Forum Discussion

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

IO discrepancy between CIC 10.1 and 13.1

I'm upgrading IP and I see an IO mismatch between the output data widths of the CIC core generated in 10.1 versus the one from 13.1. The documentation for both shows the same formula for calculating output width:

Bin + Nlog2(RM) - log2(R)

For me, this comes out to 45.68. The 10.1 design generates 47 bits and the 13.1 design generates 46 bits. Although the 13.1 design seems more accurate, I've already verified the design with the 10.1 core. I've generated other cores and have noticed that the 10.1 compiler consistently adds 1 bit to the output. Can anyone tell me if I need to drop a bit or shift a bit to match the previous behavior? Or did the algorithm get updated, and I can no longer expect a bit accurate match to my previous design?

Alex

3 Replies

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

    could it be the bitwidth you are referring to depends on cic being decimator or interpolator.

    The equation you have given is for interpolator. For decimator you don't subtract log2(R)
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    11 years ago

    --- Quote Start ---

    The design is interpolator for BOTH the 10.1 core and 13.1 core.

    --- Quote End ---

    in that case 46 bits represents the maximum bitwidth required at last integrator stage. any extra bit can be ignored and should not affect the result as you will truncate LSBs

    to get your final gain.