Homec4science
Diffusion EMPoWER (master) Browse

R1252/simulation-code/cvx/examples/circuit_design/html/simple_NAND2_gate_design.html
de096106547amaster

Last Change
Raw File

simple_NAND2_gate_design.html
View Options

<!DOCTYPE HTML>
<html>
<head>
<meta charset="UTF-8">
<title>Two-input NAND gate sizing (GP)</title>
<link rel="canonical" href="http://cvxr.com/cvx/examples/circuit_design/html/simple_NAND2_gate_design.html">
<link rel="stylesheet" href="../../examples.css" type="text/css">
</head>
<body>
<div id="header">
<h1>Two-input NAND gate sizing (GP)</h1>
Jump to:&nbsp;&nbsp;&nbsp;&nbsp;
<a href="#source">Source code</a>&nbsp;&nbsp;&nbsp;&nbsp;
<a href="#output">Text output</a>
&nbsp;&nbsp;&nbsp;&nbsp;
<a href="#plots">Plots</a>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="../../index.html">Library index</a>
</div>
<div id="content">
<a id="source"></a>
<pre class="codeinput">
<span class="comment">% Boyd, Kim, Patil, and Horowitz, "Digital circuit optimization</span>
<span class="comment">% via geometric programming"</span>
<span class="comment">% Written for CVX by Almir Mutapcic 02/08/06</span>
<span class="comment">% (a figure is generated)</span>
<span class="comment">%</span>
<span class="comment">% This is an example taken directly from the paper:</span>
<span class="comment">%</span>
<span class="comment">% Digital circuit optimization via geometrical programming</span>
<span class="comment">% by Boyd, Kim, Patil, and Horowitz</span>
<span class="comment">% Operations Research 53(6): 899-932, 2005.</span>
<span class="comment">%</span>
<span class="comment">% Solves the problem of choosing device widths w_i for the given</span>
<span class="comment">% NAND2 gate in order to achive minimum Elmore delay for different</span>
<span class="comment">% gate transitions, subject to limits on the device widths,</span>
<span class="comment">% gate area, power, and so on. The problem is a GP:</span>
<span class="comment">%</span>
<span class="comment">% minimize D = max( D_1, ..., D_k ) for k transitions</span>
<span class="comment">% s.t. w_min &lt;= w &lt;= w_max</span>
<span class="comment">% A &lt;= Amax, etc.</span>
<span class="comment">%</span>
<span class="comment">% where variables are widths w.</span>
<span class="comment">%</span>
<span class="comment">% This code is specific to the NAND2 gate shown in figure 19</span>
<span class="comment">% (page 926) of the paper. All the constraints and the objective</span>
<span class="comment">% are hard-coded for this particular circuit.</span>
<span class="comment">%********************************************************************</span>
<span class="comment">% problem data and hard-coded GP specs (evaluate all transitions)</span>
<span class="comment">%********************************************************************</span>
N = 4; <span class="comment">% number of devices</span>
Cload = 12; <span class="comment">% load capacitance</span>
Vdd = 1.5; <span class="comment">% voltage</span>
<span class="comment">% device specs</span>
NMOS = struct(<span class="string">'R'</span>,0.4831, <span class="string">'Cdb'</span>,0.6, <span class="string">'Csb'</span>,0.6, <span class="string">'Cgb'</span>,1, <span class="string">'Cgs'</span>,1);
PMOS = struct(<span class="string">'R'</span>,2*0.4831, <span class="string">'Cdb'</span>,0.6, <span class="string">'Csb'</span>,0.6, <span class="string">'Cgb'</span>,1, <span class="string">'Cgs'</span>,1);
<span class="comment">% maximum area and power specification</span>
Amax = 24;
wmin = 1;
<span class="comment">% varying parameters for the tradeoff curve</span>
Npoints = 25;
Amax = linspace(5,45,Npoints);
Dopt = [];
disp(<span class="string">'Generating the optimal tradeoff curve...'</span>)
<span class="keyword">for</span> k = 1:Npoints
fprintf(1,<span class="string">' Amax = %5.2f:'</span>, Amax(k));
cvx_begin <span class="string">gp</span> <span class="string">quiet</span>
<span class="comment">% device width variables</span>
variable <span class="string">w(N)</span>
<span class="comment">% device specs</span>
device(1:2) = PMOS; device(3:4) = NMOS;
<span class="keyword">for</span> num = 1:N
device(num).R = device(num).R/w(num);
device(num).Cdb = device(num).Cdb*w(num);
device(num).Csb = device(num).Csb*w(num);
device(num).Cgb = device(num).Cgb*w(num);
device(num).Cgs = device(num).Cgs*w(num);
<span class="keyword">end</span>
<span class="comment">% capacitances</span>
C1 = sum([device(1:3).Cdb]) + Cload;
C2 = device(3).Csb + device(4).Cdb;
<span class="comment">% input capacitances</span>
Cin_A = sum([ device([2 3]).Cgb ]) + sum([ device([2 3]).Cgs ]);
Cin_B = sum([ device([1 4]).Cgb ]) + sum([ device([1 4]).Cgs ]);
<span class="comment">% resistances</span>
R = [device.R]';
<span class="comment">% area definition</span>
area = sum(w);
<span class="comment">% delays and dissipated energies for all six possible transitions</span>
<span class="comment">% transition 1 is A: 1-&gt;1, B: 1-&gt;0, Z: 0-&gt;1</span>
D1 = R(1)*(C1 + C2);
E1 = (C1 + C2)*Vdd^2/2;
<span class="comment">% transition 2 is A: 1-&gt;0, B: 1-&gt;1, Z: 0-&gt;1</span>
D2 = R(2)*C1;
E2 = C1*Vdd^2/2;
<span class="comment">% transition 3 is A: 1-&gt;0, B: 1-&gt;0, Z: 0-&gt;1</span>
<span class="comment">% D3 = C1*R(1)*R(2)/(R(1) + R(2)); % not a posynomial</span>
E3 = C1*Vdd^2/2;
<span class="comment">% transition 4 is A: 1-&gt;1, B: 0-&gt;1, Z: 1-&gt;0</span>
D4 = C1*R(3) + R(4)*(C1 + C2);
E4 = (C1 + C2)*Vdd^2/2;
<span class="comment">% transition 5 is A: 0-&gt;1, B: 1-&gt;1, Z: 1-&gt;0</span>
D5 = C1*(R(3) + R(4));
E5 = (C1 + C2)*Vdd^2/2;
<span class="comment">% transition 6 is A: 0-&gt;1, B: 0-&gt;1, Z: 1-&gt;0</span>
D6 = C1*R(3) + R(4)*(C1 + C2);
E6 = (C1 + C2)*Vdd^2/2;
<span class="comment">% objective is the worst-case delay</span>
minimize( max( [D1 D2 D4] ) )
subject <span class="string">to</span>
area &lt;= Amax(k);
w &gt;= wmin;
cvx_end
<span class="comment">% display and store computed values</span>
fprintf(1,<span class="string">' delay = %3.2f\n'</span>,cvx_optval);
Dopt = [Dopt cvx_optval];
<span class="keyword">end</span>
<span class="comment">% plot the tradeoff curve</span>
plot(Dopt,Amax);
xlabel(<span class="string">'Dmin'</span>); ylabel(<span class="string">'Amax'</span>);
disp(<span class="string">'Optimal tradeoff curve plotted.'</span>)
</pre>
<a id="output"></a>
<pre class="codeoutput">
Generating the optimal tradeoff curve...
Amax = 5.00: delay = 11.56
Amax = 6.67: delay = 9.23
Amax = 8.33: delay = 7.84
Amax = 10.00: delay = 6.90
Amax = 11.67: delay = 6.23
Amax = 13.33: delay = 5.73
Amax = 15.00: delay = 5.34
Amax = 16.67: delay = 5.03
Amax = 18.33: delay = 4.77
Amax = 20.00: delay = 4.55
Amax = 21.67: delay = 4.37
Amax = 23.33: delay = 4.22
Amax = 25.00: delay = 4.08
Amax = 26.67: delay = 3.96
Amax = 28.33: delay = 3.86
Amax = 30.00: delay = 3.76
Amax = 31.67: delay = 3.68
Amax = 33.33: delay = 3.60
Amax = 35.00: delay = 3.54
Amax = 36.67: delay = 3.47
Amax = 38.33: delay = 3.42
Amax = 40.00: delay = 3.36
Amax = 41.67: delay = 3.32
Amax = 43.33: delay = 3.27
Amax = 45.00: delay = 3.23
Optimal tradeoff curve plotted.
</pre>
<a id="plots"></a>
<div id="plotoutput">
<img src="simple_NAND2_gate_design__01.png" alt="">
</div>
</div>
</body>
</html>
c4science · Help