Dynamic Torque Measurement Background multiplication can’t occur. Nonetheless, the frequency response Rotary power sources and absorbers have discrete poles and/ defined by equations [1] and [2] is always higher than that of a or pistons and/or gear meshes, etc. As a result, they develop drive with damping. Thus, these equations will quickly reveal and absorb torque in a pulsating rather than a smooth manner. the measurement bandwidth and natural frequency limits. Furthermore, a driveline consists of several inertias and torsion 1 springs which resonate at one or more frequencies. Finally, even Pump Input Torque Jm when running at “constant” average speed, every drive has some Motor Output Torque = 1 + 1 - 4π2 f2 1 angular acceleration which, in combination with shaft inertias, Jm Jp k generate dynamic inertia torques. Setting the denominator of equation [1] to zero and solving Thus, real world driveline torque is never constant. Instead, for frequency yields Fr. Fr is in hertz when Jm and Jp have in-lbf it consists of an average torque with superimposed oscillatory s2 units, and k has in lbf-in/rad units. components. The oscillatory components can excite driveline resonance(s). These effects are exacerbated during transient load conditions. Accurate measurement and/or control of
Fr = 1 2π √ (Jm + Jp ) k (Jm Jp ) 2 dynamic driveline torque requires an understanding of how the Figure 1 is a plot of the undamped frequency response of the Torquemeter and other drive components interact. This note Motor-Torquemeter-Pump drive with: provides insight into those interactions from theoretical and Jm = 100 in-lbf s2 practical viewpoints. Jp = 172 in-lbf s2 k = 309,000,000 lbf-in/rad for an MCRT 87007V(25-3) Torque Measurements Under Steady State Conditions Torquemeter, and Consider a drive consisting of a Motor, a Torquemeter and a k = 28,412,000 lbf-in/rad (3,210 kNm/rad) for a competitive Bearingless Torquemeter Pump. Assume the couplings and shafts are infinitely stiff, and Motor Torque = 20,000 lbf-in the Torquemeter and couplings have negligible inertia. Then, the shaft network consists of the Motor Inertia (Jm), the Pump The stiffer MCRT device has a significantly higher Inertia (Jp), the Torquemeter spring constant (k), and system undamped natural frequency than the competitive device. damping (Ω); damping is assumed to be viscous. Frequencies above about 1.5Fr are attenuated. At 6 kHz the attenuation is 290 times for the MCRT Torquemeter and 3,160 Undamped Frequency Response times for the competitive device. With damping present, the With no damping present, equation [1] defines the drive’s drive’s resonant frequency and measurement bandwidth will frequency response, i.e., the ratio of pump input torque to motor be lower; see following discussion. output torque. At the drives natural frequency (Fr) the output When torque is input to the undamped network, no power torque becomes infinite. When damping is present, infinite is delivered or absorbed. As a result, the drive will accelerate the
Application Note 221101D S. Himmelstein and Company
shaft system to the motor’s no-load speed. Since both inertias is greater than critical it is Overdamped. Figures 2 and 3 are subjected to the same acceleration, equation [3] describes plot critical, 0.2 times critical and 5 times critical responses. the frequency response to about 20 percent of Fr. See Tech Memo 221201 for details on computing damped Output Torque Jp system response. Input Torque = (Jp + Jm ) 3 A device’s measurement bandwidth is commonly defined as the frequency at which the response is down 3 DB or, has For the case being discussed, equation [3] confirms the fallen to 70.79% of the start value. The following table lists drive 0.632 ratio observed in Figure 1. performance with each Torquemeter installed versus network damping. For completeness the -3 DB frequency as well as the Frequency Response With Damping Present frequencies at which the response falls to 1%, 0.1% and 0.01% Figures 2 and 3 show the frequency response of the drive with are included. damping present. Since torque is input directly to the motor Whether you accept the 3 DB criteria or a more stringent inertia (Jm)—think of it as the developed motor torque — and definition, the stiffer MCRT 87007V Torquemeter clearly power is absorbed by the pump, then the ratio of output to provides the widest measurement bandwidth. Its bandwidth input torques is unity well below resonance. is about 3.3 times higher than the other device. Note 3.3 is Depending on damping, three response types are the square root of the two Torquemeters’ stiffness ratio. possible. We define Critical Damping as the lowest damping At low frequencies, the effect of damping is relatively that avoids oscillatory torques and overshoots when the drive small. At intermediate frequencies, torque can be multiplied is excited by a step function. A network is Underdamped if or attenuated depending on system damping. Since damping its damping is less than critical. Oscillations and overshoots is usually unknown, the precise torque value is indetermi- will occur when the drive is underdamped. When damping nate in this frequency region. Furthermore, data in this
Figure 3. Response with competitive 3kNm torquemeter installed.
Application Note 221101D S. Himmelstein and Company
frequency region only reflects the dynamic driveline many applications. During transient conditions three system response if the Torquemeter is permanently installed. responses are possible. Which occurs depends on whether High frequencies are significantly reduced, asymptotically system damping is equal to, greater or less than Critical approaching 40 DB per decade. Assuming critical damping, Damping. with the MCRT 87007V installed, 3 kHz signals are reduced Remember, Critical Damping (Ωc) is the lowest that by 115 times. At 3 kHz the reduction is 1,250 times with the avoids ringing and overshoots when driven by step function. competitive device rendering any data useless. At 6 kHz System response is fastest when damping is critical. When signal reduction is 460 times for the MCRT 87007V(25-3) damping is below critical, oscillatory torques will be and 4,998 times for the more compliant competitive device. generated which can result in large torque overshoots and With either device installed, signals of 1 kHz are suspect, increase the time to reach equilibrium. When damping is 6 kHz and higher torque signals are useless. greater than critical, oscillations don’t occur, but the Despite both Torquemeters’ wide signal chain bandwidth, response time is increased. actual installed measurement bandwidth is much less. It is Figures 4 and 5 show the drive response to a 20,000 primarily determined by the drive and load inertias and the lbf-in step input. The damping used is critical, five times Torquemeter stiffness and, to a lesser extent, by system critical and one fifth of critical. The much stiffer MCRT damping. As long as the signal chain bandwidth is higher 87007V(25-3) has significantly faster response than the than the drives’ natural frequency, it has no effect on competitive product. The table that follows lists pertinent measurement bandwidth. response times. When the damping is 20% of critical, the peak overshoot Torque Measurement During Transient Conditions is within the Mechanical Overload and Electrical Over- Starting, stopping, reversing, impact loads, etc. are transient range ratings of both Torquemeters, but the competitive occurrences that must be monitored and/or controlled in device is marginal on Electrical Overrange. The MCRT
Figure 5. Step response with competitive 3kNm torquemeter installed.
Application Note 221101D S. Himmelstein and Company
has much higher margins and is truly linear in both Torquemeter Length Mechanical Overload and Electrical Overrange. Since Bearingless Flanged Torquemeters provide the highest stiff- the competitive device is only repeatable in Electrical ness, lowest deflection, shortest overall length and highest Overrange, it will have unknown errors in that region. natural frequency. However, you should be aware a very When damping is lower, torque peaks increase short length carries an accuracy/performance penalty. exacerbating that situation. See Tech Memo 221201 for That is, a very short sensor sacrifices isolation from the details on computing damped system response. attachment bolts to the torsion element. The problem doesn’t To summarize, like the steady state case, transient show up during static calibration. Nonetheless, in a typical response speed is primarily a function of Torquemeter installation, dynamic bending and thrust loads will induce Stiffness and the coupled Drive and Load Inertias. error signals in very short devices. Because of this, even Network damping plays a secondary but vital role. though a shorter length lowers material and machining Provided it is higher than the drive’s natural frequency, costs, Himmelstein products are designed with enough signal chain bandwidth has no effect on the measurement length to provide inherent isolation during operation, of transient phenomena. and to provide the other characteristics needed for accurate static and dynamic measurements.
Other Important Matters Ambient Electrical Noise
Torquemeter Electrical Overrange Dynamic torque measurements can be corrupted by Without sufficient Electrical Overrange, when torque electrical noise often present in industrial environments. signals are large, torque peaks are clipped. That results in Pulse Width Modulation based adjustable speed drives, large errors in reported average torque and, also generates which use carrier frequencies between 1 and 10 kHz, are output frequencies not present on the drive. Modern a common noise source. Without proper grounding and Himmelstein Torquemeters have high (150 to 300%) adequate shielding , magnetic and capacitive coupling and Electrical Overrange to avoid these errors. Moreover, to leakage currents can induce high frequency noise in the eliminate signal distortion and amplitude errors, we guaran- torque instrumentation and data cabling. Himmelstein tee maximum nonlinearity of 0.1% in the Electrical Torquemeters and Readouts include selectable low pass Overrange region. Many competitive Torquemeters have filters with at least ten cutoff frequencies. Those filters can virtually no Electrical Overrange. Others, such as the one optimize dynamic measurements by providing the widest listed, have insufficient Electrical Overrange and don’t useable bandwidth while eliminating high frequency noise. specify nonlinearity in Overrange; they merely guarantee repeatability. See Application Note 20805B for more Torquemeter Natural Frequency information on the critical importance of Overrange. For accurate dynamic measurements, the Torquemeter must have a natural frequency (F n) above the drivelines’ natural Signal Chain Bandwidth frequency. Himmelstein production Torquemeters have The signal chain, rotor and stator electronics and rotor to stator natural frequencies from several hundred Hz to above 10 signal transmission, should have a bandwidth well above the kHz. The resonance of practical drivelines virtually always highest measurable shaft frequency. As a practical matter, falls between 5 and 500 Hz. their bandwidth is constant for any Torquemeter Series. Flanged Torquemeters have the highest natural frequency For modern Himmelstein Torquemeters, it is well above the (F n). You can estimate a Torquemeters’ F n using equation [2] driveline torsional natural frequency. Signal chain bandwidth and published values for Stiffness (k) and Inertia. Him- never determines the maximum measurement bandwidth melstein Torquemeters are axially symmetrical. That is, the when a well-designed Torquemeter is installed. It is primarily input and output inertias are equal. Thus the Torquemeter is determined by the drive and load inertias and the a torsion spring with half the total Inertia at each end; Torquemeter Torsional Stiffness. see relevant specifications.
