Michel Ghilardi and Stéphane Rollier of LEM explore Class R Transducer for Trains
Transducers that are capable of measuring thousands of Amps with Class accuracy R help keep the trains running across country borders. By Michel Ghilardi, R&D Project Manager for LEM, and Stéphane Rollier, Product & MarCom manager at LEM.Meas
Along with that capability comes the need for accurate billing. The traction units consuming energy in several European countries want to check they really pay what they consume as energy in all the countries traversed, and eventually need a recognised way to contest any abnormal consumption. The rail networks that supply the power demand an accurate log of energy consumed while the train is within their system and a European standard is now being finalised (EN 50463) that formalises those requirements.
The detail contained within EN 50463, which relates specifically to energy meters of Class R accuracy for on-board measurement, is a subject in its own right; however, it imposes not only high levels of accuracy at nominal current values but also at different percentages of the nominal current. For example, measuring a current at just 1% of the full rated primary current, the standard still requires an accuracy of 5% of the measured value to achieve a defined Class 0.5R accuracy; the standard also imposes stringent limits of acceptable variation with temperature, through an overall temperature span of -40 to +75 °C.
To address this market, and the problem of current measurement in both traction and industry applications, LEM has conceived and designed a new series of current transducers. Three models measure a current (DC or AC) up to 4000 Amps RMS (6000 Amps peak) in vehicles that are supplied with energy from networks up to 3000 V. The transmission of power and signals from a high voltage environment to a low voltage environment requires specific insulation features depending on the application, and a fundamental aspect of the design of the ITC series was to meet all the necessary standards to enable these features. The transducer provides extra margin on the Class 1R (‘R’ indicating rail-traction) specification recognised by EN 50463 achieving Class 0.5R accuracy.
Figure 1: Closed loop current transducer principle
Initial design studies confirmed that to reach this level of performance a transducer based on closed-loop fluxgate technology would be needed. The basic principle of the closed loop measurement technique is shown in Figure 1. The primary (load) current IP flows in a conductor that passes through the loop or core (i.e. into or out of the paper in Figure 1), giving rise to a current linkage ΘP. For accurate measurement of DC currents, the method consists of compensating (balancing or nulling) IP by an opposing current linkage ΘS created by a current Is flowing through a known number of turns NS, to obtain:
ΘP − ΘS = 0 or NP•IP − NS•IS = 0
Where NP and NS are the numbers of primary and secondary turns, respectively. To obtain an accurate measurement, it is necessary to have a highly accurate device to measure the condition Θ = 0 precisely. The aim is to obtain a current transducer with the following characteristics: excellent linearity; outstanding long-term stability; low residual noise; high frequency response; and high reliability.
To achieve accurate compensation of the two opposing current linkages (ΘP, ΘS), a detector capable of accurately measuring Θ = 0 must be available, which means that the detector must be very sensitive to small values of a residual magnetic flux Ψ (created by the current linkage Θ) in order to achieve the greatest possible detector output signal.
Figure 2: Magnetic cores hysteresis cycles
Fluxgate detectors rely on the property of many magnetic materials to exhibit a non-linear relationship between the magnetic field strength H and the flux density B. The hysteresis cycles of the magnetic cores have a form comparable to the one represented in Figure 2, that is, approximately square (depending on the type of material used).
If we look at the hysteresis diagram plotting B = f(H) on the magnetisation curve, we can see that for a given field strength H1 a small change ΔH1 gives rise to a flux density variation ΔB1. However, further along the cycle at field strength H2, where the gradient of the line is less, for a variation of the same magnitude ΔB2 = ΔB1, the variation in the field strength, ΔH2, must be much greater. The detection of the zero flux condition (Ψ = 0) is based on this phenomenon.
Figure 3: Full bridge driving the Fluxgate (saturable inductor)
A full bridge (Figure 3) is used to drive the Fluxgate with a square wave signal creating a current flowing alternatively through the switches S1/S4 and S2/S3, to finish through the measuring resistor. When applying a square wave voltage (yellow [top] signal in Figure 4) to a saturable inductor until its magnetic core starts to saturate, a current is created. Without primary current, IP =0, this current is symmetric.
Figure 4: [from top] Square wave voltage; Current created; Asymmetry of the created current
When a DC current flows through the aperture of the core, the curve of the hysteresis cycle is then shifted, causing asymmetry of the current produced by the square wave voltage (red trace in Figure 4). This current is then measured using an accurate resistor and the asymmetry is used to adjust the secondary current in the compensation winding so that it exactly compensates the primary current.
The accuracy of the measurement will not only depend on the accuracy of the resistor measuring the Fluxgate current, but also strongly on the sensitivity of the flux detector. However, in spite of the DC measurement function accuracy, there are some drawbacks to this DC measurement system, leading to the arrangement shown in Figure 5.
Figure 5: Adding a second flux detector provides a solution to voltage peak re-injection
As the winding ‘D’ of the flux detector is coupled with the compensation winding ’S’, the applied square wave voltage is re-injected into the compensation winding and creates a parasitic current in the measurement resistor.
However, the square wave voltage induced in the S winding by this flux may be practically cancelled out when a second D’ winding is mounted on a second detector core (identical to D). The compensation winding S encloses both detector cores. The residual flux (the sum of the opposed fluxes in D and D’) will create very small voltage peaks correlated with the fluxgate excitation.
Figure 6: Block schematic of the main functions in the ITC transducer
Figure 6 shows the major functional blocks in the ITC transducer design. Fluxgates are self-oscillating, their currents are measured by the microcontroller (MCU) through an A/D converter, by reading the voltage across the bridge resistor. The MCU provides: synchronous rectification of the fluxgate signal; low pass filtering; compensation of offset and reduction of offset drift (carried out digitally rather than by analogue circuits); regulation of secondary current; management of protection functions such as monitoring supplies, and detecting over-voltage and overload; driving the output stages; and Fluxgate management. The output of secondary current regulator is converted into an analogue value using a D/A converter that gives the reference to the PWM generator for the output stage. To reduce losses and avoid any requirement for an external heatsink, the output amplifier operates in Class D; an LC filter removes harmonics to keep output noise low.
A second switch mode output stage is added only for balancing supply currents between positive and negative power supply (subject of a LEM patent, Figure 7). ITC models provide a current output generated by the Class D amplifier; with the double Class D output current generator, the supply currents are balanced and reduced versus a transducer using a Class AB amplifier.
Figure 7: Class D output current generator
The supply current is always positive and almost equal on both supplies: 0.3A at nominal output current on each power supply polarity; electronic losses are less than 6W when supplied with +/- 24V. By contrast, a design that used a Class AB amplifier would have required a supply current of 1.6A and would have created 30W of losses at nominal current.
However, Class D operation is known to create a ripple of about (in this case) 25mA.t (mA.t: Number of mA seen on the primary side for one turn done with the primary conductor into the transducer aperture) in an 80kHz bandwidth, and 3A.t in a 200kHz bandwidth, which can readily be filtered.
Traction applications can give rise to very high primary current overload conditions, and the ITC series has been designed to tolerate overloads of up to 100kA for 100mSec. This translates to a secondary peak current of 40A with a high potential to destroy some internal components and customer burden resistance, as well as printed circuit tracks. Thus a special dedicated electronic circuit has been designed in the ITC models, as indicated by the ‘Overload Protection’ block in Figure 6.
If the primary current exceeds the measuring range, the electronics cannot maintain the flux compensation: if this state persists, the fluxgate detector becomes completely saturated and unable to measure the flux error. When this happens, the transducer stops for 300 to 500mSec and then sweeps the output current to find the point at which compensation is correct again and the normal function can resume.
An additional logic output called ‘/VALID’ is provided indicating if the transducer is operating in its normal range or not. There is also a test winding to inject a current that simulates a primary current of 10% of nominal maximum, for a re-verification of the transducer in the application.
Bandwidth for the ITC 4000-S is 80kHz (3dB point) and the transducers’ apertures allow the insertion of large conductors: 102mm diameter (for the ITC 4000-S) or 63mm diameter (for the ITC 2000-S). The units are not sensitive to off-axis positioning of smaller-diameter conductors.
In three models, to measure up to 1, 2 or 4kARMS (nominal), the ITC series meets the rail-traction industry sector's needs by reaching the Class accuracy 0.5R defined in the prEN 50463 standard for on-board energy monitoring operating over the temperature range –40 to +85°C. They are equally applicable to any situation in which kA-level current measurement accuracy of 0.5 % from 5 % to 120 % of the nominal current is required.