Using ADA4945-1 to Capture Low-Frequency PWM / CAN Signals with AD9467 — Flat Output After Signal Conditioning

Hello Analog Devices Team,

I am working with the AD9467 (16-bit, 250 MSPS) ADC and using ADA4945-1 (full-power mode) as the front-end differential driver. My goal is to capture low-frequency digital signals such as:

• PWM signals from Arduino / Raspberry Pi (kHz to a few MHz)

• CAN bus physical-layer waveforms

Current Signal Chain

• Source: Arduino / Raspberry Pi PWM or CAN signal (single-ended)

• Conditioning: ADA4945-1 configured as single-ended to differential

• ADC: AD9467 FMC board

• ADC input connected via SMA (J100)

Board Modifications

To support DC-coupled operation, I modified the AD9467 FMC front end:

• Removed input transformers and AC-coupling capacitors

• Initially removed R111 / R112 / C107 (ADC VCM bias network)

• Later re-installed R111 / R112 / C107

• Tested multiple configurations:

  • OUT+ only (single-ended into ADC)

  • OUT+ → VIN+, OUT− → VIN− (true differential)

Observed Behavior

• At higher-frequency sine waves (tens of MHz), the ADC behaves as expected

• For low-frequency PWM / square signals, the ADC output becomes:

  • Flat line

  • Or shows sporadic spikes / unstable levels

• ADA4945 outputs look correct on an external oscilloscope

• Output common-mode is ~2.15 V, matching AD9467 requirements

Key Question

Is it architecturally valid to use:

• ADA4945-1
  to extract low-frequency PWM / digital signals (kHz–MHz range) from microcontrollers and CAN bus sources when driving the AD9467?

Or is the AD9467 + high-speed differential driver not suitable for reliable capture of:

• DC-coupled

• Low-frequency

• Large duty-cycle digital waveforms?

Specifically:

1. Is the flat ADC output expected behavior due to sampling capacitor charge settling at low frequencies?

2. Is a different ADC architecture (e.g., SAR ADC) recommended for such signals?

3. Is there a recommended front-end topology for mixed-signal (PWM + CAN + analog) capture using AD9467?

Any guidance on whether this use case is supported, discouraged, or requires additional circuitry would be greatly appreciated.

Thank you for your time and support.

Best regards,
Sai Teja