NVGate CBT principle and settings
Constant Band Tracking feature consists in Order analysis calculation based on energy summation from FFT data. Summation band (in Hz) defined by user will be fixed over time, through run up or down condition of measurement. Provided tach signal is connected to FFT, energy summation around multiple (integer or not) of rotating frequency will be calculated in real-time.
FFT Spectrum giving discrete frequency lines, energy summation will be bases on summation of spectral lines values. Minimum number of lines to be taken into account is related to windowing function used for time block filtering. Up to 8 different orders may be extracted per channels. Order selection may be different from one channel to another.
Setting Constant Band Tacking Parameters
a/ Order amplitude extraction
Order amplitude extraction from FFT spectra is highly dependant on FFT speed itself. High resolution in FFT setting (delta f very small) will result in long trigger block and consequently slow FFT refresh. Fast run or run down conditions (quickly changing vibration signature) may therefore be incompatible with slow FFT calculation. Several ways to go round this problem may be to reduce Spectrum resolution to an acceptable value and add Overlap (see picture 1) in FFT calculation.
Number of averages is also a parameter to look after as too many averages may smooth run up/down vibration signature. Contrarily too few averages may result in a noisy looking order amplitude curve.
Picture 1: Overlap setting
b/ Frequency bandwidth definition
Frequency band on which order amplitude may be extracted is to be defined by user for each channel (see picture 2). Too wide frequency band definition may lead to measurement error, as harmonics from rotation speed may be very close at low speed. For instance at 1000 RPM, Order 1 will be at 16.6Hz, Order 2 at 33.2Hz, defining a frequency band for energy summation of 50Hz for instance for order 1 will result in integrating Order 1 and Order 2 in this case. More generally speaking the amplitude value of Order 1 calculated with Constant Band Tracking technique may often be overestimated during low speed analysis process.
Picture 2: frequency bandwidth definition
Minimum frequency bandwidth available
Minimum band available is directly related to weighting window type used in FFT calculation.
Due to window shape, energy of a pure tone sine wave for instance may be split on 3 to more spectral line. Consequently a minimum frequency band will be proposed in the setting window.
Example of Hanning window:
From the curve shown in Picture 3, it is to be seen that applying Hanning weighting window on a pure tone sine wave will result in energy spread between the 2 side spectral lines (-1 and +1 on Picture 3) of sine frequency. Energy on the next spectral lines (-2 and +2 on Picture 3) will be reduced to 0 amplitude.
Picture 3: Hanning weighting window
In this case, if f is the spectral resolution of FFT, f0 the frequency of a vibration phenomenon, minimum energy calculation for Constant Band tracking should be done at spectral lines:
[(f0-f) + f0 + (f0+f)]
that is to say a frequency bandwidth of
| CBT min band | = (f0+f)-(f0-f) | ||
| = 2*f | |||
NOTE: when f0 is not matching exactly one spectral line (f0 is not a multiple of f) leakage phenomenon may happen. In this case energy of vibration will be spread on next closest spectral lines (on the left and right). Energy picked up at spectral lines (f0+f) or (f0-f) corresponds itself to the energy component measured on a frequency band of [f0+f-; f0+f+] or [f0-f-; f0-f+], where correspond to leakage bandwidth. In OROS OR2x, was considered to be equivalent to f/2.
c/ Phase calculation in Constant Band Tracking
In order to extract precise phase information from Constant Band Tracking, measurement has to be synchronised with rotational speed. Acquisitions of FFT blocks (trigger block) will need to be synchronised with Tach pulse in order to guaranty exact start condition in time (see Picture 4)
Picture 4: Trigger setup for CBT phase
Increase of phase precision with averaging domain.
Provided the FFT trigger condition is set to Tach pulses, averaging domain be can be set to Time. This will have for consequence to activate Time Shift Resampling (see User Note: S015-081-1 Time shift resampling) in time domain and therefore increase phase accuracy of trigger block by a factor of 16.(see Picture 5)
Picture 5: Time domain averaging setting
Similarly, to guaranty that core information of Trigger Block (and therefore phase information) is not filtered out by weighting window, it is recommended to center time Trigger Bock by shifting it half a block. This is made possible through feature Start Delay by adjusting it to a negative value of half the size of a trigger block (equivalent to 1/[2*f], where f is the spectral resolution). (see picture 6)
Picture 6: Trigger delay setting in CBT