imx8m plus isp camera calibration

[chp0423]

Hello, I am using imx8m plus to drive an ov2740 sensor, and using isp tuning and calibration tool for the imagae quality tuning. I have the following questions:

1. there are some sample images in the calibration tool, which are .raw images. However, the isp tuning tool can only capture .pgm and .yuv images. Can I directly use the pgm images to calibrate the ISP? I have done BLC calibration using the captured pgm images, but the results seem not to be right, because the black level in the ov2740 datasheet is 64, while my calculation results is 172. Do I need to transform the .pgm to .raw? When capturing the BLC images, I disabled all the functions of the ISP.

2. In the isp tuning tool, the manual exposure tab always present 0 for the exposure, in a result I can't adjust the exposure time. However I have set the maximum integration time in the driver function and my s_exp function is correctly written. what was wrong with the exposure slider?

3. I was testing the auto exposure function of the ISP. However I found that when I adjusted light intensity, there was only gain value changing, and the exposure time was constant (this exposure value can be set to CMOS without error). please see the following picture. Do you have any idea why the integration time was not changing during AEC control?

[joanxie]

where did you get the calibration tools? this should be offered by the third party company

[chp0423]

Our company has signed a NDA with NXP, with the help of NXP distributor. Therefore I can access the secure files. Is there any forum location I can get technical help? because in the public forum I can't post pictures as they are confidential.

[joanxie]

I got reply from marketing team:

|"Customer can engage with partner for tuning their camera module. They need to negotiate the price which works best for them."

if I get any further update, I will let you know it

[chp0423]

thank you!

[joanxie]

this is what I got:

"

 Please check the comment in tuning tool.

“For this demonstration system, the convergence speed of auto exposure is not optimized for low frame rates

(which can occur either due to manual setting or because of low light conditions and enabled AFPS). There is

significant headroom for acceleration of the convergence speed of the Automatic Exposure Control (AEC) software

for products. To avoid flicker, the AEC controls the integration time and gain. If the AEC is disabled and the

integration time is set manually, this may result in flicker. The “Exposure” slider in this dialog shows the current

exposure value, but it cannot be used to manually control the exposure. If the auto exposure mode (Level 3) is

disabled, this slider in this dialog can be used to control the exposure manually.”

So if Auto exposure control is enabled one cannot use manual control to set the parameter in tuning tool. Yes they can set the maximum integration time in the driver function though.

"

[chp0423]

Hello, I encountered a problem with calibration tool XML generator, could you please help me forward it to the technical team?

After I finished all calibrations, I created my OV2740 sensor folders and put in my calibration files (.txt) according to the guideline. When I run the XML generator, the tool returned error with "index exceeds matrix dimensions", please see the following figure. After some experiments I found it was my ColorReproduction files that caused the problem, because when I replaced my ColorReproduction files with the sample OV2775 files, the tool can run normally. I have compared my ColorReproduction file (they were generated by isp calibration tool) with the sample OV2775 file and didn't find any dimension difference. Could you please tell me what's the problem of my generated color calibration files?

[joanxie]

What is the resolution setting of the folders name?

[chp0423]

I seem to find the problem. The txt files in the ColorReproduction folder must be named to "cc_para_100.txt", like the following figure shows. Other naming will cause the tool returning error

[joanxie]

good to know this