IIR LPF filter coefficients and test

[iHapenciuc]

Hi,

I'm a big fan of LabVIEW.

I am having an issue now. I am trying to implement an IIR LPF on a 16bit stucture.

To get the coeficients I am using the "IIR filter design v1.vi" with the posibility to compute the coeficients.

To test the IIR structure I have added in the enumerator up left the "TEST" tab wich should create my desired IIR LPF using the previously computet coeficients using "Inverse Chebashev".

But the result on the "TEST" is way different from what I want and is indicated on "Inv Chebashev". Values I want are default in the vi.

Why?

Please help!

Andrei

[johnsold]

Andrei,

1. The Inverse Chebyshev Filter.vi uses the IIR Cascade Filter.vi, not the IIR Filter.vi which you used. The Cascade Filter.vi produces results closer to what you expect, but still not exactly the same.

2. The coefficients you apply to the Test case are not identical to the coefficients generated by the Inv Chebyshev Coefficients .vi. Change the display to show 12 digits of precision. The first Reverse Coefficient is -0.991763327393 while the first value sent to Test is -0.991800000000.

3. Integer arithmetic for filters is a specialized field because of the limited precision, rollover errors, and roundoff errors. I assume this what you are referring to when you mentioned a 16-bit structure. How should you scale and represent -0.991763327393 on a 16-bit system? Check with your NI rep about the FPGA devices. I think they may have some filter designs for integer math.

Lynn

[iHapenciuc]

I like a lot this forum because I always get straight answers to my questions.

Thank you a lot Johnsold. I have fix the issue. I am aware about the effect of truncation caused by the limited representation. I use dsPIC33 and all maths is fractional. Yet I will have to play around with the coefficients to get the response function I need.

Thank you very much

Andrei