Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
13 years ago

VIP Suite Scaler II Megacore function

Hi Alterians,

My design consists of a Test patter generator followed by a Scaler ii followed by a clocked video output.

My aim is to scale down the input resolution (1280*720) to a standard VGA resolution (640*480) to test the capabilities of the core.

All my cores are configured on real time with NiosII commands.

I am having hard time outputting the correct test pattern onto my std screen.

The thing is that it works perfectly when the resolutions are closer to each other. Such as from SVGA 800*600 to 640*480.

The problem as it seems is related to stall behavior as specified in ScalerII chapter in VIP Suite Handbook. The output stalls for the time needed to read one or more input lines. Which makes me think to increase the clock between the Test Pattern Gen and the Scaler ii. But SOPC doesn't allow that, and I believe it is a better practice to drive the video datapath component with the same clock, but feed the Clocked output by the clock standard I willing to use.

Does anybody have other suggestions of what might be the problem and how to solve it?

PS: I have checked some Altera video reference designs and there is always a Frame Buffer placed right before the CVO. Does it help? It seems only for rate conversion.

1 Reply

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    13 years ago

    The framebuffer in the example designs is to even out the stalls in the video pipeline that are likely caused by the scalar.

    If your video pipeline looks like this:

    pattern gen -> scalar -> frame buffer -> clocked video output

    Then the scalar can stall as much as it wants and not cause the clocked video output to starve

    because it can keep reading from the frame buffer the same frame over and over until the scalar

    finishes.

    What you might try to do if you cannot afford to use a frame buffer is to increase the clock speed

    that your scalar is running at so that the overall amount of stall time in the pipeline is reduced

    and you don't starve the CVO. You could also add some large fifos to help smooth out

    the stalls in the pipeline so that CVO doesn't starve.