• Resolved

Compiler/OPT9221: Calibration pattern/setup for nonlinearity calibration

Intellectual 615 points

Replies: 9

Views: 159

Part Number: OPT9221

Tool/software: TI C/C++ Compiler

Hello TI!

About 1 year ago I asked you some questions about nonlinearity calibration.

We produced little series devices base on your chip OPT9221 and we are going start new one.

At current step we try to understand what is better for calibration process.

Main question is:

Do you have some prototype/ready solution/or ... for phase calibration setup?

We have made raw design of setup for nonlinearity calibration because it the most difficult from point of view measurements.

It consist from 16 positions in accordance with chosen  unambiguous range and one position for next 3 calibration.

I think it will able work properly but maybe your can advice us something useful for our goals.

(in attachment image of raw design )

  • Hi Semenov, 

    Such a setup would work, but it seems that you have to move the camera manually. Something like an optical rail would help you better, with motors connected to the target/camera programmed to move a fixed distance for these 16 positions. 

    For pixel-wise phase offset and common phase offset calibrations, you can use one known distance from the full range. 

  • In reply to Suramya Gupta:

    Suramya thank you for your answer!

    One more question

  • In reply to Suramya Gupta:

    Suramya thank you for your answer!

    One more question:

    Is it possible to measure phase values at for example 50 equal points within unambiguous range and after calculate interpolation like in python scripts?

    (for do not adjust positions  for R/16*i i = 0..15 special distances)

  • In reply to Semen Semenov:

    Semenov, 

    If you're taking 50 equal points, it would be better to go for software non-linearity correction. The calibration steps will change, but you can implement an LUT in software, with measured phase and expected phase, and correct for the phase accordingly. This would give you a better resolution than the 16 points on the OPT9221. 

    Suramya

  • In reply to Suramya Gupta:

    Suramya thank you for your answer!

    If use 50 or more equal points, make LUT, evaluate phase values for 16 points and write to respective registers it will be better than using 16 equal points?

    If using exactly 16 points what tolerance is appropriate for distance/position?

  • In reply to Semen Semenov:

    Semenov, 

    Yes, using 50 or more equal points would make the interpolation function around these 16 points better (more points means a better curve fit). However, the correction is limited by the number of points available for the LUT, so, even though there will be an improvement (due to a better fit of the straight line), it won't be as good as having a 50-point LUT in software. 

    Suramya

  • In reply to Suramya Gupta:

    Suramya thank you for your answer!

    So,

    using 50 or more equals points and writing it to chip have not sufficiently better accuracy than using exactly 16 points.

    Using 50 or more equals points and applying software calibration by software lead to sufficiently better result than in last line.

    Is it right?

    What is appropriate accuracy for distance measurement during calibration?

    For example, using 20 MHz => R = 7.5 m - unambiguous range,

    7.5 is not divisible exactly to 16. Is using step = 47 cm enough for good measurement?

  • In reply to Semen Semenov:

    Semenov, 

    Using 50 points will slightly improve accuracy, as the straight line model would be a better fit. 

    Yes, using 50 points in software would give a much better result. 

    For 7.5 m, 47 cm should work fine. 

    Suramya

  • In reply to Suramya Gupta:

    Suramya thank you for your answer!