05-14-2014 04:54 PM
I am using xilinx image enhancement v8.0 in my porjects. Everything is OK except setting register 0x0020 via AXI interface.
The register is needed to be set when input video resolution changed. It seems the active h size value can't be set.
I found xilinx has a solution about this issues(AR# 60173):
This is a known issue with the Image Enhancement core v8.0 (Rev. 1), v8.0 (Rev. 2), v8.0 (Rev. 3).
The user can work around this issue by asserting the sclr input pin after changing the columns.
This issue will be resolved in a future release of the Image Enhancement core.
My question is where "the sclr input pin" is ?
Can I have a work around solution ?
05-15-2014 02:31 PM
I tried to pull "aresetn" low for 16ms after setting Reg 0x0020. The video totally lost.
Can I have a detail timing description ?
05-20-2014 05:16 PM
I think I have the same issue here.
I use the Enhance 8.0 core to run a 1-shot image processing.
I input say 256 lines of say 256 pixels with the same "counting value" (0 to 255).
The input of the CORE is correct (successive lines with TLAST asserted and same value for the entire line).
The output of the CORE actually pushes out the last 16 lines (i.e: 240, 241, etc...) ahead of the new lines, resulting in an image shift.
I suspect that this might be because I am using the AXI Lite port to configure the ACTIVE_SIZE register.
I also do not know what is meant by sclr. Doing a soft reset (via CONTROL register) or a hard reset, or disable, all result in the Core loosing its configuration.
Any help or a fix will be welcome.
05-21-2014 10:37 PM
Can you please help clarify which aresetn you are talking about?
Should it be S_AXI_ARESETn of the AXI Lite configuration interface, or the ARESETn input reset signal?
I think doing a ARESETn just resets all programmed values to default so I don't think that is the solution?
There is definitely an error with the enhance core whenever changing HSIZE to a value that's not the default Core value.
Some guidance on this will be greatly appreciated,
05-23-2014 03:47 PM
Hello... me again.
The fix proposed doesn't work. Asserting ARESET_N will reset the core (hence lose the ACTIVE_SIZE info), and reprogramming the line size with a value different from the default will always result in odd filtering results.
It will be great if that gets resolved soon,
05-28-2014 02:25 PM
You might want to try delaying the reset until after the TUSER_SOF following the update to the registers. This is because the registers are double buffered and do not actually take affect until after an TUSER_SOF.