Hi,
I am using Sony IMX662 sensor with Ameba Mini board. I can see red and blue lines (please check the attched one) on the image in low light conditions. I am already using the tuned iq, fcs and sensor bin files provided in github repo, but still the quality is not good in low light conditions, even though imx662 is good option for low light.
Please tell me which ISP parameters I need to change to fix the issue using ISP contorl API from application.
Thanks,
Yugandhar
Hello Yugandhar,
Hope you are doing well.
Regarding the issue you encountered, we suggest exploring the hardware power architecture first, followed by the software side .
In our experience, a similar situation can occur when the AmebaPro2 and the image sensor share the same power supply, leading to what we call “power noise .”
To help us narrow down the cause, could you please help verify the following poin ts?
(1) Power Architecture Check: Please confirm if the VDD33 of the AmebaPro2 module is also powering the image sensor. To avoid potential interference from power loads, it is generally recommended to use a 5V LDO connected to the required power supply, keeping the AmebaPro2’s power independent from external applicati ons.
(For your reference, you can find detailed power design examples on page 4 of the AMEBAPRO2_RL6601_MINI_EVB_2V0_20220817.pdf. It shows the image sensor and Pro2 module being supplied independently via V_US B.)
(2) Testing Options: You might also want to try temporarily disabling the Wi-Fi function or using UVC for direct image output. This would help us see if the situation improves under those condi tions.
Please let us know your findings, and we would be happy to discuss the next steps wi th you.
Best regards,
Sabrina
Hi @Sabrina_Lin ,
Can you provide link to the mentioned pdf file, so that I can share with my hardware team for verification.
Thanks,
Yuagandhar
Hi @yugandhar
You can download from HDK.
You can find AMEBAPRO2_RL6601_MINI_EVB_2V0_20220817.pdf in SCH folder .
We have used the same EVK design for our development and have already shared the schematics with the PM via email.
However, we observed a difference between our two EVK boards. Both boards use different LDOs for the camera supply. One board uses the G9141, and it is not working. The other board has an LDO marked MEK4, and it is functioning properly.
Could you please help us understand the reason for this difference in behavior?
Hi @Sabrina_Lin ,
I tried to build the code by disabling WIFI in platform_opts.h as below, but build failed with errors.
#define CONFIG_WLAN 0
We never tried by disabling WiFi, can you tell me how to fix these compilation errors.
Thanks,
Yugandhar
hi, Amit
Good day,
I will follow up via email regarding the circuit details. Looking forward to your feedback.
hi, @Pammy
Good day,
Regarding the software, please discuss the settings with Yugandhar.
Thank you sincerely for your assistance.
Hi @yugandhar
To disable the WiFi temporarily, you need not set CONFIG_WLAN to 0. You can set it back to 1.
Then, navigate to main.c in project/realtek_amebapro2_v0_example/src and comment out line 121 to 128 to disable WiFi.
Hi @Pammy ,
We tried as you suggested, but at runtime it is waiting for wifi connection, so we commented wifi_common_init and tested again.
We captured in dark room, the expected image should be black, but still below noisy one.
How to believe that WiFi is not consuming power by just commenting few parts of code?
In future if we want to go with Ethernet and want to completely switch off WiFi to reduce power loss, what are required code changes?
Thanks,
Yugandhar
Hi all,
Let me update some HW information about this topic.
There is a update for the Realtek AmebaPro2 HW design Guide. It has 2 pages that discussed the power noise.
Hi @yugandhar ,
Thanks for trying out the suggested method to disable WiFi. By commenting out the WiFi APIs, you should have completely switched off WiFi and you will be able to notice a drop in power consumption, you may carry out a power measurement to verify it. Based on the method suggested above, we did observe an approximate 40-50mA drop in current while we tested AMB82 EVB with the WiFi enabled and disabled firmware.
Regarding the noise issue, may we know if you have tried using UVC for direct image output? We tried to replicate your issue by running the UVCD example using another sensor similar to IMX662, we did not observe a noisy stream. We have attached a snapshot below showing the video stream in a dark room.
Thank you.
