NMI TR 17 Two-stage Transformers with Partitioned Excitation Cores Greig W Small 11 © Commonwealth of Australia 2013 First edition — August 2013 National Measurement Institute Bradfield Road, Lindfield, NSW 2070 PO Box 264, Lindfield, NSW 2070 T F W (61 2) 8467 3600 (61 2) 8467 3610 www.measurement.gov.au ii CONTENTS 1 Introduction ......................................................................................................................... 1 2 Partitioned Excitation Core ................................................................................................. 2 3 Excitation Windings with Unequal Turns .......................................................................... 4 3.1 Ratio of Unity ............................................................................................................ 4 3.2 Ratios other than Unity.............................................................................................. 5 4 Conclusion .......................................................................................................................... 7 5 References ........................................................................................................................... 7 iii ABSTRACT Partitioning the excitation core of a two-stage transformer into two or more sections with their separate excitation windings may effect a significant economy of core material. It may also allow the close approximation of non-integer transformer ratios with fewer turns than would be required in a conventional design. Application of the technique to the design of transformers with amplifier-assisted excitation and to an active coaxial choke is described. Keywords: two-stage transformers, pre-excited transformers, partitioned excitation cores, amplifier-assisted excitation, active coaxial chokes iv 1 INTRODUCTION Two-stage voltage transformers comprise a magnetic circuit, usually a toroidal core, with a winding that is connected to a voltage source to provide the excitation of the transformer, and a second core with a winding that couples both cores and is connected to a voltage source to define the input, or primary, voltage. The second core is excited by the deficit in excitation of the first core. The first core and its excitation winding may be shielded to contain any leakage flux. Conducting shields may enclose, and be connected to the respective grounds of, the primary and secondary windings to contain capacitance currents. A representation of the half cross-section of such a transformer is shown in Figure 1 [1, 2]. Electrostatic screens Excitation winding and core Magnetic shield Primary winding and core Secondary winding or windings Figure 1. Two-stage transformer with magnetic shield and inter-winding screens In some low-frequency designs the second core winding may also be connected to the excitation source, and a third core and winding added to reduce further the excitation error [3]. For the purposes of this document, these and similar arrangements are to be regarded as elaborations of two-stage transformers. The excitation core is designed to carry the total excitation flux, which is proportional to the applied voltage and inversely proportional to the turns of the excitation winding and to the signal frequency. Efficient use of the core will see that the peak flux density approaches the saturation flux density for the core material. In a toroidal core, the excitation field is inversely proportional to the distance from the core axis so that the flux density is also approximately, to the degree that the excitation is linear, inversely proportional to radial distance. To this degree the peak flux density at the outer radius, r2, of the core is less than that at the inner radius, r1, in the ratio r1/r2. For small transformers this may be inconsequential, but for larger transformers operating at lower frequencies and with requirements for higher voltages per turn the excitation core may be both expensive and massive and it becomes important to design for efficient use of core material. For a given total flux to be carried by the core, this points to a high aspect ratio of core height to radial width. Factors that limit this approach are the mechanical implications of a high aspect ratio and the greater length, and hence higher resistance and uncoupled reactance, of the excitation winding with consequent increase in excitation error. Alternatively, more efficient use of core material may be had by partitioning the core radially into two or more sections, each with its own excitation winding (Figure 2). NMI TR 17 1 Excitation winding and core Excitation winding and core Figure 2. Two-stage transformer with a partitioned excitation core 2 PARTITIONED EXCITATION CORE Consider a core of height h1, inner radius r1 and outer radius r2 with flux density Bmax at radius r1. Again, to the degree that the excitation is linear, the flux density B(r) at radius r is: B(r) = Bmax r1/r and the total flux ϕ is: ϕ = Bmax h1 r2 r1 / r dr r1 = Bmax h1 r1 loge(r2/r1) (1) Suppose the excitation were to be increased in proportion to the radius, so that the flux density is maintained at Bmax. For the same total flux ϕ, the outer radius r2' is implied by: Bmax h1 (r2'− r1) = Bmax h1 r1 loge(r2/r1) giving (r2'/r1) = 1 + loge(r2/r1) (2) The mass m' of this core relative to that, m, of the first is: (m'/m) = ((r2'/r1)2 – 1)/((r2/r1)2 – 1) (3) Values of r2'/r1 and m'/m for a range of values of r2/r1 are given in Table 1. Table 1. Hypothetical minimum masses for several values of relative width r2/r1 1.1 1.2 1.3 1.4 1.5 NMI TR 17 r2'/r1 1.095 1.182 1.262 1.336 1.405 m'/m 0.951 0.904 0.860 0.819 0.780 r2/r1 1.6 1.7 1.8 1.9 2.0 r2'/r1 1.470 1.531 1.588 1.642 1.693 m'/m 0.744 0.710 0.679 0.650 0.622 2 This result is the hypothetical limit to the reduction in core size and mass for the given height, inner radius and maximum flux density. To approach this limit, consider partitioning the core radially into n concentric cylinders of equal width, each cylinder being wound with nT turns and all windings connected in parallel to the voltage source V of excitation. The spacing between cylinders is, for the purposes of the argument, assumed to be zero, ignoring the radial space that must be occupied by the excitation windings and any structure enclosing the component cores. Clearly, with increasing values of n, the increasing numbers of turns and the increasing space needed to accommodate the windings rapidly become self-defeating. Now consider the core to be partitioned into two cores, each of which carries half the total flux, with the same values of r1, h1 and Bmax. Let the outer radius of the inner core and the inner radius of the outer core be r12, neglecting the space required by the excitation windings, and the outer radius of the outer core be r2'. The total flux ϕ1 in the inner core is: ϕ1 = Bmax h1 r1 loge(r12/r1) (4) For ϕ1 to be equal to ϕ/2: loge(r12/r1) = loge(r2/r1)/2 (r12/r1) = √(r2/r1) so that (5) The total flux ϕ2 in the outer core is: ϕ2 = Bmax h1 r12 loge(r2'/r12) (6) and for ϕ2 to be equal to ϕ1: r12 loge(r2'/r12) = r1 loge(r12/r1) (7) That is (r12/r1) loge((r2'/r1)/(r12/r1)) = loge(r12/r1) from which (r2'/r1) = exp[(1 + 1/(r12/r1)) loge(r12/r1)] (8) Some values of the relative widths r2/r1 of the single core and r2'/r1 of the partitioned core, together with their relative masses m'/m, are given in Table 2. A typical value is around 1.5, corresponding to a ratio of outer to inner diameters of 3:2; in this case the saving in material with the partitioned core is nearly 13%. Table 2. Relative widths and masses of single core and equivalent partitioned core r2/r1 1.1 1.2 1.3 1.4 1.5 r2'/r1 1.098 1.191 1.279 1.364 1.445 m'/m 0.974 0.948 0.922 0.896 0.871 r2/r1 1.6 1.7 1.8 1.9 2.0 r2'/r1 1.523 1.598 1.670 1.740 1.807 m'/m 0.846 0.822 0.799 0.777 0.755 With the pair of cores each carrying half the flux of the single core, the excitation winding on each core will require exactly twice the number of turns as that on the single core. The partitioned core with its windings connected in parallel is then functionally equivalent to the single core and its winding. NMI TR 17 3 3 EXCITATION WINDINGS WITH UNEQUAL TURNS Allowing the excitation windings to have different numbers of turns one from the other leads to interesting possibilities, including the partitioning of the cores for other than the same flux in both cores, and for effecting non-integer transformer ratios that might otherwise be approximated only with high numbers of turns. Let the numbers of turns of the excitation windings be N1 and N2, and of the common secondary winding N12. As before, the excitation windings are connected in parallel to the source of excitation voltage. The effective turns ratio R is: R = N12/N1 + N12/N2 = N12(1/N1 + 1/N2) Let N1 = 2N12 – a and N2 = 2N12 + b (9) (10) where a and b are integers. Then R = N12(N1 + N2)/(N1N2) = (4N122 + (b – a)N12)/(4N122 + (b – a)N12 + [(b – a)N12 – ab]) (11) For values of b less than or equal to a, the denominator of equation (11) is less than its numerator and R must be greater than 1. Other and more useful possibilities result from setting b greater than a, are examined. 3.1 Ratio of Unity Putting (b – a)N12 = ab in equation (11) results in R = 1 for all values of a. A ratio of 1 is required for a conventional two-stage design where the N12 winding couples the additional core and serves as the defining primary winding of the two-stage transformer. For (b – a) = 1, N12 = a(a + 1) and from equation (10), N1 = a(2a + 1) and N2 = (a + 1)(2a + 1). The first several combinations of turns are given in Table 3. Table 3. Combinations of turns with unity ratio a 1 2 3 4 5 N12 2 6 12 20 30 N1 3 10 21 36 55 N2 6 15 28 45 66 a 6 7 8 9 10 N12 42 56 72 90 110 N1 78 105 136 171 210 N2 91 120 153 190 231 Since a, (a + 1) and (2a + 1) are mutually prime in all combinations, N12, N1 and N2 are also mutually prime. Hence each of these combinations of turns may be scaled up but not down to numbers of turns appropriate to the application. For values of (b – a) other than 1: (b – a)N12 = a2 + (b – a)a (12) To avoid combinations that are multiples of some other, N12, a and b must have no factors in common. If (b – a) has any unpaired prime factor pj then, from equation (12), pj must be a NMI TR 17 4 factor of a2 and therefore of a, of b, of a2/(b – a), and hence of N12. (b – a) must therefore be a square, and a2 but not a must be divisible by (b – a). The smallest value a1 of a is therefore: a1 = √(b – a) If a is any multiple of a1 such that the multiplier m includes a prime factor pi of (b – a), then pi will also be a factor of a2/(b – a) and hence of N12 as well as of a and b. Multipliers m must therefore be mutually prime with (b – a). Combinations of turns for several values of (b – a) and a meeting these constraints are given in Table 4. Table 4. Combinations of turns with unity ratio and values of (b-a) other than 1 (b-a)→ a↓ 2 6 10 14 18 N12 3 15 35 63 99 4 N1 4 24 60 112 180 N2 12 40 84 144 220 (b-a)→ a↓ 3 6 12 15 21 N12 4 10 28 40 70 9 N1 5 14 44 65 119 N2 20 35 77 104 170 (b-a)→ a↓ 4 12 20 28 36 N12 5 21 45 77 117 16 N1 6 30 70 126 198 N2 30 70 126 198 286 It is seen that larger values of (b – a) rapidly lead to large values for the numbers of turns, while for smaller numbers of turns the proportional differences between N1 and N2 are large and probably less useful than is the case for (b – a) = 1. 3.2 Ratios other than Unity As noted above, combinations of different numbers of winding turns on two primary cores permit the approximations to ratios that otherwise might require much larger numbers of turns. For example, consider a transformer ratio bridge for the comparison of a resistor of value 100 kΩ with a resistor of value RK/2, where RK is the von Klitzing, or quantised Hall, resistance of which the presently accepted value is 25 812.807 Ω. The nominal transformer ratio is 7.748 091 87. From the continued fraction expansion of the ratio, the successive convergents (and their errors in parts per million) are: 7:1 8:1 23:3 31:4 984:127 1 015:131 (−96 552) (32 512) (−10 509) (246.27) (−7.792) (−0.034) To approximate this ratio to a few parts per million with a conventional transformer requires a primary winding of about 1000 turns. For a partitioned-core transformer with paralleled primary windings of 51 and 79 turns and a secondary winding of 4 turns, the ratio error is −1.541 parts per million. Another application is in the amplifier-assisted excitation of a two-stage transformer. In this example, a unity-gain amplifier drives the two excitation windings. A sensing winding, coupling both the excitation cores and the additional core, provides the input signal to the amplifier (Figure 3). The loop gain must be less than unity to ensure that this arrangement is stable, but close to unity so that the pre-excitation error is small. As the primary winding need NMI TR 17 5 provide only the deficit in excitation, this configuration may be advantageous where the primary voltage source is unsuitable or inconvenient as the source of pre-excitation. Let: (b – a)N12 = ab + c (13) where c is an integer. Exhaustive examination of combinations of (b – a), a and c would be unwieldy, and only the case of (b – a) = 1 will be considered here. Then: N12 = a2 + a + c, N1 = 2a2 + a + 2c N2 = 2a2 +3a + 2c + 1 and (14) Values of N12, N1, N2 and the ratio for the first several values of a and c are given in Table 5. Feedback winding and screen +1 Figure 3. Two-stage transformer with amplifier-assisted excitation Table 5. Values of turns N12, N1, N2 and ratio for (b – a) = 1 and several values of a and c c→ a↓ 1 2 3 4 5 6 1 N12 3 7 13 21 31 43 N1 5 12 23 38 57 80 2 N2 8 17 30 47 68 93 Ratio 0.975 0 0.995 1 0.998 6 0.999 4 0.999 7 0.999 9 N12 4 8 14 22 32 44 N1 7 14 25 40 59 82 3 N2 10 19 32 49 70 95 Ratio 0.971 4 0.992 5 0.997 5 0.999 0 0.999 5 0.999 7 N12 5 9 15 23 33 45 N1 9 16 27 42 61 84 N2 12 21 34 51 72 97 Ratio 0.972 2 0.991 1 0.996 7 0.998 6 0.999 3 0.999 6 It may be seen that, for example, with a = 4 and c = 2, the numbers of turns are 22, 40 and 49 (or any common multiple thereof) and the feedback gain, apart from the excitation error, is in deficit of unity by 0.001. Where the excitation core would have been partitioned for considerations of economy, this implementation avoids the need for an attenuator. In other cases (for example, amplifierNMI TR 17 6 assisted coaxial chokes in bridge circuits), the use of a resistive attenuator as an alternative may prejudicially introduce additional noise and additional impedances in series and admittances to ground. Where economy of core material is not a consideration, it may be more convenient to partition the excitation core axially rather than radially. The design of a coaxial choke of this configuration with excitation of 102 and 115 turns and feedback of 54 turns (a = 6, c = 12, in deficit by 0.001) is shown in Figure 4 [4]. +1 Coaxial cable Figure 4. Amplifier-assisted coaxial choke employing a partitioned excitation core The impedance presented to the coaxial cable by the second core is enhanced by the factor 1/δ, where δ is the deficit in excitation. For δ = 0.001 the enhancement factor is 1000, equivalent to winding 31 turns of the cable on the same core without amplifier assistance. In this configuration of active coaxial choke the feedback loop is completed and defined locally to the device and independently of the coaxial cable, as distinct from other implementations in which the feedback loop is completed by the outer circuit of the cable. 4 CONCLUSION The partitioning radially of the excitation core of a two-stage transformer into two or more separately excited sections may lead to a saving of more than 10% in core material. This may be a significant economy, particularly in transformers operating at lower frequencies and with a requirement for fewer turns per volt. Partitioning the core also introduces the possibility of unequal turns of excitation windings, allowing close approximations to awkward ratios with appreciably fewer turns than in a conventional design. A particular example is the amplifierassisted excitation of the transformer with a unity gain amplifier and a ratio of feedback turns to excitation turns fractionally less than unity. 5 REFERENCES [1] Thompson, AM and Small, GW (1971) AC bridge for platinum resistance thermometry. Proc. IEE 118, 1662–66 [2] Small, GW (1996) The Efficacy of High Permeability and Conducting Shells as Magnetic Shields on Toroidal Cores. TIP Technical Memorandum No. 144, TIP3004 (4 November) [3] Small, GW; Budovsky, IF; Gibbes, AM and Fiander, JR (2005) Precision three-stage 1000 V/50 Hz inductive voltage divider. IEEE Trans. Instrum. Meas. 54(2) 600-3 [4] Small, GW and Marais, EL (2012) An active coaxial choke for precision impedance measurements. CPEM 2012 Digest, Washington DC (1 – 6 July 2012) (submitted) NMI TR 17 7