Hi, I would take a full number of cycles (!) of the signal and FFT it. Then notch out the spectral line corresponding to the fundamental, IFFT back to time domain, and you've got the double-frequency anomaly. That is, it's a single bin as long as the frequency is stable enough. If the one-per-rotation signal causes harmonics, I think there is no way to separate it from the two-per-rotation anomaly without additional information.
Separating Harmonic Spectra
Started by ●November 30, 2007
Reply by ●December 3, 20072007-12-03
Reply by ●December 3, 20072007-12-03
On 2 Des, 19:08, Greg Berchin <gberc...@comicast.net> wrote:> On Sun, 02 Dec 2007 12:09:07 -0500, Jerry Avins <j...@ieee.org> wrote: > >Can we assume that the twice-per-revolution anomaly is actually a > >once-around for a part geared 2:1? > > No. The "once-per" is a physical situation that occurs at only one > rotation angle; the "twice-per" is a physical situation that occurs at > two different rotation angles. > > >If not, does it actually consist of > >dub dub dub dub 180 degrees apart, or might it be a syncopated lub-dub > >... lub-dub ...? (I hope that's not too cryptic!) > > Nominally 180�, but this signal is indicative of severe wear, so there > will be plenty of room for slop.This is the detail I would *try* to exploit: The 1/rev event will *probably* have a more coherent spectrum in the sense that the phase relation between harmonics is "fixed". The 2/rev event is, from what you say, a more random event where the harmonics take a more random nature. You can check this hypothesis by making a long recording and check the coherence of spectrum lines at increasing time separations. If the hypothesis is correct, the overharmonics of the 1/rev event will be a stable foundation under the more fluctuating harmonics of the 2/rev event. So the harmonics with contributions only from the 2/rev event would be expected to be far less coherent over time than the harmonics with contributions from both events. If that made sense to anyone but me... Rune
Reply by ●December 3, 20072007-12-03
On Dec 2, 1:23 pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com> wrote:> What if you integrate the signal in the two dimensions per revolution > and per half revolution?Hmmm; that's got me thinking. The fundamental and odd harmonics of the combined signal MUST come from the 1-per phenomenon. So this is really a matter of separating the even harmonics of the 1-per from the entirety of the 2-per. Even harmonics are generated by asymmetrical waveforms (asymmetry above and below the zero-amplitude line). Odd harmonics are generated by symmetrical waveforms. There's got to be a way to separate the symmetrical portions of the waveform from the asymmetrical portions of the waveform -- much like decomposing a waveform into its even- and odd-symmetric parts. Thanks, Greg
Reply by ●December 3, 20072007-12-03
On Dec 2, 12:52 pm, John O'Flaherty <quias...@yeeha.com> wrote:> Just exploring this, how do you know the 2/per isn't just a harmonic > of the 1/per?They are entirely separate physical phenomena.> If they're separate anomalies, can you eliminate either > temporarily, to find out what the harmonic structure of the other is?No, unfortunately. I have no control over them. All that I can do is observe their effects. Thanks, Greg
Reply by ●December 3, 20072007-12-03
On 3 Des, 08:55, Rune Allnor <all...@tele.ntnu.no> wrote:> If that made sense to anyone but me...Ouch! That came out a bit different than was intended! I just tried to indicate that I had a more or less clear idea about what I tried to say when I wrote the previous post, but after reading over it I could only make sense of the text because I knew what I tried to say. No sinister intentions in any way. Sorry. Rune
Reply by ●December 3, 20072007-12-03
Greg Berchin wrote:> On Dec 2, 1:23 pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com> > wrote: > >> What if you integrate the signal in the two dimensions per revolution >> and per half revolution? > > Hmmm; that's got me thinking. The fundamental and odd harmonics of > the combined signal MUST come from the 1-per phenomenon. So this is > really a matter of separating the even harmonics of the 1-per from the > entirety of the 2-per. > > Even harmonics are generated by asymmetrical waveforms (asymmetry > above and below the zero-amplitude line). > > Odd harmonics are generated by symmetrical waveforms. > > There's got to be a way to separate the symmetrical portions of the > waveform from the asymmetrical portions of the waveform -- much like > decomposing a waveform into its even- and odd-symmetric parts.Again, this assumes that the two-per anomaly is "balanced". If it sounds like tik-tik-pause, tik-tik-pause, the harmonic structure will be richer. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●December 3, 20072007-12-03
On Dec 3, 12:04 pm, Jerry Avins <j...@ieee.org> wrote:> Again, this assumes that the two-per anomaly is "balanced". If it sounds > like tik-tik-pause, tik-tik-pause, the harmonic structure will be richer.Good point. If the "tiks" are close enough together, they start to look like a double-tik event that occurs once per revolution instead of two single-tik events. Greg
Reply by ●December 3, 20072007-12-03
On Dec 2, 11:53 pm, "mnentwig" <mnent...@elisanet.fi> wrote:> I would take a full number of cycles (!) of the signal and FFT it. Then > notch out the spectral line corresponding to the fundamental, IFFT back to > time domain, and you've got the double-frequency anomaly.That assumes that the periodic phenomena has significant (or any) energy in the fundamental frequency bin. Lot's of interesting phenomena have a (nearly) missing fundamental sinusoidal component. The interesting parts might all be in the harmonics and perhaps the phase relationships between such. But how does one go about separating out the even multiple harmonics of the "once per" event from the spectra of the "twice per" event plus all of its harmonics? But the fact the Greg can hear two ticks and one thump (or something like that) in the time/angular domain says that the impulse responses might be at least partially separable in that domain. IMHO. YMMV. -- rhn A.T nicholson d.0.t C-o-M
