Robust CAN Electrical Measurements in LabVIEW — Handling Ringing and Overshoot with PXIe-5111

[sanacdac]

Hello NI Community,

I am developing an automated CAN physical layer validation system in LabVIEW using a PXIe-5111 oscilloscope. The goal is to validate CAN electrical characteristics against ISO 11898-2.

Test Setup

• Oscilloscope: NI PXIe-5111

• Driver: NI-SCOPE in labview

• CAN speed: 250 kbps (Classical CAN, ISO 11898-2)

• CH0 → CAN_H

• CH1 → CAN_L

• Probe: 10x passive probe, 1 MΩ input impedance

Acquisition Settings

• Sample rate: 10 MS/s

• Record length: 50,000 samples

• Trigger: Rising edge on CH0 at 3.0V

---

Measurements Being Performed

I am currently validating:

1. Bus Voltage Levels

• CAN_H dominant/recessive

• CAN_L dominant/recessive

• Differential voltage (Vdiff)

2. Differential Signal Characteristics

• Peak dominant Vdiff

• Recessive Vdiff floor

• Peak-to-peak differential amplitude

Thresholds are derived from ISO 11898-2:2024:

• Table 3 → recessive state

• Table 5 → dominant state

---

Current Measurement Method

My current LabVIEW flow is:

1. Initialize NI-SCOPE

2. Configure vertical/horizontal timing

3. Configure analog edge trigger

4. Acquire CAN_H and CAN_L waveforms

5. Extract Y arrays

6. Use:

   • Array Max → dominant levels

   • Array Min → recessive levels

7. Compute:

   • Vdiff = CAN_H − CAN_L

8. Compare against ISO thresholds

9. Generate PASS/FAIL result

---

Problem Encountered

The issue is that CAN edges contain ringing and overshoot artifacts.

Because I am using raw Array Max and Array Min, even a single transient sample can corrupt the measurement.

Example:

• True dominant CAN_H level ≈ 3.65V

• One ringing spike reaches 4.10V

• Array Max reports 4.10V

Similarly:

• Recessive ringing can briefly drive Vdiff slightly negative

• Array Min then produces false failures

So the measurement becomes sensitive to edge artifacts rather than representing the steady-state bus condition intended by ISO 11898-2.

---

Approaches I Am Considering

1. Percentile-Based Measurement

Use a percentile (95th/99th) instead of raw max/min.

Pros:

• Simple implementation

• Rejects isolated spikes

Concern:

• Percentile selection feels somewhat arbitrary for compliance testing

---

2. Histogram / Cluster-Based Measurement

Use histogram analysis to identify dominant and recessive voltage clusters, then calculate the mean of each cluster.

Pros:

• Physically representative

• Similar to oscilloscope statistical measurements

Concern:

• More implementation complexity

---

3. Threshold-Masked Mean (Current Preferred Approach)

Create a dominant-state mask using:

• Vdiff > 0.9V

Then:

• Extract only confirmed dominant samples

• Compute mean CAN_H / CAN_L values from masked samples

This naturally excludes:

• recessive bits

• transitions

• ringing regions

This feels closest to measuring the actual steady-state dominant bus voltage.

---

4. NI-SCOPE Built-In Measurements

Use NI-SCOPE scalar measurements such as:

• VOLTAGE_HIGH

• VOLTAGE_LOW

I understand these internally use histogram/statistical methods.

However, I am unsure whether they are appropriate for a two-state CAN waveform versus a more traditional periodic digital waveform.

---

Questions

1. For CAN electrical compliance measurements, which methodology is generally considered most appropriate?

2. Are NI-SCOPE measurements like VOLTAGE_HIGH / VOLTAGE_LOW suitable for CAN waveforms with dominant/recessive states?

3. Is the threshold-masked mean approach considered valid for ISO-oriented automated testing?

4. Has anyone implemented a similar CAN physical-layer validation system using PXI hardware and NI-SCOPE? If so, what measurement strategy did you use to avoid ringing-related false readings?

Any guidance or practical experience would be greatly appreciated.

Thank you.