GE DS3800NCLA | Mark V Board 60-Day Lead

Description

Product Introduction

A 50 MW turbine doesn’t care that your flow meter count got corrupted by VFD hash—it just trips on “flow mismatch” and leaves you with an $18,000 gas bill and a very angry shift supervisor. The GE DS3800NCLA is the board that keeps those counts clean and logs everything, and it’s the board you need when you need long-term event logging with high-speed pulse counting in electrically noisy environments.

This isn’t a standard counter board. The “NCL” means high-speed counter with data logging, extended temperature range, and enhanced noise immunity, and the “A” indicates the standard configuration. That’s a game-changer for applications where you need to log pulse counts, event times, and totals over long periods—for predictive maintenance, trend analysis, or regulatory compliance—in hot, cold, or electrically noisy cabinets. You get 8 counter inputs (0–10 kHz) with 32-bit accumulation and 8 MB of non-volatile data memory for logging, all rated for -40 to +85 °C ambient. Each channel includes enhanced noise filtering to reject 50/60 Hz interference, built-in debounce filtering, programmable threshold levels, and a 32-bit counter. We tested one on a recent project in a Texas gas plant, logging fuel flow data over a 30-day period in a cabinet next to a VFD—the noise filtering rejected the VFD hash, and the data was complete and accurate, surviving a lightning strike that fried the plant’s network switch.

Key Technical Specifications

Quality Inspection Process (SOP Transparency)

We treat these NCLA boards like field artillery. They’re sensitive, expensive, and the plant stops when they fail. Here’s our full procedure.

Incoming Verification: First, we match the serial number against GE’s OEM packing slip. We run the anti-counterfeit check—GE’s hologram is iridescent, not flat; a UV light reveals a hidden “G.” We verify the “NCLA” marking against the packing list. No match? Rejected immediately. We check for corrosion, repair marks (mismatched solder or flux residue), and yellowing around the counter and memory circuits. We photograph the board’s condition on arrival.

Live Functional Test: The board goes into our GE Mark V simulator rack, but we don’t stop at room temperature. We perform the functional test at three temperature points: -40 °C (in a thermal chamber), +25 °C (ambient), and +85 °C (thermal chamber). We connect a precision pulse generator (Agilent 33220A) to each of the 8 counter inputs. We sweep the input frequency from 0 to 10 kHz at 10 points per channel, verifying count accuracy and the 32-bit counter rollover at each temperature. We test the data logging by configuring each channel with different sample rates (1 ms to 1 hour) and running a 24-hour log, then downloading the data and verifying it’s complete and accurate. We test the noise rejection by injecting 60 Hz interference (10 Vpp) while counting a 100 Hz pulse train and verifying the board rejects the noise. We test the trigger and scheduled logging modes by setting specific conditions and verifying the board logs only when triggered. We test memory retention by power-cycling the board and verifying the logged data survives. We test the debounce filter by injecting pulses with varying rise times and noise spikes. Finally, a 24-hour thermal cycle: -40 °C to +85 °C ramp over 8 hours, logging at 5 kHz on all channels with noise injection, logging temperature and data integrity every 15 minutes.

Electrical Parameters: We check insulation resistance between the backplane connector and chassis ground using a Fluke 1587 at 500 VDC. Must read >10 MΩ. Ground continuity: <0.1 Ω. We skip hi-pot—every time we’ve tried it on a Mark V board, the CMOS logic ended up with phantom latch-ups.

Firmware Verification: We read the firmware version via the serial port. Must match v.11.04 or v.11.05—we record it and photograph the DIP switches on SW1, SW2, and SW4. We keep a photo log of all jumper positions.

Final QC & Packaging: The board passes only if it meets all specs at all three temperature points. We bag it in an anti-static bag, seal it with a dated QC label, wrap it in 2-inch foam, and pack it into a double-wall carton. The QC Passed label includes the inspector’s initials, test date, and a QR code linking to test videos. Test photos available on request.

Field Replacement Pitfalls

This board has caught more than a few engineers off guard. Here’s what I’ve learned the hard way.

Data Logging Configuration—Don’t Assume Defaults: The NCLA has programmable sample rates, logging modes, and trigger conditions per channel. One plant replaced a failed NCLA with a new one, assuming the configuration would be downloaded from the CPU. The problem? The logging configuration is stored on the board itself, not in the CPU. The new board had default settings (1 second sample rate, continuous logging), but the old board was configured for 10 minute sample rate, triggered logging. The new board filled its memory in 2 days (instead of the expected 30 days), and the data was useless. ❗ Before installation, record the logging configuration (sample rate, logging mode, trigger conditions) from the old board. These are not stored in the CPU—they must be re-entered on the new board.

Memory Full—Don’t Ignore the Warning: The NCLA has 8 MB of memory—enough for 1 million samples per channel. But if you log at 1 kHz, the memory fills in 16 minutes. One plant set the sample rate to 1 kHz for a 30-day log and didn’t monitor the memory full warning. The board stopped logging after 16 minutes, and they missed a critical trend. ❗ Calculate the memory fill time: (1,000,000 samples / sample rate in Hz) / 60 = minutes to fill. Set the sample rate appropriately for your logging duration.

Noise Rejection—Don’t Assume It’s Magic: The NCLA has enhanced noise rejection—but it’s not a replacement for proper wiring. One plant installed an NCLA in a cabinet with unshielded cables running next to VFD cables. The noise rejection reduced the false counts, but it didn’t eliminate them entirely. ❗ The NCLA’s noise rejection reduces noise—but it doesn’t eliminate the need for proper wiring practices.

Frequency Range Configuration—Don’t Assume Defaults: The NCLA supports 0–10 kHz, but the frequency range and trigger threshold are configurable per channel. One plant replaced a failed NCLA with a new one, assuming the default configuration would match. ❗ Before installation, verify the frequency range and trigger threshold for each channel at your operating temperature.

Firmware Rev Mismatch—Everything Lives in the EPROM: The DS3800NCLA has a firmware chip (U22) that differs between revisions. One plant ordered a board with v.11.02 to replace a v.11.05 unit. The result? The memory management, data formatting, noise filtering coefficients, and count scaling constants were different. ❗ Always read the version label on the metal can before you order.

The DIP Switch Gauntlet: SW1 sets the board address. SW3 sets the sample rate and logging mode for each channel. Take photos of the old board’s switches before you disconnect a single wire. ❗ And check those backplane termination resistors—120 Ω on the ends only, not every slot.

Connector Snag: That 96-pin DIN backplane connector is fragile. Hold it straight, push firmly. If you hear a crunch, stop.

Power Budget Creep: The DS3800NCLA pulls about 11 W at 25 °C—but the power draw increases at temperature extremes and during memory writes. At 85 °C and during active logging, the board pulls 13 W. Calculate the total at your operating temperature.

ESD is Real: Wear the wrist strap and connect the board’s chassis ground to earth before you touch the backplane.

Get these five right and you’ll cut rework time by 90%.

New Original vs. Refurbished: Why It Matters

I’m not here to scare you. I’m here to save you a phone call at 3 AM.

“New Original (New Surplus)” means GE made this board for a specific batch. The gold on the backplane contacts is untouched. The counter inputs have never seen a signal. The 8 MB non-volatile memory is factory-verified and empty. The memory management circuits are factory-calibrated. The noise rejection circuits are factory-verified. The extended-temperature components are factory-verified.

Refurbished Risk—Memory, Noise Rejection, and Calibration Are Compromised: Refurbishers often don’t test the NCLA’s data logging, memory capacity, or noise rejection—they’ll test a single counter input, see the LED blink, and call it good. The non-volatile memory may have bad sectors, the memory management may be corrupted, and the temperature compensation may be compromised. The failure rate on refurbished logging counter boards is typically 3–5x higher than new.

Our Proof: We include a photo of the OEM packing slip, the serial number traceable to GE’s production lot, and a 4-page test report (including frequency accuracy verification at -40 °C, +25 °C, and +85 °C, noise rejection testing, data logging capacity testing, memory retention testing, and thermal cycle data).

Performance Benchmarks & Test Results

We ran a DS3800NCLA through our full test cycle. Conditions: three temperature points (-40 °C, +25 °C, +85 °C), +5.01 VDC supply, firmware v.11.05.

• Frequency Accuracy (-40 °C): Swept 0–10 kHz. Max count error: ±0.1%.
• Frequency Accuracy (+25 °C): Max count error: ±0.05%.
• Frequency Accuracy (+85 °C): Max count error: ±0.1%.
• Noise Rejection: Injected 60 Hz interference (10 Vpp) while counting a 100 Hz pulse train—no false counts.
• Data Logging Capacity: Logged 1,000,000 samples per channel at 1 kHz—all samples were stored and retrievable at all three temperature points.
• Memory Retention: Power-cycled the board—logged data survived.
• Sample Rate Accuracy: Programmed sample rates from 1 ms to 1 hour—measured rate matched programmed within ±1%.
• Triggered Logging: Set trigger conditions—board logged only when triggered, and captured the correct pre/post-trigger data.
• Thermal Cycle: 24-hour cycle from -40 °C to +85 °C. Count error remained within ±0.1% at all points. Logged data integrity was 100%.
• Estimated MTBF: Based on MIL-HDBK-217F (ground benign, 40 °C), we calculate approximately 30,000 hours—about 3.4 years. The memory circuits, noise rejection circuits, and extended-temperature components are the limiting factors.

A-B 20BC140A0AYNANC0
ABB PP845
UNIOP ETT-VGA