Strange Spectral Peaks around 0 hz? 3

[rasmumm]

I had been having trouble with taking measurements using my 2120 CSI analyzer. on some readings the overall measurements have been much higher than the last. The difference in the spectrum being that there was a peak between 0 and 2.5hz that had a ski slope effect. This peak did not show up everytime that I took readings on that component, and for some months I believed it to be a problem with the meter. Recently I put another meter on live mode to monitor the spectrum and discovered that on somewhat of a random basis the "ski slope" would appear, and then completely go away only to appear a few moments later. I can rule out the meter since I have used a number of others with the same result Does anyone have any idea what could be the cause of the peak? What could cause this?

 

[FredGarvin]

Rasmumm,

You could be seeing alisaing of the data. If your sampling rate on your analyizer is less than twice the highest frequency you are interested in observing, some of that high frequency data can rollover into your low frequency data and give you a huge mess at the low frequencies. Or, along the same lines, you could have a high frequency component you're not expecting showing up. Your analyizer should have an anti-alias filter in an effort to reduce this effect. Whether you are interested in the high frequency information or not, if it's there, you will get aliasing if your set up is not correct.

 

[plawrence]

According to RBMware help:

Aliasing
An effect that results in erroneous frequency spectra when the frequency of the signal being sampled is more than 0.5 times the sampling rate. CSI analyzers include anti-aliasing filters that eliminate these errors.

Also, if you're using the wizard to build your points, you should have a 2.0 Hz low-end cutoff (see Low Frequency Signal Conditioning Limit) in your analysis parameter set that elimates this trash from the calculations.

I've also seen this kind of behaviour if the accel goes through a big temperature swing.

Out or curiosity, what's the running speed and application?

Patrick

 

[GregLocock]

I doubt that it is antialiasing, as the unknown signal would have to be almost exactly equal to the sampling rate for it to manifest itself as a near DC signal. Possible but unlikeley.

It is almost invariably an instrumentation issue.

Depending on the type of accelerometer (I'll assume it is charge coupled) the following are possible causes

1) You've forgotten to switch high pass filtering on as Peter suggested. This is often referred to as AC/DC on the setting, rather inaccurately.

2) The lead is waving around when you are making the measurement. Microdot leads are very sensitive to this. Leads should be taped down.

3) You may have damaged the cable - typically the outer core breaks away from the plug.

4) Water in the lead

There are a few others, but those normally sort it out.


Cheers

Greg Locock

 

[FredGarvin]

That is definitely true about the cable. I forgot about that. That would also help with the intermittent part of the problem.

 

[anands78]

Greglocock, I was of the opinion that if the microdot cable is bad, then the red "Fault" light in the Signal conditioner unit turns on. I have only used PCB accelerometers; is this true in other units also? Rasmumm are you doing any sort of processing (say integration) on the signal? I recently did measurements with an accmeter and integratd it to get displacements and found huge low freq peaks(mine was static tests though..)

 

[GregLocock]

Not reliably. The little PCB charge amps are just detecting continuity and shorts, as I remember, whereas nasty signals can be caused by cable faults that pass these simple tests.
I've used 3 different industrial vibration analysers, none of them checked the cables directly.

Just thinking about that, they tended to use ICP rather than charge coupling, I'm not sure but they should be less sensitive to tribolitic (cable whipping) noise.
Cheers

Greg Locock

 

[MachineryWatch]

Rasmumm,
I have seen this behavior when using a 100 millivolt/g accelerometer on a screw compressor oil pump (gear type). The pump was worn and was creating acceleration levels over 50 g's. This was saturating the amps in the ICP accelerometer. A quick look at the specs for the sensor indicated 50 g's was it's limit. I tried several other manufacturers units and got the same results. I switched to a 10 millivolt/g sensitivity accelerometer which can take higher g levels and had no more problems getting good data at that position.

Skip Hartman

http://www.machinerywatch.com

 

[rasmumm]

Thank you all, this information has been very helpful!

 

[DanT]

With IRD Datapak I can set up a measuring a point to watch a running overall for a few seconds, and then require another hit of the button to start the data collection.
Slopes are pretty much a thing of the past, although I don't think I hve measured a real nasty bearing in a while.

With our old TEC smart meters >>one<< of the ways I could induce a significant ski-slope was to fidget and sway, so the cable would swing a little. After a couple of hours of data collecting it is hard to keep my mind from a-wandering.

 

[jmack]

We also use the IRD Datapak as well as the Enwatch system and I personally see &quot;Ski Slopes&quot; on a regular basis. We see them in flooded dryer situations where the saturation effect is prominant, we in fact use this as an indicator of flooding! This can be saturation where you are collecting a signal well outside the maximum frequency range for the transducer.