Sensor streaming fails in Non-continuous clock mode in L4T 24.2 & L4T 24.2.1

Robotics & Edge Computing Jetson Systems Jetson TX1
1

Hi All,

My sensor (RAW10, FullHD, 4 Lane) configured in non-continuous clock mode works well in L4T 24.1, but fails in L4T 24.2 & L4T 24.2.1.

I am partially getting the image, but it blinks and most of the region is overlapped with vertical lines.

Below are VI/CSI Status register values for L4T 24.2 & L4T 24.2.1:

TEGRA_VI_CSI_ERROR_STATUS       : 0x00000001
TEGRA_CSI_PIXEL_PARSER_STATUS   : 0x00000034
TEGRA_CSI_CIL_STATUS            : 0x00000000
TEGRA_CSI_CILX_STATUS           : 0x00000000

I tried changing “discontinuous_clk” property in the dtb, but noticed that this property is no where used in the BSP (I might be wrong though, needs confirmation).

If you could let me know which register setting decides between Non-continuous & continuous clock mode, I can dump the register and confirm the current setting too.

Any hints on what could be my issue or where should I focus to fix the issue?

Regards,
Rejeesh

2

@Rejeesh
What command/tools you are using to grab the sensor image?
If you are using the yavata/v4l2-ctl you can try modify the kernel/drivers/media/platform/tegra/csi/csi2_fops.c
the REG TEGRA_CSI_CIL_PHY_CONTROL from 0xA to 0x4A, you can download the TRM from download center to check the detail.

3

Hi Shane.

I am using V4L2 tool.

I already tried setting both 0xA & 0x4A to TEGRA_CSI_CIL_PHY_CONTROL register, but no change in behavior.

In fact, 0x4A is the default in 24.2 & 24.2.1.

Regards,
Rejeesh

4

Hi Shane,

I am running out of ideas, any other suggestions?

Regards,
Rejeesh

5

@Rejeesh
From the log it could be the resolution you define is smaller than the sensor real output size.

External Media

6

Hi Shane,

Thanks for taking it up.

It might not be the case, since the same sensor with same TX1 side resolution settings (1920x1080) is working in 24.1.

I also tried changing CSI2_PIXEL_STREAM_A_CONTROL0_0 from 0x2A0301F0 to 0x280301F0 (24.1 value , PAD short lines). The output pattern changed but its still not the correct one.

Any other register to look into (specifically related to continuous/ discontinuous mode settings)?

Regards,
Rejeesh

7

Any thoughts Shane?

8

Hi Shane & others,

Just wanted to update that there are no issues with dis-continuous mode in L4T28.1, hence decided to proceed with 28.1.

Hopefully, I will not have to revert to 24.2.1 or 24.1 later. :)

Regards,
Rejeesh