Prev: Difference between pi/4 QPSK and pi/4 DQPSK
Next: Achieving 'equal loudness' for A/B testing
From: BobG on 27 May 2010 14:14
Whats the rule of thumb for how far to detune the two outside filters
for third octave bandwidth? Is the middle filter supposed to have a
little less gain to flatten the passband out? Is the Q of all 3
filters the same? Do the skirts roll off at 18dB per octave? (I
searched for 'triple tuned' but no hits, so I hope I dont get flamed
for asking)
From: Tim Wescott on 27 May 2010 15:21
On 05/27/2010 11:14 AM, BobG wrote:
> Whats the rule of thumb for how far to detune the two outside filters
> for third octave bandwidth? Is the middle filter supposed to have a
> little less gain to flatten the passband out? Is the Q of all 3
> filters the same? Do the skirts roll off at 18dB per octave? (I
> searched for 'triple tuned' but no hits, so I hope I dont get flamed
> for asking)

Well, not a flame perhaps, but:

It sounds like you're trying to make a filter in 2010 using design rules
from 1940. There's nothing inherently bad in this, but there are
certainly faster ways.

Back up a bit and tell us what you're really trying to do. These days
-- at least if this is a DSP project -- one would normally think of this
in terms of a 6th-order bandpass filter, one would specify it by
passband ripple, 3dB (or 6dB) bandwidth, shape factor, and ultimate
attenuation. Do that, and most of your questions become moot, because
the answer to all of them is "whatever the filter design coughs up".

Generally with filter chains like this the filters are cascaded, not
paralleled. So the middle filter gain neither makes the center of the
passband rise or fall -- the Q of the middle filter does.

Depending on your filter design the Q of the filters may or may not be
the same, but I _think_ that for a 'normal' Butterworth or Chebychev
filter you can expect the 'end' filters to have higher Q.

The skirts will roll off at 18dB per octave far from the passband, but
close to the passband the more ripple you accept the sharper the skirts
will be. Someone who lives and breaths filter design would be able to
quote you figures at this point, all I can say is that it happens.

Try searching on "bandpass filter" and "IIR". In Altavista the syntax
for the search would be +"bandpass filter" +IIR -- I don't know how you
could coerce Google to find what you want.

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
From: Jerry Avins on 27 May 2010 15:25
On 5/27/2010 2:14 PM, BobG wrote:
> Whats the rule of thumb for how far to detune the two outside filters
> for third octave bandwidth? Is the middle filter supposed to have a
> little less gain to flatten the passband out? Is the Q of all 3
> filters the same? Do the skirts roll off at 18dB per octave? (I
> searched for 'triple tuned' but no hits, so I hope I dont get flamed
> for asking)

Do you perhaps have analog filters in mind? The design procedures for
digital filters don't usually go in that direction.

I remember doing the math for stagger-tuned IF strips in the early
1050s. It wasn't all that hard, but I don't remember off hand. There's
an old paper in the Proceedings of the IRE.

Jerry
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
From: Tim Wescott on 27 May 2010 15:29
On 05/27/2010 12:25 PM, Jerry Avins wrote:
> On 5/27/2010 2:14 PM, BobG wrote:
>> Whats the rule of thumb for how far to detune the two outside filters
>> for third octave bandwidth? Is the middle filter supposed to have a
>> little less gain to flatten the passband out? Is the Q of all 3
>> filters the same? Do the skirts roll off at 18dB per octave? (I
>> searched for 'triple tuned' but no hits, so I hope I dont get flamed
>> for asking)
>
> Do you perhaps have analog filters in mind? The design procedures for
> digital filters don't usually go in that direction.
>
> I remember doing the math for stagger-tuned IF strips in the early
> 1050s. It wasn't all that hard, but I don't remember off hand. There's
> an old paper in the Proceedings of the IRE.

Whoa! Jerry, I knew you were old, but I didn't realize you were _that_
old. For that matter, I didn't realize that they were building IF
strips in the middle ages.

Were you in England for the Battle of Hastings? What was it like?

--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
From: Rune Allnor on 27 May 2010 15:30
On 27 Mai, 21:25, Jerry Avins <j...(a)ieee.org> wrote:

> I remember doing the math for stagger-tuned IF strips in the early
> 1050s.

Er...?

Rune
 |  Next  |  Last
Pages: 1 2 3
Prev: Difference between pi/4 QPSK and pi/4 DQPSK
Next: Achieving 'equal loudness' for A/B testing