f



group delay time?

It is known, that non-uniformity of the frequency response of the channel
influences level of ISI. And what kind of distortions is characteristic for
non-uniform group delay time? 
0
alex65111 (38)
2/5/2009 8:08:31 AM
comp.dsp 20333 articles. 1 followers. allnor (8509) is leader. Post Follow

15 Replies
470 Views

Similar Articles

[PageSpeed] 31

On Feb 5, 3:08=A0am, "alex65111" <alex65...@list.ru> wrote:
> It is known, that non-uniformity of the frequency response of the channel
> influences level of ISI. And what kind of distortions is characteristic f=
or
> non-uniform group delay time?

A non-constant group delay vs frequency (non-linear phase vs
frequency) distorts the pulse shape. This causes ISI.

John
0
sampson164 (501)
2/5/2009 10:50:07 AM
>On Feb 5, 3:08=A0am, "alex65111" <alex65...@list.ru> wrote:
>> It is known, that non-uniformity of the frequency response of the
channel
>> influences level of ISI. And what kind of distortions is characteristic
f=
>or
>> non-uniform group delay time?
>
>A non-constant group delay vs frequency (non-linear phase vs
>frequency) distorts the pulse shape. This causes ISI.
>
>John
>


As on level of pulsations group delay it is possible to estimate effective
duration of the impulse response?
0
alex65111 (38)
2/5/2009 8:47:48 PM
On Feb 5, 3:47=A0pm, "alex65111" <alex65...@list.ru> wrote:
> >On Feb 5, 3:08=3DA0am, "alex65111" <alex65...@list.ru> wrote:
> >> It is known, that non-uniformity of the frequency response of the
> channel
> >> influences level of ISI. And what kind of distortions is characteristi=
c
> f=3D
> >or
> >> non-uniform group delay time?
>
> >A non-constant group delay vs frequency (non-linear phase vs
> >frequency) distorts the pulse shape. This causes ISI.
>
> >John
>
> As on level of pulsations group delay it is possible to estimate effectiv=
e
> duration of the impulse response?

If you have a model for your channel, then it's trivial to find its
impulse response. Look at it and determine how much of it has a
significantly-large magnitude.

Jason
0
cincydsp (353)
2/5/2009 9:30:11 PM
On 5 Feb, 21:47, "alex65111" <alex65...@list.ru> wrote:

> As on level of pulsations group delay it is possible to estimate effective
> duration of the impulse response?

I can't see that there is a relation between group delay
and the duration of the impulse response. For IIR filters
the group delay can be finite for all frequencies [*].
Yet the duration of the impulse response is infinitely long.

Rune

[*] I'm pretty sure that's the case for 1st order Butterworth
    filters.
0
allnor (8509)
2/5/2009 9:53:25 PM
Rune Allnor wrote:
> On 5 Feb, 21:47, "alex65111" <alex65...@list.ru> wrote:
> 
>> As on level of pulsations group delay it is possible to estimate effective
>> duration of the impulse response?
> 
> I can't see that there is a relation between group delay
> and the duration of the impulse response. For IIR filters
> the group delay can be finite for all frequencies [*].
> Yet the duration of the impulse response is infinitely long.
> 
> Rune
> 
> [*] I'm pretty sure that's the case for 1st order Butterworth
>     filters.

C'mon, Rune. What's the difference between a first-order Butterworth and 
a first-order IIR od any class?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
0
jya (12871)
2/5/2009 11:12:49 PM
On Feb 5, 4:53=A0pm, Rune Allnor <all...@tele.ntnu.no> wrote:
> On 5 Feb, 21:47, "alex65111" <alex65...@list.ru> wrote:
>
> > As on level of pulsations group delay it is possible to estimate effect=
ive
> > duration of the impulse response?
>
> I can't see that there is a relation between group delay
> and the duration of the impulse response.

well, i think they are linearly related, for a filter of fixed shape
and adjustable resonant frequency, f0.  lower that frequency and the
effective impulse response length (however "effective" is defined) and
the mean group delay (however "mean" is defined) both go up in a 1/f0
manner.

> For IIR filters
> the group delay can be finite for all frequencies [*].
> Yet the duration of the impulse response is infinitely long.

how about the "effective duration"?

i thought that someone had a result that related

   integral{ tau(w) * |H(w)|^2 dw}

to

   sum{ n * |h[n]|^2 }

for a general filter.

i dunno.

r b-j



r b-j
0
rbj (4086)
2/6/2009 12:39:57 AM
On 6 Feb, 00:12, Jerry Avins <j...@ieee.org> wrote:
> Rune Allnor wrote:
> > On 5 Feb, 21:47, "alex65111" <alex65...@list.ru> wrote:
>
> >> As on level of pulsations group delay it is possible to estimate effec=
tive
> >> duration of the impulse response?
>
> > I can't see that there is a relation between group delay
> > and the duration of the impulse response. For IIR filters
> > the group delay can be finite for all frequencies [*].
> > Yet the duration of the impulse response is infinitely long.
>
> > Rune
>
> > [*] I'm pretty sure that's the case for 1st order Butterworth
> > =A0 =A0 filters.
>
> C'mon, Rune. What's the difference between a first-order Butterworth and
> a first-order IIR od any class?

Excluding the order, elliptic filters have stuff going on
so that the unwrapped phase responses have near-infinite
slopes.

Rune
0
allnor (8509)
2/6/2009 8:37:43 AM
On 6 Feb, 01:39, robert bristow-johnson <r...@audioimagination.com>
wrote:
> On Feb 5, 4:53=A0pm, Rune Allnor <all...@tele.ntnu.no> wrote:
>
> > On 5 Feb, 21:47, "alex65111" <alex65...@list.ru> wrote:
>
> > > As on level of pulsations group delay it is possible to estimate effe=
ctive
> > > duration of the impulse response?
>
> > I can't see that there is a relation between group delay
> > and the duration of the impulse response.
>
> well, i think they are linearly related, for a filter of fixed shape
> and adjustable resonant frequency, f0. =A0lower that frequency and the
> effective impulse response length (however "effective" is defined) and
> the mean group delay (however "mean" is defined) both go up in a 1/f0
> manner.
>
> > For IIR filters
> > the group delay can be finite for all frequencies [*].
> > Yet the duration of the impulse response is infinitely long.
>
> how about the "effective duration"?
>
> i thought that someone had a result that related
>
> =A0 =A0integral{ tau(w) * |H(w)|^2 dw}
>
> to
>
> =A0 =A0sum{ n * |h[n]|^2 }
>
> for a general filter.
>
> i dunno.

I haven't seen too many discussions of the group delay,
so you might be right.

My interpretation is that the group delay says something
about the time it takes for a response to occur on the
output of the filter, not how long the response lasts.

If you want to look at durations of the IR, the locations
of the poles are better indicators: The IR of resonant
systems last for long times, and these systems have poles
close to the unit circle.

Rune
0
allnor (8509)
2/6/2009 8:41:49 AM
Let's assume, that is available two variants - 
1) ripple of group delay - 10ns and ripple of frequency response in
passband - 3dB, 
2) ripple of group delay - 20ns but ripple of frequency response in
passband - 1dB. 
What variant of the filter is better for choosing for bandpass 10MHz,
5МHz?
0
alex65111 (38)
2/6/2009 7:10:03 PM
Rune Allnor wrote:
> On 6 Feb, 00:12, Jerry Avins <j...@ieee.org> wrote:
>> Rune Allnor wrote:
>>> On 5 Feb, 21:47, "alex65111" <alex65...@list.ru> wrote:
>>>> As on level of pulsations group delay it is possible to estimate effective
>>>> duration of the impulse response?
>>> I can't see that there is a relation between group delay
>>> and the duration of the impulse response. For IIR filters
>>> the group delay can be finite for all frequencies [*].
>>> Yet the duration of the impulse response is infinitely long.
>>> Rune
>>> [*] I'm pretty sure that's the case for 1st order Butterworth
>>>     filters.
>> C'mon, Rune. What's the difference between a first-order Butterworth and
>> a first-order IIR od any class?
> 
> Excluding the order, elliptic filters have stuff going on
> so that the unwrapped phase responses have near-infinite
> slopes.

Can one determine from the coefficients the difference between a 
first-order Butterworth and a first-order elliptic? Does a first-order 
filter have enough degrees of freedom to define a class?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
0
jya (12871)
2/6/2009 9:48:00 PM
alex65111 wrote:
> Let's assume, that is available two variants - 
> 1) ripple of group delay - 10ns and ripple of frequency response in
> passband - 3dB, 
> 2) ripple of group delay - 20ns but ripple of frequency response in
> passband - 1dB. 
> What variant of the filter is better for choosing for bandpass 10MHz,
> 5МHz?

Better for what purpose? Which is better, a bicycle or a truck?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
0
jya (12871)
2/6/2009 9:49:54 PM
On 6 Feb, 22:48, Jerry Avins <j...@ieee.org> wrote:

> Can one determine from the coefficients the difference between a
> first-order Butterworth and a first-order elliptic?

Nope.

>  Does a first-order
> filter have enough degrees of freedom to define a class?

Nope.

I was just playing it safe: Knowing the guys around here
I wanted to maek sure no one came up with a filter I had
missed that has infinite group delay. Of course it was the
safety feature that boomeranged right back in my face... ;)

Rune
0
allnor (8509)
2/6/2009 10:24:08 PM
Rune Allnor wrote:
> On 6 Feb, 22:48, Jerry Avins <j...@ieee.org> wrote:
> 
>> Can one determine from the coefficients the difference between a
>> first-order Butterworth and a first-order elliptic?
> 
> Nope.
> 
>>  Does a first-order
>> filter have enough degrees of freedom to define a class?
> 
> Nope.
> 
> I was just playing it safe: Knowing the guys around here
> I wanted to maek sure no one came up with a filter I had
> missed that has infinite group delay. Of course it was the
> safety feature that boomeranged right back in my face... ;)

I suppose a filter with all zero coefficients has infinite group delay, 
or at least seems to. Since the output isn't connected to the input, 
when you put something in, nothing comes out. :-)

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
0
jya (12871)
2/6/2009 10:54:26 PM
>alex65111 wrote:
>> Let's assume, that is available two variants - 
>> 1) ripple of group delay - 10ns and ripple of frequency response in
>> passband - 3dB, 
>> 2) ripple of group delay - 20ns but ripple of frequency response in
>> passband - 1dB. 
>> What variant of the filter is better for choosing for bandpass 10MHz,
>> 5МHz?
>
>Better for what purpose? Which is better, a bicycle or a truck?
>
>Jerry
>-- 
>Engineering is the art of making what you want from things you can get.
>???????????????????????????????????????????????????????????????????????
>


Better for the analog IF filter of the digital receiver intended for
demodulation of QAM or PSK or OFDM or DSSS or ...
0
alex65111 (38)
2/7/2009 6:48:16 AM
On Fri, 06 Feb 2009 17:54:26 -0500, Jerry Avins <jya@ieee.org> wrote:

>Rune Allnor wrote:
>> On 6 Feb, 22:48, Jerry Avins <j...@ieee.org> wrote:
>> 
>>> Can one determine from the coefficients the difference between a
>>> first-order Butterworth and a first-order elliptic?
>> 
>> Nope.
>> 
>>>  Does a first-order
>>> filter have enough degrees of freedom to define a class?
>> 
>> Nope.
>> 
>> I was just playing it safe: Knowing the guys around here
>> I wanted to maek sure no one came up with a filter I had
>> missed that has infinite group delay. Of course it was the
>> safety feature that boomeranged right back in my face... ;)
>
>I suppose a filter with all zero coefficients has infinite group delay, 
>or at least seems to. Since the output isn't connected to the input, 
>when you put something in, nothing comes out. :-)
>
>Jerry

Or zero group delay and infinite attentuation?   ;)

I guess that filter has an infinite number of arbitrary combinations
of infinite group delay and infinite attenuation.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
0
eric.jacobsen (2636)
2/7/2009 8:01:56 PM
Reply: