linear phase iir filters

robert bristow-johnson <rbj@surfglobal.net> wrote in message news:<BB719CCA.33D5%rbj@surfglobal.net>...
> >> While I haven't checked out the Clements & Pease paper, I don't understand
> >> what RBJ is talking about in the mentioned link, when he says "Truncated
> >> IIR". Isn't that, by definition, a FIR structure?
> 
> well, it *is* an FIR.  what you mean be "structure" might affect my answer.

I only meant "no feedback loop".

> the term "Truncated IIR" is from Julius Smith as referened by Peter below.
> a TIIR is an FIR that has some structure that would remind one of an IIR.
> 
> the simplest example of a TIIR is the one of a moving average filter.  there
> are two ways of implementing it.  one is with the straight forward FIR
> structure in which all coefs are 1/N (N is the number of samples getting
> averaged).  the other structure is with a digital integrator but where the
> input has a delay line and adder that will subtract x[n-N] from the
> integrator input.
> 
> not every TIIR is linear phase.  some of them require cascading with one
> that has the opposite sense of decay or it requires segmentation,
> time-reversing the segment, TIIRing with an identical filter, flipping the
> result back around in time, and overlap-adding the result.  it's a mess, but
> it's how you make a nearly IIR filter act with linear phase in a real-time
> app.
> 
> >> For that matter, why would anyone want to make a linear phase IIR?
> 
> i guess so that all frequencies are delayed equally.
> 
> >> Isn't it way easier to stick with the good ol' FIR?
> 
> usually it is.  but if the number of taps are large, it's not always cheaper
> than the Truncated IIR approach.

OK... it seems as a gain in run-time and/or lower power consumption is 
achieved at the expense of a serious drain on brain power... 

Neat stuff! 

Rune

In article f56893ae.0308271007.348925d5@posting.google.com, Rune Allnor at
allnor@tele.ntnu.no wrote on 08/27/2003 14:07:

> robert bristow-johnson <rbj@surfglobal.net> wrote in message
> news:<BB719CCA.33D5%rbj@surfglobal.net>...
>>>> While I haven't checked out the Clements & Pease paper, I don't understand
>>>> what RBJ is talking about in the mentioned link, when he says "Truncated
>>>> IIR". Isn't that, by definition, a FIR structure?
>> 
>> well, it *is* an FIR.  what you mean be "structure" might affect my answer.
> 
> I only meant "no feedback loop".

well, this kind of FIR filter (implemented as a TIIR) definitely does have a
feedback loop or two.  but it is still an FIR (the reason why is that there
is pole-zero cancellation of the IIR poles that are not at the origin).

>> the term "Truncated IIR" is from Julius Smith as referened by Peter below.
>> a TIIR is an FIR that has some structure that would remind one of an IIR.
>> 
...
> 
> OK... it seems as a gain in run-time and/or lower power consumption is
> achieved at the expense of a serious drain on brain power...

that's about right.  i have never implemented a TIIR linear phase filter
other than the moving average filter but it is a little bitchy particularly
if you are doing the time-reversal step.  emptying out the buffer and doing
overlap add is really a bitch.

> Neat stuff! 

it *is* interesting.

r b-j

an2or@mailcircuit.com (Andor) wrote in message news:<ce45f9ed.0308260255.46fe1c3a@posting.google.com>...
> shamganth wrote:
> > sir,
> > iam doing my postgraduation in communication in south india at Anna
> > University. Iam doing my project in Linear phase IIR filters. I have
> > the base materials for Linear phase iir filters. I wish to implement
> > the Linear phase IIR filter system in Matlab. so yet i haven't get an
> > idea how to implement the system. 
> 
> I've got a good idea: use the filtfilt() function.
> 
> Regards,
> Andor

I think that deals with zero phase IIR filter. Although zero phase is
a special
kind of linear phase filter it will be difficult to meet few important
purpose:

In general, a zero phase impulse response is defined as
h(n)=h(-n) and It is also not causal.

Most common FIR filter in practice is causal,symmetric and linear
phase,
It can be shown for such system,
H(z)=z^-(N-1)/2HZP(z)
where H(z) is  transfer function of linear phase filter and HZP(z) is
that of
a zero phase filter i.e linear phase filter is a delayed version of
zero phase filter.

Now if the original poster has a requirement of causal system then the
above
MATLAB function can not serve the purpose.

Please correct if I am missing some point here.
Regards,
Santosh

Jerry Avins <jya@ieee.org> writes:

> It's obviously a disk with binoculars, but the seeing hasn't beem good
> enough for me to see the polar caps with my 5" scope. Not likely
> tonight, either, but maybe tomorrow.

Here in Brisbane last night it was glorious astronomy weather.  The
city lights are a little bright for "serious" astronomy, but for
seeing Mars it was great.

Ciao,

Peter K.

-- 
Peter J. Kootsookos

"Na, na na na na na na, na na na na"
- 'Hey Jude', Lennon/McCartney

robert bristow-johnson <rbj@surfglobal.net> writes:

> you should meet Julius sometime.  he's a pretty cool dude with enough
> knowledge of computer music and audio to scare the pants offa me.

I haven't met him, but have e-corresponded occassionally (though not for
some time).  His PhD thesis was pretty impressive, and he certainly
knows his stuff with respect to computer music!!

Ciao,

Peter K.

-- 
Peter J. Kootsookos

"Na, na na na na na na, na na na na"
- 'Hey Jude', Lennon/McCartney

Jerry Avins <jya@ieee.org> wrote in message news:<3F4CD230.BE265766@ieee.org>...
> "Peter J. Kootsookos" wrote:
> > 
> > robert bristow-johnson <rbj@surfglobal.net> writes:
> > 
> > > can you guys down-under see Mars as well as us guys?  it's even a little
> > > cloudy and i can see it really well.
> > 
> > Yup!  It's pretty impressive. :-)
> > 
> > Ciao,
> > 
> > Peter K.
> > 
> > --
> > Peter J. Kootsookos
> > 
> > "Na, na na na na na na, na na na na"
> > - 'Hey Jude', Lennon/McCartney
> 
> It's obviously a disk with binoculars, but the seeing hasn't beem good
> enough for me to see the polar caps with my 5" scope. Not likely
> tonight, either, but maybe tomorrow.

I've seen it once. Due to the latitude (I'm 100 km north of the arctic
circle) it's only a couple of weeks since it became dark enough to see 
stars and planets. And now the autumn weather has started, which means 
fog and lots of rain...

The one time I did see Mars it came up just above the mountains to the 
south. The seeing was poor, but at 60x it looked clearly red, like a 
lump of red-hot coal glowing in the sky. 

Rune

santosh nath wrote:
...
> It can be shown for such system,
> H(z)=z^-(N-1)/2HZP(z)
> where H(z) is  transfer function of linear phase filter and HZP(z) is
> that of
> a zero phase filter i.e linear phase filter is a delayed version of
> zero phase filter.
> 
> Now if the original poster has a requirement of causal system then the
> above
> MATLAB function can not serve the purpose.

There is no causal linear-phase IIR filter (except for the identity) -
just approximations. I think the OP is looking for the time-reversal
algorithm, as he explained himself in another post. This can
implemented using the filtfilt() function.

For real-time applications (ie. where you do not have the whole buffer
of samples available to run through the filter forwards and
backwards), some kind of overlap-add or -save scheme is required. This
induces a delay, which is where the zero-phase filter becomes linear
phase, as you have stated above.

Regards,
Andor

Jerry Avins wrote:

> > robert bristow-johnson <rbj@surfglobal.net> writes:
> > 
> > > can you guys down-under see Mars as well as us guys?  it's even a little
> > > cloudy and i can see it really well.
> > 
> > Yup!  It's pretty impressive. :-)
> > 
> > Ciao,
> > 
> > Peter K.
> > 
> > --
> > Peter J. Kootsookos
> > 
> > "Na, na na na na na na, na na na na"
> > - 'Hey Jude', Lennon/McCartney
> 
> It's obviously a disk with binoculars, but the seeing hasn't beem good
> enough for me to see the polar caps with my 5" scope. Not likely
> tonight, either, but maybe tomorrow.
> 
> Jerry

My astronomy program tells me that seeing should be better in early
September, when Mars will rise higher into the sky.

Perhaps you will even be able to see the canali with the binoculars
then :).

Regards,
Andor

On Wed, 27 Aug 2003 14:03:15 -0400, robert bristow-johnson
<rbj@surfglobal.net> wrote:

>
>Rick,
>
>you should meet Julius sometime.  he's a pretty cool dude with enough
>knowledge of computer music and audio to scare the pants offa me.
>
>r b-j
>

Hi R B-J,
  yep,, I didn't mean to sound so negative.

Smith has posted an *AMAZING* amount of DSP 
material on the web - to the benefit of 
of us all.  I'm sure he's forgotten more DSP 
than I'll ever know.

See Ya,
[-Rick-]
 

On 28 Aug 2003 13:17:36 +1000, p.kootsookos@remove.ieee.org (Peter J.
Kootsookos) wrote:

>Jerry Avins <jya@ieee.org> writes:
>
>> It's obviously a disk with binoculars, but the seeing hasn't beem good
>> enough for me to see the polar caps with my 5" scope. Not likely
>> tonight, either, but maybe tomorrow.
>
>Here in Brisbane last night it was glorious astronomy weather.  The
>city lights are a little bright for "serious" astronomy, but for
>seeing Mars it was great.
>
>Ciao,
>
>Peter K.

Hi Dr. K,

   Northern California had good viewing last night.

When I first saw Mars I thought it was an 
airplane, so I looked in another direction.
Then I noticed the 'airplane' wasn't moving.
Wow!

Now I wished I hadn't donated my home-built 
6-inch reflector to the local high school.
I wonder if I could have seen Mars' 
polar caps with a 6" reflector and my old eyepieces.

See Ya,
[-Rick-]