Help with multi-taps delay

[jetset]

I've designed a multi taps delay using a non-floating point DSProcessor.
My question is, can I implement feedback gain of the taps using just integers operations? You know feedback gains should be less than 1
(for example 0.5, 0.6, 0.7....) but I don't know how to do this just using integer operations.

 

[dora]

Hi jetset,

I am not expert in the DSP implementation area but will share with you what i know.
The idea behind the integer representation is to not have floating point calculation which are very slow on the
fixed point DSP (done by software library). So you have to transfer the float calculation with integer. Solution
Multiply your filter coefficients by 2^n (it is left shift by n steps), scale your float input data also (lat say *2^m) to the
integer one. Then take the integer part.
Note that in the above conversion some round off error is introduced so n and m should be selected carefully. then perform
the calculation using int calculation and then rescale by the /2^(m+n) - right shift. Note that this shiftings sometimes should
be done on each input sample (for IIR filters) but can done only ones at the beginning and at the end of filter calculation for the
FIR filters.

Good luck!
dora

 

[brmadhukar]

Hi,
In a fixed point processor you have enough instructions to calculate the FIR filter. You look at the the instruction set and architetcture carefully. If you are using ADSP PM me.
B R M

 

[jetset]

ok Dora you gave me an idea,,, thanks