No IR Image, and broken Depth Image on Four SR300

[LBill]

Folks,

I left four of my realsense SR300 running for 16 hours, only streaming Depth images, and nothing else. Now I am checking the images, and my Depth image on all four sensors are messed up. When I run the cpp-capture.cpp example, I cannot see my IR image, and the Depth is weird. This is happening on all four sensors equally, as I just test each one individually, turned off, and then back on (full shutdown) my computer, and unplugged, and plugged back in each sensor. Why is this happening? I bought the sensors this week, and started using them since Tuesday, February 21st, 2017 (two days ago). See image Below (IR is completely dark, and Depth has that weird pink color. Only RGB is image is working well on all four sensors):

[MartyG]

I have not come across this problem before so it is hard to give a reliable answer. When you mentioned the pink image though, it reminded me of an issue in LED television sets where the picture can go pink if the LED screen's backlight fails.

The RealSense cameras have an IR emitter component for the IR camera that helps them with lighting exposure. It makes me wonder if the image may be being pinkified because the IR emitters in the cameras have been affected. If this is the case (and it's a big if, because I have nothing much to back up my speculation with), the cameras should still be under warranty as you've only just bought them, and returnable for replacements. Contact mailto:click.support@intel.com click.support@intel.com

[idata]

Hello Pototo,

I was wondering if you decided to go on with the replacement process through Click Support (mailto:click.support@intel.com click.support@intel.com).

If you have any other question, or need any type of assistance, don't hesitate to contact us.

Regards,

Andres V.

[LBill]

It might be a computer issue. I tired one sensor on a new computer, with same software, and it worked. I still need to try all sensors at the same time, though. If it does not work today, then we'll return the sensors.

[LBill]

Do you know what kind of issue might be happening with my USB 3.0 ports? Maybe not enough power (even though I use a powered USB 3.0 hub for all my USB sensors)?

[MartyG]

I am wondering, did you have all 4 cameras attached to the same computer? If you are using cpp-capture.cpp, I guess you are using the Librealsense SDK?

I mention this because whilst the Librealsense SDK supports multiple cameras, the official RealSense SDK does not - although you can have multiple cams attached, only one cam will work in the official SDK. So a multi-cam setup on the same machine would only work with Librealsense.

[LBill]

I see the problem. I had to redo the Video4Linux backend preparation to apply the kernel patches. And that solved my problem. I wonder why, though, is this not a persistent thing? I did this step the very first time, which is why the depth cameras worked, but this went away somehow. That is scary. I'll let the cameras run continuously for the next three days to see if this happens again.

[MartyG]

I remember there was a person in 2014 who wanted to run his Real sense app for 3 days but it crashed after about 16 hours because the PC ran out of memory (a couple of the SDK depth instructions have memory leaks unless you remember to Release() memory at the end of the script).

[LBill]

That makes sense. I just used the multicam example, which does not release the memory (but I thought it was using smart pointers or another method behind the scenes). But I saw another example that has memory release at the end. I guess I'll use that last example instead.

Thanks

[LBill]

By the way, using the latest librealsense I managed to run one SR300 on Linux for 96 continuous hours (4 days). That's pretty good progress. But now, when I run three realsense, they only run for 1.5 hours before one of them gets a frame timeout (wait_for_frame()). Much better than the time when I asked this question, though.

[MartyG]

Thanks for the update! It's funny you mention wait_for_frame, as I answered a question involving that only yesterday. A couple of people have suggested in the past that when they had wait_for_frame timeouts, they got better stability if using the camera with a mains-powered USB hub.

[LBill]

I used this industrial USB 3.0 hub, and now I ran multiple realsense cameras for 15 hours. I guess USB ports performance matters with this crashing issue:

https://www.baslerweb.com/en/products/vision-components/additional-accessories/usb-3-0-hub-4-ports-side-by-side/ USB 3.0 Hub, 4 Ports side by side - Cards, Hubs, Switches

[MartyG]

Thanks for the update! I'm glad a hub gave you the stability you needed.