From: Vladimir Vassilevsky on

The presentation of Rick Lyons "Improving FIR Filter
Coefficient Precision" was added:

http://www.compdsp.com/presentations/Lyons/Lyons-Precise%20FIR%20Coefficients.pdf

COMP.DSP 2010 presentations:

http://www.compdsp.com/presentations.htm


//-----------------

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
From: Manny on
On Jun 28, 8:16 pm, Vladimir Vassilevsky <nos...(a)nowhere.com> wrote:
>   The presentation of Rick Lyons "Improving FIR Filter
> Coefficient Precision" was added:
>
> http://www.compdsp.com/presentations/Lyons/Lyons-Precise%20FIR%20Coef...
>
> COMP.DSP 2010 presentations:
>
> http://www.compdsp.com/presentations.htm
>
> //-----------------
>
> Vladimir Vassilevsky
> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com

Not to take away anything from the author, but as far as I can tell,
people have been doing this for ages. In fact, the serial method could
easily implement a parallel method with a device known in the digital
world as "barrel shifter." In such a case, you'd store an unsigned
value with the coefficient akin to an "instruction" in ISA to tell you
how much you should barrel-shift per loop iteration.

Regards,
-Momo
From: steveu on
>On Jun 28, 8:16=A0pm, Vladimir Vassilevsky <nos...(a)nowhere.com> wrote:
>> =A0 The presentation of Rick Lyons "Improving FIR Filter
>> Coefficient Precision" was added:
>>
>>
http://www.compdsp.com/presentations/Lyons/Lyons-Precise%20FIR%20Coef...
>>
>> COMP.DSP 2010 presentations:
>>
>> http://www.compdsp.com/presentations.htm
>>
>> //-----------------
>>
>> Vladimir Vassilevsky
>> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com
>
>Not to take away anything from the author, but as far as I can tell,
>people have been doing this for ages. In fact, the serial method could
>easily implement a parallel method with a device known in the digital
>world as "barrel shifter." In such a case, you'd store an unsigned
>value with the coefficient akin to an "instruction" in ISA to tell you
>how much you should barrel-shift per loop iteration.
>
>Regards,
>-Momo

Yep. Looks like another example of relearning a forgotten art of doing
things efficiently. :-)

Steve

From: Manny on
On Jun 29, 1:58 am, "steveu" <steveu(a)n_o_s_p_a_m.coppice.org> wrote:
> >On Jun 28, 8:16=A0pm, Vladimir Vassilevsky <nos...(a)nowhere.com> wrote:
> >> =A0 The presentation of Rick Lyons "Improving FIR Filter
> >> Coefficient Precision" was added:
>
> http://www.compdsp.com/presentations/Lyons/Lyons-Precise%20FIR%20Coef...
>
>
>
>
>
> >> COMP.DSP 2010 presentations:
>
> >>http://www.compdsp.com/presentations.htm
>
> >> //-----------------
>
> >> Vladimir Vassilevsky
> >> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com
>
> >Not to take away anything from the author, but as far as I can tell,
> >people have been doing this for ages. In fact, the serial method could
> >easily implement a parallel method with a device known in the digital
> >world as "barrel shifter." In such a case, you'd store an unsigned
> >value with the coefficient akin to an "instruction" in ISA to tell you
> >how much you should barrel-shift per loop iteration.
>
> >Regards,
> >-Momo
>
> Yep. Looks like another example of relearning a forgotten art of doing
> things efficiently. :-)
>
> Steve

Indeed. Think of it as a hard-coded floating-point operation.

-Momo
From: HardySpicer on
On Jun 29, 1:49 pm, Manny <mlou...(a)hotmail.com> wrote:
> On Jun 29, 1:58 am, "steveu" <steveu(a)n_o_s_p_a_m.coppice.org> wrote:
>
>
>
> > >On Jun 28, 8:16=A0pm, Vladimir Vassilevsky <nos...(a)nowhere.com> wrote:
> > >> =A0 The presentation of Rick Lyons "Improving FIR Filter
> > >> Coefficient Precision" was added:
>
> >http://www.compdsp.com/presentations/Lyons/Lyons-Precise%20FIR%20Coef...
>
> > >> COMP.DSP 2010 presentations:
>
> > >>http://www.compdsp.com/presentations.htm
>
> > >> //-----------------
>
> > >> Vladimir Vassilevsky
> > >> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com
>
> > >Not to take away anything from the author, but as far as I can tell,
> > >people have been doing this for ages. In fact, the serial method could
> > >easily implement a parallel method with a device known in the digital
> > >world as "barrel shifter." In such a case, you'd store an unsigned
> > >value with the coefficient akin to an "instruction" in ISA to tell you
> > >how much you should barrel-shift per loop iteration.
>
> > >Regards,
> > >-Momo
>
> > Yep. Looks like another example of relearning a forgotten art of doing
> > things efficiently. :-)
>
> > Steve
>
> Indeed. Think of it as a hard-coded floating-point operation.
>
> -Momo

Two adaptive filters in tandem one. Some nice analysis there. You need
to get all the papers in the same format if you have a conference.


Hardy