GigE vision on NI cRIO-9068

[Robst]

Hello:

We are planning a project in which we are intending to use GigE vision cameras and embedded vision processing. Our client is interested in the cRIO-9068 as the embedded platform.

The question is: Is there support for GigE Vision on the cRIO-9068?

I cannot find a document that says so, then I guess no official support is provided. If this is the case, I would like to know why.  I believe that VxWorks targets do not support GigE, but this cRIO runs Linux. From the outside, I think it would be possible to provide GigE support for Linux.  The only devices that seem to support GigE run Phar Lap ETS.

Thanks in advance for your kind reply.

[BruceAmmons]

By embedded vision processing, do you mean using the FPGA for preprocessing the image, or just running on an embedded computer?  If you want to use the FPGA for image processing, you would need to acquire the image on the RT side, then send it to the FPGA for processing, then send the results back.  In a case like this, you would be better off using a PCIe-8237R, which feeds the incoming image directly to the FPGA.  If you are planning on doing all the processing on the RT side, you just need to know if the drivers will work, which is a good question I am hoping NI will answer.

Bruce

[Robst]

Hello Bruce:

The application will be in a rough environment, in an industrial setting, that is why we want a rugged controller.  We are planning to use the FPGA as a coprocessing unit of the RT processor, so we may have to pass the image to the FPGA and get it back afterwards as you say. The GigE requirement is important, because the cameras will be widespread, and the maximum cable length with other buses is an issue.

Best regards.

[BruceAmmons]

Have you considered the EVS-1464RT?  It is a few years old, but probably still faster than a cRIO.

I made a mistake earlier.  I thought the FPGA GigE board had a full blown FPGA, but it is just for customizing the DIO.  I doubt you could do much vision processing on it.  Unless you are doing heavy duty processing, the RT system should be able to keep up.

Bruce

[Robst]

Hello Bruce:

I actually was considering the CVS-1457RT as an alternative, but, yes, the EVS-1464RT is certainly also a very good option, I can't believe I did not consider it before, thank you for the suggestion!  Let's see what the client thinks about the new alternatives.

[BruceAmmons]

I think the EVS is much more powerful than the CVS.  The CVS is based on an older design and has a fairly slow processor (single core atom), while the EVS has a dual core processor.  The CVS would be good for low speed requirements with just one, maybe two cameras.  High speed analysis will require more than the CVS can offer.

Bruce

[Robst]

Thank you for the feedback Bruce, I was thinking that the Atom processor inside the CVS was dual core.  I will update that, and provide the feedback to the team.

[BruceAmmons]

Not only single core, but slow.  The Atom is an energy efficient chip that sacrifices speed for low power usage.  It is used in some lightweight notepads, perhaps phones...  I saw on one website it got a score of 300 while current PC CPUs got about 3000.  That means it is going to run at about 1/10 the speed of your desktop or laptop.

Bruce

[BlueCheese · Accepted solution]

Just to address there original question, the 9068 does not support GigE Vision. The Ethernet MAC embedded on the Zynq chip it uses does not meet the demands of GigE Vision (similar to the older VxWorks/PPC cRIOs).

The 9068 does support USB3 Vision cameras via its USB 2.0 port though, and many GigE Vision cameras have identical models in USB variants.

Eric

[Robst]

Hello Eric:

Thank you for the information! So it seems that the cRIO-9068 will not be suitable for our application. Yes, we are aware of the fact that many vendors have an USB-equivalent of their GigE models, however, in this application, the maximum wiring lenght is an issue. We'll turn to the EVS that Bruce suggested (hopefully) or to the CVS if we are unlucky.

Best regards.