Hello, I would just like to clarify what people mean when they say a signal is colored? My assumption is it means white gaussian noise is added to the signal, but I just want this to be clarified. Thank you.
What does colored signal mean?
Started by ●November 2, 2010
Reply by ●November 2, 20102010-11-02
On 11/02/2010 10:40 AM, pacman101 wrote:> Hello, > > I would just like to clarify what people mean when they say a signal is > colored? > > My assumption is it means white gaussian noise is added to the signal, but > I just want this to be clarified."Colored" in this context means that the signal spectrum is shaded to one side or another. In audio usage folks will actually attach more specific adjectives: "red" noise has a 20dB/decade dropoff, as if white noise were run through an integrator; "pink" noise has a 10dB/decade dropoff, simulating flicker noise. Sometimes you'll see other 'colors' defined, but that's getting out there. I've not seen the term applied to a signal -- usually you see it applied to noise, and it just means that the autocorrelation of the noise has some extent in time, where white noise has an autocorrelation function that is just an impulse. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●November 2, 20102010-11-02
>On 11/02/2010 10:40 AM, pacman101 wrote: >> Hello, >> >> I would just like to clarify what people mean when they say a signal is >> colored? >> >> My assumption is it means white gaussian noise is added to the signal,but>> I just want this to be clarified. > >"Colored" in this context means that the signal spectrum is shaded to >one side or another. In audio usage folks will actually attach more >specific adjectives: "red" noise has a 20dB/decade dropoff, as if white >noise were run through an integrator; "pink" noise has a 10dB/decade >dropoff, simulating flicker noise. Sometimes you'll see other 'colors' >defined, but that's getting out there. > >I've not seen the term applied to a signal -- usually you see it applied >to noise, and it just means that the autocorrelation of the noise has >some extent in time, where white noise has an autocorrelation function >that is just an impulse.Whitening and pre-whitening filters abound, and they are usually whitening a signal, and not its associated noise. If we are whitening something, we must have considered it coloured to begin with. Steve
Reply by ●November 2, 20102010-11-02
Tim Wescott <tim@seemywebsite.com> wrote:> On 11/02/2010 10:40 AM, pacman101 wrote:>> I would just like to clarify what people mean when they say >> a signal is colored?(snip)> "Colored" in this context means that the signal spectrum is shaded to > one side or another. In audio usage folks will actually attach more > specific adjectives: "red" noise has a 20dB/decade dropoff, as if white > noise were run through an integrator; "pink" noise has a 10dB/decade > dropoff, simulating flicker noise. Sometimes you'll see other 'colors' > defined, but that's getting out there.> I've not seen the term applied to a signal -- usually you see it applied > to noise, and it just means that the autocorrelation of the noise has > some extent in time, where white noise has an autocorrelation function > that is just an impulse.I believe I have seen the term applied to audio signals, though I am not so sure when or where. One might use the term uncolored for an audio signal/system with appropriately flat frequency response. That is, the signal is uncolored when the system is flat. (Especially as the signal being flat has a different meaning.) -- glen
Reply by ●November 2, 20102010-11-02
On Nov 3, 6:40�am, "pacman101" <janpac01@n_o_s_p_a_m.yahoo.com> wrote:> Hello, > > I would just like to clarify what people mean when they say a signal is > colored? � > > My assumption is it means white gaussian noise is added to the signal, but > I just want this to be clarified. > > Thank you.It means they have spent coloured wrongly. Hardy
Reply by ●November 2, 20102010-11-02
On Nov 2, 2:57�pm, HardySpicer <gyansor...@gmail.com> wrote:> On Nov 3, 6:40�am, "pacman101" <janpac01@n_o_s_p_a_m.yahoo.com> wrote: > > > Hello, > > > I would just like to clarify what people mean when they say a signal is > > colored? � > > > My assumption is it means white gaussian noise is added to the signal, but > > I just want this to be clarified. > > > Thank you. > > It means they have spent coloured wrongly. > > HardyIrony? :) To address the OP, white is used to describe signals with a flat power spectral density throughout the frequency range of interest. The term white of course comes from our understanding (somewhat of a misconception) of visible light, where perfectly white light is a superposition of all visible light frequencies with equal weighting (that is equal amplitude not necessarily phase). Of course in reality, while what I just said is true, we only need to have the three different types of cones in our eyes stimulated equally to see white light, so there are theoretically infinite number of combinations of frequencies that can produce white light. But I digress... Colored, in the sense of "colored noise" or a "coloring filter" is anything that does not have a flat power spectral density, or a filter that takes a white signal in, and produces a non-white output. This distortion would be considered "coloring" the signal.
Reply by ●November 3, 20102010-11-03
>On Nov 2, 2:57=A0pm, HardySpicer <gyansor...@gmail.com> wrote: >> On Nov 3, 6:40=A0am, "pacman101" <janpac01@n_o_s_p_a_m.yahoo.com>wrote:>> >> > Hello, >> >> > I would just like to clarify what people mean when they say a signalis>> > colored? =A0 >> >> > My assumption is it means white gaussian noise is added to the signal,=>but >> > I just want this to be clarified. >> >> > Thank you. >> >> It means they have spent coloured wrongly. >> >> Hardy > >Irony? :) > >To address the OP, white is used to describe signals with a flat power >spectral density throughout the frequency range of interest. The term >white of course comes from our understanding (somewhat of a >misconception) of visible light, where perfectly white light is a >superposition of all visible light frequencies with equal weighting >(that is equal amplitude not necessarily phase). Of course in reality, >while what I just said is true, we only need to have the three >different types of cones in our eyes stimulated equally to see white >light, so there are theoretically infinite number of combinations of >frequencies that can produce white light. But I digress... > >Colored, in the sense of "colored noise" or a "coloring filter" is >anything that does not have a flat power spectral density, or a filter >that takes a white signal in, and produces a non-white output. This >distortion would be considered "coloring" the signal. >I am trying to relate what colored signals mean in RF. One of the reasons this question is important to me is that literature states that in adaptive filtering, when colored signals are the signals passing through the reference channel, the performance of LMS and NLMS is poor. I am trying noise cancellation in RF and I noticed that when I put an attenuator in my reference channel, the performance is worse. As of right now I am using bandlimited white gaussian noise as the interferer to cancel out. So can coloring be applied to RF?
Reply by ●November 3, 20102010-11-03
>>On Nov 2, 2:57=A0pm, HardySpicer <gyansor...@gmail.com> wrote: >>> On Nov 3, 6:40=A0am, "pacman101" <janpac01@n_o_s_p_a_m.yahoo.com> >wrote: >>> >>> > Hello, >>> >>> > I would just like to clarify what people mean when they say a signal >is >>> > colored? =A0 >>> >>> > My assumption is it means white gaussian noise is added to thesignal,>= >>but >>> > I just want this to be clarified. >>> >>> > Thank you. >>> >>> It means they have spent coloured wrongly. >>> >>> Hardy >> >>Irony? :) >> >>To address the OP, white is used to describe signals with a flat power >>spectral density throughout the frequency range of interest. The term >>white of course comes from our understanding (somewhat of a >>misconception) of visible light, where perfectly white light is a >>superposition of all visible light frequencies with equal weighting >>(that is equal amplitude not necessarily phase). Of course in reality, >>while what I just said is true, we only need to have the three >>different types of cones in our eyes stimulated equally to see white >>light, so there are theoretically infinite number of combinations of >>frequencies that can produce white light. But I digress... >> >>Colored, in the sense of "colored noise" or a "coloring filter" is >>anything that does not have a flat power spectral density, or a filter >>that takes a white signal in, and produces a non-white output. This >>distortion would be considered "coloring" the signal. >> > >I am trying to relate what colored signals mean in RF. One of thereasons>this question is important to me is that literature states that inadaptive>filtering, when colored signals are the signals passing through the >reference channel, the performance of LMS and NLMS is poor. I am trying >noise cancellation in RF and I noticed that when I put an attenuator inmy>reference channel, the performance is worse. As of right now I am using >bandlimited white gaussian noise as the interferer to cancel out. So can >coloring be applied to RF?Yes. Usually when applied to a comms channel, and the flatness of the signal spectral across that channel. If you have been reading about LMS and NLMS you should have already found why a reasonably flat spectrum around the adaption loop matters. Steve Steve
Reply by ●November 3, 20102010-11-03
>>>On Nov 2, 2:57=A0pm, HardySpicer <gyansor...@gmail.com> wrote: >>>> On Nov 3, 6:40=A0am, "pacman101" <janpac01@n_o_s_p_a_m.yahoo.com> >>wrote: >>>> >>>> > Hello, >>>> >>>> > I would just like to clarify what people mean when they say asignal>>is >>>> > colored? =A0 >>>> >>>> > My assumption is it means white gaussian noise is added to the >signal, >>= >>>but >>>> > I just want this to be clarified. >>>> >>>> > Thank you. >>>> >>>> It means they have spent coloured wrongly. >>>> >>>> Hardy >>> >>>Irony? :) >>> >>>To address the OP, white is used to describe signals with a flat power >>>spectral density throughout the frequency range of interest. The term >>>white of course comes from our understanding (somewhat of a >>>misconception) of visible light, where perfectly white light is a >>>superposition of all visible light frequencies with equal weighting >>>(that is equal amplitude not necessarily phase). Of course in reality, >>>while what I just said is true, we only need to have the three >>>different types of cones in our eyes stimulated equally to see white >>>light, so there are theoretically infinite number of combinations of >>>frequencies that can produce white light. But I digress... >>> >>>Colored, in the sense of "colored noise" or a "coloring filter" is >>>anything that does not have a flat power spectral density, or a filter >>>that takes a white signal in, and produces a non-white output. This >>>distortion would be considered "coloring" the signal. >>>> >Yes. Usually when applied to a comms channel, and the flatness of the >signal spectral across that channel. > >If you have been reading about LMS and NLMS you should have already found >why a reasonably flat spectrum around the adaption loop matters. > >Steve > > >Steve >So would it be better if I go with Affine Projection? How about Gradient-adaptive learning algorithms? I haven't thoroughly studied these more complex algorithms yet.
Reply by ●November 3, 20102010-11-03
On Nov 3, 6:28�am, "pacman101" <janpac01@n_o_s_p_a_m.yahoo.com> wrote:> >On Nov 2, 2:57=A0pm, HardySpicer <gyansor...@gmail.com> wrote: > >> On Nov 3, 6:40=A0am, "pacman101" <janpac01@n_o_s_p_a_m.yahoo.com> > wrote: > > >> > Hello, > > >> > I would just like to clarify what people mean when they say a signal > is > >> > colored? =A0 > > >> > My assumption is it means white gaussian noise is added to the signal, > = > >but > >> > I just want this to be clarified. > > >> > Thank you. > > >> It means they have spent coloured wrongly. > > >> Hardy > > >Irony? :) > > >To address the OP, white is used to describe signals with a flat power > >spectral density throughout the frequency range of interest. The term > >white of course comes from our understanding (somewhat of a > >misconception) of visible light, where perfectly white light is a > >superposition of all visible light frequencies with equal weighting > >(that is equal amplitude not necessarily phase). Of course in reality, > >while what I just said is true, we only need to have the three > >different types of cones in our eyes stimulated equally to see white > >light, so there are theoretically infinite number of combinations of > >frequencies that can produce white light. But I digress... > > >Colored, in the sense of "colored noise" or a "coloring filter" is > >anything that does not have a flat power spectral density, or a filter > >that takes a white signal in, and produces a non-white output. This > >distortion would be considered "coloring" the signal. > > I am trying to relate what colored signals mean in RF. �One of the reasons > this question is important to me is that literature states that in adaptive > filtering, when colored signals are the signals passing through the > reference channel, the performance of LMS and NLMS is poor. �I am trying > noise cancellation in RF and I noticed that when I put an attenuator in my > reference channel, the performance is worse. �As of right now I am using > bandlimited white gaussian noise as the interferer to cancel out. �So can > coloring be applied to RF?Why does it make a difference if it's "RF"? RF is just a blanket term to describe (very loosely) a particular frequency range. What I described above made no mention of specific frequency ranges, only power spectral density shape.
