 Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me. |
Forum Index : Microcontroller and PC projects : DOS MMBasic Feature Request- Single Precision Math
| Author | Message | ||||
jwettroth Regular Member  Joined: 02/08/2011 Location: United StatesPosts: 70 |
I love DOS MMBasic and use it a lot- the ability to access serial ports is just great. The ease of putting together little calculation programs quickly is awesome. Degrading the accuracy of the floating point may seem like a strange request- I'm not trying to save memory, I am trying to develop a program on the PC and get an idea of how it will work when ported down to a Micromite. I'm building a project that plays bird calls from 40 minutes before sunset to sunset. For the prototype, I calculated the times off line by inputting my latitude and longitude from the web. I then had an array of alarm times for each day during the play season- about 8 weeks-t worked fine but wasn't very elegant. Over the winter, I did a little research and wrote a DOS MMBASIC app that will do all the trig calculations given your position. It works well and agrees with the my web resources. Its not very polished but I'd be happy to upload it. I'm planning to port all of this back to a Micromite II for the final system. My DOS program and my Micromite program don't agree because the mite uses single precision. Would it be possible to have a data type or an option in DOS MMBASIC to use single precision floating point like the mites. Perhaps also, maybe there is a work around that I'm not seeing. I was thinking about doing something like rounding the mantissa by scaling or similar. Any assistance appreciated. Regards All John Wettroth |
||||
| matherp Guru  Joined: 11/12/2012 Location: United KingdomPosts: 8570 |
If you download the source from MMBASIC.COM and the free WATCOM C compiler you can rebuild the DOS version for single precision. It is just one change in the file configuration.h. Geoff has provided a batch file to do the build and it really is very easy. #define MMFLOAT double // precision of all floating point operations #define STR_SIG_DIGITS 9 // number of significant digits to use when converting MMFLOAT to a string becomes #define MMFLOAT float // precision of all floating point operations #define STR_SIG_DIGITS 5 // number of significant digits to use when converting MMFLOAT to a string |
||||
| jwettroth Regular Member Joined: 02/08/2011 Location: United StatesPosts: 70 |
Thank you- that's a great solution. I've always wanted to delve into compiling the source too. Thanks JW John Wettroth |
||||
CaptainBoing Guru Joined: 07/09/2016 Location: United KingdomPosts: 1985 |
you might want to consider this: http://www.fruitoftheshed.com/MMBasic.Quick-and-Dirty-Daylight-Indicator.ashx it uses key sun up/sundown times for your location (stored in the case statements) then interpolates between the two based on the actual date. It looks quite rough but is surprisingly accurate - to within five minutes or so - so much so you don't notice. I use it for numerous dusk-to-dawn light switches and I never have to bother with the lights - they just come on as twilight starts and go off as twilight ends in the morning. Saves all the headache of a "proper" sunrise/set calculations using julian dates etc. jus' sayin' Edited 2021-04-10 05:19 by CaptainBoing |
||||
| jwettroth Regular Member Joined: 02/08/2011 Location: United StatesPosts: 70 |
CaptainBoing- Thanks I'll check it out. The real calcuation is kind of ridiculous- got the algorithm from the US Naval Observatory. John Wettroth |
||||
TassyJim Guru Joined: 07/08/2011 Location: AustraliaPosts: 5886 |
With most calculations there will be negligible difference between single an double precision. Using this program, dating back to when MMBasic used line numbers. 5 DIM A(2), D(2) 10' Sunrise-Sunset adapted by TassyJim TBS 20 GOSUB 300 30 INPUT "Lat-deg="; B5 32 INPUT "Long-deg="; L5 40 INPUT "Time zone-hrs="; H 50 L5 = L5 / 360: Z0 = -H / 24 60 GOSUB 1170: T = (J - 2451545) + F 70 TT = T / 36525 + 1 ' TT = centuries 80 ' from 1900.0 90 GOSUB 410: T = T + Z0 100 ' 110 ' Get Sun's Position 120 GOSUB 910: A(1) = A5: D(1) = D5 130 T = T + 1 140 GOSUB 910: A(2) = A5: D(2) = D5 150 IF A(2) < A(1) THEN A(2) = A(2) + P2 160 Z1 = DR * 90.833 ' Zenith dist. 170 S = SIN(B5 * DR): C = COS(B5 * DR) 180 Z = COS(Z1): M8 = 0: W8 = 0: PRINT 190 A0 = A(1): D0 = D(1) 200 DA = A(2) - A(1): DD = D(2) - D(1) 210 FOR C0 = 0 TO 23 220 P = (C0 + 1) / 24 230 A2 = A(1) + P * DA: D2 = D(1) + P * DD 240 GOSUB 490 250 A0 = A2: D0 = D2: V0 = V2 260 NEXT 270 GOSUB 820 ' Special msg? 280 END 290 ' 300 ' Constants 320 P1 = 3.14159265 322 P2 = 2 * P1 330 DR = P1 / 180 332 K1 = 15 * DR * 1.0027379 340 Ss$ = "Sunset at " 350 sR$ = "Sunrise at " 360 M1t$ = "No sunrise this date" 370 M2t$ = "No sunset this date" 380 M3t$ = "Sun down all day" 390 M4t$ = "Sun up all day" 400 RETURN 410 ' LST at 0h zone time 420 T0 = T / 36525 430 S = 24110.5 + 8640184.8 * T0 440 S = S + 86636.6 * Z0 + 86400 * L5 450 S = S / 86400: S = S - INT(S) 460 T0 = S * 360 * DR 470 RETURN 480 ' 490 ' Test an hour for an event 500 L0 = T0 + C0 * K1: L2 = L0 + K1 510 H0 = L0 - A0: H2 = L2 - A2 520 H1 = (H2 + H0) / 2 ' Hour angle, 530 D1 = (D2 + D0) / 2 ' declination, 540 ' at half hour 550 IF C0 > 0 THEN 570 560 V0 = S * SIN(D0) + C * COS(D0) * COS(H0) - Z 570 V2 = S * SIN(D2) + C * COS(D2) * COS(H2) - Z 580 IF SGN(V0) = SGN(V2) THEN 800 590 V1 = S * SIN(D1) + C * COS(D1) * COS(H1) - Z 600 Ax = 2 * V2 - 4 * V1 + 2 * V0: B = 4 * V1 - 3 * V0 - V2 610 Dy = B * B - 4 * Ax * V0: IF Dy < 0 THEN 800 620 Dy = SQR(Dy) 630 IF V0 < 0 AND V2 > 0 THEN PRINT sR$; 640 IF V0 < 0 AND V2 > 0 THEN M8 = 1 650 IF V0 > 0 AND V2 < 0 THEN PRINT Ss$; 660 IF V0 > 0 AND V2 < 0 THEN W8 = 1 670 E = (0-B + Dy) / (2 * Ax) 680 IF E > 1 OR E < 0 THEN E = (0-B - Dy) / (2 * Ax) 690 T3 = C0 + E + 1 / 120 ' Round off 700 H3 = INT(T3): M3 = INT((T3 - H3) * 60) 705 Hr$=STR$(H3):Hr$=RIGHT$("0"+Hr$,2):Mn$=STR$(M3):Mn$=RIGHT$(" 0"+Mn$,2) 710 PRINT Hr$;":";Mn$; 720 H7 = H0 + E * (H2 - H0) 730 N7 =0 -COS(D1) * SIN(H7) 740 D7 = C * SIN(D1) - S * COS(D1) * COS(H7) 750 AZ = ATN(N7 / D7) / DR 760 IF D7 < 0 THEN AZ = AZ + 180 770 IF AZ < 0 THEN AZ = AZ + 360 780 IF AZ > 360 THEN AZ = AZ - 360 782 PRINT ", azimuth "; 790 PRINT AZ 800 RETURN 810 ' 820 ' Special-message routine 830 IF M8 = 0 AND W8 = 0 THEN 870 840 IF M8 = 0 THEN PRINT M1t$ 850 IF W8 = 0 THEN PRINT M2t$ 860 GOTO 890 870 IF V2 < 0 THEN PRINT M3t$ 880 IF V2 > 0 THEN PRINT M4t$ 890 RETURN 900 ' 910 ' Fundamental arguments 920 ' (Van Flandern & 930 ' Pulkkinen, 1979) 940 L = .779072 + .00273790931 * T 950 G = .993126 + .0027377785 * T 960 L = L - INT(L): G = G - INT(G) 970 L = L * P2: G = G * P2 980 V = .39785 * SIN(L) 990 V = V - .01 * SIN(L - G) 1000 V = V + .00333 * SIN(L + G) 1010 V = V - .00021 * TT * SIN(L) 1020 U = 1 - .03349 * COS(G) 1030 U = U - .00014 * COS(2 * L) 1040 U = U + .00008 * COS(L) 1050 W = -.0001 - .04129 * SIN(2 * L) 1060 W = W + .03211 * SIN(G) 1070 W = W + .00104 * SIN(2 * L - G) 1080 W = W - .00035 * SIN(2 * L + G) 1090 W = W - .00008 * TT * SIN(G) 1100 ' 1110 ' Compute Sun's RA and Dec 1120 S = W / SQR(U - V * V) 1130 A5 = L + ATN(S / SQR(1 - S * S)) 1140 S = V / SQR(U): D5 = ATN(S / SQR(1 - S * S)) 1150 R5 = 1.00021 * SQR(U) 1160 RETURN 1165 ' 1170 ' Calendar --> JD 1180 INPUT "Year= "; Y 1183 INPUT "Month= "; M 1186 INPUT "Day= "; Dy 1190 G = 1: IF Y < 1583 THEN G = 0 1200 D1 = INT(Dy): F = Dy - D1 - .5 1210 J =0 -INT(7 * (INT((M + 9) / 12) + Y) / 4) 1220 IF G = 0 THEN 1260 1230 S = SGN(M - 9): Ax = ABS(M - 9) 1240 J3 = INT(Y + S * INT(Ax / 7)) 1250 J3 =0 -INT((INT(J3 / 100) + 1) * 3 / 4) 1260 J = J + INT(275 * M / 9) + D1 + G * J3 1270 J = J + 1721027 + 2 * G + 367 * Y 1280 IF F >= 0 THEN 1300 1290 F = F + 1: J = J - 1 1300 RETURN 1310 ' 1320 ' This program by Roger W. Sinnott calculates the times of sunrise 1330 ' and sunset on any date, accurate to the minute within several 1340 ' centuries of the present. It correctly describes what happens in the 1350 ' arctic and antarctic regions, where the Sun may not rise or set on 1360 ' a given date. Enter north latitudes positive, west longitudes 1370 ' negative. For the time zone, enter the number of hours west of 1380 ' Greenwich (e.g., -5 for EST, -4 for EDT). The calculation is 1390 ' discussed in Sky & Telescope for August 1994, page 84. The results are: I really must get rid of the line numbers. Jim VK7JH MMedit MMBasic Help |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 8570 |
But not all  Cdeagle's solar eclipse program gives no result at all with his reference example in single precision but works perfectly in double |
||||
| jwettroth Regular Member Joined: 02/08/2011 Location: United StatesPosts: 70 |
Below is my code- heavily borrowed from sources in listing and pretty unpolished. >>>> 'Sunrise/Sunset Algorithm ' john wettroth oct 2020 ' algorithm and theory from Almanac for Computers USNO Washington DC 20392 ' and heavily cribbed from edwilliams.org (great resource) ' Written for MMBASIC for DOS- ' consider early alpha- basic operation verified for my location but ' no real testing done and no edge cases looked at. Proceed at your own risk. ' 'Inputs ' day month year 'latitude longitude outpu't 'zenith- type of sunrise/sunset 'official 90 deg 50' 'civil 96 deg 'nautical 102 deg 'astronomical 108 deg 'Notes ' 1. longitude is positive east ' 2. angles in degrees ' '******************************************** 'constants and inputs- these are hard coded here for testing 'could make them an input statement, etc ' mon%=9 'october day%=16 '16 year%=2020 'lat and log of location (Raleigh, NC)- positive is east, negative west latitude= 35.7796 longitude= -78.6382 'PI/180 convert degrees to radians PR=Pi/180 'BEGIN CODE ' day of year calculation n1=Int(275* mon%/9) n2=Int((mon%+9)/12) n3=(1+Int((year%-4 * Int(year%/4) +2)/2)) n=n1-(n2*n3)+day% - 30 Print "Day of Year for ";Str$(mon%,2)+"/"+Str$(day%,2)+"/"+Right$(Str$(year%,4),2)+" is ";n 'convert our longitude to approximate time lnghour=longitude /15 'rise time is N+ ((6-lnghour)/24) 'seTime at= N+ ((18-lnghour)/24) Print "Tvalue for Sunset =";at 'sun mean anomaly M=(.9856*at)-3.289 Print "Mean Solar Anomaly=";M 'sun true longitude L= M+ (1.916 * Sin(M*PR)) + (.020 * Sin(2*M*PR)) + 282.634 L=L-(360*(L\360)) Print "sun true longitude=";L 'sun right ascension RA=(180/Pi)*Atn(.91764 * Tan(pr*L)) Print "right ascension=";ra 'ra adjusted to l quadrant LQUAD= 90* Int(L/90): Print "LQUAD =";LQUAD RQUAD= 90* Int(RA/90):Print "RQUAD =";RQUAD RA=RA+(LQUAD-RQUAD) Print "RA Adjusted=";RA 'CONVERT RA TO HOURS RA=RA/15 Print "RA IN HOURS";RA 'DECLINATION SDEC= .39782*Sin(L*PR) CDEC= Cos(pr*(ASin(SDEC)/pr)) Print "SINDEC =";SDEC Print "COSDEC =";CDEC 'SUN LOCAL HOUR ANGLE COSH = (Cos(90.833*PR) - (SDEC*Sin(LATITUDE*PR)))/(CDEC*Cos(LATITUDE*PR)) If COSH > 1 Then Print "SUN NEVER RISES AT THIS LOCATION" If COSH < -1 Then Print "SUN NEVER SETS HERE" Print "Cos of Local Hour Angle =";cosh 'convert cosh to hours H=180*ACos(cosh)/Pi Print "Hour Angle in Degrees =";H H=H/15 Print "Hour angle in time =";H 'local mean time T=H+RA-(.06571*at)-6.622 If t>24 Then t=t-24 If t<0 Then T=t+24 Print "SUNSET LOCAL MEAN TIME=";T 'adj to utc and subtract lnghour T=T-LNGHOUR Print "UTC TIME FOR SUNSET";T LT=T-5+1 Print "TIME OF SUNSET IS ";Int(LT);" HOURS AND "; Int(60*(LT-Int(LT))); " MINUTES" End John Wettroth |
||||
 Print this page |
