Hi, I've been playing with wave lattice filters for a little while and only yesterday realized that they were just a structure used to implement the same good old transfer function that maps nicely onto the Direct Form I, II, transposed etc. For some reason I thought they were a different beast. Now I wonder how do people come up with new structures. The mapping from the transfer function to a lattice wave filter is not obvious at all! Let's say that I couldn't just start from DF1 and come up with a lattice structure... Are there still new structures being found yielding power or area advantages? Are there any book or research on this topic? Thanks! Diego
Still new digital filter structures to be found?
Started by ●March 5, 2010
Reply by ●March 5, 20102010-03-05
On 5 Mar, 14:45, "gretzteam" <gretzt...@yahoo.com> wrote:> Hi, > I've been playing with wave lattice filters for a little while and only > yesterday realized that they were just a structure used to implement the > same good old transfer function that maps nicely onto the Direct Form I, > II, transposed etc. For some reason I thought they were a different beast. > > Now I wonder how do people come up with new structures. The mapping from > the transfer function to a lattice wave filter is not obvious at all! Let's > say that I couldn't just start from DF1 and come up with a lattice > structure... > > Are there still new structures being found yielding power or area > advantages? > > Are there any book or research on this topic?There are two lines of thought on this topic: 1) There is *always* something new anexciting to be discovered. 2) Most of the useful stuff has already been discovered. I must admit that I subscribe to view 2) - that most of the useful DSP stuff is already ot there. This view is based on the premise that 1) There are only so many ways to compute the same quantity 2) A lot of ridiculously smart people have reviewed the subject in the past, which leave very little to be discovered in the future I know some of the regulars here disagree fiercly with me, but it's up to them, not me, to argue in defence of their views. Rune
Reply by ●March 5, 20102010-03-05
Rune Allnor wrote:> On 5 Mar, 14:45, "gretzteam" <gretzt...@yahoo.com> wrote: >> Hi, >> I've been playing with wave lattice filters for a little while and only >> yesterday realized that they were just a structure used to implement the >> same good old transfer function that maps nicely onto the Direct Form I, >> II, transposed etc. For some reason I thought they were a different beast. >> >> Now I wonder how do people come up with new structures. The mapping from >> the transfer function to a lattice wave filter is not obvious at all! Let's >> say that I couldn't just start from DF1 and come up with a lattice >> structure... >> >> Are there still new structures being found yielding power or area >> advantages? >> >> Are there any book or research on this topic? > > There are two lines of thought on this topic: > > 1) There is *always* something new anexciting to be discovered. > 2) Most of the useful stuff has already been discovered. > > I must admit that I subscribe to view 2) - that most of the > useful DSP stuff is already ot there. This view is based on > the premise that > > 1) There are only so many ways to compute the same quantity > 2) A lot of ridiculously smart people have reviewed the > subject in the past, which leave very little to be > discovered in the future > > I know some of the regulars here disagree fiercly with me, > but it's up to them, not me, to argue in defence of their > views.Of course there are new things to be discovered. The hard question is, are there new *useful* things to be discovered? Sometimes, a slightly inferior solution is useful for circumventing a patent. Jerry -- Blaise Pascal: Men never do evil so completely and cheerfully as when they do it from religious conviction. ��������������������������������������������������������������
Reply by ●March 5, 20102010-03-05
Jerry Avins wrote:> Rune Allnor wrote: >> On 5 Mar, 14:45, "gretzteam" <gretzt...@yahoo.com> wrote: >>> Hi, >>> I've been playing with wave lattice filters for a little while and only >>> yesterday realized that they were just a structure used to implement the >>> same good old transfer function that maps nicely onto the Direct Form I, >>> II, transposed etc. For some reason I thought they were a different >>> beast. >>> >>> Now I wonder how do people come up with new structures. The mapping from >>> the transfer function to a lattice wave filter is not obvious at all! >>> Let's >>> say that I couldn't just start from DF1 and come up with a lattice >>> structure... >>> >>> Are there still new structures being found yielding power or area >>> advantages? >>> >>> Are there any book or research on this topic? >> >> There are two lines of thought on this topic: >> >> 1) There is *always* something new anexciting to be discovered. >> 2) Most of the useful stuff has already been discovered. >> >> I must admit that I subscribe to view 2) - that most of the >> useful DSP stuff is already ot there. This view is based on >> the premise that >> >> 1) There are only so many ways to compute the same quantity >> 2) A lot of ridiculously smart people have reviewed the >> subject in the past, which leave very little to be >> discovered in the future >> >> I know some of the regulars here disagree fiercly with me, >> but it's up to them, not me, to argue in defence of their >> views. > > Of course there are new things to be discovered. The hard question is, > are there new *useful* things to be discovered? Sometimes, a slightly > inferior solution is useful for circumventing a patent. > > JerrySometimes, an effort to circumvent a patent leads to a superior solution (I wish I had an example to hand). -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Reply by ●March 5, 20102010-03-05
Rune Allnor wrote:> On 5 Mar, 14:45, "gretzteam" <gretzt...@yahoo.com> wrote: >> Hi, >> I've been playing with wave lattice filters for a little while and only >> yesterday realized that they were just a structure used to implement the >> same good old transfer function that maps nicely onto the Direct Form I, >> II, transposed etc. For some reason I thought they were a different beast. >> >> Now I wonder how do people come up with new structures. The mapping from >> the transfer function to a lattice wave filter is not obvious at all! Let's >> say that I couldn't just start from DF1 and come up with a lattice >> structure... >> >> Are there still new structures being found yielding power or area >> advantages? >> >> Are there any book or research on this topic? > > There are two lines of thought on this topic: > > 1) There is *always* something new and exciting to be discovered. > 2) Most of the useful stuff has already been discovered. > > I must admit that I subscribe to view 2) - that most of the > useful DSP stuff is already ot there. This view is based on > the premise that > > 1) There are only so many ways to compute the same quantity > 2) A lot of ridiculously smart people have reviewed the > subject in the past, which leave very little to be > discovered in the future > > I know some of the regulars here disagree fiercely with me, > but it's up to them, not me, to argue in defense of their > views.I think that Rune's cynicism has a lot of evidence to back it up, but I'm still cynical of it. Most of the new stuff being discovered is just a variation on a theme, and as much of it is done in pursuit of PhDs or tenure it tends to get published with an excess of fanfare. Having said that, I think that there is room for new techniques to be discovered, or to be adopted by practitioners. Particularly if you're working with an ASIC or an FPGA and you're looking to clever ways to reduce logic and/or power I think there's room for incremental improvement. So I don't think there are any techniques out there that are going to offer astounding improvements, and techniques that _do_ offer improvements are going to do so at the cost of easy understandability. Keep in mind that the "direct form" filters have the coefficients of the polynomials as gains -- I'm pretty sure that's why they're called "direct form". There may be filters out there that are significantly 'better' in some technical way, but I can pretty much guarantee you that you wouldn't be able to look at the code that generates them and just write down the filter transfer function. -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Reply by ●March 5, 20102010-03-05
gretzteam wrote:> Hi, > I've been playing with wave lattice filters for a little while and only > yesterday realized that they were just a structure used to implement the > same good old transfer function that maps nicely onto the Direct Form I, > II, transposed etc. For some reason I thought they were a different beast.It usually helps to read the textbook first.> Now I wonder how do people come up with new structures. The mapping from > the transfer function to a lattice wave filter is not obvious at all! Let's > say that I couldn't just start from DF1 and come up with a lattice > structure...Sometimes it could be useful to mimic the topology of analog filters. BTW, lattice filter originates from analog filter, too.> Are there still new structures being found yielding power or area > advantages?Well, in some cases you can get by lesser numeric precision or avoid some multiplications in favor of shifts and additions.> Are there any book or research on this topic?Any book on filter design has examples of different structures. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●March 5, 20102010-03-05
Tim Wescott <tim@seemywebsite.now> wrote: (snip)> Sometimes, an effort to circumvent a patent leads to a > superior solution (I wish I had an example to hand).Superior or not, the only one I remember is cobalt doped ferric oxide to circumvent the patent on chromium dioxide. I do remember having a real BASF cassette with real CrO2, and that it wasn't as good as the other type II cassettes that I was using at the time. -- glen
Reply by ●March 5, 20102010-03-05
>So I don't think there are any techniques out there that are going to >offer astounding improvements, and techniques that _do_ offer >improvements are going to do so at the cost of easy understandability. >Keep in mind that the "direct form" filters have the coefficients of the >polynomials as gains -- I'm pretty sure that's why they're called >"direct form". There may be filters out there that are significantly >'better' in some technical way, but I can pretty much guarantee you that >you wouldn't be able to look at the code that generates them and just >write down the filter transfer function. > >-- >Tim Wescott >Control system and signal processing consulting >www.wescottdesign.com >Hi all, Thanks for the 'philosophical' answers. Now does anybody know of a group doing active research on finding new structures? Thanks
Reply by ●March 5, 20102010-03-05
On 5 Mar, 19:54, "gretzteam" <gretzt...@yahoo.com> wrote:> Now does anybody know of a group doing active research on finding new > structures?Why would anyone want to? The lattice / ladder structures date back at least to the '60s / '70s; possibly a lot further. If there is anything at all going on, it would be in the realm of Kalman'ish filters, like uncented KFs, H_inf or particle filters. This stuff on filter structures is *ancient*. Rune
Reply by ●March 5, 20102010-03-05
Rune Allnor wrote:> On 5 Mar, 19:54, "gretzteam" <gretzt...@yahoo.com> wrote: > >> Now does anybody know of a group doing active research on finding new >> structures? > > Why would anyone want to? The lattice / ladder structures > date back at least to the '60s / '70s; possibly a lot further. > If there is anything at all going on, it would be in the realm > of Kalman'ish filters, like uncented KFs, H_inf or particle > filters. > > This stuff on filter structures is *ancient*.You still see papers in the IEEE Circuits & Systems transactions, mainly having to do with clever ways to implement them in full-custom silicon. Perhaps the OP should be asking if there's a good _review_. A Kalman filter is orthogonal to what (I think) the OP wants -- the Kalman design methodology finds the best behavioral model of a filter given some system description and optimization rule; the OP seems to be looking for different filter topologies to realize a known transfer function. For that matter, you could easily want a filter that is both Kalman and is structured to best use the available numerical precision. In that case one may want to take a Kalman filter and rearrange it's terms to better leverage the numeric precision of the processor. -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
