GB923373A - Improvements in and relating to voltage or current dividing networks - Google Patents

Abstract

923,373. Resistance networks. E. R. WIGAN. March 2, 1960 [Dec. 2, 1958; May 27, 1959], Nos. 38762/58 and 18022/59. Class 40 (8). [Also in Group XXXVI] A voltage or current-dividing network comprises resistances, or reactances all of one kind, arranged in series and shunt branches, and switching means for transferring a conductance (or susceptance) from a series to a shunt branch for varying the ratio between current or voltage at the input and output terminals whilst keeping the output impedance constant. Leaving aside reactances, the networks consist effectively of two serially-connected resistances, of values, say, a and b, the output voltage being taken across R b . the dividing ratio for no load being given by N=b/(a+b) and the output impedance R<SP>1</SP> (input terminals short-circuited) being given by 1/R<SP>1</SP>=1/a+1/b. Thus to vary N in steps equal to 0.1, ten equal resistors r 1 , r 2 , &c., Fig. 12, each of value 10R<SP>1</SP> are used and a tenposition switch is arranged to include a number of these in parallel in the series branch and the remainder in parallel in the shunt branch. To vary N in ten steps of 0.01, resistances of 100R<SP>1</SP> are required, and so on. The switch shown comprises insulated contact-rings Z 1 , Z 2 which move together and effect transfer of the resistors r 1 &c. between series and shunt branches. The switch is set for a maximum value of N=0.9, further decade switches, connected in parallel, providing further places of decimals adding up to the remaining 0.1 fraction. In the generalized form of the network shown in Fig. 1, P and Q represent the decade resistances. The load resistance may be made equal to R<SP>1</SP> in which case E 2 /E 1 =N/2. If the source has a finite internal resistance, variation of N will affect the absolute values of E 1 and E 2 . To remove this effect, a compensating resistance R s , shunting the input, is designed to provide a constant input impedance which is designated R 1 when R s is infinite and R 2 when R s is finite. R 1 is given by 1/R 1 =(N-N<SP>2</SP>/2)/R<SP>1</SP>. For an ideal constant value of R 2 equal to R<SP>1</SP>, R s must have an ideal value R s 1 given by R<SP>1</SP>/R<SP>1</SP>s= 1 - N+N<SP>2</SP>/2. Other equations for an ideal compensating resistance may be set up for values of load impedance and input impedance other than R<SP>1</SP>. In one form, R s comprises serially connected resistances to a total value of R<SP>1</SP>(1+N). In this case 1/R s = 1/(1+N)= (1-N +N<SP>2</SP>-N<SP>3</SP>. . .) which crudely approximates the ideal equation (the last equation is normalized for R<SP>1</SP>=1). The compensating networks of Figs. 2, 3 are intended for use in a three-decade network, the switch in the first decade (first decimal point) having two extra contacts which are arranged to select values of A and B whilst the other switches carry one contact each which together vary the resistance C. The latter is made to vary linearly with the value of the associated switches up to value of e.g. 130 #. A is chosen according to 1/A= 1 - N a +N a <SP>2</SP>/2 where N a is the setting of the first switch. B is then chosen such that 1/(A+#C)= 1 - (N a +0.1)+(Na+0.1)<SP>2</SP>/2 where #C is the value of B and C in parallel when C has its maximum value. It is stated that these values will hold R 2 to within about 0.4% of the total value. Similar rules are given for the choice of values in the compensating network of Fig. 3 and a table of values is set out. Errors occurring at specific periodic values of N (e.g. multiples of N=0.05) may be compensated by the addition of further resistances in series and shunt with C operated by contacts associated with the last two decade switches.