Bessel Filter (Digital)

On Thu, 18 Nov 2010 08:47:20 -0800 (PST), robert bristow-johnson
<rbj@audioimagination.com> wrote:

>On Nov 18, 11:42=A0am, eric.jacob...@ieee.org (Eric Jacobsen) wrote:
>
>> ... =A0I miss the old days.
>
>Eric, *these* are the good old days.
>
>r b-j

Yeah, unfortunately, that's always the case.

;)  or :'(



Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com

robert bristow-johnson wrote:

> On Nov 17, 12:33 pm, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
> 
>>robert bristow-johnson wrote:
>>
>>>On Nov 17, 11:48 am, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
>>
>>>>>>Everybody complains about "ringing" and overshoot, but nobody could
>>>>>>explain why it is bad.
>>
>>>>>i think the control systems people (like Tim) might be able to explain
>>>>>it.
>>
>>>>How this is relevant to Bessel/Butterwoth/etc?
>>
>>>in designing your controller, you might want a net transfer function
>>>(of plant with feedback and controller) that isn't a Butterworth, but
>>>more like a Bessel.
>>
>>AFAIK they have many other criteria for loop optimization and they don't
>>set matching Bessel response as the specific goal of the design.
>>
>>Also, as Tim said, having a Bessel filter inside a loop is not any
>>condition for the whole loop to be ringing or not ringing.
> 
> 
> it wasn't about just the controller inside, it was about the whole
> thing (didn't i state it clearly?).  with enough independent
> parameters of the controller ( H(s) ) to tweak, it seems possible to
> me that a controller can send the poles of the whole (H*G)/(1+H*G) to
> wherever you need them to be for stability and perhaps some other
> design goals.  you otta be able to put the zeros wherever you wish (at
> least i think you can in the digital case).
> 
> so, what else does the Bessel filter have as an interesting or
> desirable quality regarding this servo-mechanism?  symmetrical edges?
> does another filter have as fast step response while possessing one or
> both qualities?
> 
> so then, given G, why wouldn't you try to design H so that (H*G)/
> (1+H*G) is a Bessel filter or something similar (if there are other
> constraints)?

Yes, it could be possible to place poles (and sometimes zeroes) as you 
wish (*); however the control systems theory is about optimizing 
something with respect to something. For the texbook case of the LTI 
system of the 2nd order, Bessel response would be optimal if the 
criteria is quickest settling to zero as t -> infinity. If you take 
least squares as criteria, you will get classic 1/4 damping result. My 
point: there is nothing implicitly good about Bessel response with 
respect to control systems.

* Consult with Tim Wescott

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

raj malhotra <f2007327@n_o_s_p_a_m.bits-goa.ac.in> wrote:
(big snip)

> I can observe that people on this forum don't give solutions to the
> problems being asked. They just fight like little kids on the problem and
> discuss it and try to prove that they are too smart. But actually , i sense
> that they are noobs and might be possible that they copy paste the stuff
> they post here . And moreover, its not a big deal to master a subject if
> you are working on that subject from soo long. 

Most of us still have some things to learn here, though we might
wish that we knew it all.  There are some who answer requests for
homework solutions with misleading answers.  If someone copies the
answer without even trying to understand it, then they get what they
deserve.  If someone shows that they have tried, but still don't
understand, then they will usually get useful replies.
 
> Earlier , i used this forum to ask a doubt related to FFTW and a blocking
> mechanism , and that time too the condition and the responses were similar.
> They never help , just wander around the topic. I am actually laughing at
> this situation right now. Nevertheless, thanks guys!

I am not sure what a blocking mechanism is, so I can't answer that one.

There are FFT implementations that do block floating point, with one
exponent for the whole array.  That makes a lot of sense for FFT,
though not so much for other problems.  

If you mean blocking and the effects of caching, then it is probably
a little off topic for most in this group.  That doesn't mean that
it is off-topic for the group, but many here don't work on those
problems.  For questions about implementations of FFT, a numerical
analysis group would be a better choice.

-- glen

>
>
>raj malhotra wrote:
>
>> I can observe that people on this forum don't give solutions to the
>> problems being asked. They just fight like little kids on the problem
and
>> discuss it and try to prove that they are too smart.
>
>The questions are dumb. That 2 x 2 = 4 stuff is not interesing.
>If you need a solution, pay for it. Or be happy with what you are
getting.


Again by writing "2*2=4 stuff" you are just trying to show you are smart.
ROFL!! And moreover, even if i pay you twice the amount, you'll be unable
to answer i'm pretty sure now :) 



>> But actually , i sense
>> that they are noobs and might be possible that they copy paste the
stuff
>> they post here . 
>
>:))))))) Is that what you generally do?
>JFYI, this newsgroup is a place for marketing and leisure of industry 
>experts.

If this newsgroup is a place for marketing and leisure of industry experts,
than what are you doing here on this newsgroup????



>> And moreover, its not a big deal to master a subject if
>> you are working on that subject from soo long. 
>
>Of course. Work towards your mastership.

I too wish you luck to work towards it :D But i don't think my wishes will
work , coz the ego problem in you is greater than your will to learn.
Unless one becomes humble, all his work goes in vain.


>> Earlier , i used this forum to ask a doubt related to FFTW and a
blocking
>> mechanism , and that time too the condition and the responses were
similar.
>> They never help ,
>
>You never pay, and never will. Then what is the point of explaining the 
>trivial things to you?
>> just wander around the topic. I am actually laughing at
>> this situation right now. Nevertheless, thanks guys!
>
>Clever man needs just a hint; he works the details himself. You've got a 
>lot of hints, still can't work the details :)))))


Wish you good luck to work on your ego! 

Raj!

On Nov 18, 3:14&#4294967295;pm, "gretzteam" <gretzteam@n_o_s_p_a_m.yahoo.com>
wrote:
> >On Nov 18, 1:23=A0pm, "raj malhotra" <f2007327@n_o_s_p_a_m.bits-
> >goa.ac.in> wrote:
>
> >> I can observe that people on this forum don't give solutions to the
> >> problems being asked.
>
> >You suppose that any question deserves an answer.
> >Like it or not, but engineering is about progress.
> >The Bessel filter might have had a purpose some
> >time in the past, but that doesn't mean it still
> >has.
>
> >Once upon a time steam engines was the main technology.
> >These days they are mere curiosities.
>
> >Rune
>
> Then I'd like to know how you would design the following filter. Maybe
> there are tricks I'm not aware of...
>
> -fs = 2kHz
> -fc = 40Hz
> -Attenuation about 18dB/octave more or less (similar to 3rd order
> Butterworth)
> -maximum ringing ~1%

What do you mean by 'ringing'?

> -low computational complexity (3 to 5 non-trivial multiplications
> maximum).

I haven't seen very many specs where 'ringing' is mentioned.
I assume you mean 'transients in time domain'. If so, I'd say
that one would need to prioritize one or the other: *Either*
use the spec in frequency domain and accept whatever happens
in time domain, *or* use the spec in time domain and accept
whatever happens in frequency domain.

As I am sure you know, the two domains are interlinked such
that any gain one wants in one, imposes some sort of penalty
in the other. Usually at some computational cost.

Only a fool will spec up something that is severely constrained
in both. Or all three domains, if one counts in computational
load as well.

Rune