Olov

Can you answer my question with a simple "yes" or "no", with plain English proof, that the data comes from or does not come from between 1000 to 1040 Hz?

If it does not, how does it decide which data to use, in the 40 Hz spectrum and waveform that is between the HP filter and the 40000 Hz max range?

quote:

but the highest peak within those 1000 values in the 40Hz band.

Looks like this quote from your reply says that this is where it comes from, 1000 to 1040 Hz with a HP filter of 1K.

Hi Ralph,
I think the answer is the waveform has data from 1Khz to 40Khz (it is sampling at 2.56*40Khz just like a regular acceleration waveform is sampled at 2.56* the f-max except PeakVue doesn't employ anti-aliasing filters so the Peakvue waveform is sampled at a rate of 102,400 samples per second). The high passed PeakVue waveform containing all data above the filter setting then has a FFT performed on it, just like a regular waveform has. Forget now that it is a PeakVue waveform, just look at as a regular waveform to see why the frequencies are seen in the spectrum. In a regular waveform if you have peaks spaced at a time period of every .0333 seconds you will expect to see a peak in the spectrum at 30 Hz. With Peakvue you will see the same thing, a peak at 30 Hz. When a FFT is done on a waveform the peaks seen in the spectrum are generated from the period between the repeating peaks in the waveform. If the waveform peaks are not periodic there will be no frequency seen in the spectrum. If you have a regular waveform from a bearing whth an inner race defect you will have impacting at the period of thr inner race defect , it will usually be amplitude modulated at the period of turning speed. You would expect to see the bearing inner race defect frequency with sidebands of turning speed in the spectrum, with peakvue you would see the same thing because the waveform has the high frequency energy generated at the period of the defect and at higher amplitude when it is in the load zone. You are not necessarily seeing the impact of the frequency (if the actual frequency is below the HP filter) but each time it impacts or generates a friction event (high frequency energy above the HP filter range will be generated) it will be repeated in the waveform at the period of the defect.
This is why you see frequencies in the spectrum that are below the HP filter. You are not actually seeing that frequency as in regular vibration data but you are seeing the period of the the high frequency energy generated each time the impact or friction event occurs. It doesn't have to be discrete peaks at this frequency but can be a "packet" of high frequency energy repeated each time the impact of friction occurs and you will still see the frequency that is the inverse of the period between the "packets" of energy in the spectrum.

Hopefully Dr. Robinson will see this and correct any mistakes or misconceptions I have about the way this works.

Jim Crowe

Ralph - The region roughly around 2 KHz +/-, depending on the bearing dimensions, contains most for the bearing outer race friction mode resonances and the energy in that is what you see most of if it's included in the passband before demodulation

Ralph,
I am also a bit curious about that 1000-1040 Hz bit but the quote was from Mr Robinson. So I can´t say simple yes or no I would lean on no since if you made a direct FFT of bandpass filtered data 1000-1040Hz you would not get the same result. Basically because the data is sampled to be OK and true up to 40KHz according to Mr. Nyquist´s rule. What you do is picking out the highest values from this large amount of data so how can this selection be made?
So we sample 102 000 times/s and only need 102 sample/s for 40Hz Fmax
so for each sample we need there are 1000 samples to pick from so from
these 1000 we take the one with the highest peak value and from the next 1000 samples we take the next value to put in the PV TWF as many as required for the resolution that is set, so for the 1024 samples required for 400 lines it takes 1024x1000 samples at 102 000/s so it takes 10s and that give the PV TWF that then can be made to a FFT.
So the measurement time would be the same as sampling 1024 times at 102 sampes/s but you had 10x102 000 samples to select from so the risk having a transient sneaking btw. the samples and getting by not noticed is very small. So compared to slow sampling or filtering or filtering and demod, there is a difference. So you can say that PV is a major over sampling procedure.
Sorry not so short and not yes or no but this is the way I tried to get it make sense to me, hope it´s not so far from the real deal. I have no inside on this part from reading what have been publicised. Olov

Ralph, the answer to your question regarding the 100 to 1040 Hz range is "no".

The mechanics of what PeakVue is doing is well described by Olov in his post on 6/6/05 at 07:08 AM is correct. Like wise, Jim Crowe's response is correct. PeakVue spectra identifies the rate at which a 'stress wave initating event' is occuring. The PeakVue TWF provides data to assist in severity assesment (how large are the impacts).

A point of clarification is that when the analogue signal is converted to a digital signal employing a 40 Khz bandwidth (a sampling rate of 102,400 S/S), the signal is passed through an antialiasing filter set at 40,000 Hz.

Jim Robinson Emerson Process Management

Is there an explanation of why there is a "tapered" down pattern in these spectrums of 500 Hz HP thru 20K Hz HP filters on the same bearing without ever moving the transducer, rather than using all the data that is available from the 40000 Hz sampled area and not "dividing" it by the "fmax". Is all this data in the waveform that is assigned to the 500 Hz spectrum? Or did we lose it when we divided by 100?


Sorry I drifted so far away from the original by Alan, but this question has bugged me for years.

One final question though, If I take a normal spectrum/waveform with a 2K Hz Fmax and then take a Peakvue with a 1K HP filter and 1K Hz Fmax, would I see the same thing, to a certain degree in the Peakvue that I see in the area from 1000 Hz to my defined 2000 Hz Fmax, if there is anything "happening between 1K and 2K? Same question, worded differently. Try it sometime.

Ralph - take a look at the the acceleration (black) curves in the graphs at:

http://vibrotek.com/articles/factors/index.htm

Note also that that the signal roll-off from each element is considerably different as frequency increases.

Just to make it complete, this a patent covering some of the things:
http://164.195.100.11/netacgi/nph-Parser?Sect1=PTO1&Sec...895857&RS=PN/5895857

Olov