Machine Vision

cancel
Showing results for 
Search instead for 
Did you mean: 

Benchmark Speed of CVS-1456

I am using a CVS-1456 with VBAI. While benchmarking the application, my toolset time is approximately 20ms.

The overall inspection time is between 35 - 40 ms and framerates are <25fps.  What is the extra time being used for? Can I reduce overhead? 

I would like to get to 30fps.

 

The camera is set up for 60fps and is working very well. Acquisition is <2ms.

 

Any ideas?

 

 

Thanks in advance for your efforts.

 

Rich

0 Kudos
Message 1 of 7
(4,135 Views)

Can you provide your inspection? Or at least a text file with the timing details of the steps used and the amount of time for steps? If you don't want to post the inspection on the public forum, you can send the inspection to

vbai.support@gmail.com

 

What version of VBAI are you using?

Thanks,

Brad

0 Kudos
Message 2 of 7
(4,122 Views)

This portion of the issue has been cleared up.

I formatted and re-installed the system software (Ver. 2009) into the CVS.

I have configured Cam0 and Cam1 for 640 x 480 mono8, 30fps. Each camera has inspection tasks at different angles on the part being inspected. First camera 0 inspects while the part moves into alignment. Then, the second camera, Cam1, inspects.

Everything is running at 30fps without trouble...while it is running.

When the system comes out of VBAI program mode into the run mode, or when entering or leaving the Inspection Interface, the first and second inspection states both acquire from Cam0. I never mess with the camera acquisition parameters, so I am not doing anything obvious to cause this.

Then, a reboot is required to correct the camera acquisitions.

Has anyone seen this before?

 

 

Sincerely,

 

Rich

0 Kudos
Message 3 of 7
(4,084 Views)

Vision Builder AI 2009 SP1 should fix this problem. It should be available on the web by the end of the week or early next week. The readme says the following:

 

ID    Fix Issue:

187595Acquire Image (IEEE 1394) steps select the default camera in an inspection that uses multiple Acquire Image (IEEE 1394) steps to acquire images from multiple cameras connected to a remote target after saving and reloading the inspection or when editing each Acquire Image (IEEE 1394) step.

Let me know if this is still a problem after installing Vision Builder AI 2009 SP1 and updating the target with SP1.

 

Thanks,

Brad

0 Kudos
Message 4 of 7
(4,074 Views)

Thanks Brad.

I will download the fix as it becomes available.

One more thing....there is always one more thing.

The application has morphed slightly in that the customer requirement will call for a 60fps camera and a 30fps camera. The firewire bandwidth is there for it. That isn't the issue.

 

The problem is that the system ran at 60 fps for a while and then dropped to 34-40fps. The cause (shown in the benchmark utility) is that the display time jumped from <5ms to over 16ms. I don't recall adjusting it. Is there any way to optimize (shorten) the display time?

 

Thanks.

 

Rich

0 Kudos
Message 5 of 7
(4,027 Views)

See if one of your cameras changed from 8-bit to something 10-bit or something with more data. You can also go to the menu option Target>>Target Options>>Video Mode and change the color quality to something lower (this will also improve the amount of memory available). Make sure the screen resolution is also not bigger than the camera so you're not spending time displaying extra pixels you don't need.

 

Are the two cameras the same resolution? Are you switching back and forth between displaying images of different resolution? Just wondering if that might also contribute to slower display times.

 

Hope this helps,

Brad

0 Kudos
Message 6 of 7
(4,023 Views)

Brad,

 

I will check the items you pointed out.  

To answer your question, yes, the cameras are set differently and they switch from camera0 to camera1.

The first camera is 640 x 480, mono8, 30fps. The second is 640 x 480, mono8, 60fps but may be switched to Format 7, 400 x 200 if it works faster.

 

Rich

0 Kudos
Message 7 of 7
(4,005 Views)