AGX Orin FPD-Link™ III Interface Card 16x Support

[d4l3k]

Hi, it doesn't appear that this repo has support for the 16x FPDLink III Orin AGX card. Where can I find an example dts/dtb for it?

Relevant page: https://www.d3engineering.com/product/designcore-nvidia-jetson-agx-orin-fpd-link-iii-interface-card/

Thanks!

cc @D3-growe, @D3-jwatts @D3-aavery

[D3-jkiggins]

Hi,

We haven't made support for that hardware configuration public due to an overhaul of our code-base internally. You should have received a reply to your support email with the info you're looking for.

Let me know if that works as a solution here, thanks!

[d4l3k]

@jkiggins-d3 thanks for sharing it! Just downloaded it, very appreciated

[d4l3k]

@jkiggins-d3 I tried that and am running into the same issue as my homebrewed dtbs.

It appears that the i2c communication with pcal953@70 and ub953 has issues. When I run i2cdetect there's no response at i2c address 0x70.

[   32.335533] ub960 2-0030: csi speed:1472-1664 lanes:4 continuous-clock:yes
[   32.343099] ub960 2-0030: could not find regulator: avdd
[   32.357274] ub960 2-0030: revision mask id: 0x04
[   32.362039] ub960 2-0030: detected: DS90UB960-Q1 A0
[   32.585713] ub960 2-0032: csi speed:1472-1664 lanes:4 continuous-clock:yes
[   32.593088] ub960 2-0032: could not find regulator: avdd
[   32.607234] ub960 2-0032: revision mask id: 0x04
[   32.611995] ub960 2-0032: detected: DS90UB960-Q1 A0
[   32.848807] ub960 2-0034: csi speed:1472-1664 lanes:4 continuous-clock:yes
[   32.856272] ub960 2-0034: could not find regulator: avdd
[   32.870453] ub960 2-0034: revision mask id: 0x04
[   32.875212] ub960 2-0034: detected: DS90UB960-Q1 A0
[   33.112470] ub960 2-0036: csi speed:1472-1664 lanes:4 continuous-clock:yes
[   33.121745] ub960 2-0036: could not find regulator: avdd
[   33.140932] ub960 2-0036: revision mask id: 0x04
[   33.145700] ub960 2-0036: detected: DS90UB960-Q1 A0
[   34.212804] ub960 2-0032: [2] Found child node: ub953@18
[   34.338045] tegra-i2c 3180000.i2c: I2C transfer timed out
[   34.344015] ub953: probe of 2-0046 failed with error -110
[   34.355546] ub960 2-0032: [2] Found child node: pcal953@70
[   34.462913] tegra-i2c 3180000.i2c: I2C transfer timed out
[   34.485421] ub960 2-0032: [2] Failed to create i2c device for pcal953@70: -110
...
[   44.119307] ub960 2-0032: WARNING: Only detected activity on 1 ports. Are all sensors connected?

Output from i2cdetect:

rice@orin:~$ sudo i2cdetect -l
i2c-3	i2c       	3190000.i2c                     	I2C adapter
i2c-1	i2c       	c240000.i2c                     	I2C adapter
i2c-8	i2c       	31e0000.i2c                     	I2C adapter
i2c-6	i2c       	31c0000.i2c                     	I2C adapter
i2c-4	i2c       	Tegra BPMP I2C adapter          	I2C adapter
i2c-2	i2c       	3180000.i2c                     	I2C adapter
i2c-0	i2c       	3160000.i2c                     	I2C adapter
i2c-7	i2c       	c250000.i2c                     	I2C adapter
i2c-5	i2c       	31b0000.i2c                     	I2C adapter

rice@orin:~$ sudo i2cdetect -y -r 2
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          -- -- -- -- -- -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- 18 -- -- -- -- -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: UU -- UU -- UU -- UU -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- -- -- 48 -- -- -- -- -- -- --
50: UU UU -- -- -- -- -- -- -- -- -- -- -- -- -- --
60: -- -- -- -- -- -- -- -- 68 -- -- -- -- -- -- --
70: -- -- -- -- UU UU UU --

[d4l3k]

@jkiggins-d3 is there a technical doc with all the i2c addresses for the orin board?

[d4l3k]

Ok, I've figured out that these are expected errors as there's no i2c bus or eeprom on my camera modules. The serializer is also ub913 not ub953 so that explains the serializer errors

[D3-jkiggins]

@d4l3k, it sounds like you've figured things out in the short term. In-case this is still helpful, here are the I2C addresses for devices on D3's FPD III interface board.

I2C ADDRESS TABLE - I2C_XAVIER_SCL/SDA

DEVICE	ADDRESS(BIN/HEX)
24AA1026	0b101000X/0x50 and 0x51
DS90UB960	0b0110000/0x30
DS90UB960	0b0110010/0x32
DS90UB960	0b0110100/0x34
DS90UB960	0b0110110/0x36
PCAL9539A	0b1110100/0x74
PCAL9539A	0b1110101/0x75
PCAL9539A	0b1110110/0x76

I2C ADDRESS TABLE - I2C_XAVIER_GP2_SCL/SDA

DEVICE	ADDRESS(BIN/HEX)
SC18IS602B	0b0101000/0x28
PCAL9539A	0b1110111/0x77

[d4l3k]

@jkiggins-d3 thanks! Do you have a list of all the test pads? The nvidia csi driver is throwing errors about line lengths being incorrect (either too long or too low) and I want to probe the test pad with a logic analyzer and oscilloscope to get the raw CSI lines to help debug further.

There doesn't seem to be any way to get the closed source nvidia drivers to write out the decoded csi data so I'm having issues debugging further. Are the CSI lines exposed anywhere that's accessible?

[D3-jkiggins]

It sounds like you already have trace turned on, but just in case, there are some good instructions here: https://elinux.org/Jetson/l4t/Camera_BringUp.

You can also look at some of the status registers on the UB960/953. There is a LINE_LEN field for example, which reports the line length observed over FPDLink.

There doesn't seem to be any way to get the closed source nvidia drivers to write out the decoded csi data

This has been our experience as well.

I'll reply to your support email with the CSI debug pad details.

[d4l3k]

@jkiggins-d3 I've been looking at the traces (and decoding the error messages) but not completely sure what's wrong. I suspect there's some instability or mismatched clock speeds causing issues. I've checked the status registers on the UB960/913 and everything seems to be stable and the error bits stay cleared once I read them.

One weird thing is that the line length seems to be stable when read but when rebooting the LINE_LEN register switches between a couple of different values. It's very strange since the line length change register isn't set so it means that it's stable during running but on reboot it changes to a different value and sticks?

Makes me wonder if there's something odd about the clock frequencies -- but it is weird that there's no errors reported from ub960/913 indicating any instability.

Appreciate your help with the debug pad details, thanks!