DSPRelated.com
Forums
Online · Live + on demand
February 9-11, 2027
Price goes up Jan 1: register now to lock in today's rate
Register Now →

Greg Berchin's "Precise Filter Design" in IEEE Signal Processing Magazine

Started by Randy Yates ●December 19, 2006
As a long standing member of IEEE and IEEE signal processing society, I
do not think paying 420 USD / year to download 25x12 articles is too
expensive, a rip off or in any other way financially oppresive. Think
about it. For that price one receives:

* The results of research I did not have to conduct or pay for.
* The expertise not just of university professors but those in
industry.
* The convenience of browsing and downloading papers dating at least as
far back as 1988 and often as far back as the 1960's (classic papers).
* Remember when you had to go to a special library for these articles?
Now they are available at your desktop with a search capability.
Someone had to scan these into a PDF and meta tag them according to
topic, title, author, date, abstract, etc. Someone has to maintain the
web site and adminster the service. Those people need to eat and be
clothed just like you.

The only suggestions I have regarding the IEEE explore service is
* That my quota of 25 papers not reset every month. I would very much
like to "carry over" the number of articles I don't download to future
months when I may need more allowance.
* A few publications do not appear covered by the subscription (e.g.
IEEE Letters). Let's include all the IEEE publications.

So, If it is of use to you and has value, then pay for it.    Just pay
for it. 

-V, rank-and-file engineer

Viognier wrote:
> As a long standing member of IEEE and IEEE signal processing society, I > do not think paying 420 USD / year to download 25x12 articles is too > expensive, a rip off or in any other way financially oppresive. Think > about it. For that price one receives: > > * The results of research I did not have to conduct or pay for. > * The expertise not just of university professors but those in > industry. > * The convenience of browsing and downloading papers dating at least as > far back as 1988 and often as far back as the 1960's (classic papers). > * Remember when you had to go to a special library for these articles? > Now they are available at your desktop with a search capability. > Someone had to scan these into a PDF and meta tag them according to > topic, title, author, date, abstract, etc. Someone has to maintain the > web site and adminster the service. Those people need to eat and be > clothed just like you. > > The only suggestions I have regarding the IEEE explore service is > * That my quota of 25 papers not reset every month. I would very much > like to "carry over" the number of articles I don't download to future > months when I may need more allowance. > * A few publications do not appear covered by the subscription (e.g. > IEEE Letters). Let's include all the IEEE publications. > > So, If it is of use to you and has value, then pay for it. Just pay > for it. > > -V, rank-and-file engineer >
I'd still like to see something in between the $10+/paper (I'm pretty sure I paid $13 for a paper last month) and 300 papers/year for $420. That 300 papers/year works out to less than $1.50 per paper. I can't read that many papers in a year and still get my job done. If I download one paper a month, its $42/paper with the subscription. It'd be nice to have an intermediate level, or a lower fee for the low volume user. Surely if a volume user can get the papers for $1.50, the cost to deliver to a low volume user can be lower.
If I suscribe to the IEEE Member Digital Library (MDL) for USD 420/year
(I do) and download 300 papers (I don't), that gives me a cost of $1.40
per paper, which I too consider reasonable.  But 300 papers/year
(actually a max of 25/month) is quite a lot, and does not take into
account the actual needs of working engineers.  For those who need much
less, the cost/paper is much higher.
   - My own needs vary according to project phases.  In many months, I
need none; in others I need 1-4, and some months I've downloaded as
many as 16.  But the subscription payment of $35/month is required in
order to maintain this access, so I can't pay for just the months I
need to research the Library.  (There's a 6 month minimum subscription,
and thereafter, it's month-to-month.)
   - But I'll guess that my needs are more research-oriented than most
engineers and that many more would have a need for, say 20/year or
fewer.  A lower-cost, lower-paper-count, annual (not per-month)
subscription would be of value to many more engineers than the current
MDL. (See NetFlix for a model.  They have subscriptions geared to DVD
renters from occasional viewers to junkies.)
   - I like Rick's proposal (several papers/month free with IEEE
membership, more at modest cost) even better.  That sounds too good to
be practical, but I support changes to make this great online resource
available at more reasonable costs.

MY MDL subscription is for my personal use only.  I can't share a
downloaded paper with colleagues.  IEEE Enterprise (company
subscriptions) pricing probably works for large or even medium-sized
companies, but is just of reach for a small company.
   - My current employer has just 28 employees total, of whom 17 are
engineers, of whom perhaps 5 would ever want to use Xplore, and no more
than 3 would use it very much at all.  The smallest Enterprise
subscription is USD 2995/year.
   - And I've worked for smaller companies than that!

cheers,
  jerry

Rick Lyons wrote:

> I AGREE COMPLETELY. > > The IEEE charges *FAR FAR* too much money for their > articles. (That's a shame too because the > the authors, the editors, and the article reviewers > are *ALL* volunteers!) > > IEEE members should be able to download five articles > per month for free. Beyond that an article should cost, > say, a dollar.
That sounds like what I'm after. Five free downloads a month would be perfect for me.
> As far as I can tell, IEEE publications are predominantly > controlled by university professors. Their > articles and papers are overwhelmingly written > by, and for, university professors. That's why, > years ago, I fought to create the "DSP Tips & Tricks" > column for the IEEE Sig. Proc. magazine (SPM).
If I did resign from the IEEE, your "DSP Tips & Tricks" column is the only one I would really miss.
> I wanted to see articles published that would be > useful to working DSP engineers, instead of the > standard "Isomorphic Decomposition in Tight Gabor > Frames" stuff that the IEEE typically publishes.
IEEE Transactions on Signal Processing does have some articles of interest fro practical working engineers, but you're right, the majority are completely academic with little or no practical application.
> If you guys are actually willing to DO SOMETHING > to influence the IEEE to reduce their "article > download" cost, then we should mount a massive > E-mail campaign. I can find the appropriate > IEEE Officers' E-mail addresses if you wish. > Then again, if you're too lazy to do anything > to influence the IEEE then you get what you get.
I'm in. This has annoyed me for far too long to ignore it any longer. Erik -- +-----------------------------------------------------------+ Erik de Castro Lopo +-----------------------------------------------------------+ "You can have quality software, or you can have pointer arithmetic; but you cannot have both at the same time." -- Bertrand Meyer, 1989
Viognier wrote:

> As a long standing member of IEEE and IEEE signal processing society, I > do not think paying 420 USD / year to download 25x12 articles is too > expensive, a rip off or in any other way financially oppresive.
Yep, 300 papers for $420 is bloody cheap. If I could pay $1.40 per paper for the ones I want to read I would be over the moon happy. However, I'm a working engineer who does plain old software with a small amount of DSP. When my work mainly involves standard software I can go for 6-9 months without even thinking about accessing the IEEE and then I start a new project that involves some DSP and suddenly I want access to 20-30 papers over a 3 month period. The IEEE currently offers two options: - $420 for 300 papers - $10 per paper and neither of these fits my needs at all well and I suspect there are a huge number of IEEE members for which this is also true. What I am asking for is a more options that would allow engineers like myself to access this resource.
> Think > about it. For that price one receives: > > * The results of research I did not have to conduct or pay for. > * The expertise not just of university professors but those in > industry.
Yep, and you seem to forget that large numbers of those papers are completely unfathomable and its not always easy to tell just from the abstract. One of the main reasons I am unhappy about paying $10 per paper is because some papers turn out to be completely useless for my purposes.
> * The convenience of browsing and downloading papers dating at least as > far back as 1988 and often as far back as the 1960's (classic papers).
I have paid $10 for some of the classics and been very happy to do so.
> * Remember when you had to go to a special library for these articles?
Yep, my university had them. All I had to do was walk to the library. As a working engineer I don't have that luxury any more.
> Now they are available at your desktop with a search capability. > Someone had to scan these into a PDF and meta tag them according to > topic, title, author, date, abstract, etc. Someone has to maintain the > web site and adminster the service. Those people need to eat and be > clothed just like you.
Firstly, the authors of the paper work for free, the reviewers of the paper work for free and the editors of the journals work for free. No, I don't expect the web site maintainers and administrators to work for free but running a site the size of IEEE Xplore shouldn't require more than 5-10 people. Secondly, I know that University I went to was paying over $10000 a year for subscriptions to IEEE publications (that figure from a decade ago). Multiply that by a couple of hundred universities around the world plus the thousands of engineers that pay the $420 and theres a lot of money. Where does all that money go? My last point on this topic is that if the IEEE halves the price of the current yearly subscription they may well get twice as many subscribers. That would be good for the IEEE, good for the IEEE members and good for our profession. Erik -- +-----------------------------------------------------------+ Erik de Castro Lopo +-----------------------------------------------------------+ "A typical software project can present more opportunities to learn from mistakes than some people get in a lifetime." -- Steve McConnell (in Rapid Development)
On Wed, 20 Dec 2006 17:47:29 -0500, Randy Yates <yates@ieee.org> wrote:

>After my iniial post, I began wondering about that. I would take it >as an opportunity to investigate the algorithm further and perhaps >gain some new insights.
Okay; I figured out what caused the instability. It was not anything unusual about the transfer function, and it was not the fault of the algorithm. It was the fault of the programmer. And this was not one of those bugs where I programmed the right thing the wrong way. No, this was one where I programmed the wrong thing the right way. The original FDLS program was written in Fortran. Many years ago I rewrote it in C. A few months ago I rewrote it again in Matlab. In the Matlab rewrite I optimized-out something that I shouldn't have. I had to go all the way back to my master's thesis to figure out why the optimization can't work. Anyway, anybody who wants a copy of the corrected Matlab program can email me -- remove the "i" from "comicast" in my email address. To everyone who received copies of the program in the past, I will send a corrected copy to you if I can still find your email address. Anything before version 2.00 is now obsolete. Rick Lyons, if you're reading this; you'll receive a new copy, too. The change does not affect either of the articles. I expect to do this over the next day or so. -- Greg
Greg Berchin <gberchin@comicast.net> wrote in 
news:ug1mo2pptopg7301r6oskrv2uhf8u66vu8@4ax.com:

> On Wed, 20 Dec 2006 17:47:29 -0500, Randy Yates <yates@ieee.org> wrote: > >>After my iniial post, I began wondering about that. I would take it >>as an opportunity to investigate the algorithm further and perhaps >>gain some new insights. > > Okay; I figured out what caused the instability. It was not anything > unusual about the transfer function, and it was not the fault of the > algorithm. It was the fault of the programmer. And this was not one
of
> those bugs where I programmed the right thing the wrong way. No, this > was one where I programmed the wrong thing the right way. > > The original FDLS program was written in Fortran. Many years ago I > rewrote it in C. A few months ago I rewrote it again in Matlab. In
the
> Matlab rewrite I optimized-out something that I shouldn't have. I had > to go all the way back to my master's thesis to figure out why the > optimization can't work. > > Anyway, anybody who wants a copy of the corrected Matlab program can > email me -- remove the "i" from "comicast" in my email address. To > everyone who received copies of the program in the past, I will send a > corrected copy to you if I can still find your email address. Anything > before version 2.00 is now obsolete. > > Rick Lyons, if you're reading this; you'll receive a new copy, too.
The
> change does not affect either of the articles. > > I expect to do this over the next day or so. > > -- Greg >
I just received my copy of the Tips & Tricks article today and I also downloaded the program from the link provided in the article. Are these files correct? I think it would be interesting to see as an example, the A weight filter you just posted. This is such as classic use of a program like yours since the desired function is defined by standard in the s domain. We are expanding the DSP resources on our site and we would be glad to post your files if you like. -- Al Clark Danville Signal Processing, Inc. -------------------------------------------------------------------- Purveyors of Fine DSP Hardware and other Cool Stuff Available at http://www.danvillesignal.com
On Fri, 22 Dec 2006 00:41:14 GMT, Al Clark <dsp@danvillesignal.com>
wrote:

>I just received my copy of the Tips & Tricks article today and I also >downloaded the program from the link provided in the article. > >Are these files correct?
You get version 1.03 from the download site. That version is fine as long as the "artificial delay" parameter is kept to zero. Otherwise it will lead to instability. Version 2.00 solves that problem. I'll send you a copy in the next day or so.
>I think it would be interesting to see as an example, the A weight filter >you just posted. This is such as classic use of a program like yours >since the desired function is defined by standard in the s domain.
I'll try to make that happen. It may have to wait until after the holidays.
>We are expanding the DSP resources on our site and we would be glad to >post your files if you like.
Thanks. -- Greg
Erik de Castro Lopo wrote:
> Yep, 300 papers for $420 is bloody cheap....
Erik, Point taken. I see now that perhaps there should be other pay-as-you-go options. I too have downloaded "dud" articles based on the abstract and at 10USD a pop, I can see how that could be expensive and frustrating. Like you, my work has me downloading articles in bursts and there are many a month where the 25 article quota lies mostly untapped. The underlying theme of my previous post was to combat a pervasive Internet-culture sentiment that everything electronic (software, music, articles, information, source code) should be free. Though I don't subscribe at all to an Atlas-Shrugged philosophy, I do believe that ".. a worker is worth his keep" and that we should be careful to give credit where it is due. I hope you would agree. -V
"Viognier" <viognier@cox.net> writes:
> [...] > The underlying theme of my previous post was to combat a pervasive > Internet-culture sentiment that everything electronic (software, music, > articles, information, source code) should be free.
I agree with the sentiment, but our IEEE membership, and the societies we subscribe to, aren't free! I'm currently paying around $300/year. Isn't that worth more than a monthly subscription to Spectrum and Signal Processing Magazine? I would be happy with this for general membership: the first 10 papers a year, free. After that, $3.00/download. Perhaps membership in a society should entitle more for the papers of that society. -- % Randy Yates % "Maybe one day I'll feel her cold embrace, %% Fuquay-Varina, NC % and kiss her interface, %%% 919-577-9882 % til then, I'll leave her alone." %%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO http://home.earthlink.net/~yatescr