I have I and Q data that I want to low pass filter. From what I have read I would impliment the same LPF twice, one on I and one on Q. Is that the only way to do this, or is there some trick out there I have not found out about? Also I would assume my LPF must have linear phase? Thanks. -JF
Quadrature low pass filter question
Started by ●May 10, 2010
Reply by ●May 10, 20102010-05-10
On 10 Mai, 18:24, "jacobfenton" <jacob.fenton@n_o_s_p_a_m.gmail.com> wrote:> I have I and Q data that I want to low pass filter. From what I have read I > would impliment the same LPF twice, one on I and one on Q.Yes.> Is that the only > way to do this,Probably not.> or is there some trick out there I have not found out > about?There might be, but using the same filter on both the I and Q channels is both simple and convenient.> Also I would assume my LPF must have linear phase?Why? Rune
Reply by ●May 10, 20102010-05-10
Rune Allnor wrote:> On 10 Mai, 18:24, "jacobfenton" <jacob.fenton@n_o_s_p_a_m.gmail.com> > wrote: >> I have I and Q data that I want to low pass filter. From what I have read I >> would impliment the same LPF twice, one on I and one on Q. > > Yes. > >> Is that the only >> way to do this, > > Probably not.Probably not, but I can't imagine an easier or lower cost one.>> or is there some trick out there I have not found out >> about? > > There might be, but using the same filter on both the I and Q > channels is both simple and convenient. > >> Also I would assume my LPF must have linear phase? > > Why? > > RuneTell us about your application and we'll help you with answering Rune's last question. Sometimes it is vital that one's filters have linear phase, sometimes it matters not at all, and sometimes it's good but not necessary. -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Reply by ●May 10, 20102010-05-10
On May 10, 11:24=A0am, "jacobfenton" <jacob.fenton@n_o_s_p_a_m.gmail.com> wrote:>............ Is that the only way to do this,What you are getting into is multidimensional (n-dimensional filter) MIMO (I and Q channels) signal processing. I would suggest you look at Rajagopal, Ravikiran's dissertation from Ohio State, "An Algebraic View of Multidimensional Multiple-Input, Multiple-Output Finite Impulse Response Equalizers". He gives a good background overview, and has a good set of references. You may want to just implement two separate filters. Maurice Givens
Reply by ●May 11, 20102010-05-11
On 10 Mai, 22:11, maury <maury...@core.com> wrote:> On May 10, 11:24=A0am, "jacobfenton" > > <jacob.fenton@n_o_s_p_a_m.gmail.com> wrote: > >............ =A0 =A0Is that the only way to do this, > > What you are getting into is multidimensional (n-dimensional filter) > MIMO (I and Q channels) signal processing.I would agree with your suggestion that one gets into MIMO stuff if one accepts the quadrature filter as *both* an implementation of the complex-valued arithmetics of a complex-valued filter, *and* that the I and Q components are independent. In that case one needs to handle the cross terms between the 'real' and 'imaginary' components to emulate the complex-valued filter, which means one need to get into some sort of MIMO structure. Or use complex-valued arithmetics. But the quadrature filter contains two real-valued sequences that usually (always?) are not independent, but have been derived as (emulations for) the real and imaginary components of a Hibert transform of one real-valued signal. Each of these real-valued signals must then be processed under the conditions of a real- valued signals. So far so good. As I understand it, the fact that the I and Q component are related through the HT, implicitly takes care of the cross-terms in the complex-valued arithmetics. Which means that one, simple, SISO real-valued filter is sufficient to handle the 'complex'- valued quadrature signal. The fact that the I and Q channels are processed individually as real-valued data, also have some implictaions. Quadrature filters are actually restricted by Nyquist's limit, Fs > 2B, whereas complex-valued filters are restricted by the far more forgiving Fs > B. Or have I misunderstood or missed out on something? Rune
Reply by ●May 11, 20102010-05-11
If you just don't want to implement the LPF twice, you can double your LPF system clock to complete the I and Q's LP process in only one filter. "jacobfenton" <jacob.fenton@n_o_s_p_a_m.gmail.com> д����Ϣ����:DrKdnXQKFfXUrnXWnZ2dnUVZ_qCdnZ2d@giganews.com...>I have I and Q data that I want to low pass filter. From what I have read I > would impliment the same LPF twice, one on I and one on Q. Is that the > only > way to do this, or is there some trick out there I have not found out > about? Also I would assume my LPF must have linear phase? > > Thanks. > > -JF
Reply by ●May 11, 20102010-05-11
On 5/11/2010 5:09 AM, Rune Allnor wrote: ...> The fact that the I and Q channels are processed individually as > real-valued data, also have some implictaions. Quadrature filters > are actually restricted by Nyquist's limit, Fs> 2B, whereas > complex-valued filters are restricted by the far more forgiving > Fs> B. > > Or have I misunderstood or missed out on something?It's the same thing. For complex signals, there are two samples per sample period: one real, the other imaginary. That is effectively sampling twice as fast, so the bandwidth is doubled. If one wants to get mathematically hoity-toity, one can say that real sampling covers the range from -fs/2 to +fs/2, while I/Q sampling covers the range from 0 to fs. Personally, I think that obscures the truth that there really /are/ twice as many samples. The second set of samples needn't be Q. d(Real)/dt, for example, would serve as well. Jerry -- "I view the progress of science as ... the slow erosion of the tendency to dichotomize." --Barbara Smuts, U. Mich. �����������������������������������������������������������������������
Reply by ●May 11, 20102010-05-11
On 5/11/2010 5:21 AM, Zhi.Shen wrote:> If you just don't want to implement the LPF twice, you can double your LPF > system clock to complete the I and Q's LP process in only one filter.Oh? How do you deal with perpetual start-up "transients"?> "jacobfenton"<jacob.fenton@n_o_s_p_a_m.gmail.com> д����Ϣ����:DrKdnXQKFfXUrnXWnZ2dnUVZ_qCdnZ2d@giganews.com... >> I have I and Q data that I want to low pass filter. From what I have read I >> would impliment the same LPF twice, one on I and one on Q. Is that the >> only >> way to do this, or is there some trick out there I have not found out >> about? Also I would assume my LPF must have linear phase? >> >> Thanks. >> >> -JFJerry -- "I view the progress of science as ... the slow erosion of the tendency to dichotomize." --Barbara Smuts, U. Mich. �����������������������������������������������������������������������
Reply by ●May 11, 20102010-05-11
Jerry Avins <jya@ieee.org> writes:> [...] > It's the same thing.So, e.g., using complex sampling doesn't allow you to gather more bandwidth using today's technology in ADCs than real sampling does? -- Randy Yates % "Maybe one day I'll feel her cold embrace, Digital Signal Labs % and kiss her interface, mailto://yates@ieee.org % til then, I'll leave her alone." http://www.digitalsignallabs.com % 'Yours Truly, 2095', *Time*, ELO
Reply by ●May 11, 20102010-05-11
"jacobfenton" <jacob.fenton@n_o_s_p_a_m.gmail.com> writes:> I have I and Q data that I want to low pass filter. From what I have read I > would impliment the same LPF twice, one on I and one on Q.That is true if your lowpass filter is a real filter. If it is a complex filter, you'd have more work to do.> Is that the only way to do this, or is there some trick out there I > have not found out about?Mathematically, it's the only way to get from here to there. However, there are infinite choices in how you design your filter and perform the convolutions.> Also I would assume my LPF must have linear phase?Depends on whether or not you need a linear-phase filter. If not, it doesn't have to be linear phase. Let me put it this way: there is nothing inherent in quadrature lowpass filtering that requires a linear-phase filter. -- Randy Yates % "She has an IQ of 1001, she has a jumpsuit Digital Signal Labs % on, and she's also a telephone." mailto://yates@ieee.org % http://www.digitalsignallabs.com % 'Yours Truly, 2095', *Time*, ELO
