> "Randy Yates" <yates@ieee.org> wrote in message
> news:m3zltg9dem.fsf@ieee.org...
>> "VelociChicken" <bob@yahoob.com> writes:
>> I've never heard of "two-path" polyphase filters, but the operation of
>> polyphase decimation filters is very straightforward: since you're going
>> to throw away M-1 of M samples when decimating by M, simply don't
>> compute them. The rest is bookkeeping.
>> --
>
> Exactly. Polyphase is only about the internal implementation of some
> filters and not about filter design - as in coefficient selection.
It's a different and idiosyncratic use of "polyphase". Mr. Chicken was
misled by the author 0f the URL he quoted. Don't cluck at him.
Jerry
--
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������
Reply by Fred Marshall●March 3, 20082008-03-03
"VelociChicken" <bob@yahoob.com> wrote in message
news:uUYyj.83377$jH4.73721@fe1.news.blueyonder.co.uk...
>>> If you google 'polyphase filter' you'll find better
>>> explanations than I can give. I've optimised this
>>> codehttp://www.musicdsp.org/archive.php?classid=3#39which states that it
>>> can
>>> achieve up to '150db rejection' with a transition band of .05
>
>>150 dB attenuation, that's on the order of 10e-7, right?
>>I'm not sure I believe that - one is pushing the limits
>>of what can be done with single-precision floating point
>>numbers...
>>Rune
>
> Single-precision floating point? *pfft* - nobody uses those anymore!
> : )
>
After reading through your references I wonder if it might not be a good
idea for you to tell us what you think a "halfband filter" is. Because, it
isn't simply a filter with cutoff at fs/4.
In order to get a halfband filter of any reasonable length and 150dB
rejection, the transition would have to be awfully wide I would think.
Fred
Reply by Fred Marshall●March 3, 20082008-03-03
"Randy Yates" <yates@ieee.org> wrote in message
news:m3zltg9dem.fsf@ieee.org...
> "VelociChicken" <bob@yahoob.com> writes:
> I've never heard of "two-path" polyphase filters, but the operation of
> polyphase decimation filters is very straightforward: since you're going
> to throw away M-1 of M samples when decimating by M, simply don't
> compute them. The rest is bookkeeping.
> --
Exactly. Polyphase is only about the internal implementation of some
filters and not about filter design - as in coefficient selection.
Fred
Reply by Fred Marshall●March 3, 20082008-03-03
"DSPGURU" <kruptan@gmail.com> wrote in message
news:0f4608f3-0df6-4634-968a-5a725f74b750@m36g2000hse.googlegroups.com...
On Mar 1, 3:00 pm, "VelociChicken" <b...@yahoob.com> wrote:
> >> My question to you good folks - is there a clever 'trick' to cut
> >> frequencies
> >> off above this specific (SR / 4) point using time domain techniques? Or
> >> a
> >> more efficient frequency domain method?
>
> >> I've tried a cascade of Butterworth filters, but it just isn't steep
> >> enough
> >> to consider using it in my app.
>
> >> I know is only a small chance there is a trick for this, but I've got
> >> nothing to lose asking! : )
>
> >> Thanks,
>
> >> Dave
>
> > Hi Dave,
>
> > Google half-band filters. Every other coefficient (except the middle
> > one) is zero.
>
> Hello, thanks Randy that looks likely to be perfect! I hope I can keep the
> coef lengths short...
>
> D- Hide quoted text -
>
> - Show quoted text -
Hello Mr. Chicken,
If you read about halfband filter (books or old posts in comp.dsp) you
see they have really bad transition and are EXACTLY NOT what you want.
Keep looking.
Hmmmmm.... in my world, the transition width of a halfband filter is
something that you specify. So, I'm curious why you say this or where,
exactly, you read it.
Fred
Reply by Fred Marshall●March 3, 20082008-03-03
"DSPGURU" <kruptan@gmail.com> wrote in message
news:0f4608f3-0df6-4634-968a-5a725f74b750@m36g2000hse.googlegroups.com...
On Mar 1, 3:00 pm, "VelociChicken" <b...@yahoob.com> wrote:
> >> My question to you good folks - is there a clever 'trick' to cut
> >> frequencies
> >> off above this specific (SR / 4) point using time domain techniques? Or
> >> a
> >> more efficient frequency domain method?
>
> >> I've tried a cascade of Butterworth filters, but it just isn't steep
> >> enough
> >> to consider using it in my app.
>
> >> I know is only a small chance there is a trick for this, but I've got
> >> nothing to lose asking! : )
>
> >> Thanks,
>
> >> Dave
>
> > Hi Dave,
>
> > Google half-band filters. Every other coefficient (except the middle
> > one) is zero.
>
> Hello, thanks Randy that looks likely to be perfect! I hope I can keep the
> coef lengths short...
>
> D- Hide quoted text -
>
> - Show quoted text -
Hello Mr. Chicken,
If you read about halfband filter (books or old posts in comp.dsp) you
see they have really bad transition and are EXACTLY NOT what you want.
Keep looking.
Hmmmmm.... in my world, the transition width of a halfband filter is
something that you specify. So, I'm curious why you say this or where,
exactly, you read it.
Fred
Reply by Ron N.●March 3, 20082008-03-03
On Mar 1, 10:02 am, "VelociChicken" <b...@yahoob.com> wrote:
> Hello, I'm currently using an FFT to zereo out all the frequences above half
> Nyquist for my application. It makes perfect cutoff, ...
Unfortunately, zeroing the "upper" bins of an FFT does
not make a perfect cutoff filter in the general case.
This seems to be a common misconception (why?).
The problem is that any sharp transistions between bins may
cause the all the frequencies between nearby bins to have
wildly varying gains, and for the ripple in your frequency
response to not only cover your entire frequency range,
but to wrap around (unless you use really long FFTs in
relation to your allowed S/N or quantization error).
IMHO. YMMV.
--
rhn A.T nicholson d.0.t C-o-M
Reply by VelociChicken●March 3, 20082008-03-03
>> If you google 'polyphase filter' you'll find better
>> explanations than I can give. I've optimised this
>> codehttp://www.musicdsp.org/archive.php?classid=3#39which states that it
>> can
>> achieve up to '150db rejection' with a transition band of .05
>150 dB attenuation, that's on the order of 10e-7, right?
>I'm not sure I believe that - one is pushing the limits
>of what can be done with single-precision floating point
>numbers...
>Rune
On Mar 3, 9:24 am, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:
...
> It shoud be really really fast and it should do a very very good job on
> filtering. What else do you need to know? Only geeks care about the
> numbers.
>
> Vladimir Vassilevsky
> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com
And only trolls need fear the light of numbers:)
Dale B. Dalrymple
Reply by Rune Allnor●March 3, 20082008-03-03
On 3 Mar, 18:36, "VelociChicken" <b...@yahoob.com> wrote:
> >> Hey guys, I'm using this
> >>http://users.cscs.wmin.ac.uk/~krukowa/Poly.html
> >> it's perfect for the job, and it's really, really fast!
>
> >> Cheers,
>
> >> VelociChicken
>
> > So, if that is a good approach, would you please tell us what are your
> > passband width, passband tolerance, stopband width and stopband
> > attenuation requirements and what design did you get from that
> > approach to meet them?
>
> > Real numbers clarify the discussion faster than 'really, really',
> > really, really.
>
> > Dale B. Dalrymple
> >http://dbdimages.com
>
> What I meant by fast was, computation wise, it out performed all the other
> filters I've tried.
I know of people who emphasized 'computationally fast' in just
about everything they tried to do. While their programs certainly
were *faster* than eveybody else's, it took them more than 15
years to get their programs to compute the *correct* numbers...
> If you google 'polyphase filter' you'll find better
> explanations than I can give. I've optimised this codehttp://www.musicdsp.org/archive.php?classid=3#39which states that it can
> achieve up to �'150db rejection' with a transition band of .05
150 dB attenuation, that's on the order of 10e-7, right?
I'm not sure I believe that - one is pushing the limits
of what can be done with single-precision floating point
numbers...
Rune
Reply by VelociChicken●March 3, 20082008-03-03
>>
>> Hey guys, I'm using this
>> http://users.cscs.wmin.ac.uk/~krukowa/Poly.html
>> it's perfect for the job, and it's really, really fast!
>>
>> Cheers,
>>
>> VelociChicken
>
> So, if that is a good approach, would you please tell us what are your
> passband width, passband tolerance, stopband width and stopband
> attenuation requirements and what design did you get from that
> approach to meet them?
>
> Real numbers clarify the discussion faster than 'really, really',
> really, really.
>
> Dale B. Dalrymple
> http://dbdimages.com
What I meant by fast was, computation wise, it out performed all the other
filters I've tried. If you google 'polyphase filter' you'll find better
explanations than I can give. I've optimised this code
http://www.musicdsp.org/archive.php?classid=3#39 which states that it can
achieve up to '150db rejection' with a transition band of .05