[BUG REPORT] High order shelving filters have a weird dB curve.



The Holters filter implementation seem to have a bit of a problem. Here are two similar filters, one is set to +5dB, the other to -5dB. You can see on the probed graph that they do not boost and cut in a symmetric way.

Where it gets very weird is when you actually notice that the dB boost/cut is spot on when you set the cursor to the max independently of what the max value is.
Here are some examples with the cursor set at max opposite.

Here's another example showing the difference when the cursor is not at the max.



Would be very nice to get this fixed or to have a workaround. Cheers.



edit: Changed the title to BUG REPORT
[edited by: CopperMaze at 11:56 AM (GMT -4) on 19 Sep 2020]
  • Also, it is looking like the Band shelves doesn't have this problem.

  • Hello,

    It's been one month. Did someone tried to replicate? I would be glad to be wrong, really.

    •  Analog Employees 
    on Oct 27, 2020 7:42 PM 1 month ago in reply to CopperMaze

    Hello CopperMaze,

    Things have been rather busy so I have fallen behind on my EngineerZone support. You have not been singled out. 

    I checked our bug reports and I had reported a bug with this cell a while back and it was regarding having a negative input to the control port. I do not think it was this exact issue. My bug report was worked on and it has been fixed but it has not been released yet. It is due for the Rev 4.6 release. 

    Can you upload a project that shows the issue you are seeing to save me some time? I would like to send it to the programmer to see if he fixed this issue when he fixed the issue I complained about. There is a good chance that both bugs were related since they both revolve around negative gain settings. 

    Thanks,

    Dave T

  • Hi Dave, thanks for the answer.

    I've joined a project for the ADAU1452 showing the issue.

    The two bugs may be related, but the problem I'm seeing here is affecting both positive and negative boosts. I'm thinking about a lookup table that may not have exactly the right curve or is too coarse.
    The band shelves seem to be less affected by this but they are not perfect neither, especially in frequencies under their FL parameter where they seem to apply some slight gains. I didn't tested them as thoroughly as the shelves so I cannot tell really, but it wouldn't surprise me if they were suffering from the same problem.

    TEST_ADAU1452_HighOrderShelvingFilters_v0.1.dspproj.zip

    Thanks,