Reply by Tim Wescott November 12, 20082008-11-12
On Tue, 11 Nov 2008 09:02:21 -0600, Vladimir Vassilevsky wrote:

> Tim Wescott wrote: >> On Mon, 10 Nov 2008 19:58:47 -0600, Vladimir Vassilevsky wrote: >> >> >> >>>No point to do an adaptive filter. The random fluctuations of the >>>waveform smear the spectrum. >>> >>> >> I would disagree on the last two points. If you're running off of a >> continent-wide grid then the frequency variations are very small >> indeed. But if you're making a product that may ever have to run off of >> a generator, or on a small grid (and there are some, even in >> out-of-the-way places in developed countries) then this isn't going to >> hold true. >> >> And if the frequency varies, then a fixed-frequency comb filter isn't >> going to help much. > > Some time ago I developed VLF instruments. The harmonics of the power > frequency are the major source of interference in VLF; typically they > are dominating over the ambience up to the 1000th harmonic orders. Of > course, I took advantage of the comb filtering, and got a substantial > improvement. However the random variations of the power waveform due to > the loads switched on/off, etc. are the natural limitation to that > improvement, and this limitation happens much earlier then the frequency > instability comes into play. So, I don't see much benefit in making the > filter frequency adaptive, unless someone is running from the autonomous > alternator. BTW, those cheap alternators are terrible: the frequency can > be off by as much as 10 Hz or so; I had issues with that on the > different project :)
I wouldn't expect any modern power grid that's bigger than a small city to have enough variation to cause a hint of a problem. But after 20 years of saying "the customer did _what_?!?", sighing, and starting in on a design modification, I've gotten more willing to be defensive in my engineering. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by Jerry Avins November 11, 20082008-11-11
HardySpicer wrote:
> On Nov 11, 2:26 pm, "DJT" <JanTorgrims...@hotmail.com> wrote: >> Hello everybody! >> >> I'm trying to remove power line hum from a signal and have two questions >> concerning this. >> >> 1. How high do the harmonics go? (In general, round numbers) >> >> 2. How much frequency variations (+/-) should I expect in the fundamental >> >> 50 Hz component? >> >> The reason for asking these questions is that I am having a hard time >> choosing between a fixed comb-notch filter and something that tracks the >> frequency variations by adapting... >> >> Thanks in advance >> >> JT > > I am always surprised that people want to filter out hum when the best > solution is to avoid it in the first place. > Often a bad layout in a board,earthing etc has a lot to do with it. > Mains hum is everywhere, but if you are careful with > layout,shielding,grounding you can minimise it.
In humid weather, my favorite AM radio station is almost completely jammed by power-line harmonics when I am near certain poles. further away, the interference is merely annoying. (When it's very bad, I note the pole numbers and alert the power company. Spraying the insulator with WD-40 helps considerably.) Can you suggest how I can avoid this? Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
Reply by HardySpicer November 11, 20082008-11-11
On Nov 11, 2:26&#2013266080;pm, "DJT" <JanTorgrims...@hotmail.com> wrote:
> Hello everybody! > > I'm trying to remove power line hum from a signal and have two questions > concerning this. > > 1. How high do the harmonics go? (In general, round numbers) > > 2. How much frequency variations (+/-) should I expect in the fundamental > > &#2013266080; &#2013266080;50 Hz component? > > The reason for asking these questions is that I am having a hard time > choosing between a fixed comb-notch filter and something that tracks the > frequency variations by adapting... > > Thanks in advance > > JT
I am always surprised that people want to filter out hum when the best solution is to avoid it in the first place. Often a bad layout in a board,earthing etc has a lot to do with it. Mains hum is everywhere, but if you are careful with layout,shielding,grounding you can minimise it. Hardy
Reply by Eric Jacobsen November 11, 20082008-11-11
On Mon, 10 Nov 2008 18:26:44 -0800 (PST), Robert Adams
<robert.adams@analog.com> wrote:

>On Nov 10, 8:58&#2013266080;pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com> >wrote: >> DJT wrote: >> > Hello everybody! >> >> > I'm trying to remove power line hum from a signal and have two questions >> > concerning this. >> >> > 1. How high do the harmonics go? (In general, round numbers) >> >> Infinitely high. &#2013266080;The rolloff factor ballpark is ~10dB per octave. >> >> > 2. How much frequency variations (+/-) should I expect in the fundamental >> > 50 Hz component? >> >> Very accurate in the long term; short term accuracy ~ 1e-4. >> >> > The reason for asking these questions is that I am having a hard time >> > choosing between a fixed comb-notch filter and something that tracks the >> > frequency variations by adapting... >> >> No point to do an adaptive filter. The random fluctuations of the >> waveform smear the spectrum. >> >> Vladimir Vassilevsky >> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com > >On a slightly related note; did you know that if you are analyzing a >recording and wish to know when and where it was made, you can extract >the hum on the recording, track the tiny fluctuations in frequency >over the duration of the recording, and correlate these with the >frequency variations that occur in the power grids of various >countries? These fluctuations are recorded (for just this purpose), so >you can tell both the time it was recorded as well as the country >(just search the database for a correlation peak). Since all >generators are syncronously linked, the frequency variations are valid >for the entire country (or continent, even). > >This doesn't work if you are running off a battery, and are living in >a cave ..... > >Bob Adams
Interesting. So the next obvious step would be to have each power generating station modulate their output very slightly with stuff like data and time, spread with a unique spreading code that would identify the station. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
Reply by Vladimir Vassilevsky November 11, 20082008-11-11

Tim Wescott wrote:
> On Mon, 10 Nov 2008 19:58:47 -0600, Vladimir Vassilevsky wrote: > >
>>No point to do an adaptive filter. The random fluctuations of the >>waveform smear the spectrum. >> > > I would disagree on the last two points. If you're running off of a > continent-wide grid then the frequency variations are very small indeed. > But if you're making a product that may ever have to run off of a > generator, or on a small grid (and there are some, even in out-of-the-way > places in developed countries) then this isn't going to hold true. > > And if the frequency varies, then a fixed-frequency comb filter isn't > going to help much.
Some time ago I developed VLF instruments. The harmonics of the power frequency are the major source of interference in VLF; typically they are dominating over the ambience up to the 1000th harmonic orders. Of course, I took advantage of the comb filtering, and got a substantial improvement. However the random variations of the power waveform due to the loads switched on/off, etc. are the natural limitation to that improvement, and this limitation happens much earlier then the frequency instability comes into play. So, I don't see much benefit in making the filter frequency adaptive, unless someone is running from the autonomous alternator. BTW, those cheap alternators are terrible: the frequency can be off by as much as 10 Hz or so; I had issues with that on the different project :) Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by DJT November 11, 20082008-11-11
>On Mon, 10 Nov 2008 19:58:47 -0600, Vladimir Vassilevsky wrote: > >> DJT wrote: >> >>> Hello everybody! >>> >>> I'm trying to remove power line hum from a signal and have two >>> questions concerning this. >>> >>> 1. How high do the harmonics go? (In general, round numbers) >> >> Infinitely high. The rolloff factor ballpark is ~10dB per octave. > >And you can't get more general (or round!) than infinity. >> >>> 2. How much frequency variations (+/-) should I expect in the >>> fundamental 50 Hz component? >> >> Very accurate in the long term; short term accuracy ~ 1e-4. >> >>> The reason for asking these questions is that I am having a hard time >>> choosing between a fixed comb-notch filter and something that tracks >>> the frequency variations by adapting... >> >> No point to do an adaptive filter. The random fluctuations of the >> waveform smear the spectrum. >> >I would disagree on the last two points. If you're running off of a >continent-wide grid then the frequency variations are very small indeed.
>But if you're making a product that may ever have to run off of a >generator, or on a small grid (and there are some, even in out-of-the-way
>places in developed countries) then this isn't going to hold true. > >And if the frequency varies, then a fixed-frequency comb filter isn't >going to help much. > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >Do you need to implement control loops in software? >"Applied Control Theory for Embedded Systems" gives you just what it
says.
>See details at http://www.wescottdesign.com/actfes/actfes.html >
Found the thread and I think this idea looks interesting assuming the frequency varies significantly... "If there's a need to preserve the signal much above the line rate and if you have a sample of the line frequency available, an cheap adaptive filter that acquires the line frequency and uses that to set the spacing on a comb filter would work well." However I don't really understand one of the answers I got concerning my idea to use an adaptive filter (I was thinking NLMS)... "No point to do an adaptive filter. The random fluctuations of the waveform smear the spectrum." I can buy this, but still is it not a little bit harsh to say that there is no point to use an adaptive filter... I mean this should be higly dependent on the application, the characteristics of the signals the amount of "smearing" that can be tolerable and so on. Am I completely lost here or? JT
Reply by Glen Herrmannsfeldt November 11, 20082008-11-11
Robert Adams wrote:
(snip)

> On a slightly related note; did you know that if you are analyzing a > recording and wish to know when and where it was made, you can extract > the hum on the recording, track the tiny fluctuations in frequency > over the duration of the recording, and correlate these with the > frequency variations that occur in the power grids of various > countries? These fluctuations are recorded (for just this purpose), so > you can tell both the time it was recorded as well as the country > (just search the database for a correlation peak). Since all > generators are syncronously linked, the frequency variations are valid > for the entire country (or continent, even).
> This doesn't work if you are running off a battery, and are living in > a cave .....
Or if your recorder has a synchronous motor run off the power line. Well, if you had a constant frequency signal in the recording other than the power line (assumed not constant) then you could do it directly from a synchronous motor powered recording even without hum. -- glen
Reply by Tim Wescott November 11, 20082008-11-11
On Mon, 10 Nov 2008 19:58:47 -0600, Vladimir Vassilevsky wrote:

> DJT wrote: > >> Hello everybody! >> >> I'm trying to remove power line hum from a signal and have two >> questions concerning this. >> >> 1. How high do the harmonics go? (In general, round numbers) > > Infinitely high. The rolloff factor ballpark is ~10dB per octave.
And you can't get more general (or round!) than infinity.
> >> 2. How much frequency variations (+/-) should I expect in the >> fundamental 50 Hz component? > > Very accurate in the long term; short term accuracy ~ 1e-4. > >> The reason for asking these questions is that I am having a hard time >> choosing between a fixed comb-notch filter and something that tracks >> the frequency variations by adapting... > > No point to do an adaptive filter. The random fluctuations of the > waveform smear the spectrum. >
I would disagree on the last two points. If you're running off of a continent-wide grid then the frequency variations are very small indeed. But if you're making a product that may ever have to run off of a generator, or on a small grid (and there are some, even in out-of-the-way places in developed countries) then this isn't going to hold true. And if the frequency varies, then a fixed-frequency comb filter isn't going to help much. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by Tim Wescott November 11, 20082008-11-11
On Mon, 10 Nov 2008 20:35:12 -0500, Jerry Avins wrote:

> DJT wrote: >> Hello everybody! >> >> I'm trying to remove power line hum from a signal and have two >> questions concerning this. >> >> 1. How high do the harmonics go? (In general, round numbers) >> >> 2. How much frequency variations (+/-) should I expect in the >> fundamental >> >> 50 Hz component? >> >> The reason for asking these questions is that I am having a hard time >> choosing between a fixed comb-notch filter and something that tracks >> the frequency variations by adapting... >> >> Thanks in advance > > The harmonic structure depends on the source. A "dirty" SCR controller > or an arking power-line insulator can be heard on an AM radio. The > frequency stability depende on the grid. 60 years ago, electric clocks > in New York City would lose as much as three seconds during the day and > evening and make it up overnight. It has been much better than that for > many years now. > > Jerry
Jerry, you forgot to tell him to eliminate the hum at the source! (I _think_ this is a different guy than the last one). It's still good advise. OP: There's been a thread on this recently, if you google for it you'll see lots of discussion. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" gives you just what it says. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by Vladimir Vassilevsky November 11, 20082008-11-11
"Robert Adams" <robert.adams@analog.com> wrote in message
news:8e9934f0-451b-4b2f-a673-9259b24e37d3@c36g2000prc.googlegroups.com...

> On a slightly related note; did you know that if you are analyzing a > recording and wish to know when and where it was made, you can extract > the hum on the recording, track the tiny fluctuations in frequency > over the duration of the recording, and correlate these with the > frequency variations that occur in the power grids of various > countries? These fluctuations are recorded (for just this purpose), so > you can tell both the time it was recorded as well as the country > (just search the database for a correlation peak). Since all > generators are syncronously linked, the frequency variations are valid > for the entire country (or continent, even).
I have doubts about the feasibility. The frequency measurement will produce something like one useful bit per minute; so the recording has to be too long to be practical. Also, the fluctuations of the speed of the recording are likely to be of the same order as the power frequency variations. Plus phase shifts, propagation delays, waveform distortions, etc... Vladimir Vassilevsky DSP and Mixed Signal Design Consultan http://www.abvolt.com