Hi Everybody, I am looking for Information about implementing a Digital-hp-Filter to eliminate all frecuencies under 2 Hz. The sample frecuency is 100kHz. We want to analyze signals between 2Hz an 10kHz. I do not know if it is possible to design such a filter. Does anybody have any idea or suggestion how to find a solution to this problem? thanks in advance - Albert
designing hp at 2Hz, Fs = 100kHz
Started by ●September 11, 2007
Reply by ●September 11, 20072007-09-11
"albert" <eoyaga@gmx.de> a �crit dans le message de news: 1189503905.222954.299930@r29g2000hsg.googlegroups.com...> Hi Everybody, > I am looking for Information about implementing a Digital-hp-Filter to > eliminate all frecuencies under 2 Hz. The sample frecuency is 100kHz. > We want to analyze signals between 2Hz an 10kHz. I do not know if it > is possible to design such a filter. > Does anybody have any idea or suggestion how to find a solution to > this problem? > thanks in advance > > - AlbertI've never actually tested this method but a strong decimation would I think make sense. Assume that S is your original signal, at 100Ksps, something like that : S1=lowpass(S, 10KHz) S1D=downsampling(S1, 30Ksps), giving DC-10KHz at 30Ksps S2=lowpass(S1D,100Hz) S2D=downsampling(S2,300sps), giving DC-100Hz at 300sps S3=lowpass(S2D,2Hz) S2U=upsampling(S3,30Ksps); giving DC-2Hz at 30Ksps S4=S1D-S3 S4 should be roughly your desired 2Hz-10KHz signal, and all filters are quite easy to implement (FIR for example). Of course this would need to be studied in mode details, on the phase relationship side in particular, but that's just an idea... Yours, Robert Lacoste www.alciom.com
Reply by ●September 11, 20072007-09-11
albert wrote:> Hi Everybody, > I am looking for Information about implementing a Digital-hp-Filter to > eliminate all frecuencies under 2 Hz. The sample frecuency is 100kHz. > We want to analyze signals between 2Hz an 10kHz. I do not know if it > is possible to design such a filter. > Does anybody have any idea or suggestion how to find a solution to > this problem? > thanks in advance > > - Albert >Be sure the frequency response of your data converter goes all the way to DC (or 2 Hz if such a thing exists). You'll need a "special" A/D converter to do that. Also make sure that whoever is doing the signal chain is in tune with what you're doing so they don't shoot you in the foot by inserting a DC-blocking cap or something like that. What sort of analysis do you plan to do? You could potentially decimate by 10 to get a 10 kHz signal and then do a 10k point FFT to get 1 Hz resolution. Will that work? If you plan to do any filtering down at 2 Hz you'll need a really long filter! Brad
Reply by ●September 11, 20072007-09-11
"albert" <eoyaga@gmx.de> wrote in message news:1189503905.222954.299930@r29g2000hsg.googlegroups.com...> Hi Everybody, > I am looking for Information about implementing a Digital-hp-Filter to > eliminate all frecuencies under 2 Hz. The sample frecuency is 100kHz. > We want to analyze signals between 2Hz an 10kHz. I do not know if it > is possible to design such a filter.The design of such a filter is no problem at all: a trivial IIR highpass. The implementation is little tricky: you will need a 64bit int or at least double precision float computations. Vladimir Vassilevsky DSP and Mixed Signal Consultant www.abvolt.com
Reply by ●September 11, 20072007-09-11
On Tue, 11 Sep 2007 06:57:55 -0500, Brad Griffis wrote:> albert wrote: >> Hi Everybody, >> I am looking for Information about implementing a Digital-hp-Filter to >> eliminate all frecuencies under 2 Hz. The sample frecuency is 100kHz. >> We want to analyze signals between 2Hz an 10kHz. I do not know if it >> is possible to design such a filter. >> Does anybody have any idea or suggestion how to find a solution to >> this problem? >> thanks in advance >> >> - Albert >> > > Be sure the frequency response of your data converter goes all the way > to DC (or 2 Hz if such a thing exists). You'll need a "special" A/D > converter to do that. Also make sure that whoever is doing the signal > chain is in tune with what you're doing so they don't shoot you in the > foot by inserting a DC-blocking cap or something like that.Eh? You must work with audio a lot. Just about all but audio A/D converters work just fine down to DC, with the exception of some DC offset problems. -- Tim Wescott Control systems and communications consulting http://www.wescottdesign.com Need to learn how to apply control theory in your embedded system? "Applied Control Theory for Embedded Systems" by Tim Wescott Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html
Reply by ●September 11, 20072007-09-11
On Tue, 11 Sep 2007 02:45:05 -0700, albert wrote:> Hi Everybody, > I am looking for Information about implementing a Digital-hp-Filter to > eliminate all frecuencies under 2 Hz. The sample frecuency is 100kHz. > We want to analyze signals between 2Hz an 10kHz. I do not know if it > is possible to design such a filter. > Does anybody have any idea or suggestion how to find a solution to > this problem? > thanks in advance > > - AlbertIf you plan on eliminating _all_ energy below 2Hz you can expect to have a filter that takes an infinitely long time to settle. Me, I'd figure out just how much signal I could stand down there, and design a filter that does that and nothing more. -- Tim Wescott Control systems and communications consulting http://www.wescottdesign.com Need to learn how to apply control theory in your embedded system? "Applied Control Theory for Embedded Systems" by Tim Wescott Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html
Reply by ●September 11, 20072007-09-11
"Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in news:guwFi.1948$4J3.1740@newssvr22.news.prodigy.net:> > "albert" <eoyaga@gmx.de> wrote in message > news:1189503905.222954.299930@r29g2000hsg.googlegroups.com... >> Hi Everybody, >> I am looking for Information about implementing a Digital-hp-Filter >> to eliminate all frecuencies under 2 Hz. The sample frecuency is >> 100kHz. We want to analyze signals between 2Hz an 10kHz. I do not >> know if it is possible to design such a filter. > > The design of such a filter is no problem at all: a trivial IIR > highpass. The implementation is little tricky: you will need a 64bit > int or at least double precision float computations. > > Vladimir Vassilevsky > DSP and Mixed Signal Consultant > www.abvolt.com > >Won't this be a fairly high-order filter?? <2Hz with a 100kHz sample frequency would have to drop like a hot rock! An FIR filter like that would have some unattractive properties, like ringing, and an IIR filter would have huge phase problems. If you don't need to implement this in real time, I'd filter and decimate, all the way down to around 200 Hz if you can. High-Pass filter this to give you your low-pass component. Take that result, interpolate it back up to the original sample rate, and subtract it from the original signal. -- Scott Reverse name to reply
Reply by ●September 11, 20072007-09-11
Scott Seidman wrote:>>>I am looking for Information about implementing a Digital-hp-Filter >>>to eliminate all frecuencies under 2 Hz. The sample frecuency is >>>100kHz.>>The design of such a filter is no problem at all: a trivial IIR >>highpass. The implementation is little tricky: you will need a 64bit >>int or at least double precision float computations. >> > > Won't this be a fairly high-order filter?? <2Hz with a 100kHz sample > frequency would have to drop like a hot rock!For IIR, the order is determined by how many dB of attenuation per octave is required. The sample rate is irrelevant. An FIR filter like that> would have some unattractive properties, like ringing, and an IIR filter > would have huge phase problems. > > If you don't need to implement this in real time, I'd filter and > decimate, all the way down to around 200 Hz if you can. High-Pass filter > this to give you your low-pass component. Take that result, interpolate > it back up to the original sample rate, and subtract it from the original > signal.Overcomplication. VLV
Reply by ●September 11, 20072007-09-11
On Tue, 11 Sep 2007 12:26:57 -0500, Vladimir Vassilevsky wrote:> Scott Seidman wrote: > >>>>I am looking for Information about implementing a Digital-hp-Filter >>>>to eliminate all frecuencies under 2 Hz. The sample frecuency is >>>>100kHz. > > >>>The design of such a filter is no problem at all: a trivial IIR >>>highpass. The implementation is little tricky: you will need a 64bit >>>int or at least double precision float computations. >>> >> >> Won't this be a fairly high-order filter?? <2Hz with a 100kHz sample >> frequency would have to drop like a hot rock! > > For IIR, the order is determined by how many dB of attenuation per > octave is required. The sample rate is irrelevant. >Well, the sample rate affects the precision you'll need to keep from underflow, and it'll have to be substantial in this case.> > An FIR filter like that >> would have some unattractive properties, like ringing, and an IIR >> filter would have huge phase problems. >> >> If you don't need to implement this in real time, I'd filter and >> decimate, all the way down to around 200 Hz if you can. High-Pass >> filter this to give you your low-pass component. Take that result, >> interpolate it back up to the original sample rate, and subtract it >> from the original signal. > > Overcomplication. >I wouldn't reject it out of hand -- I'd see how much computational resources were being dedicated to the filtering with an "uncomplicated" approach that met my spec, then I'd check to see if I could buy them back using the complicated method. -- Tim Wescott Control systems and communications consulting http://www.wescottdesign.com Need to learn how to apply control theory in your embedded system? "Applied Control Theory for Embedded Systems" by Tim Wescott Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html
Reply by ●September 11, 20072007-09-11
On Tue, 11 Sep 2007 12:26:57 -0500, Vladimir Vassilevsky wrote:> Scott Seidman wrote: > >>>>I am looking for Information about implementing a Digital-hp-Filter >>>>to eliminate all frecuencies under 2 Hz. The sample frecuency is >>>>100kHz. > > >>>The design of such a filter is no problem at all: a trivial IIR >>>highpass. The implementation is little tricky: you will need a 64bit >>>int or at least double precision float computations. >>> >> >> Won't this be a fairly high-order filter?? <2Hz with a 100kHz sample >> frequency would have to drop like a hot rock! > > For IIR, the order is determined by how many dB of attenuation per > octave is required. The sample rate is irrelevant. >Well, the sample rate affects the precision you'll need to keep from underflow, and it'll have to be substantial in this case.> > An FIR filter like that >> would have some unattractive properties, like ringing, and an IIR >> filter would have huge phase problems. >> >> If you don't need to implement this in real time, I'd filter and >> decimate, all the way down to around 200 Hz if you can. High-Pass >> filter this to give you your low-pass component. Take that result, >> interpolate it back up to the original sample rate, and subtract it >> from the original signal. > > Overcomplication. >I wouldn't reject it out of hand -- I'd see how much computational resources were being dedicated to the filtering with an "uncomplicated" approach that met my spec, then I'd check to see if I could buy them back using the complicated method. -- Tim Wescott Control systems and communications consulting http://www.wescottdesign.com Need to learn how to apply control theory in your embedded system? "Applied Control Theory for Embedded Systems" by Tim Wescott Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html
