On Sat, 22 Jun 2013 19:05:35 -0500, "Ro_ny" <95309@dsprelated> wrote:>Guys, I am ready to accept the opinion of each one of you, but only with >the above basic example, otherwise these are words in the air. > >Rony J.To be honest this sounds like you want someone to do the work for you. There are plenty of examples of code for filters, etc., in many tutorials and libraries on the internet in all sorts of programming languages. A bit of searching should turn them up. FWIW, there are also LOTS of other good tools for filter design. I don't do any filter design in Matlab or Octave or SciLab. I do all that in MathCAD because I find it far more useful to do so, but that's just my personal preference. You haven't even specified what sort of filter parameters are useful to you, so even if somebody did post filter design code, it may not meet your needs. Why should somebody go out of their way to do work that may not even be useful to you because you didn't specify what you needed? And please don't make your next step being a tighter spec on what you want. Please go do your own research now that people have been kind enough to get you started. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Octave or Scilab
Started by ●June 22, 2013
Reply by ●June 22, 20132013-06-22
Reply by ●June 22, 20132013-06-22
"Ro_ny" <95309@dsprelated> writes:> Guys, I am ready to accept the opinion of each one of you, but only with > the above basic example, otherwise these are words in the air.I'm with Eric. Besides, doing this basic example in each yourself will allow you to form your own opinion on which one you like better. -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●June 22, 20132013-06-22
On 6/22/2013 3:31 PM, Ro_ny wrote:> Hello people, > > For designing pulsed shaping and other filters on Windows 7 for Fixed-point > MCUs, which is better to download (the latest version of): Octave or > Scilab? > > And both export to a file the coefficients normalized to the needed format > e.g. Qm.n? > > PS: Octave is claimed to be much more compatible with Matlab. > > Rony J.A young man and young woman came to rabbi to ask for his advice. Man: I got some money. What should I do with it? Should I put it into bank? Invest into something? Or buy gold or real estate? Woman: Tomorrow is my first night with fiancee. What night dress should I put on? Should it be a long dress, or a short dress, or something open, or something closed? Rabbi: Dear children, I have one answer for both of you: whatever you do, you will be fucked. VLV
Reply by ●June 23, 20132013-06-23
On Sat, 22 Jun 2013 16:16:01 -0500, Ro_ny wrote:> Let’s go for Scilab. Can you show what to type there in order to have > the coefficients of a Gaussian FIR filter into a text file? (I’ll change > the constants afterwards)You're asking to be spoon-fed here, but since you're trying to assess a new program and not demanding that we do your homework for you*, I guess I'll go along. This is a copy of a session where I'm just typing at the command line; I'm not trying to write a program or anything. The result is the coefficients for a Gaussian FIR filter that's 81 taps long: -->ix = (-4:0.1:4); // Make an index -->x = exp(-ix^2); // Make the actual Gaussian filter -->x = x / sum(x); // Adjust for a DC gain of 1 -->clf; plot2d(x) // Always make a check plot -->fd = mopen("gauss.txt", 'w'); // Open a file for writing -->mfprintf(fd, "%f\n", x'); // Print the numbers as floating point. // The "x'" makes x into a column // vector, which causes mfprintf // to print the whole thing in one go, // which spares me from writing // a 'for' loop -->mclose(fd); // tidy up. This is almost identical to what you would do in Octave or Matlab to get the same result, except that I think those will automatically print all the elements of a row vector, as opposed to a column vector. This does _not_ get you your desired Qn.m format: that is left as an exercise for the reader -- 'cause I'm not going to spoon-feed you _that_ much unless you pay me. Note that you really should round the coefficients to match your filter precision (using round, floor, or ceil), then you should look at its response in the frequency domain (using either bode or the FFT). * Or if you are demanding that we do your homework for you, you're being very clever about it. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●June 23, 20132013-06-23
Thanks I'll try to find the Qm.n however in order to assess I need to see how for example GSM parameters (BT, O, N) are integrated here. Rony J. _____________________________ Posted through www.DSPRelated.com
Reply by ●June 23, 20132013-06-23
On Sun, 23 Jun 2013 04:13:51 -0500, Ro_ny wrote:> Thanks I'll try to find the Qm.n however in order to assess I need to > see how for example GSM parameters (BT, O, N) are integrated here.I think you're misunderstanding the utility of these packages. They're MATH packages. You're doing DSP. DSP is not math, but it includes a bunch of applied math -- and you need to do the part where you apply that math. Once you have done that, then Scilab or Octave or whatever will make the actual computations easy. What they won't do is your thinking for you. -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Reply by ●June 23, 20132013-06-23
On Sat, 22 Jun 2013 15:40:54 -0500, Tim Wescott <tim@seemywebsite.please> wrote:>On Sat, 22 Jun 2013 15:31:40 -0500, Ro_ny wrote: > >> Hello people, >> >> For designing pulsed shaping and other filters on Windows 7 for >> Fixed-point MCUs, which is better to download (the latest version of): >> Octave or Scilab? >> >> And both export to a file the coefficients normalized to the needed >> format e.g. Qm.n? >> >> PS: Octave is claimed to be much more compatible with Matlab. > >Octave is, indeed, much more compatible with Matlab. If this is an issue >it should weight your choices. > >I vastly prefer Scilab -- it's easier to use that Octave, and has some >object-oriented-ish features that are nice. But it's not trying to be a >Matlab clone. > >Either one will do what you want. They have the same basic functions >that would let you design your filters, and they both have formatted >print functions that will (with a bit of effort on your part) let you >print your values to a file just about any way you want to.Hi Tim, maybe someone's already said this, but one advantage of Octave is that it allows the re-use (perhaps with some modification) of an awful lot of signal processing Matlab code that can be downloaded from the web. [-Rick-]
Reply by ●June 23, 20132013-06-23
On Sun, 23 Jun 2013 10:12:55 -0700, Rick Lyons wrote:> On Sat, 22 Jun 2013 15:40:54 -0500, Tim Wescott > <tim@seemywebsite.please> wrote: > >>On Sat, 22 Jun 2013 15:31:40 -0500, Ro_ny wrote: >> >>> Hello people, >>> >>> For designing pulsed shaping and other filters on Windows 7 for >>> Fixed-point MCUs, which is better to download (the latest version of): >>> Octave or Scilab? >>> >>> And both export to a file the coefficients normalized to the needed >>> format e.g. Qm.n? >>> >>> PS: Octave is claimed to be much more compatible with Matlab. >> >>Octave is, indeed, much more compatible with Matlab. If this is an >>issue it should weight your choices. >> >>I vastly prefer Scilab -- it's easier to use that Octave, and has some >>object-oriented-ish features that are nice. But it's not trying to be a >>Matlab clone. >> >>Either one will do what you want. They have the same basic functions >>that would let you design your filters, and they both have formatted >>print functions that will (with a bit of effort on your part) let you >>print your values to a file just about any way you want to. > > Hi Tim, > maybe someone's already said this, but one > advantage of Octave is that it allows the re-use (perhaps with some > modification) of an awful lot of signal processing Matlab code that can > be downloaded from the web. > > [-Rick-]Good point. Scilab is supposed to have a conversion utility, but I don't have a great deal of faith in it. Mostly I do control stuff, and Scilab is mostly a lot better for that. If I'm going to be useful to a customer doing comms stuff, it's got to be some tiny part of some larger job, or it's got to be for some bizarre channel for which little prior art exists -- if it's a lot of ordinary stuff, I'm too hobbled by lack of familiarity with all the standards. (I did do a couple of jobs pertaining to cell phones -- but in that case it was extracting unusual information from the channels.) -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Reply by ●June 23, 20132013-06-23
>I think you're misunderstanding the utility of these packages. They're >MATH packages. You're doing DSP.I think this forum is destined first of all for DSP?:) Secondly, Matlab seems to have a command of gaussfir(bt,n,o) - what’s the equivalent of Scilab (and also Octave if should be similar) for it? Rony J. _____________________________ Posted through www.DSPRelated.com
Reply by ●June 23, 20132013-06-23
On Sun, 23 Jun 2013 16:08:29 -0500, Ro_ny wrote:>>I think you're misunderstanding the utility of these packages. They're >>MATH packages. You're doing DSP. > > I think this forum is destined first of all for DSP?:)Yes. And we get a number of people who confuse memorizing the function calls in the Matlab DSP toolbox with having a deep knowledge of DSP.> Secondly, Matlab seems to have a command of gaussfir(bt,n,o) - what’s > the equivalent of Scilab (and also Octave if should be similar) for it?I dunno -- but the code that I gave you makes a Gaussian FIR. Try loading Scilab and typing "help Gaussian". See what happens. Really, if you can't understand how to do the math for yourself, then you're going to have a steep uphill climb trying to do any real DSP work. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
