What you can try is : If you know the frequency of the hum perfectly and if it is stable, - just take the FFT of the signal - delete the hum completely (E.g. 60Hz): - reconstruct this frequency by a Heuristic (E.g. make it equal to the mean of the close higher frequency and close lower frequency) - take the IFFT This will have very good result if the heuristic is good.
60 Hz Hum removal
Started by ●July 1, 2004
Reply by ●July 8, 20042004-07-08
Reply by ●July 8, 20042004-07-08
"lakie" <just_me_myself_and_i1@hotmail.com> wrote in message news:1089278685.13175@seven.kulnet.kuleuven.ac.be...> What you can try is : > > If you know the frequency of the hum perfectly and if it is stable, > - just take the FFT of the signal > - delete the hum completely (E.g. 60Hz): > - reconstruct this frequency by a Heuristic (E.g. make it equal to themean> of the close higher frequency and close lower frequency) > - take the IFFT > > This will have very good result if the heuristic is good. > >Well, that doesn't match the OP's problem. He said there is good information around 60Hz. Any static filter will remove that information along with the interference. I should note that there are more than one kind of adaptive filter or canceller and they are best described as a "system": You start with the desired "input" signal to the system. You want a cleaned up version at the "output" of the system. An adaptive filter is a component of the system that is there to achieve the cleanup. The input to the adaptive filter is not necessarily the "input" to the system mentioned above and the output of the adaptive filter is not necessarily the "output" of the system. Here are some examples: In general, broadband noise can't be cancelled but it can be filtered - because noise subtracted from noise results in the same noise energy as would noise added to noise ... unless they are corrrelated, either operation results in more noise. Most of these systems add/subtract the output of the adaptive filter from the system input. Whether the result of this addition/subtraction is viewed as the system output depends..... In a noise canceller, the adaptive filter part of the system receives an input that attempts to be a reasonable replica of the interference. The input is *not* the same as the "signal" or system "input". This replica doesn't have to be perfect in amplitude, phase or harmonic content. The adaptive filter adjusts the amplitude and phase of the replica so that it can be subtracted from the "signal" - thus removing the interference perhaps perfectly and not removing any of the desired signal ... unless the desired signal has components that are correlated with the interference. Here, the result of the addition/subtraction is the system output. If the input to the adaptive filter is the same as the system input then this structure becomes a "dynamic line subtractor" and all you get out is noise. Noticing this, gets us to the line enhancer: In a line enhancer, the adaptive filter part of the system receives a delayed version of the system input. The delay serves to decorrelate the broadband noise input to the filter as compared to the system input. Because the noise is uncorrelated, the best thing the filter can do is shut off so that noise isn't added to noise at the output of the adder/subtractor. In addition, it will cancel lines as mentioned above. Because the adaptive filter shuts off where there is only noise and because its output includes narrowband sinusoids, the output of the adaptive filter itself is the system output. Here the result of the addition/subtraction is *not* the system output but only an internal signal used for adaptation. The line enhancer then really looks like any other filter in that there is a single input and a single output. It's just that it's adaptive and serves to eliminate broadband noise. At any moment, it's just a filter. The noise canceller is more "complicated" because, as a system, it has two inputs. It's not a "filter" in the normal sense because it really is an adaptive subtractor - the output is not a linear transformation of one input as we normally think of a filter. The output is the superposition of two inputs - one of which is adapted to the output to minimize the output. So, if the reference input can be obtained reasonably, then a noise canceller can do a very good job. And, it doesn't remove signal energy at all - even if it's at exactly the same frequency as the interference for a brief time. (Signal dynamics are important in all this). Obviously if the interference and the signal of interest are at exactly the same frequency (thus relative phase) all the time, and are of the same relative amplitude all the time, then the canceller will cancel everything. Another way to say this is that the interference and the signal are highly correlated. A really simple example is a signal that is the sum of two sinusoids that are very close together in frequency - one of which is deemed interference. We have a good replica of the interference but don't know how its amplitude and phase relate to the actual interference in the signal. The adaptive filter will try to push the adder/subtractor output to zero using the reference. So, it will adjust the amplitude and phase of the reference so that it matches the amplitude and phase of the interference in the signal. When these are subtracted, the signal is free from the interference (except for dynamics in the interference that the filter can't track) and there is no impact on the signal at all. Fred
Reply by ●July 23, 20042004-07-23
"Jerry Avins" <jya@ieee.org> wrote in message news:40e4cee3$0$23332$61fed72c@news.rcn.com...> Tim Wescott wrote: > > ... > > > But I've become a luddite in my old age. > > You're learning. The only excuse for filtering is that the signal is > already acquired and you can't go back and do it over. An adaptive > filter may be useful if the interfering frequency drifts slowly, but > power lines rarely do. >Seems to me that an adaptive filter is just what you need - the LMS algorithm with the hum being fed to the reference input. I would use normailised LMS though instead of ordinanry LMS. For audio applications I would just high pass filter. Tom
Reply by ●August 6, 20042004-08-06
There was an article around 1970 that described a Widrow LMS adaptive filter for hum cancelation that claimed 105 dB of hum rejection. It was an analog filter and was simple to build and not critical in parts. I don't have a pointer to the article. Sorry. In article <1090573371.34587@radsrv1.tranzpeer.net>, "Tom" <somebody@knowherex.netgx> wrote:> >"Jerry Avins" <jya@ieee.org> wrote in message >news:40e4cee3$0$23332$61fed72c@news.rcn.com... >> Tim Wescott wrote: >> >> ... >> >> > But I've become a luddite in my old age. >> >> You're learning. The only excuse for filtering is that the signal is >> already acquired and you can't go back and do it over. An adaptive >> filter may be useful if the interfering frequency drifts slowly, but >> power lines rarely do. >> > >Seems to me that an adaptive filter is just what you need - the LMS >algorithm with the hum being fed to the reference input. I would use >normailised LMS though instead of ordinanry LMS. For audio applications I >would just high pass filter. > > > > >Tom > >
Reply by ●August 6, 20042004-08-06
George Bush wrote:> There was an article around 1970 that described a Widrow LMS adaptive filter > for hum cancelation that claimed 105 dB of hum rejection. It was an analog > filter and was simple to build and not critical in parts. I don't have a > pointer to the article. Sorry.Fascinating! Motor-driven capacitors and inductors? 105 dB is an excellent spec for any analog filter. It takes good shielding to achieve it. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●August 9, 20042004-08-09
On Fri, 06 Aug 2004 13:03:26 -0400, Jerry Avins <jya@ieee.org> wrote:>George Bush wrote: > >> There was an article around 1970 that described a Widrow LMS adaptive filter >> for hum cancelation that claimed 105 dB of hum rejection. It was an analog >> filter and was simple to build and not critical in parts. I don't have a >> pointer to the article. Sorry. > >Fascinating! Motor-driven capacitors and inductors? 105 dB is an >excellent spec for any analog filter. It takes good shielding to achieve >it.105dB is not achievable. It would require that the mains hum being cancelled was stable to a few ppm, a rather unrealistic assumption. Regards, Allan.
Reply by ●August 9, 20042004-08-09
Allan Herriman wrote:> On Fri, 06 Aug 2004 13:03:26 -0400, Jerry Avins <jya@ieee.org> wrote: > > >>George Bush wrote: >> >> >>>There was an article around 1970 that described a Widrow LMS adaptive filter >>>for hum cancelation that claimed 105 dB of hum rejection. It was an analog >>>filter and was simple to build and not critical in parts. I don't have a >>>pointer to the article. Sorry. >> >>Fascinating! Motor-driven capacitors and inductors? 105 dB is an >>excellent spec for any analog filter. It takes good shielding to achieve >>it. > > > 105dB is not achievable. It would require that the mains hum being > cancelled was stable to a few ppm, a rather unrealistic assumption. > > Regards, > Allan.Yeah. Considering the political implications, I didn't want to just come straight out and declare that George Bush was practicing mass deception. ("Fascinating!" often replaces "Bullshit!" in polite discourse.) Jerry -- ... the worst possible design that just meets the specification - almost a definition of practical engineering. .. Chris Bore ������������������������������������������������������������������������
Reply by ●August 10, 20042004-08-10
105 dB was what was claimed. I didn't test the circuit. If I remember correctly, it was suggested for hum calcellation in EKG/EEG applications. I think it was in conference proceedings and not a peer reviewed joiurnal. It's been almost thirty years since I read the article. I don;t recall any motor powered capacitors, it was in the primitive days of electronics. I chose the alias, George W. Bush, because he is, byoyond question and without question, the worst president in my memory. (Carter is a distant second.) My memory only goes back to Truman so I can't comment on some of the earlier presidents. I don't suppose you would believe in a design for an analog Kalman filter? In article <4117863c$0$5910$61fed72c@news.rcn.com>, Jerry Avins <jya@ieee.org> wrote:>Allan Herriman wrote: > >> On Fri, 06 Aug 2004 13:03:26 -0400, Jerry Avins <jya@ieee.org> wrote: >> >> >>>George Bush wrote: >>> >>> >>>>There was an article around 1970 that described a Widrow LMS adaptive filter > >>>>for hum cancelation that claimed 105 dB of hum rejection. It was an analog >>>>filter and was simple to build and not critical in parts. I don't have a >>>>pointer to the article. Sorry. >>> >>>Fascinating! Motor-driven capacitors and inductors? 105 dB is an >>>excellent spec for any analog filter. It takes good shielding to achieve >>>it. >> >> >> 105dB is not achievable. It would require that the mains hum being >> cancelled was stable to a few ppm, a rather unrealistic assumption. >> >> Regards, >> Allan. > >Yeah. Considering the political implications, I didn't want to just come >straight out and declare that George Bush was practicing mass deception. >("Fascinating!" often replaces "Bullshit!" in polite discourse.) > >Jerry
Reply by ●August 10, 20042004-08-10
George Bush wrote:> 105 dB was what was claimed. I didn't test the circuit. If I remember > correctly, it was suggested for hum calcellation in EKG/EEG applications. I > think it was in conference proceedings and not a peer reviewed joiurnal. It's > been almost thirty years since I read the article.It's been over 50 years since L. Ron Hubbard foisted Dianetics on us, so you're forgiven. Figure out, sometime, what a brightness 105 dB brighter than a firefly represents. That should explain our skepticism. (A lighthouse doesn't come close.) Jerry -- ... the worst possible design that just meets the specification - almost a definition of practical engineering. .. Chris Bore ������������������������������������������������������������������������
Reply by ●August 11, 20042004-08-11
Jerry Avins wrote:> George Bush wrote: > >> 105 dB was what was claimed. I didn't test the circuit. If I >> remember correctly, it was suggested for hum calcellation in EKG/EEG >> applications. I think it was in conference proceedings and not a peer >> reviewed joiurnal. It's been almost thirty years since I read the >> article. > > > It's been over 50 years since L. Ron Hubbard foisted Dianetics on us, so > you're forgiven. Figure out, sometime, what a brightness 105 dB brighter > than a firefly represents. That should explain our skepticism. (A > lighthouse doesn't come close.) > > JerryWhen I see a big dB number like that and assuming the author is honest, its often a case of a broad band ratio given that one of the components is narrow band. Signal power is typically proportional to 20 Log10 of bandwidth in Hz. Noise power is proportional to 10 log10 of bandwidth in Hz. The author might have meant that the power line noise is 105dB down referenced to the signal output power for the full band. There is likely some confusion here about the context of the 105dB number. It does not violate the bounds of credulity in all contexts.
