OV7740 camera on I.MX6 using CSI0 in BT.656 mode

[robert_lim]

I am using an imx.6 Sabrelite (PN:BD-S1-imx6) from BoundaryDevices to develop a camera system. The camera sensor is the ov7740 from omnivision, attached to Sabrelite via CSI0 (8 data lines, pixelclock and camera system clock). As I do not have the possibility to connect the hsync and vsync signal this all has to work in BT.656 mode.

I am having issue getting images from the driver. The video for linux drivers hangs waiting for data and after a timeout breaks down the video4 linux pipeline.

The steps I have taken so far:

1. Started with working kernel (boundary-imx_3.10.17_1.0.0_beta) and
2. Made sure that the camera is detected and can be configured over I2C.
3. Built a driver, OV7740.c, based on OV5642.c with the following changes:
   1. Replaced modes  to generate a 640x480 30 fps data on the OV7740     
   2. Changed the chip_id     
   3. Changed ov7740_data.pix.pixelformat to V4L2_PIX_FMT_YVU420     
   4. Added V4L2 required IOCTL functions

1. Sensor is correctly detected on I2C
2. PCLK and XCLK are correct
3. Use tool: mxc_v4l2_capture.out and gstreamer to try to capture sensor data

The test encounter DQBUF failed during capturing, the error is as follow:

ERROR: v4l2 capture: mxc_v4l_dqueue timeout enc_counter 0 VIDIOC_DQBUF failed.

When trying to decipher the video for linux kernel module, it seems that this occurs when the data queue hasn't been filled until a certain timeout. The queue should be filled through the CSI interrupt handler. As I can see it now, it seems that the interrupt is never raised.

I have taken the following link which have similar problem using BT656 mode to capture in progressive mode as a reference:

https://community.freescale.com/thread/309869

https://community.freescale.com/docs/DOC-95412

For more info, below summarize our setting and dumped info to get the CSI BT656 without HSYNC/VSYNC working:

1. function ioctl_g_ifparm() sets p->u.bt656.bt_sync_correct = 0:
2. modify the code in mxc_v4l2_s_param(), let csi_param.clk_mode = IPU_CSI_CLK_MODE_CCIR656_PROGRESSIVE
3. The CSI0 connects parallel, IOMUXC_GPR1 = 0x48693005 and IPU_CONF = 0x00000785

I think this is due to the CSI/IPU not setup properly for BT.656 mode without hsync/vsync signals. Please give some support on how to configure the CSI/IPU with this camera?

Thanks in advance.

[aurelienbouin]

Hello Thomas,

I am sorry I don't have more advice ...

What I knwo is that when CSI0 was not working with my hardware there was at least the CSI test mode working ...

It seems weird that for you it is not ...

Regards,

Aurelien

[aurelienbouin]

Nobody ?

[aurelienbouin]

I finally figured out what was wrong !!!

There is in the source linux file from freescale arch/arm/mach-imx/mach-imx6q.c a wonderful line : else if (of_machine_is_compatible("fsl,imx6dl-sabresd") ||

So if you try to use your own device tree file you won't get IOMUXC_GPR13 configured correctly !

By default IOMUXC_GPR13 has the value 0x00000008 and should be 0x0000000c for csi parallel use case ...

In your first line of your device tree you actually need to add compatible = XXX, "fsl,imx6dl-sabresd"; and then it will work ...

Hope it will help ...

[thomasquartier]

Hi,

I'm trying to interface the OV7740 in HREF mode (ov7740 camera with custom i.MX6DL board using CSI0 in HREF (gated clock) mode)).

Did you make any other changes to the ov7740 / mxc_v4l2_capture / ipu drivers?

I having the same error:

     No interrupts on IPU

     camera_callback function is never called

     cam->enc_counter is never incremented

     VIDIOC_DQBUF ioctl fails

Regards,

Thomas

[aureliencourrie]

Hi Thomas,

Did you solved your problem ?? I've the same thing with my ov7740 :-(...

Thanks

Wee Do

[thomasquartier]

Hi,

Sorry for the late response.

Our problem is solved, it was a hardware configuration on our custom board.

Because of the wrong configuration we didn't receive the pixel clock on the i.MX6.

Regards,

Thomas

[aurelienbouin]

Hello Thomas,

Are you sure you add in your dts the line

compatible = "yourcompany,imx6dl-yourboard", "fsl,imx6dl", "fsl,imx6dl-sabresd";

Because it calls something magical in arch/arm/mach-imx/mach-imx6q.c

if (of_machine_is_compatible("fsl,imx6dl-sabresd")

...

regmap_update_bits(gpr, IOMUXC_GPR13, 0x3F, 0x0C);

Which enable CS0 instead of MIPI ...

Else have you try the CSI test mode ?

With that line you should enable it :

devregs -c imx6dls 0x02630000 0x4008818; devregs -c imx6dls; devregs -c imx6dls  0x02630010 0x1000000;devregs -c imx6dls 0x02630004 0x04ff09ff

You should find the tool devregs on my github : aurelihein/devregs · GitHub

It is really usefull to set on the go registers ...

Hope it will help you

Aurelien

[thomasquartier]

Hi Aurelien,

Thanks for your reply!

The "fsl,imx6dl-sabresd" option is added to the list of compatible devices and the IOMUXC_GPR13 registers is set correct during boot.

I tried the camera in test mode and I have still the same error.

I used devregs to change registers while avconv was running (waiting for the DQBUF ioctl) but it didn't change anything.

More advice is welcome!

Regards,

Thomas

[aureliencourrie]

Hi Aurélien,

I have to use the ov7740 on custom board (with an imx6dl), so could you share your dts file you use for ov7740, please.

Thanks in adance

[aurelienbouin]

Hi,

No problem

Do not forget in the compatible line to add "imx6dl-sabresd"

Like this

compatible = "yourcompany,imx6dl-yourboard", "fsl,imx6dl", "fsl,imx6dl-sabresd";

Since the HTML use here is %/!£$# use this link : dts_ov7740.dts - Pastebin.com

Sincerely,

Aurelien BOUIN

[aureliencourrie]

Hi,

Thank for your quick answer :-).

Unfortunately, I can't try it for now, i've not received my ov7740.

I'll try when I received it and keep you in touch.

Regards

[YixingKong]

Robert

Has the issue resolved? If yes, please close this DI and other one with the same title.

Thanks,

Yixing

[robert_lim]

Hi Yixing and QiangLi,

Thanks to all the advices, I have followed the suggestions but it was still not working on my system. Instead of keeping trial and error that behind my schedule, we added Vsync to make it works so as to go to the next work. But it just a workaround so that we can start with subsequent works, the product is stick to use BT656 without Vsync/Hsync, I will have to solve this problem in the future, it would be great if the issue is still open.

Thanks,

Robert

[qiang_li-mpu_se]

Hi Robert, you can also reference to that link: https://community.freescale.com/thread/295157

In that link, we verified iMX53 CSI without VSYNC and HSYNC connected (720x480P with BT1120). Currently we have no iMX6 reference board which can support BT656 progressive input, but the iMX6 and iMX5 CSI are same.

[robert_lim]

Thanks QiangLi,

Currently I am using VSYNC and HSYNC connected for BT656. I will try this and update status here afterwards :smileyhappy:

Appreciate your help.

Regards,

Robert

[aurelienbouin]

Hello Robert,

I do have the same OV7740 sensor, and even with VSYNC and HSYNC connected I am unable to get any picture, i get the same error "mxc_v4l_dqueue timeout enc_counter 0 VIDIOC_DQBUF failed"

Could you share your source code it could really help me

Thank you for your help

Aurelien

[aurelienbouin]

hello,

I am really unable to get any data from my sensor ... I use this configuration in dts :

			MX6QDL_PAD_CSI0_DAT12__IPU1_CSI0_DATA12	0x80000000
			MX6QDL_PAD_CSI0_DAT13__IPU1_CSI0_DATA13	0x80000000
			MX6QDL_PAD_CSI0_DAT14__IPU1_CSI0_DATA14	0x80000000
			MX6QDL_PAD_CSI0_DAT15__IPU1_CSI0_DATA15	0x80000000
			MX6QDL_PAD_CSI0_DAT16__IPU1_CSI0_DATA16	0x80000000
			MX6QDL_PAD_CSI0_DAT17__IPU1_CSI0_DATA17	0x80000000
			MX6QDL_PAD_CSI0_DAT18__IPU1_CSI0_DATA18	0x80000000
			MX6QDL_PAD_CSI0_DAT19__IPU1_CSI0_DATA19	0x80000000
			MX6QDL_PAD_CSI0_DATA_EN__IPU1_CSI0_DATA_EN	0x80000000
			MX6QDL_PAD_CSI0_PIXCLK__IPU1_CSI0_PIXCLK	0x80000000
			MX6QDL_PAD_CSI0_MCLK__IPU1_CSI0_HSYNC	0x80000000
			MX6QDL_PAD_CSI0_VSYNC__IPU1_CSI0_VSYNC	0x80000000

Here is my driver for ov7740 : ov7740.c - Pastebin.com

I keep getting : ERROR: v4l2 capture: mxc_v4l_dqueue timeout enc_counter 0

Here is a log if it can help solved the problem : http://pastebin.com/328M7EMK

I checked on the pins signals everything seemsnormal ... pixclock is here

Thank you for your support,

Sincerely

Aurelien BOUIN

[pchevalier]

can you update ?

I have the same problem !

Thanks :smileyhappy:

[qiang_li-mpu_se]

Hi Robert, if there is no EAV/SAV embedded on date line, the VSYNC and HSYNC pins are needed for BT656 input.

[robert_lim]

Hi Qiang Li,

The EAV/SAV was embedded on data line, as shown below:

Thanks,

Robert