forked from testaco/DCP6
-
Notifications
You must be signed in to change notification settings - Fork 1
/
cordic16s16.v
220 lines (220 loc) · 7.81 KB
/
cordic16s16.v
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
//
// cordic16s16.v - 16-bit Input/16-bit Output Serial CORDIC Processing Engine
//
// (C) Copyright 2009-2013 John B. Stephensen
//
// This Verilog source file and all its derivatives are licensed only for
// personal non-profit educational use in the Amateur Radio Service and
// the license is not transferrable. The information is provided as-is for
// experimental purposes and the author does not warranty its freedom
// from defects or its suitability for any specific application.
//
// This module implements the CORDIC algorithm with 24-bit internal resolution. The X
// and Y inputs are rotated into the first and fourth quadrants before microrotations.
// Side effects during rotation multiply the accumulated X and Y values by a factor of
// 1.64676 which is corrected with multipliers on the X and Y outputs. The accumulated
// Z value is also corrected by a factor of +/-Pi radians using a 2-bit adder on the
// output.
//
// The CORDIC engine may be operated in vector or rotation mode for polar to rectangular
// or rectangular to polar conversion, respectively. The selection is made using the
// rotation input and must be maintained throughout the conversion process. Vector mode
// rotates X and Y until Y is zero. Rotate mode rotates X and Y until Z is zero. 20
// rotations are used for 16-bit accuracy. Accuracy will decrease when both the X and Y
// inputs are less than 1/4 of full scale. Accuracy is 14 bits at -48 dBFS, 11-bits at
// -48 dBFS and 7 bits at -72 dBFS. 20 clock cycles are required for each conversion.
//
// S6: 439 LUTs and 288 registers are used. 107 MHz maximum clock rate.
//
// Normal Warnings:
// Signal <z0_14> is unconnected.
// Signals <xr_0>, <xr_1>, <xr_2>, <xr_3>, <yr_0>, <yr_1>, <yr_2> & <yr_3> are unconnected.
//
// History:
// 4-22-09 moved input left by 1 bit as only two X and Y overflow bits are needed.
// 4-23-09 altered coarse rotation logic and shrunk correction factor to prevent overflow
// 4-24-09 use input coarse rotation for rotate mode (mi=1)
// 5-18-10 18->16-bit X and Y inputs (was 232 slices)
// 8-18-11 use serial multipliers to correct magnitude growth (was 212 slices and 2 mult.)
// 11-7-11 use 24-bit acc. and 20 iterations for reduced distortion (was 353 LUT 231 reg 118 MHz)
// 11-16-11 use 22x18=22 bit multiplier and round 20 to 16 bits (was 376 LUT 239 reg 107 MHz)
// 2-24-13 change angle accumulator and ROM to 24Q23 format, adding 1 bit of precision
//
module cordic16s16(
input [15:0] xi, // 16-bit X input
input [15:0] yi, // 16-bit Y input
input [15:0] zi, // 16-bit Z input
input mi, // 0 = vector mode, 1 = rotation mode
input load, // load X, Y and Z input data
output [15:0] xo, // 16-bit X output
output [15:0] yo, // 16-bit Y output
output [15:0] zo, // 16-bit Z output
output mo, // 0 = vector mode, 1 = rotation mode
output rdy, // X, Y and Z output data ready
input clk, // master clock
input rst // reset X, Y and Z accumulators
);
// internal signals
wire cmpx,cmpy,swap; // coarse rotation controls
reg start,rotate; // start interpolation and vector/rotate mode
wire [15:0] xx,yy; // input coarse rotation complementors
reg [15:0] x0,y0; // coarse rotation output
reg [15:0] z0; // angle
reg xs,ys; // sign bits
wire [23:0] xa,xb,x,ya,yb,y,za,zb,z;// adder/subtractor I/O
wire subxz; // 0 = add, 1 = subtract
reg [4:0] i; // iteration number
wire ce,nz,tc; // counter enable, non-zero and terminal count
reg v; // microrotaion engine data valid signal
wire vc; // correction multiplier data valid signal
wire [21:0] xc,yc; // corrected X and Y outputs
reg [15:0] zc; // corrected phase output (rounded up)
reg [19:0] xr,yr; // rounded X and Y outputs
reg vr; // rounded output valid signal
// vector mode: exchange and invert X and Y inputs to move inputs into 1st and 4th quadrants
// Q X Y X0 Y0 Cor.
// 1 + + X Y 0
// 2 - + Y -X +90
// 3 - - -Y X -90
// 4 + - X Y 0
// rotate mode: exchange and invert X and Y inputs to implement first +/-90 degrees of rotation
// Range Xo Yo Cor. z15 z14
// +90/+180 Y -X +90 1 0
// -90/-180 -Y X -90 0 1
// true/complement logic inverts inputs when necessary
assign cmpx = mi ? zi[15] & ~zi[14] : xi[15] & ~yi[15];
assign cmpy = mi ? zi[14] & ~zi[15] : xi[15] & yi[15];
tc16a tcix (
.D(xi),
.CMP(cmpx), // -90
.Y(xx)
);
tc16a tciy (
.D(yi),
.CMP(cmpy), // +90
.Y(yy)
);
// save mode and sign bits for output correction
// swap X and Y axis when necessary
// delay Z input and start signal
assign swap = mi ? zi[15] ^ zi[14] : xi[15];
always @ (posedge clk)
begin
if (load) xs <= xi[15]; // save signs
if (load) ys <= yi[15];
if (load) x0 <= swap ? yy : xx; // 0 or +/-90 rotation
if (load) y0 <= swap ? xx : yy;
if (load) z0 <= zi; // delay
if (load) rotate <= mi;
start <= load;
end
// Output accuracy is 16 bits.
// The accumulators have 3 overflow bits and 5 underflow bits.
// This results in increased accuracy at low signal levels.
// X = X +/- Y/2^1
assign xa = start ? {x0[15],x0[15],x0[15],x0,5'b00000} : x;
shift24x24 shiftyz (
.D(ya),
.Y(xb),
.S(i)
);
addsub24s accx (
.A(xa),
.B(xb),
.Q(x),
.SUB(subxz),
.CE(ce),
.CLK(clk),
.RST(rst)
);
// Y = Y +/- X/2^1
assign ya = start ? {y0[15],y0[15],y0[15],y0,5'b00000} : y;
shift24x24 shiftzy (
.D(xa),
.Y(yb),
.S(i)
);
addsub24s accy (
.A(ya),
.B(yb),
.Q(y),
.SUB(~subxz),
.CE(ce),
.CLK(clk),
.RST(rst)
);
// Z = Z +/- atan(1)
// drop bit 14 so input is between +Pi/2 and -Pi/2
assign za = start ? {z0[15],z0[15],z0[13:0],8'h00} : z;
addsub24s accz (
.A(za),
.B(zb), // add/subtract this amount
.Q(z), // accumulator output
.SUB(subxz), // 0 = add, 1 = subtract
.CE(ce), // clock enable
.CLK(clk),
.RST(rst)
);
// ROM containing arctan(2^-i)/360 * 2^24
rom32x24cordic rom (
.a(i),
.spo(zb)
);
// iteration counter counts up until terminal count reached
always @ (posedge clk)
begin
if (rst) i <= 0;
else if (ce) i <= i + 1'b1;
v <= tc;
end
assign ce = start | nz; // enable clock when loading or not zero
assign nz = |i; // counter not zero
assign tc = (i == 19); // terminal count
// select rotation (0) or vector (1) mode:
// rotate mode drives Z towards zero by watching sign bit.
// vector mode drives Y towards zero by watching sign bit.
// vec accumulator rot
// Y X Y Z Z
// + ADD SUB ADD -
// - SUB ADD SUB +
// Vector mode (0) is used for rectangular to polar coordinate conversion.
// Rotate mode (1) is used to rotate the X and Y inputs for frequency
// translation or for polar to rectangular conversion.
assign subxz = rotate ? ~za[23] : ya[23];
// correct magnitude growth in CORDIC algorithm
// multiply by 0.6072530 = 79593.86/131072 = 136E9.D/40000
mul22c17u22c xyadj (
.dix(x[22:1]), // 22-bit signed CORDIC outputs
.diy(y[22:1]),
.diz(17'h136E9),
.iv(v),
.dox(xc),
.doy(yc),
.ov(vc),
.clk(clk)
);
// round outputs from 20 bits to 16 bits using unbiased rounding
// add 7 if negative or 8 if positive and drop bottom 4 bits of sum
always @ (posedge clk)
begin
if (vc) xr <= xc[19:0] + {16'h0000,~xc[21],xc[21],xc[21],xc[21]};
if (vc) yr <= yc[19:0] + {16'h0000,~yc[21],yc[21],yc[21],yc[21]};
vr <= vc;
end
// correct phase output with +/-Pi/2 adjustment, drop bottom 8 bits and round up
// Q X Y X0 Y0 Cor.
// 1 + + X Y 00
// 2 - + Y -X 01 (+90)
// 3 - - -Y X 11 (-90)
// 4 + - X Y 00
always @ (posedge clk) if (v) zc <= {(z[23:22] + (rotate ? 2'b00 : {(xs&ys),xs})),z[21:8]} + z[7];
// connect outputs - 16-bit X-axis/magnitude, Y-axis and phase angle outputs
// Corrected X and Y outputs drop one bit due to multiplication by 0.6072530 and
// maximum value of 1.4142 * 1.6468 = 2.3289 and one bit due to replicated sign on
// multiplier output. A third bit may be dropped if inputs stay in the unit circle.
assign xo = xr[19:4];
assign yo = yr[19:4];
assign zo = zc;
assign mo = rotate;
assign rdy = vr;
endmodule