EECS 40 Spring 2003 Lecture 24 S Ross

  • Slides: 16
Download presentation
EECS 40 Spring 2003 Lecture 24 S. Ross Lecture 24 Today we will •

EECS 40 Spring 2003 Lecture 24 S. Ross Lecture 24 Today we will • Review charging of output capacitance (origin of gate delay) • Calculate output capacitance • Discuss fan-out • Discuss “complementary” nature of CMOS

EECS 40 Spring 2003 Lecture 24 S. Ross ORIGIN OF GATE DELAY When the

EECS 40 Spring 2003 Lecture 24 S. Ross ORIGIN OF GATE DELAY When the inputs A and B change such that the output F changes, VDD A S S PMOS 1 PMOS 2 F B NMOS 1 S NMOS 2 S the output cannot change instantaneously; the output capacitance must be charged or discharged. This is GATE DELAY.

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: PULL-DOWN DEVICES In our logic

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: PULL-DOWN DEVICES In our logic circuits, the NMOS transistors have • Gate terminal connected to VIN • Source terminal connected to ground directly, or through another NMOS This means • When VIN is high, NMOS transistors are “on”. They help pull-down VOUT to ground, by conducting current to discharge the output capacitance. • When VIN is low, NMOS transistors are “off”. They act as open circuits from source to drain.

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: PULL-UP DEVICES In our logic

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: PULL-UP DEVICES In our logic circuits, the PMOS transistors have • Gate terminal connected to VIN • Source terminal connected to VDD directly, or through another PMOS This means • When VIN is low, PMOS transistors are “on”. They help pull-up VOUT to VDD, by conducting current to charge the output capacitance. • When VIN is high, PMOS transistors are “off”. They act as open circuits from source to drain.

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: MODEL FOR GATE DELAY ANALYSIS

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: MODEL FOR GATE DELAY ANALYSIS There is a model for the behavior of transistors in a CMOS logic circuit to analyze charging/discharging of the output capacitance. VIN = 0 (for NMOS) VIN = VDD (for PMOS) D G R S The resistance R is the effective resistance for the device during the first half of the transition. Each device can have different R! VIN = VDD (for NMOS) VIN = 0 (for PMOS) D G R S

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: CALCULATING EFFECTIVE RESISTANCE ID(N) Consider

EECS 40 Spring 2003 Lecture 24 S. Ross REVIEW: CALCULATING EFFECTIVE RESISTANCE ID(N) Consider pull-down, when VOUT must go from VDD to 0 V. VDD t=0 t=tp ID(N) ≈ IDSAT(N) RP VOUT VDD/2 VDD VDS(N) We calculate RN by averaging the values of VDS(N)/ID(N) at the beginning and ending of delay. RN = ¾ VDD / IDSAT(N) RP = -¾ VDD / IDSAT(P) RN

EECS 40 Spring 2003 Lecture 24 S. Ross CALCULATING OUTPUT CAPACITANCE Two major sources

EECS 40 Spring 2003 Lecture 24 S. Ross CALCULATING OUTPUT CAPACITANCE Two major sources of capacitance: 1. The transistor gates in the next stage 2. The metal connection to the next stage Both can be computed using the parallel plate capacitor formula: A is the area of the plates, k is the dielectric constant of the insulator in between the plates, e 0 is the permittivity of free space, and d is the distance in between the plates. We denote since this is fixed by fabrication process.

EECS 40 Spring 2003 Lecture 24 S. Ross CALCULATING OUTPUT CAPACITANCE Each transistor gate

EECS 40 Spring 2003 Lecture 24 S. Ross CALCULATING OUTPUT CAPACITANCE Each transistor gate terminal attached to the output contributes a gate capacitance where W and L are the channel dimensions and COX is the capacitance of the gate per unit area (parameters from ID vs. VDS). Each metal connection at the output contributes a capacitance also given by the parallel plate capacitor formula, but with different length, width, and capacitance per unit area.

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE VDD S S S VOUT

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE VDD S S S VOUT 1 VOUT 2 VIN S Suppose that VIN was logic 1 for a long time, and then switches to logic 0 at t = 0. Find the propagation delay through the inverter.

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE Use VDD = 5 V

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE Use VDD = 5 V VTH(N) = -VTH(P) = 1 V COX = 5 f. F/µm 2 for both transistors L = 2 µm for both transistors W = 2 µm for both transistors l = 0 for both transistors µN= 50000 mm 2 / (V s) Calculate the effective µP= 25000 mm 2 / (V s) resistance and total output capacitance due to gate and WI = 2 µm interconnect capacitance. LI = 200 µm COX(I) = 0. 1 f. F/µm 2

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE Since VIN is now low,

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE Since VIN is now low, VOUT 1 must go from low to high. Pull-up VDD RP VDD S VOUT VDD VIN =0 V S RN RP is the resistance involved in the charging.

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE RP = - ¾ VDD

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE RP = - ¾ VDD / IDSAT(P) RP = - ¾ (5 V) / (W/L m. P COX (VGS(P) – VTH(P))2 ) RP = - ¾ (5 V) / (2 mm/2 mm 25000 mm 2/Vs 5 f. F/mm 2 (-5 V – -1 V)2 ) RP = 1. 875 k. W Now calculate COUT: There are 4 transistor gates attached to inverter output, and one wire connecting the inverter output to the NAND input. COUT = 4 CG + CI

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE CG = W L COX

EECS 40 Spring 2003 Lecture 24 S. Ross EXAMPLE CG = W L COX = (2 mm)(5 f. F/mm 2) = 20 f. F CI = WI LI COX(I) = (2 mm)(200 mm)(0. 1 f. F/mm 2) = 40 f. F COUT = 4 CG + CI = (4)20 f. F + 40 f. F = 100 f. F t. P = 0. 69 RP COUT = 0. 69 (1. 875 k. W) (100 f. F) = 129 ps

EECS 40 Spring 2003 Lecture 24 S. Ross FAN-OUT Consider our previous example. Suppose

EECS 40 Spring 2003 Lecture 24 S. Ross FAN-OUT Consider our previous example. Suppose that we connected N NAND gates to the output of the inverter. Each NAND gate adds 4 more gate capacitances and another interconnect capacitance. COUT = N(4 CG + CI) = 100 N f. F tp = 129 N ps The fan-out, or number of logic gates that can be attached to an output, is limited by propagation delay considerations.

EECS 40 Spring 2003 Lecture 24 S. Ross LOOKING AT CMOS CIRCUITS One can

EECS 40 Spring 2003 Lecture 24 S. Ross LOOKING AT CMOS CIRCUITS One can often “see” the logical operation in a CMOS circuit by looking at either the top or bottom half of the circuit. VDD A B C F B A For example, looking at the top half of this circuit, we see that the output will be connected to VDD (F is high) when: (B OR C is low) AND A is low C F = (B + C) A

EECS 40 Spring 2003 Lecture 24 S. Ross LOOKING AT CMOS CIRCUITS The bottom

EECS 40 Spring 2003 Lecture 24 S. Ross LOOKING AT CMOS CIRCUITS The bottom half of the circuit always results in the same equation as the top half, just rewritten via De. Morgan (that’s why output is always defined). VDD A B C F B A C Looking at the bottom half of this circuit, we see that the output will be connected to ground (F is low) when: (B AND C are high) OR A is high F=BC+A