Reply by Vladimir Vassilevsky ●October 3, 20032003-10-03

Curl wrote:

> 
> Hello all !
> 
> Considering IIR filtering with fixed-point DSP :
> How to arrange biquads to reduce roundoff or quantization noise ?
> 
> If anyone has resources, advices or links !


I would recommend a book:

Dietrich Schlichtharle. Digital Filters: Basics and Design. 
Springer ISBN 3-540-66841-1

Basically, the rule is to combine poles with zeroes and move the
sections with high Q to the front so the noise of high Q sections will
be filtered out by low Q sections.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com
Reply by Curl ●October 2, 20032003-10-02
"Curl" <Mr.Bilou@microsoft.fr> a &#4294967295;crit

| Rule No. 1 already respected ;o)

oops.. Too fast
The exact aswer is : I'm using a cascade form of 2nd order Form 1 IIR
filter
The poles of a biquad are combined whith the zeros of the next
biquads.

Thank you for your answers, I'm currently looking at SPRA454 (Extended
precision IIR filter Design on the TMS320C54x DSP).

Since I have 5 biquads, I think i will try different arrangements (may
be the 5!) and see the difference with a spectrum analyzer
Reply by Jon Harris ●October 1, 20032003-10-01
If you still have trouble with quanitization with DF1, then you can try some
of the lattice/ladder implementations.  Is this fixed or floating point?
What kind of numerical precision do you have (e.g. 16, 24, 32-bit)?

"Curl" <Mr.Bilou@microsoft.fr> wrote in message
news:3f7a9a73$0$28876$626a54ce@news.free.fr...

> | Al Clark a ecrit :
> | Rule No. 1:
> |  Use Direct Form 1, not Direct Form 2.
> |
> Rule No. 1 already respected ;o)
>
> What is the No.2 ?
>
Reply by Bernhard Holzmayer ●October 1, 20032003-10-01
Curl wrote:


> Hello all !
> 
> Considering IIR filtering with fixed-point DSP :
> How to arrange biquads to reduce roundoff or quantization noise ?
> 
> If anyone has resources, advices or links !
> 
> Thank you.


Investigate the Mathworks web site. 
Using their search dialog, ask for 'ss2sos'.
The first entry should lead you to the description of this 
conversion method.
It uses 'order' and 'scale'. 
Following the links should lead you to the principles behind.
If you use Matlab, signal processing toolbox is your friend.

Bernhard
Reply by Curl ●October 1, 20032003-10-01
| Al Clark a ecrit :
| Rule No. 1:
|  Use Direct Form 1, not Direct Form 2.
|
Rule No. 1 already respected ;o)

What is the No.2 ?
Reply by Al Clark ●September 30, 20032003-09-30
"Curl" <Mr.Bilou@microsoft.fr> wrote in news:3f79a987$0$28905
$626a54ce@news.free.fr:


> Hello all !
> 
> Considering IIR filtering with fixed-point DSP :
> How to arrange biquads to reduce roundoff or quantization noise ?
> 
> If anyone has resources, advices or links !
> 
> Thank you.
> 


Rule No. 1:

Use Direct Form 1, not Direct Form 2.


-- 
Al Clark
Danville Signal Processing, Inc.
--------------------------------------------------------------------
Purveyors of Fine DSP Hardware and other Cool Stuff
Available at http://www.danvillesignal.com
Reply by Curl ●September 30, 20032003-09-30
Hello all !

Considering IIR filtering with fixed-point DSP :
How to arrange biquads to reduce roundoff or quantization noise ?

If anyone has resources, advices or links !

Thank you.