Logical Effort Basics from Bart Zeydel
Logical Effort Components Template Template Width Scaled by a Input Capacitance increases by a · Ctemplate Resistance decreases by Rtemplate / a Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort Input Capacitance Cox eox / tox (unit: F/m2) Leff = L – 2xd Cgate = CoxWLeff = CoxWL(1 – 2xd / L) k1 = Cox(1 – 2xd / L) Ctemplate(inv) = k1WnLn + k1WpLp Cin(inv) = a · Ctemplate(inv) Input Capacitance is the sum of the gate capacitances Using LE, a denotes size in multiples of the template Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort Resistance R = (r / t) (L / W) ohms R = Rsheet (L / W) ohms Rchannel = Rsheet (L / W) ohms Rsheet = 1 / ( mCox ( Vgs – Vt )) ohms m = surface mobility k2 = Cox (Vgs – Vt) Rchannel = L / (k2 m W) ohms Rup(inv) = Rup-template(inv) / a Rdown(inv) = Rdown-template(inv) / a Resistance is dependent on channel width and process Larger values of a result in lower resistance. Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort Parasitics Cja junction capacitance per m2 Cjp periphery capacitance per m W width of diffusion region m Ldiff length of diffusion region m Larger values of a result in increased parasitic cap, however R decreases at same rate, thus constant RC delay. Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
RC Model for CMOS logic gate Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
LE Delay derivation for step input Vout Id C Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
LE Model derivation (cont.) Substituting in R and C to obtain k Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort Gate Delay Model Gate Template a Scaled Gate Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort Gate Delay Model Simplify Analysis by Normalizing to Inverter Template Doesn’t change with a Doesn’t change with a Same derivation can be performed for tr Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Estimating g and p of Gates Inverter NAND2 NOR2 2 2 4 2 4 2 1 2 1 1 g = (2+1)/(2+1) = 1 p = Cp-inv/Cinv = pinv g = (2+2)/(2+1) = 4/3 p = [(2+2+2)/(2+1)]pinv = 2pinv g = (4+1)/(2+1) = 5/3 p = [(4+1+1)/(2+1)]pinv = 2pinv Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Prof. V. G. Oklobdzija: High-Performance System Design 130nm Delay of Gates vs. h NOR2 NAND2 Inverter Slope = t = 7.3ps Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Normalized Delay of Gates vs. h t = 7.3ps NOR2: g=1.57, p=1.89 NAND2: g=1.14, p=1.45 Inverter: g=1, p=0.93 Effort Delay Parasitic Delay Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Normalized Delay Estimate of Gates vs. h Note: To simplify analysis Assume Cp-inv = Cinv NOR2: g=5/3, p=2 NAND2: g=4/3, p=2 Inverter: g=1, p=1 Effort Delay Parasitic Delay Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Prof. V. G. Oklobdzija: High-Performance System Design LE Path Delay Gate 1 Gate 2 Gate 3 Cout Input Capacitance Cin C2 C3 Logical Effort: g1 g2 g3 Parasitic Delay p1 p2 p3 Stage Effort: f1 f2 f3 Any n-stage path can be described using Logical Effort Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
LE Path Delay Optimization Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
LE Path Delay Optimization (cont.) By Definition, Cin and Cout are fixed. Solve for C2 and C3: Minimum delay occurs when stage efforts are equal Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Simplified Path Optimization We want the effort of each stage to be equal. = Path Effort = Stage Effort To quickly solve for F: = Logical Effort of the path Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Delay Optimization Example Cin = 1 C2 C3 C4 g f = gh Ci Gate 1 1 Gate 2 1 Gate 3 1 Gate 4 1 81 Total Delay = (84 + 4pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Delay Optimization Example Cin = 1 C2 C3 C4 g f = gh Ci Gate 1 1 Gate 2 1 Gate 3 1 40.5 Gate 4 1 2 40.5 Total Delay = (44.5 + 4pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Delay Optimization Example Cin = 1 C2 C3 C4 g f = gh Ci Gate 1 1 Gate 2 1 10.1 Gate 3 1 4 10.1 Gate 4 1 2 40.5 Total Delay = (17.1 + 4pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Delay Optimization Example Cin = 1 C2 C3 C4 g f = gh Ci Gate 1 1 10.1 Gate 2 1 10.1 Gate 3 1 4 10.1 Gate 4 1 2 40.5 Total Delay = (17.1 + 4pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Optimal Sizing for Delay Cin = 1 C2 C3 C4 g f = gh Ci Gate 1 1 3 Gate 2 1 3 Gate 3 1 3 9 Gate 4 1 3 27 Optimal Delay = (12 + 4pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Delay Optimization and Sizing Example Cin = 1 C2 C3 C4 Use Logical Effort to optimize sizes for Delay Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Prof. V. G. Oklobdzija: High-Performance System Design Delay Optimization and Sizing Example Cin = 1 C2 C3 C4 Size from output to input using fopt Delay Estimate Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Example 2: Path Optimization C2 C4 Cin = 1 C3 g f = gh Ci Size Gate 1 1 Gate 2 5/3 5/3*3/5 1 Gate 3 4/3 4/3*3/4 1 Gate 4 1 81 Total Delay = (85 + 6pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Example 2: Path Optimization Cin = 1 C2 C3 C4 g f = gh Ci Size Gate 1 1 3.66 Gate 2 5/3 3.66 2.2 Gate 3 4/3 3.66 8.06 6.04 Gate 4 1 3.66 22.1 Optimal Delay = (14.64 + 6pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Prof. V. G. Oklobdzija: High-Performance System Design Example 2: LE Solution Cin = 1 C2 C3 C4 Size from output to input using fopt S1 = 1, S2 = 2.2, S3 = 6.04, S4 = 22.1 Delay Estimate Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort for Multi-path From LE, and Minimum delay occurs when Da = Db or Fa = Fb (ignoring parasitics) Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort for Multi-path (cont.) Branching: Ratio of total capacitance to on-path capacitance Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Logical Effort for Multi-path (cont) Substituting C0 and C3 Since Minimum Delay occurs when Da = Db or Fa = Fb Similarly Branching = 2 when Ga = Gb and Cout1 = Cout2 Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Prof. V. G. Oklobdzija: High-Performance System Design Example 3: Uniform Branching C6 C5 Cout2 = C5+C3 Cin = 1 C2 C4 C3 g f = gh Ci b Gate 1 1 Gate 2 5/3 10/3 1 2 Gate 3,5 4/3 1 Gate 4,6 1 81 Total Delay = (86.67 + 6pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Prof. V. G. Oklobdzija: High-Performance System Design Example 3: Uniform Branching C6 C5 Cout2 = C5+C3 Cin = 1 C2 C4 C3 g f = gh Ci b Gate 1 1 4.36 Gate 2 5/3 4.36 2 Gate 3,5 4/3 4.36 5.69 1 Gate 4,6 1 4.36 18.6 Optimal Delay = (17.44 + 6pinv)t Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Prof. V. G. Oklobdzija: High-Performance System Design Example 3: LE Solution C6 Cout2 = C5+C3 C5 Cin = 1 C2 C3 C4 Size from output to input using fopt Delay Estimate Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design
Complex Multi-path Optimization If each path has internal branching, ba and bb are as follows Note: This solution and previous solution differ from that described in LE book (which is incorrect) Fall 2004 Prof. V. G. Oklobdzija: High-Performance System Design