EECS 40 Spring 2003 Lecture 27 S Ross

  • Slides: 12
Download presentation
EECS 40 Spring 2003 Lecture 27 S. Ross LECTURE 27 Today we will •

EECS 40 Spring 2003 Lecture 27 S. Ross LECTURE 27 Today we will • Put a twist on our normally linear operational amplifier circuits to make them perform nonlinear computations • Make a linear circuit model for the nonlinear NMOS transistor (Preview of EE 105) Next time we will • Show we can design a pipelined computer datapath at the transistor level • Use a relay to design an analog circuit that counts Trying to expose you to various complicated circuits/topics to use the tools you’ve developed and prepare you for final exam…

EECS 40 Spring 2003 Lecture 27 S. Ross NONLINEAR OPERATIONAL AMPLIFIERS When I put

EECS 40 Spring 2003 Lecture 27 S. Ross NONLINEAR OPERATIONAL AMPLIFIERS When I put a nonlinear device in an operational amplifier circuit, I can compute a nonlinear function. Consider the following circuit using the realistic diode model: VIN R + VOUT

EECS 40 Spring 2003 Lecture 27 S. Ross NONLINEAR OPERATIONAL AMPLIFIERS VIN ID R

EECS 40 Spring 2003 Lecture 27 S. Ross NONLINEAR OPERATIONAL AMPLIFIERS VIN ID R VOUT + Computes an exponential function of VIN !

EECS 40 Spring 2003 Lecture 27 S. Ross NONLINEAR OPERATIONAL AMPLIFIERS What if I

EECS 40 Spring 2003 Lecture 27 S. Ross NONLINEAR OPERATIONAL AMPLIFIERS What if I switch the positions of the resistor and the diode (and make sure VIN ≥ 0 V)? VIN R ID Computes a natural logarithm function of VIN ! Changing the position of the elements inverted the function performed! + VOUT

EECS 40 Spring 2003 Lecture 27 S. Ross TRANSISTOR AS CURRENT SOURCE But this

EECS 40 Spring 2003 Lecture 27 S. Ross TRANSISTOR AS CURRENT SOURCE But this hides the fact that This circuit acts like a IDSAT depends on VGS; it is constant current source, as really a voltage-dependent long as the transistor current source! remains in saturation mode. Load D VDD IDSAT + _ VGS If VGS is not constant, the model fails. What if VGS changes? What if there is noise in the circuit? G IDSAT + _ S Load

EECS 40 Spring 2003 Lecture 27 S. Ross THE EFFECT OF A SMALL SIGNAL

EECS 40 Spring 2003 Lecture 27 S. Ross THE EFFECT OF A SMALL SIGNAL ID If VGS changes a little bit, so does ID. VGS = 2. 1 V VGS = 2 V VGS = 1. 9 V VDS

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL Let’s include the

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL Let’s include the effect of noise in VGS. Suppose we have tried to set VGS to some value VGS, DC with a fixed voltage source, but some noise DVGS gets added in. VGS = VGS, DC + DVGS G + VGS - D DIDSAT = g(DVGS) IDSAT, DC = f(VGS, DC) S We get the predicted IDSAT, DC plus a change due to noise, DIDSAT. No current flows into or out of the gate because of the opening.

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL To be even

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL To be even more accurate, we could add in the effect of l. When l is nonzero, ID increases linearly with VDS in saturation. We can model this with a resistor from drain to source: G + VGS - DIDSAT = g(DVGS) IDSAT, DC = f(VGS, DC) ro D S The resistor will make more current flow from drain to source as VDS increases.

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL How do we

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL How do we find the values for the model? IDSAT, DC = ½ W/L m. N COX (VGS, DC – VTH)2 This is a constant depending on VGS, DC. This is a first-order Taylor series approximation which works out to DIDSAT = W/L m. N COX (VGS, DC – VTH)DVGS We often refer to W/L m. N COX (VGS, DC – VTH) as gm, so DIDSAT = gm DVGS.

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL Including the effect

EECS 40 Spring 2003 Lecture 27 S. Ross THE SMALL-SIGNAL MODEL Including the effect of l via ro, the added current contributed by the resistor is Ir 0 = ½ W/L m. N COX (VGS – VTH)2 l. VDS To make things much easier, since the l effect is small anyway, we neglect the effect of DVGS in the resistance, so the current is Ir 0 ≈ ½ W/L m. N COX (VGS, DC – VTH)2 l. VDS = IDSAT, DC l VDS This leads to r 0 = VDS / Ir 0 = (l IDSAT, DC)-1

EECS 40 Spring 2003 Lecture 27 S. Ross EXAMPLE Revisit the example of Lecture

EECS 40 Spring 2003 Lecture 27 S. Ross EXAMPLE Revisit the example of Lecture 20, but now the 3 V source can have noise up to ± 0. 1 V. VTH(N) = 1 V, W/L mn. COX = 500 m A/V 2, Find the range of variation for ID. l = 0 V-1. We figured out that saturation is the correct mode in Lecture 20. 1. 5 k. W G + IDSAT, DC = ½ W/L m. N COX (VGS, DC – VTH)2 S = 250 m. A/V 2 (3 V – 1 V)2 G = 1 m. A _ V 3 V + _ The “noiseless” value of ID is VDS _ + 4 V ID + _ D S

EECS 40 Spring 2003 Lecture 27 S. Ross EXAMPLE The variation in ID due

EECS 40 Spring 2003 Lecture 27 S. Ross EXAMPLE The variation in ID due to noise: 1. 5 k. W DIDSAT = W/L m. N COX (VGS, DC – VTH)DVGS D G G S = 100 m. A _ V 3 V + _ + VDS _ + 4 V ID + _ = 500 m. A/V 2 (3 V – 1 V) 0. 1 V S So ID could vary between 1. 1 m. A and 0. 9 m. A. Will saturation mode be maintained? Yes.