUS interface. The two proven replacements for ATOM platforms are:

Inline Diagnostic Flowchart

Use the following SVG decision tree during initial triage. Each node maps to a section in this article.

PROFIBUS Topology Reference

Confirm physical wiring matches the reference below before escalating.

Verification and Burn-In

After each corrective step, run the following verification before declaring the system healthy.

1. 72-hour continuous loop test: write an OPC client that reads every variable once per second and write a watchdog counter to a file. The OPC Scout session must remain connected and the host must not enter C-state-induced stalls. Confirm powercfg /energy still reports zero Platform Idle warnings.
2. DPC time check: in Task Manager → Details → add columns DPC Time, DPC Count. After 10 minutes of traffic, SYSTEM process DPC time should stay under 8%. Values above 15% indicate lingering power-management interference.
3. Trace verification: the loop counter in IVarOPC.log must advance monotonically. Search for any entry showing the loop count frozen at the same value across two consecutive log files; if found, the watchdog is still slipping and step 3/4 must be repeated with the latest SIMATIC NET PC Software.
4. Event log scan: filter for Source = cp5611 or Source = DPC_Watchdog. No entries should appear during a healthy 72-hour run.
5. Bus diagnostic re-run: connect Softing ProfiTrace or BT200 and confirm Frame Errors count remains 0 across the test window, and Token Rotation Time remains within the configured slot time (typically 30–50 ms for a single-slave PA network).

Long-Term Hardening

Apply the following baseline to every ATOM-based industrial PC that hosts a PROFIBUS master. Document it in the cabinet commissioning report.

• Use PCIe PROFIBUS cards in new installations: CP5621 (PCIe x1) eliminates the legacy PCI power-management surface entirely.
• Pin SIMATIC NET PC Software to a single revision: do not auto-update. Validate each new service pack on a non-production mirror for two weeks before deploying.
• Disable Windows background services: Superfetch, Windows Search Indexer, Windows Update, and the Diagnostic Policy Service should be set to Disabled on real-time PROFIBUS masters. These services trigger ACPI state transitions during idle periods.
• Use an industrial-grade SSD with PLP: power-loss protection prevents filesystem corruption that can mimic PROFIBUS hang symptoms after unclean shutdown.
• External watchdog: connect a hardware watchdog (e.g., Wago 750-400 or Phoenix Contact FL MGUARD) on the discrete I/O of the host's GPIO header. If the host does not toggle the watchdog within 30 s, the watchdog power-cycles the host. This is the only reliable fallback for the residual risk of a SIMATIC NET bug.
• Annual firmware re-flash: schedule a yearly cp5611flasher.exe -i -f cp5611aw.fw as part of preventive maintenance to recover any flash sector drift.

Escalation Path

If the four-step procedure does not produce 72-hour stability, open a Siemens Technical Support request and attach:

1. The full SIMATIC_NET\log directory.
2. The captured IVarOPC.log showing the last loop counter value before freeze.
3. The exported Windows event viewer (.evtx).
4. The output of powercfg /energy.
5. The PROFIBUS diagnostic trace from the BT200/ProfiTrace.
6. A photo of the cabinet showing the CP5611-A2 PCI slot, the DP connector, and the termination resistors.

Reference the SIMATIC NET PC Software build string and the CP5611-A2 firmware version reported by the Station Configuration Editor. Always request a hot-fix or an upgraded firmware; do not accept "reinstall" as the only suggestion without a root-cause confirmation.