E3 238 Analog VLSI Circuits Lecture 19: Feedback Gaurab Banerjee Department of Electrical Communication Engineering, Indian Institute of Science, Bangalore banerjee@ece.iisc.ernet.in Generalized Feedback Network Virtual Ground ( = A) Key Components: • Feed-forward Amplifier • Sense Mechanism (Output) • Feedback Network • Return Mechanism (Input) -> Input of H(s) = virtual ground -> signal amplitude is very small -> G(s) = β (usually frequency independent) -> β A = loop gain Closed loop transfer function Properties of Feedback -> Gain Desensitization: -> Closed loop gain determined by the feedback factor, for large A -> Large variations in A translate into small variations in Y/X -> gain desensitization -> Tradeoff between precision and closed loop gain Properties of Feedback -> Bandwidth Modification: Consider a single pole transfer function: Numerator = closed loop gain at low frequencies Denominator = Pole at (1+ β A0) ω0 -> 3 dB bandwidth increases by (1+ β A0) Properties of Feedback -> Terminal Impedance Modification: Output impedance when output voltage is measured in shunt and proportional voltage is applied in series: Voltage measured in “shunt” -> This type of feedback lowers impedance Properties of Feedback -> Input Impedance for the same configuration: Voltage applied at the input in “series” -> This type of feedback increases impedance Sense and Return Mechanisms SENSING: Voltage sensed in shunt -> “shunt feedback” Current sensed in series -> “series feedback” RETURN MECHANISM: Voltage addition -> “series” Current addition -> “shunt” Nomenclature of Feedback Type First Term (return or input applied) - Second Term (sense or output) Shunt-Series: Current Sensed, Current Applied Shunt-Shunt: Voltage Sensed, Current Applied Series-Shunt: Voltage Sensed, Voltage Applied Series-Series: Current Sensed, Voltage Applied Terminal Impedance Modification Consider the circuit for which we looked at terminal impedance modification -> voltage is measured: shunt, voltage is returned: series Series-Shunt feedback Type of feedback Input Impedance Multiplier Output Impedance Multiplier Qualitative Understanding At outputs: Series feedback measures current -> Regulates current -> “like a constant current source -> High Z Shunt feedback measures voltage -> Regulates voltage -> “like a constant voltage source” -> low Z At inputs: Series feedback applies voltage to minimize current error* -> makes current constant -> High Z Shunt feedback applies current to minimize voltage error* -> makes voltage constant -> Low Z *This is more of a “memory aid”. Actual voltage/current return at input is topology specific. An Example: Series-Series Feedback Current measured in series -> Applied at the input in series, as a voltage • This is sometimes known as “emitter feedback” or “local series feedback” • Note that an emitter follower, which looks similar, has its output voltage sampled and returned as a voltage: so, emitter follower has series-shunt feedback Calculation of Loop Gain -> Set the main input to zero -> Break the loop at some point -> Inject a test signal in the direction of feedback -> Follow the signal, obtain the signal that returns to the break point negative of the transfer function thus derived = loop gain In the figure above: First order analysis neglects: a) loading effects due to feedback network b) non-unilateral feedback Circuit Example Assume that Vx is externally biased with no impact on circuit -> KCL at node Vx: KCL at node Vout: Note that the transfer function has a RHP zero (high frequency) and a pole. At low frequencies: Circuit Example Now, consider feedback -> -> Feedback based analysis provides a good approximation -> Note that this analysis does not predict the feed-forward zero. Loading Effects Loading -> In an ideal system, we neglect the impact of the feedback network loading the open loop circuit. -> In real circuits, this may not always be negligible -> Solution: Model the feedback network in terms of Z,Y, H or G parameters and calculate the impact of loading -> Not as simple as it sounds! Sometimes, it is better to get qualitative understanding from “unloaded” feedback analysis and refine the final results with KCL/KVL or circuit simulations.