VIEWS: 4 PAGES: 14 POSTED ON: 5/27/2011
Direct Current Measurement: a NEPTUNE Power System white paper Harold Kirkham 1. Scope Question addressed This document addresses the measurement of direct current for NEPTUNE. The purpose is to review the methods that might be used for control and for protective relaying, and make recom- mendations for consideration by the Power Group. Definition There are typically two aspects to the problem of current measurement: transduction either into some useful scale or into some other quantity, and isolation from the circuit being measured. For this white paper, current measurement means the obtaining of a value at one location of the current at another, with suitable isolation from the voltages involved. Thus, it is specifically intended that the isolation function be included in these considerations. 2. Review of available current measurement methods A Taxonomy Consider first the transduction aspects of the measurement of current. A current in a wire is associated with a magnetic field (in the wire itself and in the space surrounding it) and with a volt-drop in the series resistance of the wire. Both of these effects can be used for current measurement. The magnetic field can be measured in many ways: Force (the ordinary ammeter is essentially a force measuring device that compares the force generated by the field with the strength of a spring), Magnetoresistance (materials exist whose resistance is a function of the applied magnetic field), Hall Effect (a voltage is generated in a material in proportion to the applied field and the magnitude of a control current), Magnetostriction (materials exist whose physical size is a function of the applied magnetic field), Faraday Effect (rotation of the plane of polarization of light in an optical material). If the current is ac, there is also the possibility of inductive effects, but we are concerned here with direct current. In addition to the magnetic field effects, the volt-drop across a resistor can be measured easily enough. The question of isolation is not so readily answered, however. A good current measurement requires that the series resistance be low – and this means that the volt-drop across it will be small, while the full voltage of the circuit appears as a common-mode voltage in the measuring system. Typically in the power world, if a transformer is not used, the voltages are so high that it is necessary to convert the voltage into a signal that can be transmitted via fiber optics. This conversion requires electronics operating at the full line potential – a possible but not always simple feat. For isolation in the magnetic sensors, the most common construction places a magnetic toroid around the current carrying conductor (thereby forming a single-turn primary winding). This toroid, and any secondary windings on it, are then easily isolated from the voltage of the primary. Table I summarizes the discussion as a taxonomy: Table I. A taxonomy of current measurement methods transduction method measurement effect isolation magnetic force usually results from the fact that the field extends through all magnetoresistance space Hall Effect magnetostriction magnetic field must be high, so this approach uses optical isolation Faraday Effect inherently optical electrical (volt-drop) voltage transformer or fiber optics For the purposes of the NEPTUNE power system, force-based methods can probably be ruled out. In most applications of this approach, the force is used simply as a way to obtain the deflection of a indicator, such as an instrument pointer, so the approach is not applicable. In the “force-balance” however, while the method is cumbersome, it does have some attractive attributes. Like a bridge, the deflection is zero at balance, so any aging of mechanical parts will have minimal effect. And because it works by a feedback method, it is linear over a large dynamic range. Magnetoresistance and the Hall effect are both amenable to implementations in which the measurement element is a “witness” to the magnetic field. The most accurate (and most common) approach is to place the sensor element in an air gap in a toroid, as shown in Figure 1. magnetic core sensing element air gap conductor Figure 1. Current measurement by sensing the magnetic field The approach shown in Figure 1 is applicable to the magnetoresistive, Hall, and magnetostrictive methods of current sensing. It suffers two problems because of the air gap. First, the measurement is sensitive to external fields, particularly as those caused by currents in nearby conductors. Second, the accuracy is poor, as the output depends on exactly where in the cross section of the toroid the sensing element is located. There is also some dependence on the location of the current-carrying conductor within the magnetic circuit, an effect that is exacerbated by a large air gap. A variation of the method dispenses with the magnetic circuit entirely, and uses the sensing element to “witness” the field in air. This approach has relatively low accuracy, as the output is normally quite position dependant. These defects are overcome by the methods in which the integral of the field is measured, rather than the field at a point near the conductor. The integral is, of course, around a closed loop encircling the conductor, so that Ampere’s law (I = Hdl) is implemented, and the actual shape of the closed loop is immaterial. Square loops are used if the sensing is via bulk optics. Circular ones predominate otherwise. These arrangements are shown in Figure 2. sensing optics sensing fiber light path light path conductor conductor input fiber output fiber input fiber output fiber (a) Bulk Optics (b) Fiber Optics Figure 2. Optical current sensing with the Faraday Effect The main problem with optical current measurement is it is the rotation of the plane of polarization of the optical signal that is of interest, and this cannot be measured directly. It is necessary to convert the rotation to an amplitude, and to cancel out any signal amplitude variations that are due to other effects (such as connector losses, or temperature). The components needed to accomplish this are shown in Figure 3. Light Polarizer Analyzer Source Sensing Optics Detector L Faraday Sensor D Input Output Fiber Fiber Figure 3. Typical arrangement of optical components in a Faraday sensor Optical current measurement is therefore unavoidably complex, and justified only where the cost of this complexity can be justified, for example in very high voltage systems. Isolation For the optical methods, isolation is achieved simply so long as the measurement electronics are not close to the conductor. This applies whether the optical path is in free-space, in bulk optics, or in a fiber, see Figure 2. However, it is doubtful that the complexity of an optical system can be justified on the NEPTUNE power system, so we will not dwell on the matter. For the magnetic methods, isolation by a magnetic core should be possible. In the case of magnetoresistance, Hall, and magnetostrictive transducers, the flux in a magnetic core is measured in an air-gap in the core, see Figure 1. Provided suitable clearances are arranged, isolation is simple. However, the problems caused by the air gap warrant our looking further. In an ac system, current is conventionally scaled by a transformer before it is measured. The same philosophy is used in HVdc systems, but the transformer is a little different. The primary current is) arranged to pass through a single conductor in the center of a toroid (as is typical in power systems. Actually, for dc measurements, there are two toroids. Each toroid has an excitation winding, supplied from an external source. The excitation windings are arranged in series opposition. The system direct current affects the magnetization of the two cores in such a way that the current from the ac source is affected. In some implementations, the excitation winding is also used as a sense winding. In others, a separate winding is employed. Figure 4 shows the arrangement. saturable magnetic core primary conductor excitation and sensing winding Figure 4. “Current transformer” for direct current The technique may have existed in early HVdc schemes, before the invention of the fluxgate magnetometer in WWII. The principle of operation is the same. Fault-Current Tolerance During its expected 30-year life, NEPTUNE may experience as many as a dozen cable faults, most as a result of fishing. A cable fault results in a large momentary current as the cable capacitance discharges. The peak value of the current may be as high as a few hundred amps, many times the normal current being measured, and orders of magnitude higher than the minimum current that it may be necessary to detect. Some current measurement methods may be prone to problems because of this current. It is well- known in ac measurements that toroidal-core devices (such as current transformers) are susceptible to dc magnetization: this is one of the causes of protection problems in power systems subject to quasi-static induction during solar wind events. A core that is magnetized away from the center of its B-H loop will not read accurately, and must be demagnetized by the application of a large, exponentially-decaying ac excitation. For NEPTUNE, such magnetization would be possible at all current measurement sites between the fault and the shore. In view of the cost of accessing such a potentially large number of nodes, our design for current measurement must either be immune to such problems or it must include the extra complexity of a demagnetizing scheme. Leakage Detection To implement the metering and most of the protection functions of NEPTUNE, it will be necessary to measure the current at various locations in the node and on shore. However, to implement the detection of leakage currents on the LV side of the nodes, some other approach must be used. At MBARI, the leakage detection approach taken has been to “float” the load circuit, and ground one side of the supply through a high-value resistor. In order that there should be no place in the wiring that a fault to ground could exist, each side of the supply is grounded in turn. Since the grounding resistor is around 1 M, a current as small as 1 A can be detected easily. A problem with this approach is that there is a transient charging current every time the circuit ground is moved, and this must be accounted for in the detection system. One way to do this would be to disable detection until the transient had disappeared. This is how transformer differential protection is implemented in power systems. Another way would be to compensate for the current in the detection algorithm. This is what MBARI does. Another approach would be to ground both sides of the power supply output via a resistor, and to allow the possibility of a ground fault occurring that resulted in no ground current. Statistically, this must be unlikely. An alternative method would be to directly detect the difference in currents on the two sides of the power supply. That method is considered next. 3. Review of current comparison methods An approach called the current comparator, that somewhat resembles the dc current transformer, has been used for some time in the calibration of ac current transformers. A large current is passed through two transformers, the one being calibrated and a reference one with nominally the same ratio. An additional winding is provided on the reference core to drive the flux zero. (When there is zero flux, the transformer action is described exactly by the turns ratio.) By arranging the transformers secondaries in series opposition, the flux in the reference core is nearly zero. The additional drive needed to make it exactly zero is a measure of the errors of the device under test. The principles of the fluxgate sensor and the ac current comparator can be combined to produce a dc current comparator. Such a device has been used to calibrate dc measuring equipment, for example at the National Research Council of Canada, the Canadian counterpart to NIST. The minimal configuration of a dc current comparator would consist of a magnetic core with 5 windings. The windings would be two primaries (dc, wound so that the fluxes cancel), an excitation winding (ac), a sensing winding (ac), and a feedback winding (quasi-dc). A difference in current in the two primary windings moves the average operating point of the flux up or down the B-H curve of the material. The nonlinearity that this causes produces second harmonics. These are detected in a phase-sensitive detector, and used, after filtering, to produce a quasi-dc feedback to counter the average offset. The ampere-turns necessary to restore the operating point to symmetry is a measure of the difference current in the two primaries. More elaborate versions of this arrangement would have multiple cores, often referred to as a working core and a sensing core, and may employ magnetic shielding. For the purposes of NEPTUNE, where cost is a factor, the simpler arrangement may be adequate. The implementation of a low-cost current comparator for NEPTUNE application is discussed in Appendix A. 4. Comparison of current measurement methods Table II compares the current measurement methods discussed above. Table II. A comparison of current measurement methods method accuracy frequency long-term cost response stability force good (<1%) poor (< 1 Hz) fair low (few $) magnetoresistance poor to fair (few %) good (kHz) good low (few $) Hall Effect fair to good (few %) good (kHz) fair low (few $) magnetostriction fair (few %) good (MHz) fair high (k$) Faraday Effect very good (< 1%) good (MHz) good high (k$) voltage very good (< 1%) good (MHz) good high (k$) 5. Recommendation The best method in most respects is the series resistor method, but this method is ruled out by the cost, which results from dealing with the very large common mode voltage. For NEPTUNE metering and protection applications, the selection must be made from the low cost entries in Table II. The best of the low-cost approaches is the Hall Effect. Since there are several commercial sources of Hall sensors, this method is recommended for most of the NEPTUNE current measurements. The question of magnetization by fault current must be borne in mind when specifying the equipment. For ground fault detection in the loads, a voltage detection method similar to the one implemented at MBARI is recommended. Whether it should be balanced or should rely on switching must be decided by the power group. Two areas for investigation as student projects at UW are pointed out. It is recommended that a low-cost implementation of the current comparator be investigated, and a version of the force- balance. Some ideas are presented in Appendices A and B. Appendix A Implementing current comparator electronics The current comparator is a complex device. Although capable of outstanding performance (the NRC group in Canada has achieved uncertainties of 1 in 108), cost would seem to be a barrier to widespread use in NEPTUNE. Figure A-1 shows a block diagram of the electronics needed. oscillator ÷2 frequency magnetic circuit output ÷2 excitation synch low-pass frequency drive circuit detector filter feedback currents being compared Figure A-1. Current comparator block diagram In this diagram, each of the 5 “connections” to the magnetic circuit represents a winding on the toroid. The currents being compared may constitute a “single turn” winding, made by passing the conductor once through the center of the toroid. Note that an integrator may be included in the feedback loop. It changes the steady-state error, but not the principle of operation of the system. Some of the circuitry is similar to that in a high-performance electric field meter developed at JPL in the late 1980s. Both divide the oscillator frequency by 2 just to ensure signal symmetry. Both use synchronous switching to achieve the phase-sensitive detector function, both use filtering to reduce the noise bandwidth and extend the dynamic range. The complexity of the field meter was such that a fairly large circuit board was required. The current comparator can be expected to be similar. To reduce the cost of the current comparator, I wondered if a commercial product could be “cannibalized.” The obvious choice was the fluxgate magnetometer. A fluxgate magnetometer, used to measure small magnetic fields and to make electronic compasses for navigation, uses much of the same hardware as the current comparator. The idea was first implemented in quantity in WWII as a way for low flying airplanes to detect submerged submarines, though it may have been around earlier. The biggest difference between a fluxgate magnetometer and a current comparator is that the sensing winding on the magnetometer is outside the toroid. It has to be, or the device would be insensitive to external fields. Figure A-2 shows the arrangement, drawn so as to resemble the current comparator. ÷2 oscillator frequency magnetic circuit output ÷2 excitation synch low-pass frequency drive circuit detector filter external magnetic field Figure A-2. Fluxgate magnetometer block diagram Unfortunately, while fluxgate magnetometers are commercially available, they seem to be expensive, and to use complex electronics. Although low-cost navigation compasses are available (for vehicles, for example), they use GMR (giant magnetoresistance effect) technology, not fluxgate technology. There may still be a way that some of the current comparator functions can be implemented in a commercial IC, however. There seems to be considerable overlap between the signal processing in the current comparator and that in an ordinary FM stereo receiver. In an FM stereo broadcast, the information for the two stereo channels is transmitted in such a way that a mono receiver can still perform adequately. Therefore, the mono signal, the sum of the left and right channels (L+R), is transmitted on a frequency modulated carrier, exactly as it was before the advent of stereo. In order to separate the channels, the difference (L–R) is transmitted separately, at a high enough frequency that the mono receiver will not respond to it. The L–R information is transmitted as an amplitude modulated signal on a 38-kHz subcarrier. For reasons of transmitter efficiency, the actual subcarrier is suppressed, however. Instead, a low-level pilot tone at 19 kHz (out of the passband of a mono receiver) is transmitted. This pilot tone allows stereo receivers to know that the signal is stereo even when there is no modulation on the subcarrier, and gives phase and frequency information to the AM detector. The spectrum of an FM signal therefore looks as shown in Figure A-3. 19 kHz 38 kHz pilot suppressed carrier L+R LR LR lower upper sideband sideband 0 10 20 30 40 50 60 frequency (kHz, linear scale) Figure A-3. Stereo signal spectrum As shown in Figure A-3, after the receiver’s FM demodulator, there is a baseband signal extending to 15 kHz and containing L+R information. At 19 kHz there is a low-amplitude pilot tone, unmodulated. Either side of 38 kHz are the sidebands of an AM suppressed-carrier signal with the L–R information. The amplitude is such that after demodulation, the difference signal has the same amplitude as the sum signal. The AM signal is demodulated by a synchronous detector operating at twice the frequency of the pilot tone, and locked to it in phase. By adding the resulting L–R signal to the L+R signal, a pure L signal is obtained. By subtracting them, a pure R signal is obtained. These operations are shown in Figure A-4, drawn so as to resemble Figure A-1 as far as possible. stereo decoder IC oscillator ÷2 76 kHz frequency 38 kHz 19 kHz ÷2 phase synch low-pass frequency detector detector filter R L signal feedback L low-pass sum & outputs filter dfference R input signal Figure A-4. Stereo decoder IC block diagram Again, the oscillator frequency is divided by two just to ensure symmetry. The signal is detected by a synchronous detector, operating at twice the pilot tone frequency, and then passed through a low-pass filter. In fact, the similarity between the functions in Figures A-1 and A-4 leads one to wonder whether an FM stereo circuit would function as a current comparator. While I have not done the experiment, I am inclined to think it would, with the addition of a few components. There are a few differences between this arrangement and the current comparator. In the stereo IC, the oscillator phase is locked to the pilot tone, whereas in the current comparator it is free running. (It may turn out that this is a feature we can use.) The stereo IC contains sum and difference circuits the current comparator does not need. Perhaps this will not matter. The spectrum of the current comparator signal should contain nothing in the area of the baseband L+R signal, so that only the “L–R” signal should appear in the output. In fact, it should appear identically at both outputs. Figure A-5 shows a possible use of an FM stereo IC to do the hard work in a current comparator signal chain. stereo decoder IC oscillator ÷2 76 kHz frequency 38 kHz 19 kHz ÷2 phase synch low-pass frequency detector detector filter R L signal feedback low-pass sum & output filter dfference R input signal magnetic input signal circuit 19 kHz feedback oscillator ÷2 excitation 38 kHz frequency drive circuit currents being compared Figure A-5. Current comparator implemented with stereo decoder IC The circuit has been largely subsumed into the IC. An external oscillator is still needed, and must be operated to produce excitation at the pilot tone frequency. The stereo IC locks to that signal (which will be large relative to the desired signal), and performs the current comparator’s detection functions. Difficulties may be experienced due to overload caused by the large size of the “pilot tone.” Certainly, there should be no problem obtaining lock! Another problem may be the presence in the signal of significant energy at the third harmonic. Theoretically, the synchronous detection at 2f should eliminate all odd harmonics. An estimate of the comparator spectrum is shown in Figure A-6. 19 kHz drive 57 kHz frequency third harmonic 38 kHz second harmonic 0 10 20 30 40 50 60 frequency (kHz, linear scale) Figure A-6. Probable comparator signal spectrum In addition, it may be necessary to add a driver for the feedback loop to the magnetic circuit. Nevertheless, the component count is much lower than an conventional implementation would be, so it seems to be worth the effort of trying it. Appendix B Force-balance current measurement The measurement of current by balancing forces is an ancient technique. Lord Kelvin made one in around 1882, using a modified weighing-scale and six solenoids, and Lord Raleigh had a version using coils in the Helmholz arrangement, at around the same time. The method was in use at the Standards laboratories of both England and the US until the late 1940s as the method for standardizing the ampere. The principle is illustrated in Figure B1. balance beam series-connected coils balance weight Figure B1. Current Balance (after Rayleigh) In both the Rayleigh and Kelvin versions of the current balance, the force that is balanced is that due to gravity. In the 1880s this was thought to be pretty constant from place to place, and indeed the errors that assumption caused were not deemed important for many decades. A modern variation on the theme could differ in two important ways. First, the current in some of the coils need not be the measurand, it could be a control current, scaled in some suitable way. Second the balance – the restoration to zero deflection of some mechanical system – could be effected by feedback that is measured by some conventional means and becomes the system output. There would be several advantages: the overall system could be quite robust and inexpensive, and long-term variations in the stiffness of the deflection mechanism would be relatively unimportant, affecting the system sensitivity but not the accuracy, and immunity to large current overload could be by mechanical restraint. Figure B2 shows one possible implementation. The proposed system is scarcely more complex than a Hall effect system, and while would have a very poor frequency response, it would likely confer immunity to surge-current magnetization problems. anchor block deflection current drive measurement circuit feedback current restraint block current being measured Figure B2. Possible implementation of a Current Bridge The force-balance arrangement shown in Figure B2 is described as a current bridge because it works by achieving a zero deflection in the position of the middle coil. The position of the coil can easily be measured by a strain gauge, and doubtless other approaches would also work. The drive current required to achieve balance is proportional to the measurand, the coefficient of proportionality including also the number of turns in the various coils and the mutual inductance. These factors are not likely to change. The scheme has two major defects. Probably the biggest defect in the measurement is its expected sensitivity external magnetic fields. Since the magnetics are all air-cored, shielding must be provided to reduce interference with other such sensors in the area. The other problem is sensitivity to vibration. It may be possible to make a design that has some vibration immunity. How serious a problem this will be in NEPTUNE is hard to gauge, as there will likely be no vibration except during earthquakes and volcanic eruptions .
Pages to are hidden for
"Communication options for Command and Control"Please download to view full document