Home » Archimedes archive » Archimedes World » AW-1996-03-Disc 2.adf » !AcornAns_AcornAns » MathFns/Coef

MathFns/Coef

This website contains an archive of files for the Acorn Electron, BBC Micro, Acorn Archimedes, Commodore 16 and Commodore 64 computers, which Dominic Ford has rescued from his private collection of floppy disks and cassettes.

Some of these files were originally commercial releases in the 1980s and 1990s, but they are now widely available online. I assume that copyright over them is no longer being asserted. If you own the copyright and would like files to be removed, please contact me.

Tape/disk: Home » Archimedes archive » Archimedes World » AW-1996-03-Disc 2.adf » !AcornAns_AcornAns
Filename: MathFns/Coef
Read OK:
File size: 083A bytes
Load address: 0000
Exec address: 0000
File contents
   10REM >Coef
   20REM Calculate polynomial approximations to arbitrary functions
   30REM in range [-1,+1], using Chebyshev Forced Oscillation of Error
   40REM method to derive a near minimax solution
   50REM
   60MODE MODE
   70INPUT "Please enter function to be approximated in range [-1,+1],  eg SIN(PI/4*(x+1)) ";f$
   80INPUT "Please enter degree of approximation (1 to 100) ";n%'
   90DIM f(101), c(101), t(100, 100), p(100)
  100FOR i%=0 TO n%+1
  110 x = COS(i%*PI/(n%+1))
  120 f(i%) = EVAL f$
  130NEXT
  140FOR j%=0 TO n%+1
  150 IF (j%AND1) sum = (f(0)-f(n%+1))/2 ELSE sum = (f(0)+f(n%+1))/2
  160 FOR i%=1 TO n%
  170  sum += f(i%)*COS(i%*j%*PI/(n%+1))
  180 NEXT
  190 c(j%) = sum/(n%+1)
  200 IF j%<>0 c(j%) = 2*c(j%)
  210NEXT
  220E = ABS(c(n%+1)/2)
  230PRINT '"Polynomial approximation is:"'
  240FOR i%=0 TO n%
  250 PRINT "T[";i%;"](x) *",c(i%);TAB(27);
  260 IF i%<n% PRINT "+" ELSE PRINT
  270NEXT
  280PRINT '"where T[0](x)   = 1;"
  290PRINT  "      T[1](x)   = x;"
  300PRINT  "and   T[n+1](x) = 2xT[n](x)-T[n-1](x), for n>=1"'
  310REM Now we need to gen the T polys
  320REM note we use t(i,j) to hold the coef of x^j in poly T[i]
  330t(0,0)=1
  340t(1,1)=1
  350i%=1
  360WHILE i%<n%
  370 t(i%+1,0) = -t(i%-1,0)
  380 FOR j% = 1 TO i%+1
  390  t(i%+1,j%) = 2*t(i%,j%-1)-t(i%-1,j%)
  400 NEXT
  410 i%+=1
  420ENDWHILE
  430FOR i%=0 TO n%
  440 sum = 0
  450 FOR j%=i% TO n%
  460  sum += t(j%, i%)*c(j%)
  470 NEXT
  480 p(i%) = sum
  490NEXT
  500PRINT '"Which may also be written:"'
  510FOR i%=0 TO n%
  520 PRINT "x^";i%;"  *",p(i%);TAB(27);
  530 IF i%<n% PRINT "+" ELSE PRINT
  540NEXT
  550maxerror = 0
  560FOR x=-1 TO 1 STEP 0.001
  570 v = p(n%)
  580 FOR i%=n%-1 TO 0 STEP -1
  590  v = v*x+p(i%)
  600 NEXT
  610 e = ABS(v - EVAL f$)
  620 IF e>maxerror maxerror=e
  630NEXT
  640PRINT ''"For this approximation an upper bound on error is ";maxerror
  650PRINT   "Also we now know minimax error p (the error made by the best possible order ";n%
  660PRINT   "polynomial) lies between:"'
  670PRINT E;"  and  ";maxerror;"."'
  680PRINT "This indicates how close our approximation is to it."
  690PRINT ''"Finally, if only need say 16 bit accuracy, note that our poly's coefficients"
  700PRINT   "multiplied by 2^20 & written in hex are as follows:"'
  710FOR i%=n% TO 0 STEP -1
  720 v = p(i%)
  730 IF v<0 s$="-":v=-v ELSE s$="+"
  740 v2 = v*1048576+0.5
  750 PRINT "x^";i%;"  *",s$;RIGHT$("00000000"+STR$~v2,8);TAB(27);
  760 IF i%>0 PRINT "+" ELSE PRINT
  770NEXT

� >Coef
@� Calculate polynomial approximations to arbitrary functions
C� in range [-1,+1], using Chebyshev Forced Oscillation of Error
(.� method to derive a near minimax solution
2�
<� �
FZ� "Please enter function to be approximated in range [-1,+1],  eg SIN(PI/4*(x+1)) ";f$
P<� "Please enter degree of approximation (1 to 100) ";n%'
Z)� f(101), c(101), t(100, 100), p(100)
d� i%=0 � n%+1
n x = �(i%*�/(n%+1))
x f(i%) = � f$
��
�� j%=0 � n%+1
�= � (j%�1) sum = (f(0)-f(n%+1))/2 � sum = (f(0)+f(n%+1))/2
� � i%=1 � n%
�$  sum += f(i%)*�(i%*j%*�/(n%+1))
� �
� c(j%) = sum/(n%+1)
� � j%<>0 c(j%) = 2*c(j%)
��
�E = �(c(n%+1)/2)
�&� '"Polynomial approximation is:"'
�� i%=0 � n%
�# � "T[";i%;"](x) *",c(i%);�27);
 � i%<n% � "+" � �
�
� '"where T[0](x)   = 1;"
"�  "      T[1](x)   = x;"
,9�  "and   T[n+1](x) = 2xT[n](x)-T[n-1](x), for n>=1"'
6$� Now we need to gen the T polys
@=� note we use t(i,j) to hold the coef of x^j in poly T[i]
Jt(0,0)=1
Tt(1,1)=1
^i%=1
hȕ i%<n%
r t(i%+1,0) = -t(i%-1,0)
| � j% = 1 � i%+1
�*  t(i%+1,j%) = 2*t(i%,j%-1)-t(i%-1,j%)
� �
�
 i%+=1
��
�� i%=0 � n%
� sum = 0
� � j%=i% � n%
�  sum += t(j%, i%)*c(j%)
� �
� p(i%) = sum
��
�$� '"Which may also be written:"'
�� i%=0 � n%
  � "x^";i%;"  *",p(i%);�27);
 � i%<n% � "+" � �
�
&maxerror = 0
0� x=-1 � 1 � 0.001
: v = p(n%)
D � i%=n%-1 � 0 � -1
N  v = v*x+p(i%)
X �
b e = �(v - � f$)
l � e>maxerror maxerror=e
v�
�E� ''"For this approximation an upper bound on error is ";maxerror
�Y�   "Also we now know minimax error p (the error made by the best possible order ";n%
�$�   "polynomial) lies between:"'
�� E;"  and  ";maxerror;"."'
�<� "This indicates how close our approximation is to it."
�V� ''"Finally, if only need say 16 bit accuracy, note that our poly's coefficients"
�>�   "multiplied by 2^20 & written in hex are as follows:"'
�� i%=n% � 0 � -1
� v = p(i%)
� � v<0 s$="-":v=-v � s$="+"
� v2 = v*1048576+0.5
�1 � "x^";i%;"  *",s$;�"00000000"+�~v2,8);�27);
� � i%>0 � "+" � �
�
�
00000000  0d 00 0a 0b f4 20 3e 43  6f 65 66 0d 00 14 40 f4  |..... >Coef...@.|
00000010  20 43 61 6c 63 75 6c 61  74 65 20 70 6f 6c 79 6e  | Calculate polyn|
00000020  6f 6d 69 61 6c 20 61 70  70 72 6f 78 69 6d 61 74  |omial approximat|
00000030  69 6f 6e 73 20 74 6f 20  61 72 62 69 74 72 61 72  |ions to arbitrar|
00000040  79 20 66 75 6e 63 74 69  6f 6e 73 0d 00 1e 43 f4  |y functions...C.|
00000050  20 69 6e 20 72 61 6e 67  65 20 5b 2d 31 2c 2b 31  | in range [-1,+1|
00000060  5d 2c 20 75 73 69 6e 67  20 43 68 65 62 79 73 68  |], using Chebysh|
00000070  65 76 20 46 6f 72 63 65  64 20 4f 73 63 69 6c 6c  |ev Forced Oscill|
00000080  61 74 69 6f 6e 20 6f 66  20 45 72 72 6f 72 0d 00  |ation of Error..|
00000090  28 2e f4 20 6d 65 74 68  6f 64 20 74 6f 20 64 65  |(.. method to de|
000000a0  72 69 76 65 20 61 20 6e  65 61 72 20 6d 69 6e 69  |rive a near mini|
000000b0  6d 61 78 20 73 6f 6c 75  74 69 6f 6e 0d 00 32 05  |max solution..2.|
000000c0  f4 0d 00 3c 07 eb 20 eb  0d 00 46 5a e8 20 22 50  |...<.. ...FZ. "P|
000000d0  6c 65 61 73 65 20 65 6e  74 65 72 20 66 75 6e 63  |lease enter func|
000000e0  74 69 6f 6e 20 74 6f 20  62 65 20 61 70 70 72 6f  |tion to be appro|
000000f0  78 69 6d 61 74 65 64 20  69 6e 20 72 61 6e 67 65  |ximated in range|
00000100  20 5b 2d 31 2c 2b 31 5d  2c 20 20 65 67 20 53 49  | [-1,+1],  eg SI|
00000110  4e 28 50 49 2f 34 2a 28  78 2b 31 29 29 20 22 3b  |N(PI/4*(x+1)) ";|
00000120  66 24 0d 00 50 3c e8 20  22 50 6c 65 61 73 65 20  |f$..P<. "Please |
00000130  65 6e 74 65 72 20 64 65  67 72 65 65 20 6f 66 20  |enter degree of |
00000140  61 70 70 72 6f 78 69 6d  61 74 69 6f 6e 20 28 31  |approximation (1|
00000150  20 74 6f 20 31 30 30 29  20 22 3b 6e 25 27 0d 00  | to 100) ";n%'..|
00000160  5a 29 de 20 66 28 31 30  31 29 2c 20 63 28 31 30  |Z). f(101), c(10|
00000170  31 29 2c 20 74 28 31 30  30 2c 20 31 30 30 29 2c  |1), t(100, 100),|
00000180  20 70 28 31 30 30 29 0d  00 64 11 e3 20 69 25 3d  | p(100)..d.. i%=|
00000190  30 20 b8 20 6e 25 2b 31  0d 00 6e 17 20 78 20 3d  |0 . n%+1..n. x =|
000001a0  20 9b 28 69 25 2a af 2f  28 6e 25 2b 31 29 29 0d  | .(i%*./(n%+1)).|
000001b0  00 78 11 20 66 28 69 25  29 20 3d 20 a0 20 66 24  |.x. f(i%) = . f$|
000001c0  0d 00 82 05 ed 0d 00 8c  11 e3 20 6a 25 3d 30 20  |.......... j%=0 |
000001d0  b8 20 6e 25 2b 31 0d 00  96 3d 20 e7 20 28 6a 25  |. n%+1...= . (j%|
000001e0  80 31 29 20 73 75 6d 20  3d 20 28 66 28 30 29 2d  |.1) sum = (f(0)-|
000001f0  66 28 6e 25 2b 31 29 29  2f 32 20 8b 20 73 75 6d  |f(n%+1))/2 . sum|
00000200  20 3d 20 28 66 28 30 29  2b 66 28 6e 25 2b 31 29  | = (f(0)+f(n%+1)|
00000210  29 2f 32 0d 00 a0 10 20  e3 20 69 25 3d 31 20 b8  |)/2.... . i%=1 .|
00000220  20 6e 25 0d 00 aa 24 20  20 73 75 6d 20 2b 3d 20  | n%...$  sum += |
00000230  66 28 69 25 29 2a 9b 28  69 25 2a 6a 25 2a af 2f  |f(i%)*.(i%*j%*./|
00000240  28 6e 25 2b 31 29 29 0d  00 b4 06 20 ed 0d 00 be  |(n%+1)).... ....|
00000250  17 20 63 28 6a 25 29 20  3d 20 73 75 6d 2f 28 6e  |. c(j%) = sum/(n|
00000260  25 2b 31 29 0d 00 c8 1c  20 e7 20 6a 25 3c 3e 30  |%+1).... . j%<>0|
00000270  20 63 28 6a 25 29 20 3d  20 32 2a 63 28 6a 25 29  | c(j%) = 2*c(j%)|
00000280  0d 00 d2 05 ed 0d 00 dc  14 45 20 3d 20 94 28 63  |.........E = .(c|
00000290  28 6e 25 2b 31 29 2f 32  29 0d 00 e6 26 f1 20 27  |(n%+1)/2)...&. '|
000002a0  22 50 6f 6c 79 6e 6f 6d  69 61 6c 20 61 70 70 72  |"Polynomial appr|
000002b0  6f 78 69 6d 61 74 69 6f  6e 20 69 73 3a 22 27 0d  |oximation is:"'.|
000002c0  00 f0 0f e3 20 69 25 3d  30 20 b8 20 6e 25 0d 00  |.... i%=0 . n%..|
000002d0  fa 23 20 f1 20 22 54 5b  22 3b 69 25 3b 22 5d 28  |.# . "T[";i%;"](|
000002e0  78 29 20 2a 22 2c 63 28  69 25 29 3b 8a 32 37 29  |x) *",c(i%);.27)|
000002f0  3b 0d 01 04 16 20 e7 20  69 25 3c 6e 25 20 f1 20  |;.... . i%<n% . |
00000300  22 2b 22 20 8b 20 f1 0d  01 0e 05 ed 0d 01 18 1d  |"+" . ..........|
00000310  f1 20 27 22 77 68 65 72  65 20 54 5b 30 5d 28 78  |. '"where T[0](x|
00000320  29 20 20 20 3d 20 31 3b  22 0d 01 22 1d f1 20 20  |)   = 1;".."..  |
00000330  22 20 20 20 20 20 20 54  5b 31 5d 28 78 29 20 20  |"      T[1](x)  |
00000340  20 3d 20 78 3b 22 0d 01  2c 39 f1 20 20 22 61 6e  | = x;"..,9.  "an|
00000350  64 20 20 20 54 5b 6e 2b  31 5d 28 78 29 20 3d 20  |d   T[n+1](x) = |
00000360  32 78 54 5b 6e 5d 28 78  29 2d 54 5b 6e 2d 31 5d  |2xT[n](x)-T[n-1]|
00000370  28 78 29 2c 20 66 6f 72  20 6e 3e 3d 31 22 27 0d  |(x), for n>=1"'.|
00000380  01 36 24 f4 20 4e 6f 77  20 77 65 20 6e 65 65 64  |.6$. Now we need|
00000390  20 74 6f 20 67 65 6e 20  74 68 65 20 54 20 70 6f  | to gen the T po|
000003a0  6c 79 73 0d 01 40 3d f4  20 6e 6f 74 65 20 77 65  |lys..@=. note we|
000003b0  20 75 73 65 20 74 28 69  2c 6a 29 20 74 6f 20 68  | use t(i,j) to h|
000003c0  6f 6c 64 20 74 68 65 20  63 6f 65 66 20 6f 66 20  |old the coef of |
000003d0  78 5e 6a 20 69 6e 20 70  6f 6c 79 20 54 5b 69 5d  |x^j in poly T[i]|
000003e0  0d 01 4a 0c 74 28 30 2c  30 29 3d 31 0d 01 54 0c  |..J.t(0,0)=1..T.|
000003f0  74 28 31 2c 31 29 3d 31  0d 01 5e 08 69 25 3d 31  |t(1,1)=1..^.i%=1|
00000400  0d 01 68 0c c8 95 20 69  25 3c 6e 25 0d 01 72 1b  |..h... i%<n%..r.|
00000410  20 74 28 69 25 2b 31 2c  30 29 20 3d 20 2d 74 28  | t(i%+1,0) = -t(|
00000420  69 25 2d 31 2c 30 29 0d  01 7c 14 20 e3 20 6a 25  |i%-1,0)..|. . j%|
00000430  20 3d 20 31 20 b8 20 69  25 2b 31 0d 01 86 2a 20  | = 1 . i%+1...* |
00000440  20 74 28 69 25 2b 31 2c  6a 25 29 20 3d 20 32 2a  | t(i%+1,j%) = 2*|
00000450  74 28 69 25 2c 6a 25 2d  31 29 2d 74 28 69 25 2d  |t(i%,j%-1)-t(i%-|
00000460  31 2c 6a 25 29 0d 01 90  06 20 ed 0d 01 9a 0a 20  |1,j%).... ..... |
00000470  69 25 2b 3d 31 0d 01 a4  05 ce 0d 01 ae 0f e3 20  |i%+=1.......... |
00000480  69 25 3d 30 20 b8 20 6e  25 0d 01 b8 0c 20 73 75  |i%=0 . n%.... su|
00000490  6d 20 3d 20 30 0d 01 c2  11 20 e3 20 6a 25 3d 69  |m = 0.... . j%=i|
000004a0  25 20 b8 20 6e 25 0d 01  cc 1c 20 20 73 75 6d 20  |% . n%....  sum |
000004b0  2b 3d 20 74 28 6a 25 2c  20 69 25 29 2a 63 28 6a  |+= t(j%, i%)*c(j|
000004c0  25 29 0d 01 d6 06 20 ed  0d 01 e0 10 20 70 28 69  |%).... ..... p(i|
000004d0  25 29 20 3d 20 73 75 6d  0d 01 ea 05 ed 0d 01 f4  |%) = sum........|
000004e0  24 f1 20 27 22 57 68 69  63 68 20 6d 61 79 20 61  |$. '"Which may a|
000004f0  6c 73 6f 20 62 65 20 77  72 69 74 74 65 6e 3a 22  |lso be written:"|
00000500  27 0d 01 fe 0f e3 20 69  25 3d 30 20 b8 20 6e 25  |'..... i%=0 . n%|
00000510  0d 02 08 20 20 f1 20 22  78 5e 22 3b 69 25 3b 22  |...  . "x^";i%;"|
00000520  20 20 2a 22 2c 70 28 69  25 29 3b 8a 32 37 29 3b  |  *",p(i%);.27);|
00000530  0d 02 12 16 20 e7 20 69  25 3c 6e 25 20 f1 20 22  |.... . i%<n% . "|
00000540  2b 22 20 8b 20 f1 0d 02  1c 05 ed 0d 02 26 10 6d  |+" . ........&.m|
00000550  61 78 65 72 72 6f 72 20  3d 20 30 0d 02 30 16 e3  |axerror = 0..0..|
00000560  20 78 3d 2d 31 20 b8 20  31 20 88 20 30 2e 30 30  | x=-1 . 1 . 0.00|
00000570  31 0d 02 3a 0e 20 76 20  3d 20 70 28 6e 25 29 0d  |1..:. v = p(n%).|
00000580  02 44 17 20 e3 20 69 25  3d 6e 25 2d 31 20 b8 20  |.D. . i%=n%-1 . |
00000590  30 20 88 20 2d 31 0d 02  4e 13 20 20 76 20 3d 20  |0 . -1..N.  v = |
000005a0  76 2a 78 2b 70 28 69 25  29 0d 02 58 06 20 ed 0d  |v*x+p(i%)..X. ..|
000005b0  02 62 14 20 65 20 3d 20  94 28 76 20 2d 20 a0 20  |.b. e = .(v - . |
000005c0  66 24 29 0d 02 6c 1c 20  e7 20 65 3e 6d 61 78 65  |f$)..l. . e>maxe|
000005d0  72 72 6f 72 20 6d 61 78  65 72 72 6f 72 3d 65 0d  |rror maxerror=e.|
000005e0  02 76 05 ed 0d 02 80 45  f1 20 27 27 22 46 6f 72  |.v.....E. ''"For|
000005f0  20 74 68 69 73 20 61 70  70 72 6f 78 69 6d 61 74  | this approximat|
00000600  69 6f 6e 20 61 6e 20 75  70 70 65 72 20 62 6f 75  |ion an upper bou|
00000610  6e 64 20 6f 6e 20 65 72  72 6f 72 20 69 73 20 22  |nd on error is "|
00000620  3b 6d 61 78 65 72 72 6f  72 0d 02 8a 59 f1 20 20  |;maxerror...Y.  |
00000630  20 22 41 6c 73 6f 20 77  65 20 6e 6f 77 20 6b 6e  | "Also we now kn|
00000640  6f 77 20 6d 69 6e 69 6d  61 78 20 65 72 72 6f 72  |ow minimax error|
00000650  20 70 20 28 74 68 65 20  65 72 72 6f 72 20 6d 61  | p (the error ma|
00000660  64 65 20 62 79 20 74 68  65 20 62 65 73 74 20 70  |de by the best p|
00000670  6f 73 73 69 62 6c 65 20  6f 72 64 65 72 20 22 3b  |ossible order ";|
00000680  6e 25 0d 02 94 24 f1 20  20 20 22 70 6f 6c 79 6e  |n%...$.   "polyn|
00000690  6f 6d 69 61 6c 29 20 6c  69 65 73 20 62 65 74 77  |omial) lies betw|
000006a0  65 65 6e 3a 22 27 0d 02  9e 1f f1 20 45 3b 22 20  |een:"'..... E;" |
000006b0  20 61 6e 64 20 20 22 3b  6d 61 78 65 72 72 6f 72  | and  ";maxerror|
000006c0  3b 22 2e 22 27 0d 02 a8  3c f1 20 22 54 68 69 73  |;"."'...<. "This|
000006d0  20 69 6e 64 69 63 61 74  65 73 20 68 6f 77 20 63  | indicates how c|
000006e0  6c 6f 73 65 20 6f 75 72  20 61 70 70 72 6f 78 69  |lose our approxi|
000006f0  6d 61 74 69 6f 6e 20 69  73 20 74 6f 20 69 74 2e  |mation is to it.|
00000700  22 0d 02 b2 56 f1 20 27  27 22 46 69 6e 61 6c 6c  |"...V. ''"Finall|
00000710  79 2c 20 69 66 20 6f 6e  6c 79 20 6e 65 65 64 20  |y, if only need |
00000720  73 61 79 20 31 36 20 62  69 74 20 61 63 63 75 72  |say 16 bit accur|
00000730  61 63 79 2c 20 6e 6f 74  65 20 74 68 61 74 20 6f  |acy, note that o|
00000740  75 72 20 70 6f 6c 79 27  73 20 63 6f 65 66 66 69  |ur poly's coeffi|
00000750  63 69 65 6e 74 73 22 0d  02 bc 3e f1 20 20 20 22  |cients"...>.   "|
00000760  6d 75 6c 74 69 70 6c 69  65 64 20 62 79 20 32 5e  |multiplied by 2^|
00000770  32 30 20 26 20 77 72 69  74 74 65 6e 20 69 6e 20  |20 & written in |
00000780  68 65 78 20 61 72 65 20  61 73 20 66 6f 6c 6c 6f  |hex are as follo|
00000790  77 73 3a 22 27 0d 02 c6  14 e3 20 69 25 3d 6e 25  |ws:"'..... i%=n%|
000007a0  20 b8 20 30 20 88 20 2d  31 0d 02 d0 0e 20 76 20  | . 0 . -1.... v |
000007b0  3d 20 70 28 69 25 29 0d  02 da 1f 20 e7 20 76 3c  |= p(i%).... . v<|
000007c0  30 20 73 24 3d 22 2d 22  3a 76 3d 2d 76 20 8b 20  |0 s$="-":v=-v . |
000007d0  73 24 3d 22 2b 22 0d 02  e4 17 20 76 32 20 3d 20  |s$="+".... v2 = |
000007e0  76 2a 31 30 34 38 35 37  36 2b 30 2e 35 0d 02 ee  |v*1048576+0.5...|
000007f0  31 20 f1 20 22 78 5e 22  3b 69 25 3b 22 20 20 2a  |1 . "x^";i%;"  *|
00000800  22 2c 73 24 3b c2 22 30  30 30 30 30 30 30 30 22  |",s$;."00000000"|
00000810  2b c3 7e 76 32 2c 38 29  3b 8a 32 37 29 3b 0d 02  |+.~v2,8);.27);..|
00000820  f8 15 20 e7 20 69 25 3e  30 20 f1 20 22 2b 22 20  |.. . i%>0 . "+" |
00000830  8b 20 f1 0d 03 02 05 ed  0d ff                    |. ........|
0000083a