Forum Discussion

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

FIR II Compiler V10.1 Coefficient Scaling

Hye all,

I'm implementing a 127-taps FIR for baseband shaping.

I'm using the FIRII compiler for it.

I noticed that in the 'old' FIR compiler you had the option to scale the coefficients automatically with a power of 2.

In FIRII there is only one scaling:

Y = ((2^bits-1)/max(|coefvalue|))X

I like to use the FIRII because of the coefficient reload interface it generates, but I also like either no scaling or scaling by power of two, so at the output I can easily compensate for gain.

Any idea if this is possible with FIRII and how?

Thanks!

Grtz,

Olaf

3 Replies

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

    It is really important:

    1) scaling, when coefficients reload ;

    2)I make Hilbert transformer for envelope detection: real part - as shift register, image part - as bandpass filter (FIR II 10.1). Both parts must have equal scale, but FIR II don't asks me!:cry:
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    14 years ago

    To get dc unity gain, you scale your coeffs as follows:

    coeff = round(2^n/sum(coeffs));

    in your case, you easily work around by rescaling your normalised coeffs so that your max coeff = sum of coeffs.

    what remains is the difference of 2^n & 2^n - 1 which is trivial.