Page 10 - Combined_41_OCR
P. 10

cable losses and minimizing impedance-matching   The reference value in this table was 1 g; it could
           problems.                                        as easily have been another value, such as 1 [ig
                                                             (10-6g).
           Peripheral Instrumentation                         In some instruments used for power analysis,
                                                            readout is in terms of power rather than voltage.
             Selecting Readout Parameter: Sometimes a user   Because power is proportional to E2, then a dif­
           of vibration information prefers one of the param­  ferent equation is required:
           eters (displacement, velocity, or acceleration) while
           using a pickup that senses another. For instance,                            I
           aircraft flight crews may prefer a displacement                  dB — 10 logio   a?
           readout but use velocity-sensing pickups on en­     Systems Calibration: Calibration is an orderly
           gines. The signals are electronically converted (in­  procedure for finding out how sensitivity varies
           tegrated) to represent displacement. Increasingly,   over a given range of frequency and amplitude.
           accelerometers are used in such a manner; their   Best accuracy is achieved if all elements are cali­
           signals can be integrated once, to represent ve­  brated as a system. Several general rules for cali­
           locity, or twice, to represent displacement.      bration are suggested here; literature obtainable
                                                            from the sensor manufacturer provides specific,
             Selecting Readout Format: If it is necessary to   detailed calibration instructions.
           examine the details of a motion, an oscilloscope    Displacement pickups are easy to calibrate
           or recording oscillograph is required. Meters give   statically, since relative motion between fixed and
           statistical properties of vibration such as peak or   moving parts can be generated by a lead screw.
           rms values, but measuring short, severe spikes    However, resonance in the pickup or its mounting
            (representing impact superimposed upon a vibra­  might change sensitivity in the frequency range
           tion) cannot be done with meters.                 of interest. The only way to demonstrate flat fre­
             Only in the rare condition of sinusoidal motion,   quency response is to attach the moving portion
           with all energy at one frequency, can electronic   to a motion generator.
           meters be trusted. The one exception to this rule   Strain gage, PR, and servo accelerometers can
           arises when using true-rms (trms) meters. All     be statically calibrated very simply. Hold the pick­
           other a-c-sensing electronic meters err in the com­  up on a flat, horizontal surface, and balance any
           mon situation of nonsinusoidal vibration where    electrical output to zero. Then invert the pickup
           energy exists at several frequencies. The extra   so that gravity acts in the opposite direction and
           cost of trms meters is justified, since these meters   balance the same output, again to zero. Divide
           give accurate rms indications no matter what the   the difference in output by 2 to get output/g.
           waveform. Note: A scale indicating that readout   Static accelerations on the order of ±1 g can now
           is in “volts rms” is no guarantee that the meter is   be measured. Before static accelerations signifi­
           trms.                                             cantly larger than ±1 g are measured, the accel­
                                                             erometer should be calibrated on a centrifuge.
             Linear vs. Log Scale: Normal variations of fre­   Static calibration, unfortunately, will not give
           quency or amplitude are often large enough to     dynamic sensitivity. Most crystal accelerometers
           drive the reading off of the scale of a linear-scale   (and all velocity pickups) can only be calibrated
           meter. Many meters, in short, lack sufficient dy­  dynamically on a shaker, since their output can­
           namic range. This range can be increased several   not be measured at zero frequency. For these
           orders of magnitude by using a logarithmic con­   sensors, the mechanical input should be purely
           verter. Readout of a log converter is usually in
                                                             sinusoidal so the electrical output is also sinusoidal.
           decibels, calculated from:                        This simplifies measurement of the output voltage.
                                       E
                           dB = 20 logio------
                                                               Handle With Care: The finest measurement sys­
           where E is a voltage representing displacement,   tem does not guarantee accurate measurements,
           velocity, or acceleration. E() represents an arbitrary   even when sensitivity is determined accurately by
           reference or standard value (zero dB) of that volt­  calibration. The system must be used correctly and
           age. Accelerometer output can thus be readily con­  handled carefully. The pickup must be attached
           verted from linear to logarithmic (decibel) form:  firmly to the vibrating structure, and the signal
                                                             cable protected from undue stresses. In summary,
                       Linear Output   Log Output            the instruction manuals must be read and followed.
                        (g units)       (dB)
                                                               Erroneous data caused by carelessness may
                          0.001         -60                  mean that the whole data-gathering job must be
                          0.01          -40
                          0.1           -20                  done again—if the engineer is lucky. If he is not
                          1                0                 lucky, he may continue with design, fabrication,
                          2             + 6                  and installation, never knowing until too late that
                          4             + 12                 his data are faulty.
                          10            + 20
                         100            + 40                   Next article in this series discusses data gath­
                       1,000            + 60
                                                             ering and analysis.
           124                                                                                Machine Design
   5   6   7   8   9   10