Jerry Avins wrote:> Symon wrote: > > ... > >> Check out >>http://www.saultstar.com/webapp/sitepages/content.asp?contentID=70613&catname=Local+News>> A small town in Canada where everyone's clocks gained 10 minutes. >> Cheers, Syms. > > ... > > Considering the natures of the affected clocks, my bet is on some > sort of noise. Not all clocks were affected, only those that count > zero crossings and have no flywheels. > > JerryNoise would certainly lead to different gains in different clocks, some going 7min wrong, others 10, some 2 1/2, ... If most of the clocks which were fooled, had an error of about 10min, it's probably not something which raises the noise floor. My guess: let's assume that power entered the clock not with 60Hz, but it just happened that it cycled with 70Hz between 1am and 2am. This results in 36000 additional cycles, which gives 10min. Most devices would neither detect it nor be damaged. Especially, if it had happened around midnight. The power company might not be interested to publish such an event, because it would mean that they eventually billed 16% too much from every customer during this event. Might be interesting, how power is generated/stabilized in that area. Were there other related events like power failures, brownouts around these days ??
60 Hz Hum removal
Started by ●July 1, 2004
Reply by ●July 5, 20042004-07-05
Reply by ●July 5, 20042004-07-05
Bernhard Holzmayer wrote:> Jerry Avins wrote: > > >>Symon wrote: >> >> ... >> >> >>>Check out >>> > > http://www.saultstar.com/webapp/sitepages/content.asp?contentID=70613&catname=Local+News > >>>A small town in Canada where everyone's clocks gained 10 minutes. >>>Cheers, Syms. >> >> ... >> >>Considering the natures of the affected clocks, my bet is on some >>sort of noise. Not all clocks were affected, only those that count >>zero crossings and have no flywheels. >> >>Jerry > > > Noise would certainly lead to different gains in different clocks, > some going 7min wrong, others 10, some 2 1/2, ... > > If most of the clocks which were fooled, had an error of about > 10min, it's probably not something which raises the noise floor. > > My guess: > let's assume that power entered the clock not with 60Hz, > but it just happened that it cycled with 70Hz between 1am and 2am. > This results in 36000 additional cycles, which gives 10min. > > Most devices would neither detect it nor be damaged. > Especially, if it had happened around midnight.Ohh, but there would be all sorts of alarms going off in power plant control room for that large a deviation. There system would also have been disconnected from the grid long before. Many devices would tolerate such frequency excursions. But I wonder just how spectacularly a ferroresonant constant voltage transformer would fail ;]> > The power company might not be interested to publish such an event, > because it would mean that they eventually billed 16% too much from > every customer during this event.Power would be metered correctly but the power factor for industrial customers with highly reactive loads would be strange. But that creates more problems for the power company than the customer, I believe.> > Might be interesting, how power is generated/stabilized in that > area. > Were there other related events like power failures, brownouts > around these days ?? > >
Reply by ●July 5, 20042004-07-05
Richard Owlett wrote:> Bernhard Holzmayer wrote: >> Jerry Avins wrote: >> >> >>>Symon wrote: >>>http://www.saultstar.com/webapp/sitepages/content.asp?contentID=70613&catname=Local+News>> >>>>A small town in Canada where everyone's clocks gained 10 >>>>minutes. Cheers, Syms. >>> >>> ... >>> >>>Considering the natures of the affected clocks, my bet is on some >>>sort of noise. Not all clocks were affected, only those that >>>count zero crossings and have no flywheels. >>> >>>Jerry >> >> >> Noise would certainly lead to different gains in different >> clocks, some going 7min wrong, others 10, some 2 1/2, ... >> >> If most of the clocks which were fooled, had an error of about >> 10min, it's probably not something which raises the noise floor. >> >> My guess: >> let's assume that power entered the clock not with 60Hz, >> but it just happened that it cycled with 70Hz between 1am and >> 2am. This results in 36000 additional cycles, which gives 10min. >> >> Most devices would neither detect it nor be damaged. >> Especially, if it had happened around midnight. > > Ohh, but there would be all sorts of alarms going off in power > plant control room for that large a deviation.I guess so.> There system would > also have been disconnected from the grid long before.Probably yes. But if evrything's working great, such a thing doesn't happen. They had a power fail around here, recently. A huge generator was switched to replace the power transformer station, until the defect would have been repaired. If such a generator were defect?> > Many devices would tolerate such frequency excursions. > But I wonder just how spectacularly a ferroresonant constant > voltage transformer would fail ;]Once, they searched for a broken cable in the earth. They coupled high energy high frequency pulses into the line, after they had walked to every houshold in the neighborhood telling them to take all devices from net which are usually at standby. Then, from the earth you could hear where the cable was broken. Really, without any sort of loudspeaker, transformer or so. It just gave a burst sound. After repair, the did a check on the line, and for a certain reason, they introduced such pulses again. This time they forgot to inform the neighborhood. They told me that there's a good chance that a lot of people would have to replace their VCRs, TVsets ... I guess some clocks might even have gone mad about the 12kHz. Bernhard
Reply by ●July 6, 20042004-07-06
Allan Herriman wrote:> On Sun, 04 Jul 2004 10:53:11 -0400, Carlos Moreno > <moreno_at_mochima_dot_com@xx.xxx> wrote: > > >>hel@40th.com wrote: >> >>>JA [Fri, 02 Jul 2004 10:25:56 -0400]: >>> >... Tell that to the motor on my phonograph turntable! >>> >>>Very odd since most turntables made in the last 30 years have >>>used a DC motor. >> >>But the stroboscopic light for adjusting the speed is based >>on the AC frequency -- granted, it typically has the tracks >>for both 50 and 60Hz, and for both 33.33 and 45 RPM. You'd >>have to just realize by ear that the correct adjustment is >>given by the 60Hz and not 50 as you thought (which should be >>more than obvious even for someone tone deaf :-)) > > > All of the ones I've seen have used cystal controlled strobes. > (E.g. Technics SP10, SL1200, etc.) > > Regards, > Allan.My turntable has 16 2/3 rpm for "Talking Books", 33 1/3 for LPs and 16" transcriptions, 45, and 78. The strobe disk is well viewed under fluorescent light, although a neon bulb would be better. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●July 6, 20042004-07-06
Hello. I agree that active noise cancellation is a good choice for this problem, especially if you are post processing already aquired data which is noisy. I did a similar thing for my brother who is a dialog editor in CA. He had some very noisy 60/120/240Hz mic data which I cleaned up using simple MATLAB scripts. I would use a two tap quadrature adaptive filter, they are much more responsive (adapt faster). Also, be careful any time you throw filter functions (i.e. the 60hz bandpass) within that block diagram below, you will need to use the "filtered-X" LMS version of LMS. If its voice data you are trying to clean up, you can use a voice gate to turn off/on the adaptive portion of the filter which will minimize wiping out the real signal info around 60Hz. Remember, unlike a notch filter, this adaptive approach will only perfectly remove a 60Hz signal that is the exact same phase and amplitude of the "pseudo stationary" 60Hz hum. Any of you guys going to the COMP DSP conference? We can talk more about this then. I am giving a presentation on Active Noise Control (and also one on Q-Math). Hope to see you there. -Shawn Steenhagen Applied Signal Processing. www.appliedsignalprocessing.com "Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message news:l7GdnQeKJYzheXnd4p2dnA@centurytel.net...> > "Rob Hutchinson" <rhutch7@kdsi.net> wrote in message > news:10e8scqobsj7f3@corp.supernews.com... > > What is the preferred method for removing 60 hz hum from a signalwithout> > wiping out signal info around 60 hz? A 60 hz notch filter would not be > > useful because it would attenuate the signal as well. I'm interested in > > doing this for sampled data, so all filtering would be done in thedigital> > domain. > > Subtraction is the best bet. That is, active noise cancellation. > That is, if you can establish a reference for the offending signal. > So, if the source of the 60Hz signal is available to you then you could do > something like this: > > +--------------------+ > | | > | | > | | > input------+---------------------|------------>(+)----+----> e[n] > | ^ > | | > v | > 60Hz >----------------->[LMS]------------+ > > > Adaptive canceller > LMS adaptive filter adjusts to minimize e[n] which cancels signal > and adds no additional noise. > > Now, the input to the LMS filter will be best if it's a replica of the > offending signal. Then it's subtracted from the input to get e[n]. Since > you know that the offending signal is at 60Hz, you might bandpass theinput> to the LMS filter at 60Hz. Assuming the 60Hz interference is of constant > amplitude then the LMS filter can change slowly or perhaps not at all. > > If the 60Hz interference is not of constant amplitude then you won't beable> to completely separate it from the desired "signal info around 60Hz". The > filter will work to minimize the output e[n]. If the input to the LMS > filter is bandpassed to 60Hz, then the only work the LMS filter has to dois> to adjust the amplitude and phase of the approximately 60Hz signal that is > subtracted. If the filter adjusts too quickly then it will take out moreof> the desired signal. If the filter adjusts too slowly then it will cancel > the interference less. So, the rate of change of the LMS filter may be > something you can adjust to get best performance. It will have its ownrate> limit just to work and can be made slower... > If the 60Hz information you want to keep is correlated with theinterference> then it won't help in keeping that "other 60Hz information". > > Fred > >
Reply by ●July 6, 20042004-07-06
"Tim Wescott" <tim@wescottnospamdesign.com> wrote in message news:10e9cbd9qh0t53b@corp.supernews.com... >>>> >>>>A notch can be made almost arbitrarily narrow if care is taken in the > >>>>implementation. The best bet of course is to remove the hum in the analog > >>>realm > >>>>before the signal is recorded, etc.. But if that isn't possible, then I > >>>would > >>>>say a very narrow notch filter is a decent method. If the hum is uniform > >>>in > >>>>level, another possibility is some sort of adaptive noise removal > >>>algorithm. > >>>> > >>>>"Rob Hutchinson" <rhutch7@kdsi.net> wrote in message > >>>>news:10e8scqobsj7f3@corp.supernews.com... > >>>> > >>>> > >>>>>What is the preferred method for removing 60 hz hum from a signal > >>>without > >>>>>wiping out signal info around 60 hz? A 60 hz notch filter would not be > >>>>>useful because it would attenuate the signal as well. I'm interested in > >>>>>doing this for sampled data, so all filtering would be done in the > >>>digital domain. > >> > >>If the frequency is nice and constant the best "adaptation" that you can > >>get would be a nice narrow notch. > > > > Assuming a perfectly constant hum, how about adding a 60Hz sine of the exact > > same magnitude/opposite phase? > > > Good idea. If you use a unity-gain highly resonant bandpass filter to > acquire your 60Hz sine wave and subtract it from your signal you'll have > -- a notch filter! (I've been through this before). > > I suppose that you could attempt to do this with a PLL of some sort, but > I expect that after you make it so it can identify the hum's magnitude > and phase, and correctly identify when you are operating off of battery > power in the middle of the Sahara desert and therefore don't have any > hum, and take all the other odd little corner cases into account, you'd > be better off just making a _really good_ notch filter. > > But I've become a luddite in my old age.Speaking theoretically, if you could perfectly identify the 60Hz component and it was perfectly stable in frequency, phase, and amplitude, it seems subtracting it out would be different than a notch filter. Another desired component even at exactly 60Hz would be unaffected by this process. Of course, in real life, you may be right that a notch filter is about as good as it gets, but in theory*, is seems like the approach would be superior. Shawn Steenhagen's post also seems to support this idea. -Jon *In theory, there is no difference between theory and practice. But in practice, there is!
Reply by ●July 6, 20042004-07-06
Jon Harris wrote:> "Tim Wescott" <tim@wescottnospamdesign.com> wrote in message > news:10e9cbd9qh0t53b@corp.supernews.com... > >>> > >>>>>>A notch can be made almost arbitrarily narrow if care is taken in the >>>>>>implementation. The best bet of course is to remove the hum in the analog >>>>> >>>>>realm >>>>> >>>>>>before the signal is recorded, etc.. But if that isn't possible, then I >>>>> >>>>>would >>>>> >>>>>>say a very narrow notch filter is a decent method. If the hum is uniform >>>>> >>>>>in >>>>> >>>>>>level, another possibility is some sort of adaptive noise removal >>>>> >>>>>algorithm. >>>>> >>>>>>"Rob Hutchinson" <rhutch7@kdsi.net> wrote in message >>>>>>news:10e8scqobsj7f3@corp.supernews.com... >>>>>> >>>>>> >>>>>> >>>>>>>What is the preferred method for removing 60 hz hum from a signal >>>>> >>>>>without >>>>> >>>>>>>wiping out signal info around 60 hz? A 60 hz notch filter would not be >>>>>>>useful because it would attenuate the signal as well. I'm interested in >>>>>>>doing this for sampled data, so all filtering would be done in the >>>>> >>>>>digital domain. >>>> >>>>If the frequency is nice and constant the best "adaptation" that you can >>>>get would be a nice narrow notch. >>> >>>Assuming a perfectly constant hum, how about adding a 60Hz sine of the exact >>>same magnitude/opposite phase? >>> >> >>Good idea. If you use a unity-gain highly resonant bandpass filter to >>acquire your 60Hz sine wave and subtract it from your signal you'll have >>-- a notch filter! (I've been through this before). >> >>I suppose that you could attempt to do this with a PLL of some sort, but >>I expect that after you make it so it can identify the hum's magnitude >>and phase, and correctly identify when you are operating off of battery >>power in the middle of the Sahara desert and therefore don't have any >>hum, and take all the other odd little corner cases into account, you'd >>be better off just making a _really good_ notch filter. >> >>But I've become a luddite in my old age. > > > Speaking theoretically, if you could perfectly identify the 60Hz component and > it was perfectly stable in frequency, phase, and amplitude, it seems subtracting > it out would be different than a notch filter. Another desired component even > at exactly 60Hz would be unaffected by this process. Of course, in real life, > you may be right that a notch filter is about as good as it gets, but in > theory*, is seems like the approach would be superior. Shawn Steenhagen's post > also seems to support this idea. > > -Jon > > *In theory, there is no difference between theory and practice. But in > practice, there is! > >That's two "perfectlies" in one assumption -- and that's two too many for me. Particularly since noise like this is going to change with the environment. When I originally responded I was failing to consider that the 60Hz noise you get generally isn't perfectly sinusoidal, so there will be components all over the place. In this case you can either use a bank of notch filters or an adaptive filter -- and I suspect that by the time you get up to the third or fourth harmonic you may as well just use the adaptive filter. I think I'm going to grab on to something I said elsewhere in this discussion, though: if you can eliminate the noise at the source that will be best by far. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●July 6, 20042004-07-06
The beauty of the adaptive filter is that it will track changes in amplitude and phase of the signal its tracking--even if the signal goes away. The quadrature adaptive filters are especially good at this. They can even adjust to errors in frequency between the reference signal and the error signal, in which case the LMS algorithm will actually modulate the tap weights so that the output of the adaptive filter is a different frequency than the input. You need higher step sizes to do this, in which case the "modulation effect" may also disturb nearby frequencies (and behave more like the notch filter? interesting thing to look into). If this is a process that is being tweeked by a sound engineer in a post production phase, he can control things like the adaptive step size, the reference frequency etc. One could even envision a scheme in which the weight modulation is used to tweek the frequency of the reference signal much like a PLL does to make the hum removal more automated. If the noise can't be removed in the analog domain easily, it would be nice to dedicate an extra sound track (if the luxury exists) to recording the line voltage. Now we have our exact reference signal and don't need to synthesize it. This might be a hard sell unless you can convice your customer its an insurance policy against an improperly grounded/connected piece of equipement that goes unnoticed until later in post production (again I'm referencing the movie/sound industry from my other post). -Shawn Steenhagen Applied Signal Processing. "Jon Harris" <goldentully@hotmail.com> wrote in message news:2l0cdmF7a21uU1@uni-berlin.de...> "Tim Wescott" <tim@wescottnospamdesign.com> wrote in message > news:10e9cbd9qh0t53b@corp.supernews.com... > >>> > > >>>>A notch can be made almost arbitrarily narrow if care is taken inthe> > >>>>implementation. The best bet of course is to remove the hum in theanalog> > >>>realm > > >>>>before the signal is recorded, etc.. But if that isn't possible,then I> > >>>would > > >>>>say a very narrow notch filter is a decent method. If the hum isuniform> > >>>in > > >>>>level, another possibility is some sort of adaptive noise removal > > >>>algorithm. > > >>>> > > >>>>"Rob Hutchinson" <rhutch7@kdsi.net> wrote in message > > >>>>news:10e8scqobsj7f3@corp.supernews.com... > > >>>> > > >>>> > > >>>>>What is the preferred method for removing 60 hz hum from a signal > > >>>without > > >>>>>wiping out signal info around 60 hz? A 60 hz notch filter wouldnot be> > >>>>>useful because it would attenuate the signal as well. I'minterested in> > >>>>>doing this for sampled data, so all filtering would be done in the > > >>>digital domain. > > >> > > >>If the frequency is nice and constant the best "adaptation" that youcan> > >>get would be a nice narrow notch. > > > > > > Assuming a perfectly constant hum, how about adding a 60Hz sine of theexact> > > same magnitude/opposite phase? > > > > > Good idea. If you use a unity-gain highly resonant bandpass filter to > > acquire your 60Hz sine wave and subtract it from your signal you'll have > > -- a notch filter! (I've been through this before). > > > > I suppose that you could attempt to do this with a PLL of some sort, but > > I expect that after you make it so it can identify the hum's magnitude > > and phase, and correctly identify when you are operating off of battery > > power in the middle of the Sahara desert and therefore don't have any > > hum, and take all the other odd little corner cases into account, you'd > > be better off just making a _really good_ notch filter. > > > > But I've become a luddite in my old age. > > Speaking theoretically, if you could perfectly identify the 60Hz componentand> it was perfectly stable in frequency, phase, and amplitude, it seemssubtracting> it out would be different than a notch filter. Another desired componenteven> at exactly 60Hz would be unaffected by this process. Of course, in reallife,> you may be right that a notch filter is about as good as it gets, but in > theory*, is seems like the approach would be superior. Shawn Steenhagen'spost> also seems to support this idea. > > -Jon > > *In theory, there is no difference between theory and practice. But in > practice, there is! > >
Reply by ●July 6, 20042004-07-06
Shawn Steenhagen wrote:> The beauty of the adaptive filter is that it will track changes in amplitude > and phase of the signal its tracking--even if the signal goes away. The > quadrature adaptive filters are especially good at this. They can even > adjust to errors in frequency between the reference signal and the error > signal, in which case the LMS algorithm will actually modulate the tap > weights so that the output of the adaptive filter is a different frequency > than the input. You need higher step sizes to do this, in which case the > "modulation effect" may also disturb nearby frequencies (and behave more > like the notch filter? interesting thing to look into).... So how come my adaptive spam filters keep trashing letters I want to read? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●July 6, 20042004-07-06
Shawn Steenhagen wrote:> The beauty of the adaptive filter is that it will track changes in amplitude > and phase of the signal its tracking--even if the signal goes away. The > quadrature adaptive filters are especially good at this. They can even > adjust to errors in frequency between the reference signal and the error > signal, in which case the LMS algorithm will actually modulate the tap > weights so that the output of the adaptive filter is a different frequency > than the input. You need higher step sizes to do this, in which case the > "modulation effect" may also disturb nearby frequencies (and behave more > like the notch filter? interesting thing to look into). > > If this is a process that is being tweeked by a sound engineer in a post > production phase, he can control things like the adaptive step size, the > reference frequency etc. One could even envision a scheme in which the > weight modulation is used to tweek the frequency of the reference signal > much like a PLL does to make the hum removal more automated. > > If the noise can't be removed in the analog domain easily, it would be nice > to dedicate an extra sound track (if the luxury exists) to recording the > line voltage. Now we have our exact reference signal and don't need to > synthesize it. This might be a hard sell unless you can convice your > customer its an insurance policy against an improperly grounded/connected > piece of equipement that goes unnoticed until later in post production > (again I'm referencing the movie/sound industry from my other post).... That's an interesting approach. What steps have to be taken to ensure that the recorded waveform closely resembles the stray pickup wherever it may have happened? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������