While using CSI 2120 and comparing phase difference between IB and OB bearings in "Cross Phase" mode vs. same in "Fast Bal II" mode, I have noticed significant discrepancy - about 45deg.

Now I am questioning the "Cross phase" method accuracy since I trust more the tach pulse reference mark.

Did someone else run into this problem?

Thanks,
David

Hi David--

I think what you're seeing is the difference between "relative phase" between the two channels and the "absolute" phase recorded in "FastBal" relative to location of the tach reference. In my feeble understanding, the two don't correlate at all and can't really be compared.

Hope that helps!

Tony

quote:

Originally posted by TonyW:

... the two don't correlate at all and can't really be compared.

Tony,

I think the "relative phase" (cross phase) and the DIFFERENCE of "absolute" phases recorded in FastBal relative to location of the tach reference should be the same by definition. Physically both methods measure same parameter, just in a different way.

I have to add that there was no measurable speed fluctuation during the measurement and Correlation Factor was 0.95.

David

This message has been edited. Last edited by: David_G,

David--I have to agree with you there--and I think I've run into this before. In one situation I found that if I monitored the phase reading long enough, I could see it changing/varying by a huge amount (ID'ing the source was another story). In another, however, I never did figure out the reason for the discrepancy. Hopefully someone else has some ideas for us here!

Tony

They are measuring two different things. The cross channel is measuring the difference between the waveform peaks at the frequency of the fan say. Now, for the sake of argument, don't move the accelerometers. We will use one of them to compare to the tach pulse. These two angles have no relation to each other. Take your tach tape off and move it. Viola! A new phase angle and difference appears. Does this one match? Doubtful. There is no correlation between the reference placement, wherever you put it or whatever you choose, and the high point in the reference accelerometer waveform. It is a measureable difference but is does not have a right or wrong answer.

Hope it helps.
Sean

I think David's original question is the one we're still wrestling with though. Shouldn't the phase difference measured using cross channel phase be fairly close to the difference between phase measured on the same two points using a tach? Sure, the numbers will be way different, but I would expect that the relative phase difference should be similar regardless of how it's measured... What are we missing here?

Tony

quote:

Originally posted by Sean McGrady:
...Now, for the sake of argument, don't move the accelerometers. .... Take your tach tape off and move it. Viola! A new phase angle and difference appears.

Sean,

Once you moved the tach tape, new angles will be read on both accelerometers (each reading will add/subtract same amount) but the difference will stay the same.

As I have mentioned before the discrepancy may be explained by fluctuating speed in the case of cross phase method, although in my case the correlation factor was high.

David

Ralph--

The "sloppiness" of the waveform that you describe, especially at lower resolution, makes a lot of sense to me. It seems like almost anytime I'm measuring peak spacing to confirm or ID something in the spectrum that there's almost always a little "fudge factor."

If this is true, and I have no reason to believe otherwise, we should all take note as to the somewhat lack of "precision" we're dealing with in cross-channel phase measurements!

Tony

Intuition tells me that cross-channel phase will not be as accurate as tach measurements. I only use cross-channel when there is no reference mark on the shaft that my laser tach will pickup, and when I can't easily shut the machine off.

We should all keep in mind that cross-channel only gives a "relative" phase measurement, and it appears that while that is relatively close to accurate, it's not accurate.

I applaude David_G for taking the time to actually investigate the difference. I too often don't do that. But such experimentation is an excellent way to better learn the ins-and-outs of vibration measurement, and is time well spent.

If this is the phase of the transfer function between the channels, and if you use the same window function for each channel this should work. I am not familiar with this box, but it has worked on spectrum analyzers (and software fft's) using 2 channel analysis.

The phase differences in both cases should be the same. This I have observed in all occasions. Pl check if the error is consistent every time you measure.are you close to resonance on any particular direction?

Bill,

This is a CSI 2120 box and it is my guess they derive cross phase angle using the transfer function at a given frequency. The discrepancy I have observed should not have been there in theory unless coherence was low.

Rao,

There was no resonance in sight.

Rusty,

I'll try to set up a test to verify the possible discrepancy. I'll also introduce some noise in the signal.

One test that should be useful is to find the phase between the signal and itself. This should be 0. If you can invert the signal, you can test the signal and it's negative. Integrate the signal and compare; the integrated signal should lag the original by 90 degrees.

I'll position two accelerometers on the opposite sides of a metal plate of a vibrating structure in order to verify if both tach phase difference and cross phase is 180 deg.

This should be easy to do... Will report findings.

My previous response was to a misreading of the original question. Yes, the phase difference between the two points should read the same
regardless of the method used within the setup. I think some of the others indicated that the resolution could be the issue. I would agree. With the tach, the error is probably measured in microseconds. Used to be 300microseconds for the CSI phototach. With cross channel, a lot would have to do with how well they band pass the signal to accurately read the peak. Dynamic range of the signal would also play a part in being able to read the peak accurately. The meter needs to read both an amplitude and a time from both accels, instead of just one.

Sean

quote:

Originally posted by David_G:


I have to add that there was no measurable speed fluctuation during the measurement and Correlation Factor was 0.95.

David

In my experience, a coherence of .95 in a cross channel measurement would lessen my confidence in the accuracy of the phase reading; I typically see >= .99 coherence using ~10-15 averages.

Another item to consider is that both channels have sufficient amplitude at the frequency of concern.

I've ran a little test by placing two accels next to each other. Both methods indicated 0 phase difference at 3 mil amplitude. That was encouraging, but...

Again, when accels were placed on IB and OB bearings, discrepancy between Tach and Cross Phase methods amounted to about 50 deg (respectively 200 deg and 150 deg). Trying various # of lines in Cross Phase method did not make any difference.

Coherence - 0.97. Amplitude on each of the bearings - around 3 mil. Averages - 20.

According to CSI, 0.90 or higher coherence is required for cross phase measurement.

200 degrees and 160 degrees would be complements of each other. 150 is not much removed. What happens when you switch channels for the cross phase?

Bill,

You are correct. They are complimentary. I did not notice it (I gave approximate angles, in fact, they are exactly complimentary).

Conclusion: Both methods showed the same difference which is, of course, expected.

Sorry for the confusion, but at least we are confident now that both methods produce same results.