LTC1041 UUWUAPPLICATIO S I FOR ATIO C Input Voltage Range IN (≈ 33pF) R S1 S + The input switches of the LTC1041 are capable of switching either to the V+ supply or ground. Consequently, VIN CS S2 the input voltage range includes both supply rails. This is – V– a further benefit of the sampling input structure. LTC1041 DIFFERENTIAL INPUT LTC1041 • AI01 Error SpecificationsFigure 2. Equivalent Input Circuit The only measurable errors on the LTC1041 are the R deviations from “ideal” of the upper and lower switching S • CIN. The ability to fully charge CIN from the signal source during the controller’s active time is critical in levels (Figure 1b). From a control standpoint, the error in determining errors caused by the input charging current. the SET POINT and deadband is critical. These errors may For source resistances less than 10kΩ, C be defined in terms of V IN fully charges U and VL. and no error is caused by the charging current. V VL SET POINT error U ≡ + – SET POINT 2 For RS > 10k Ω deadband error ≡ (V – VL) – 2 •DELTA For source resistances greater than 10kΩ, C U IN cannot fully charge, causing voltage errors. To minimize these errors, The specified error limits (see electrical characteristics) an input bypass capacitor, C include error due to offset, power supply variation, gain, S, should be used. Charge is shared between C time and temperature. IN and CS, causing a small voltage error. The magnitude of this error is AV = VIN • CIN (CIN + CS). This error can be made arbitrarily small by increasing C Pulsed Power (V S. P-P) Output The averaging effect of the bypass capacitor, C It is often desirable to use the LTC1041 with resistive S, causes another error term. Each time the input switches cycle networks such as bridges and voltage dividers. The power between the plus and minus inputs, C consumed by these resistive networks can far exceed that IN is charged and discharged. The average input current due to this is of the LTC1041 itself. IAVG = VIN • CIN • fS, where fS is the sampling frequency. At low sample rates the LTC1041 spends most of its time Because the input current is directly proportional to the off. A switched power output, VP-P, is provided to drive the differential input voltage, the LTC1041 can be said to have input network, reducing its average power as well. VP-P is an average input resistance of RIN = VIN/IAVG = I/(fS • CIN). switched to V+ during the controller’s active time (≈ 80µs) Since two comparator inputs are connected in parallel, R and to a high impedance (open circuit) when internal IN is one half of this value (see typical curve of R power is switched off. IN versus Sampling Frequency). This finite input resistance causes Figure 3 shows the VP-P output circuit. The VP-P output an error due to the voltage divider between RS and RIN. voltage is not precisely controlled when driving a load The input voltage error caused by both of these effects is (see typical curve of VP-P Output Voltage vs Load Current). V In spite of this, high precision can be achieved in two ways: ERROR = VIN [2CIN/(2CIN + CS) + RS/(RS + RIN)]. (1) driving ratiometric networks and (2) driving fast set- Example: assume fS = 10Hz, RS = 1M, CS = 1µF, VIN = 1V, tling references. VERROR = 1V(66µV + 660µV) = 726µV. Notice that most of the error is caused by R In ratiometric networks all the inputs are proportional to IN. If the sampling frequency is reduced to 1Hz, the voltage error from the input VP-P (Figure 4). Consequently, the absolute value of VP-P impedance effects is reduced to 136µV. does not affect accuracy. 1041fa 5