New to DSP (FIR implementation help)

Hi guys/gals

I'm new to DSP pogramming but am familiar to the high level concepts.  I 
know how the output of a FIR filter is computed and how to generate the 
coeficients and tap count.

I'm using ScopeFIR and have generated a 100 tap low pass filter.  I have 
"quantized the coefficents to 16 bits because I'm planning to use this on 
the DsPic 30F6014.  The 16bit A2D I'll be sampling outputs and outputting 0 
for -10V input and FFFF for +10V


The DsPic has C librarys that call for the coefficients to be in fraction 
format.  I assume the A2D sample also need to be in a compatable format as 
well.

What type of Coefficent export type out of ScopeFIR do I need to get the 
coefficents in the appropriate form for the DsPic C functions to be happy?

Do I need to do some kind of transform to the A2D samples to get them into 
Fraction form as well?

The output will be sent to a matching 16 bit DAQ (same +/- 10V) range and 
would like it to track the DC input to the A2D and reject the AC noise.  AA 
filter is in place. :)

thanks

Mook Johnson wrote:
> Hi guys/gals
> 
> I'm new to DSP pogramming but am familiar to the high level concepts.  I 
> know how the output of a FIR filter is computed and how to generate the 
> coeficients and tap count.
> 
> I'm using ScopeFIR and have generated a 100 tap low pass filter.  I have 
> "quantized the coefficents to 16 bits because I'm planning to use this on 
> the DsPic 30F6014.  The 16bit A2D I'll be sampling outputs and outputting 0 
> for -10V input and FFFF for +10V
> 
> 
> The DsPic has C librarys that call for the coefficients to be in fraction 
> format.  I assume the A2D sample also need to be in a compatable format as 
> well.
> 
> What type of Coefficent export type out of ScopeFIR do I need to get the 
> coefficents in the appropriate form for the DsPic C functions to be happy?
> 
> Do I need to do some kind of transform to the A2D samples to get them into 
> Fraction form as well?
> 
> The output will be sent to a matching 16 bit DAQ (same +/- 10V) range and 
> would like it to track the DC input to the A2D and reject the AC noise.  AA 
> filter is in place. :)
> 
> thanks
> 
> 
You will have to check your documentation from Microchip, but it is most 
likely that the fraction format is just a 2's complement number that is 
interpreted as a fraction -- i.e. instead of the LSB having a bit weight 
of 1 and the MSB (sign bit of a 16-bit number) having a bit weight of 
32768 a fractional number that ranges from -1 (0x8000) to almost +1 
(0xffff) will have a LSB bit weight of 1/32768 and a MSB (sign) bit 
weight of 1.

Assuming this is correct then your ADC will have to have it's MSB 
reversed so -10V results in 0x8000 and +10V (well, 9.9997V) will result 
in 0x7fff, with 0V (0x8000 out of the ADC) will be 0x0000.

I'm familiar with neither the ScopeFIR or the DsPIC, but I suspect that 
you just need to export the FIR coefficients as a vector of integers 
that range from -32768 to 32767 and make them into a C array.  You 
should be able to just tell the DsPIC to interpret them as fractional.

Assuming that the DAC format matches the ADC then you'll have to do the 
same sign bit inversion on the DAC as you did on the ADC.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

>Hi guys/gals
>
>I'm new to DSP pogramming but am familiar to the high level concepts.  I

>know how the output of a FIR filter is computed and how to generate the 
>coeficients and tap count.
>
>I'm using ScopeFIR and have generated a 100 tap low pass filter.  I have

>"quantized the coefficents to 16 bits because I'm planning to use this on

>the DsPic 30F6014.  The 16bit A2D I'll be sampling outputs and outputting
0 
>for -10V input and FFFF for +10V
>
>
>The DsPic has C librarys that call for the coefficients to be in fraction

>format.  I assume the A2D sample also need to be in a compatable format
as 
>well.
>
>What type of Coefficent export type out of ScopeFIR do I need to get the

>coefficents in the appropriate form for the DsPic C functions to be
happy?
>
>Do I need to do some kind of transform to the A2D samples to get them
into 
>Fraction form as well?
>
>The output will be sent to a matching 16 bit DAQ (same +/- 10V) range and

>would like it to track the DC input to the A2D and reject the AC noise. 
AA 
>filter is in place. :)
>
>thanks
>
>
>

ok here is what you do:
I created a FIR as example I used a tool calleddsPIC FD lite for it:

create this as .s  file


;
..............................................................................
;    File   fir.s
;
..............................................................................

		.equ firNumTaps, 49

;
..............................................................................
; Allocate and initialize filter taps

		.section .xdata,data,xmemory // see below
		.align 128
//this are the taps change as required

firTaps:
hword
	0x01F9,	0x020C,	0x021E,	0x0230,	0x0241,	0x0252,	0x0262,	0x0271,	0x0280
hword
	0x028E,	0x029C,	0x02A8,	0x02B4,	0x02BF,	0x02C9,	0x02D2,	0x02DA,	0x02E2
hword
	0x02E8,	0x02ED,	0x02F2,	0x02F5,	0x02F8,	0x02F9,	0x02FA,	0x02F9,	0x02F8
hword
	0x02F5,	0x02F2,	0x02ED,	0x02E8,	0x02E2,	0x02DA,	0x02D2,	0x02C9,	0x02BF
hword
	0x02B4,	0x02A8,	0x029C,	0x028E,	0x0280,	0x0271,	0x0262,	0x0252,	0x0241
hword 	0x0230,	0x021E,	0x020C,	0x01F9

;
..............................................................................
; Allocate delay line in (uninitialized) Y data space

		.section .ybss,bss,ymemory // see below
		.align 128

firDelay:
		.space firNumTaps*2

;
..............................................................................
; Allocate and intialize filter structure

		.section .data
		.global _firFilter

_firFilter:
hword firNumTaps
hword firTaps
hword firTaps+firNumTaps*2-1
hword 0xff00
hword firDelay
hword firDelay+firNumTaps*2-1
hword firDelay



If you use the FD lite the bit I marked see below will be wrong change it
as IK have done in here

you also need an .h file


#ifndef FIR_H
#define FIR_H

extern FIRFilterStructure firFilter;

#endif /* FIR_H */

/* The following C-code fragment demonstrates how to call the filter
routine
#include "FIR_Filter.h"
#include "fir.h"

// NUM_SAMPLES defines the number of samples in one block of input data.
// This value should be changed as needed for the application
#define NUM_SAMPLES 100

{
   // Declare input and output sample arrays.
   int  inSamples[NUM_SAMPLES], outSamples[NUM_SAMPLES];


   // Call the FIRFilterInit routine to zero out the delay line
   FIRFilterInit( &firFilter );

   // Call BlockFIRFilter for each block of input samples
   // This routine would normally be called inside a FOR or a DO-WHILE
loop
   // Only one instance has been shown
   BlockFIRFilter( &firFilter, &inSamples, &outSamples, NUM_SAMPLES );
}
*/

which is also created by FD lite and it actually does explain the rest you
have to do --- note that I called the filter "fir" (maybe not a good choice
as it confuses things) so replace that with the name of your filter.

That should do the trick




		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com