VL53L7CX performance issues in enclosed ceramic environment (toilet bowl) - "Low Confidence" during object drop

[Wjx0702]

Hi,

I am developing an object detection system using the sensor, integrated inside the inner wall of a toilet bowl to detect falling waste. I am experiencing persistent "Low Confidence" issues, and I am looking for expert advice on whether this is an inherent physical limitation or a configuration issue.

Current Setup & Environment:

• Environment: Actual toilet bowl (highly reflective curved ceramic surfaces).

• Hardware Integration: The sensor is mounted behind a cover glass, and an active NIR/Visible fill light is installed inside the bowl to assist a separate camera system.

• Configuration: VL53L7CX_RANGING_MODE_AUTONOMOUS @ 60Hz.

The Phenomenon:

• Static/Idle State: Even in this real-world environment (with cover glass and fill light), the sensor produces some "Low Confidence" data points even when nothing is happening.

• Dynamic Event (Object Drop): When an object is dropped, the frequency of "Low Confidence" reports increases significantly, often rendering the detection logic unreliable.

My Technical Questions:

1. Root Cause Analysis: Given that "Low Confidence" exists even in the static state, is it likely that the cover glass (internal cross-talk) and fill light (optical saturation/background noise) have already pushed the sensor to its SNR limit? Does the dynamic drop event simply saturate the already compromised signal?

2. Software Mitigation: With a 60Hz capture rate, is there any software-level optimization possible? Can adjusting the Timing Budget or Sharpener threshold help recover signal integrity in this high-noise environment?

3. Feasibility: Is the VL53L7CX fundamentally unsuitable for this specific application (high-reflectivity ceramic enclosure + active fill light interference)? Or is there a more robust way to process the data (e.g., histogram analysis) to differentiate the object reflection from the background noise?

4. Hardware Best Practices: Are there any recommended hardware modifications for such confined spaces, such as specific optical filters (e.g., 940nm bandpass) or mechanical sealing techniques to isolate the sensor from the fill light?

I would appreciate any guidance on whether to continue optimizing the firmware parameters or if this application exceeds the physical capabilities of the current hardware setup.

[Bin]

Hi,

What you describe is very likely a combination of cover glass crosstalk, strong multipath reflections from the curved ceramic surface, and additional optical noise from the fill light.

From the VL53L7CX datasheet and user manual, a few points are important here:

1. Cover glass can strongly affect short-range performance
   The user manual explains that histogram processing gives good crosstalk immunity beyond about 60 cm, but below that distance cover glass crosstalk can become comparable to or even larger than the real return signal. In a confined toilet-bowl geometry, this can easily generate low-confidence targets even in a static scene.

2. Autonomous mode at 60 Hz is not the most robust choice for this environment
   In continuous mode, the VCSEL is enabled during the full ranging process, which gives better ambient light immunity and better ranging robustness than autonomous mode.
   In your case, trying to keep 60 Hz autonomous may be pushing the sensor toward a low-SNR regime.

3. High-reflectivity curved ceramic surfaces are a difficult optical environment
   They can create strong internal reflections and multi-path returns, which often show up as:

   • high sigma
   • consistency failures
   • low-confidence target status
   • unstable target detection during dynamic events

4. The active fill light may also be part of the problem
   If the illumination path leaks into the sensor optical path directly or indirectly (through the cover glass or ceramic reflections), it can further reduce signal quality.

Advises

• Perform crosstalk calibration in the final mechanical assembly (with the actual cover glass and final enclosure).
• Test with 4x4 resolution instead of 8x8.
• Lower the frame rate and check behavior at 15 Hz or 10 Hz first.
• Try continuous mode to see whether low-confidence behavior improves.
• If autonomous mode is required, increase integration time and reduce the frame rate accordingly.
• Use target status + sigma as the primary confidence filters.
• Improve optical/mechanical isolation between the fill light and the ToF sensor.
• If possible, evaluate a 940 nm optical filtering / shielding approach to reduce unwanted light entering the module.

Feasibility

So this does not automatically mean the VL53L7CX is unusable, but your current setup is clearly a very challenging optical environment.
If you still see strong low-confidence behavior after:

• final crosstalk calibration,
• lower frame rate,
• 4x4 mode,
• continuous mode,
• and better light isolation,

then the limitation is likely becoming system-level / optical, not just a firmware tuning issue.

Best regards,

Bin FAN

 

[David Littell]

Just a thought: maybe abandon NIR/visible and consider actual IR (after a couple of beers, of course 😉).  To silly old me it seems like dealing with an actual/only IR image would simplify everything, starting with eliminating the need for extra illumination.