I have got a new task I'm in a little doubt if and how it can be done.. It is a mechanical setup with weighing scale that has a vibrator attached. I'm not able to change mechanical the construction. The vibrator adds a 25 hz sinusoidal to the weighing signal. Is it possible to 'remove' this unwanted component, by say a lowpass filter ? Saying 'remove' since I do not know how much damping (dB) would be needed. The actual weighing signal (of interest) is close to DC.
25hz low pass filter
Started by ●October 5, 2012
Reply by ●October 5, 20122012-10-05
On Fri, 05 Oct 2012 14:21:42 -0500, Pivskid wrote:> I have got a new task I'm in a little doubt if and how it can be done.. > > It is a mechanical setup with weighing scale that has a vibrator > attached. I'm not able to change mechanical the construction. > > The vibrator adds a 25 hz sinusoidal to the weighing signal. Is it > possible to 'remove' this unwanted component, by say a lowpass filter ? > Saying 'remove' since I do not know how much damping (dB) would be > needed. > > The actual weighing signal (of interest) is close to DC.Since there is inevitably some imperfection in measurement and filtering, you will have to determine exactly what "remove" must mean. What accuracy do you need? How big is the disturbance? Do you have any computational power available? Can you sample synchronously with the vibration? - probably others will think of more relevant questions.
Reply by ●October 5, 20122012-10-05
On Friday, October 5, 2012 3:21:42 PM UTC-4, Pivskid wrote:> I have got a new task I'm in a little doubt if and how it can be done.. > > > > It is a mechanical setup with weighing scale that has a vibrator attached. > > I'm not able to change mechanical the construction. > > > > The vibrator adds a 25 hz sinusoidal to the weighing signal. > > Is it possible to 'remove' this unwanted component, by say a lowpass filter > > ? > > Saying 'remove' since I do not know how much damping (dB) would be needed. > > > > The actual weighing signal (of interest) is close to DC.Part of this depends on how much time you allow for weighing. The period of 25 Hz is 0.04 seconds. So ideally sampling at twice the 25 Hz rate for a period that is a whole multiple of 0.04 seconds would allow simple averaging to kill the 25 Hz component in the data. Being this is mechanical, I would expect the 25 Hz to not be exact and may even vary a little bit. But I bet averaging data for 1 or more seconds (sampled maybe at 100Hz or more so you are far enough away from the Shannon limit as to not worry about aliasing). The vibration is likely to not be perfectly sinusoidal, so you can do some simple experiments. Put known weights on the scale and collect data for say a minute sampled at 100Hz. Then average subsets of the data spanning 1, 2, or more seconds and see if their averages all converge to the same value or jump all over the place. You may even collect data and then do spectral analysis to see the details of the 25Hz vibration. To get a clear picture of this you may have to sample at a rate much higher than 25Hz because nonsinusoidal vibration will/may have harmonics with significant strengths to affect the results. Once you know the nature of the interference (25 Hz and its harmonics) you can filter it out (averaging is a type of filter). You need to measure, characterize, and qualify the real nature of the problem before you can mitigate it. IHTH, Clay
Reply by ●October 5, 20122012-10-05
<clay@claysturner.com> wrote:>On Friday, October 5, 2012 3:21:42 PM UTC-4, Pivskid wrote:>> I have got a new task I'm in a little doubt if and how it can be done..>> It is a mechanical setup with weighing scale that has a vibrator attached.>> The vibrator adds a 25 hz sinusoidal to the weighing signal. >> Is it possible to 'remove' this unwanted component, by say a lowpass >> filter?>The vibration is likely to not be perfectly sinusoidal, so you can do some >simple experiments. Put known weights on the scale and >collect data for >say a minute sampled at 100Hz. Then average subsets of the data spanning 1, >2, or more seconds and see if their >averages all converge to the same >value or jump all over the place. You may even collect data and then do >spectral analysis to see >the details of the 25Hz vibration. To get a clear >picture of this you may have to sample at a rate much higher than 25Hz >because >nonsinusoidal vibration will/may have harmonics with significant >strengths to affect the results."Mechanical rectification" is usual problem with that kind of setup. The averaged value gets biased due to nonlinearity.>Once you know the nature of the interference (25 Hz and its harmonics) you >can filter it out (averaging is a type of filter). You need >to measure, >characterize, and qualify the real nature of the problem before you can >mitigate it.No. This is not interesting. Let's try wavelets, neural networks and fuzzy logic and see if it works. IHTH, Clay
Reply by ●October 5, 20122012-10-05
On Fri, 05 Oct 2012 12:56:48 -0700, clay wrote:> On Friday, October 5, 2012 3:21:42 PM UTC-4, Pivskid wrote: >> I have got a new task I'm in a little doubt if and how it can be done.. >> >> >> >> It is a mechanical setup with weighing scale that has a vibrator >> attached. >> >> I'm not able to change mechanical the construction. >> >> >> >> The vibrator adds a 25 hz sinusoidal to the weighing signal. >> >> Is it possible to 'remove' this unwanted component, by say a lowpass >> filter >> >> ? >> >> Saying 'remove' since I do not know how much damping (dB) would be >> needed. >> >> >> >> The actual weighing signal (of interest) is close to DC. > > Part of this depends on how much time you allow for weighing. The period > of 25 Hz is 0.04 seconds. So ideally sampling at twice the 25 Hz rate > for a period that is a whole multiple of 0.04 seconds would allow simple > averaging to kill the 25 Hz component in the data. Being this is > mechanical, I would expect the 25 Hz to not be exact and may even vary a > little bit. But I bet averaging data for 1 or more seconds (sampled > maybe at 100Hz or more so you are far enough away from the Shannon limit > as to not worry about aliasing). > > The vibration is likely to not be perfectly sinusoidal, so you can do > some simple experiments. Put known weights on the scale and collect data > for say a minute sampled at 100Hz. Then average subsets of the data > spanning 1, 2, or more seconds and see if their averages all converge to > the same value or jump all over the place. You may even collect data and > then do spectral analysis to see the details of the 25Hz vibration. To > get a clear picture of this you may have to sample at a rate much higher > than 25Hz because nonsinusoidal vibration will/may have harmonics with > significant strengths to affect the results. > > Once you know the nature of the interference (25 Hz and its harmonics) > you can filter it out (averaging is a type of filter). You need to > measure, characterize, and qualify the real nature of the problem before > you can mitigate it.100Hz is exactly equal to the 4th harmonic of 25Hz. You probably want to either sample fast, or to choose a sampling rate that will tend to _not_ land on any of the harmonics, and hope that when the frequency changes a bit that the harmonic you _do_ land on is high enough to not matter. Then average for a good long time. -- My liberal friends think I'm a conservative kook. My conservative friends think I'm a liberal kook. Why am I not happy that they have found common ground? Tim Wescott, Communications, Control, Circuits & Software http://www.wescottdesign.com
Reply by ●October 5, 20122012-10-05
On Fri, 05 Oct 2012 14:21:42 -0500, Pivskid wrote:> I have got a new task I'm in a little doubt if and how it can be done.. > > It is a mechanical setup with weighing scale that has a vibrator > attached. I'm not able to change mechanical the construction. > > The vibrator adds a 25 hz sinusoidal to the weighing signal. Is it > possible to 'remove' this unwanted component, by say a lowpass filter ? > Saying 'remove' since I do not know how much damping (dB) would be > needed. > > The actual weighing signal (of interest) is close to DC.It's certainly possible to reduce the unwanted component; once you have it down to the point where it's no longer your biggest source of error then you can claim that it's 'removed'. As mentioned elsewhere in the thread, if you can sample synchronously to the 25Hz that would be good. Better would be if you sourced it so you knew exactly where it was. You didn't say how quickly and how accurately you need to measure -- that's important, as it makes a difference as to how fancy your algorithm needs to be. -- My liberal friends think I'm a conservative kook. My conservative friends think I'm a liberal kook. Why am I not happy that they have found common ground? Tim Wescott, Communications, Control, Circuits & Software http://www.wescottdesign.com
Reply by ●October 6, 20122012-10-06
On 10/5/2012 5:12 PM, Vladimir Vassilevsky wrote: ... > "Mechanical rectification" is usual problem with that kind of setup. > The averaged value gets biased due to nonlinearity. Some of you may be familiar with the Hieronymus machine. en.wikipedia.org/wiki/Hieronymus_machine http://www.lifetechnology.com/hieronymus.htm The same august publication in which the original report appeared also published the details of an antigravity generator that consisted, essentially, of an electric drill driving an eccentric weight. The antigravity effect was not sufficiently developed to lift the entire apparatus, so it was suspended by cables running over pulleys and attached to counterweights. Sure enough, when the apparatus was properly proportioned and the motor turned on, it went up and the counterweights went down. Does anyone see mechanical rectification here? Bless John W. Campbell! Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●October 6, 20122012-10-06
Post an example signal, maybe. PS:>> Let's try wavelets, ...The neural network still has a hangover but they sell least-squares fit as "artificial intelligence" nowadays.
Reply by ●October 6, 20122012-10-06
On 05/10/2012 20:21, Pivskid wrote:> I have got a new task I'm in a little doubt if and how it can be done.. > > It is a mechanical setup with weighing scale that has a vibrator attached. > I'm not able to change mechanical the construction. > > The vibrator adds a 25 hz sinusoidal to the weighing signal. > Is it possible to 'remove' this unwanted component, by say a lowpass filter > ? > Saying 'remove' since I do not know how much damping (dB) would be needed. > > The actual weighing signal (of interest) is close to DC. >Is the vibrator electrically driven, and can you tap into the drive signal? If so, there are possibilities for cancelling the 25Hz component without needing long time-constant filters. Cheers -- Syd
Reply by ●October 10, 20122012-10-10
clay@claysturner.com wrote: (snip)> Hi Jerry,> You know that photons are particles with spin=1, which under > rules of quantum mechanics normally would allow their observed > angular momenta to have values of +1, 0, and -1 in units of hbar. > However relativity suppresses the "0" state so we have only > two polarizations for photons each of which has angular momentum. > I would love to know how your method converts a solitary photon > with angular momentum to one with linear momentum without > expressing it as the superposition of two anti-spining photons?This might be one of the stranger things to learn about QM, not that there aren't plenty of them. (Read the quotes at the beginning of the book "Timeline" for example.) As with many examples in this newsgroup, you can use different bases to describe the system. The way you write it, that linear polarization is the superposition of two (or more) photons, is wrong. A single photon can be in the (|+1> + |-1>)/sqrt(2) state. Much of QM doesn't work if you don't allow for that. If that isn't strange enough, consider the case of the K meson (Kaon). There are two neutral Kaons, which you can consider as being antiparticles of each other, called K and Kbar. One interacts strongly with matter, the other with antimatter. It turns out, though, that there is another basis, as with polarized light, formed by adding and subtracting (and dividing by sqrt(2)), and that those, called Klong and Kshort have a different lifetime. Note, as with the photons, the superposition exists within a single particle. A beam of K (no Kbar) particles can be considered as a beam of particles that are superpositions of Klong and Kshort. As it travels, the Kshort will decay faster (they have a shorter lifetime) leaving the Klongs. Klong is a superposition of K and Kbar. If a beam of Klong travels through matter, the K or Kbar (I forget which) will interact with the matter, resulting in a beam of the opposite one, or that can be considered, again, a superposition of Klong and Kshort. For photons, right and left are the angular momentum eigenstates, but there is nothing wrong with photons that aren't in one of those states. OK, fun experiment that you can do at home. Take two sheets of polarizer and a light source such that the two polarizers are 90 degrees apart. The beam will be mostly blocked. Now put another polarizer in between and rotate it. The first polarizer, for example absorbs the vertically polarized light, leaving only horizontal polarization. The second absorbs horizontal, leaving nothing. If the third is oriented either vertically or horizontally, no light will come through. But what happens when it is at 45 degrees? Explain the result in terms of whichever photon basis functions you like. I recommmend reading "The Feynamn Lectures on Physics", volume 3. It is not the usual introductory explanation of quantum mechanics, though written to be at (about) that level. -- glen
