Once in awhile I run into a problem of unsteady instantanious phase reading in a range of about 50-70 deg. A resonance would be one scenario when this occurs, but I believe it is not the only one.

I suspect two additional possible causes other then resonance.
1. Low vibration amplitude. Because of that the analyzer is having hard time detecting the maximum value.
2. Looseness, which results in 1x actually changing its phase randomly.

Neither of the above are mentioned in the vibration literature as far as I am aware.

Any information, experience, opinions?
Thanks,
David

David,

I have seen erratic phase measurements from both of the causes you mention. Sometimes you run into the low amplitude condition when balancing. The 1X RPM amplitude gets lower than the instrument's capability to acquire a steady phase reading.

Technical Associates of Charlotte has included both of these phase indications in their course materials.

Skip Hartman

Looseness is the first thing I look for when the phase changes. Does the amplitude change with phase? Double check you mounting of the Accel.

Also go with 1V/g hi-res and filter if necessary. The trigger is 1X snyc but you can also check phase out several places. If a signal needs cleaning, you may try a 100mF cap. I build an AL box (radio shack) w/BNC's and cap inside for an eddy current probe i've used to balance with.

quote:

1. Low vibration amplitude. Because of that the analyzer is having hard time detecting the maximum value.

Hi David,

A "trick" I have found that works when low amplitude makes my phase unstable when trying to get a balance job down to an amplitude lower than the "acceptable" amplitude is to "fool" the meter by telling it I am using a more sensitivity transducer than I actualy am.

Thank you guys for the advise.

Using a higher V/g's without telling about it to the analyzer is a sacred lie. But there is another factor affecting phase measurement accuracy and possibly stability which is Lines resolution.

When phase measurement is being performed, the analyst is not given the choice of setting up the resolution. All that is happening behind the scene. I'd think it is set up to the maximum available resolution, but it is interesting to know what is the actual resolution and what is the phase reading accuracy?

I may be way off base, but I thought that phase determination is done using time domain calculations without FFT.

I don't know for sure how it would be done... maybe compare the time from a reference event to perhaps peak of the waveform as a fraction of time between reference events. Then doing 2x phase for example would get really weird - take the reference pulse and make it occur twice as fast and repeat the process above. The results of this approach in non-sinusoidal cases would not give the same result as the FFT approach but I thought that's how it's done.

Mr. Greg makes a good point about making sure the transducer is secure. That kind of looseness definitely leads to unstable phase readings! I messed up a demonstration of comparative phase readings once because of this. It took me couple minutes to figure out what was going on and get my demo back on track. I wish I could claim it was a rookie error, but I was already too old to make that claim!

Electricpete, different instruments use different routes to determining phase. Some use the time domain approach you mentioned, others compute phase spectrums for both the tach pulse and the vibration transducers and compare the two.

I used to do phase measurements with a Wavetek real time analyzer where you set up essentially a Time Synchronous Average measurement. As the cursor was pulled across each line of resolution in the spectrum there was a phase readout as well as an amplitude readout. On that unit we would often have better success at getting a stable phase reading by using LOWER resolution in the FFT. Theory was that the lower resolution allowed the 1X RPM peak to stay in the same line of resolution during the averaging process, if speed was slightly varying.

The MAARS Model 5000 unit I use now for field diagnostics uses the waveform method for routine phase measurements and monitoring. The MAARS balance program allows the analyst to choose if he wants to use Time Synchronous Averaging to get more stable phase readings to balance to a lower amplitude.

Skip Hartman

With synchronous sampling a triggered FFT approach aught to work. There seem to be a few methods.

Comparing the FFT’s of the trigger to the signal (transfer function) has some potential issues. This method has an almost pure electronic definition. There is little energy in the trigger, usually. Slight changes in the pulse width or shape could result in significant changes to the 1X spectral content of the trigger, including its phase content. This of course would result in an error.

One method still employed dates back at least 70 years when one could use signals proportional to synchronized voltage and phase sinusoidal waves by which to multiply the signal. This causes sum and difference frequencies. Setting a filter at DC (i.e. low pass filter) will pick up components proportional to the ‘real’ and ‘imaginary’ parts of the amplitude and phase.

I have seen the sinusoidal components generated optically by having a shaft ½ dark and ½ light, thus generating a square wave. Integrate the square wave and the 1X phase component will shift 90 degrees. Do the proper scaling, and you can get the real and imaginary parts (in-phase and quadrature components for some electrical types).