OPT8241-CDK-EVM questions

A1) Calibration:

There are 4 calibrated profiles called "Metrilus Lens Only (HW)", "Metrilus Long Range (HW)", "Metrilus Short Range (HW)", and "Metrilus High Ambient (HW)" stored in the EEPROM. These are the factory calibrated profiles that should be used for evaluation, and the best one would be "Metrilus Long Range (HW)".

The other profiles ("Short Range", "Long Range", "High Ambient", "Lens Only", "No Calibration") are base profiles that are used as parent profiles during calibration. These are not calibrated profiles.

We have an accuracy (1 sigma repeatability of measurement) of about 1cm at 2m range and 3cm at 4m range on the Tintin CDK, for a 63% reflective standard target.


A2) High Ambient profile

This profile uses higher frame rates and shorter integration times and is suitable for use in higher ambient conditions.

You can use our 3DToF Estimator tool to model the range and accuracy for your use case. This is available from the following location:
www.ti.com/.../sbac124


A3) How to get all pixel data:

You can capture frames to a .vxl file (using the 'Save/Record' button on the toolbar) and export the desired data for all frames using the Voxel Exporter tool. This export feature is available from the 'File -> Export Raw Stream' menu item in Voxel Viewer.

The exported data will be a raw binary file that can be read in any program.
The exported binary fileSize is (numFrames * numCols * numRows * bytes-per-pixel).
where bytes-per-pixel is:
= 1 for ambient data (unsigned integer)
= 2 for phase and amplitude (unsigned integer)
= 4 for depth (float32)

The data is in (cols-before-rows) format. (i.e., when you reshape the raw binary file, you will do (numFrames, numCols, numRows)

We do not currently support direct export to .csv, but it is very easy to convert the raw binary files to a .csv if needed.


Alternatively, you can also interface to the camera or to .vxl files using VoxelSDK. This will allow you to programmatically control the camera parameter and acquire frames.

VoxelSDK is available here:
github.com/.../voxelsdk


A4) Z data of watch list

The Z data in the watch list is the Z from the point cloud. It is not from the depth map.

Your 1.5m (measured) vs 0.65m (actual) difference is probably because the profile was an uncalibrated profile. You can set one of the 'Metrilus * (HW)' profiles to check. (i.e., use 'Metrilus Short Range (HW)' instead of just 'Short Range')


A5) Difference between depth map and point cloud

The d from the depth map (u, v, d) gives the radial distance measured at pixel (u, v).
The point cloud is reconstructed from the depth map by using the lens intrinsic and extrinsic parameters and provides world coordinates for each point. This reconstruction also corrects for effects of lens distortion.

If the camera is facing a flat wall and is parallel to the flat wall, the Z number from the point cloud will be about the same for all pixels, but the d number will be different. The pixels near the optical center will have a smaller d than the pixels at the edge of the sensor.


A6) Different wavelengths

The OPT8241 sensor already has a narrow band optical filter around 850nm to improve the ambient performance. This filter will not allow 940 nm light to pass through.


A7) Operating temperature

I'm afraid the current products do not support the -25 to 85 C range at this time.

Takumi-san,

These are roadmap related questions and they can be answered only under NDA. Can you please mail us at ti-3dtof@list.ti.com?

Regards