Post by thread:Floating Point Math Routine (Pow) for PIC16

Yeah, but not a POW floating point routine for PIC16, only for PIC17. I really need for PIC16 family. I had not seen any POW floating point routine on the Microchip web site for PIC16... Thanks.... []4s Jeferson Martin Larry Bradley <spam_OUTlhbradleyTakeThisOuTIGS.NET> escreveu: {Quote hidden}> There are floating point routines on the Microchip web site. > > At 09:52 AM 1/9/2003 +0000, you wrote: > >Hello, > > > >Please, does anyone know where i can find a floating point > >routine in assembly for PIC16 family ? I need the POW routine > >with a precision of 32 bits. > > > >Thanks. > > > >[]4s > >Jeferson Martin > > > >-- > >http://www.piclist.com hint: The PICList is archived three different > >ways. See http://www.piclist.com/#archives for details. > > Larry Bradley > Orleans (Ottawa), Ontario, CANADA > > -- > http://www.piclist.com hint: The PICList is archived three different > ways. See http://www.piclist.com/#archives for details. > -- -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

There are two appnotes. One is the basic add, subtract stuff, the other (AN660) is for the power, log, etc. It is a lot of code - the power function requires some of the others, as I recall. I needed a power function for the system I'm working on - The exponent in my case is a constant (y=x^0.386), so I fit a 3rd degree polynomial to the curve, and used that and the standard FP routines. If the exponent is a variable, of course, you can't do that. Larry At 10:31 AM 1/9/2003 +0000, you wrote: {Quote hidden}>Yeah, but not a POW floating point routine for PIC16, only >for PIC17. I really need for PIC16 family. I had not seen >any POW floating point routine on the Microchip web site >for PIC16... > >Thanks.... > >[]4s >Jeferson Martin > > >Larry Bradley <.....lhbradleyKILLspam@spam@IGS.NET> escreveu: > > > There are floating point routines on the Microchip web site. > > > > At 09:52 AM 1/9/2003 +0000, you wrote: > > >Hello, > > > > > >Please, does anyone know where i can find a floating point > > >routine in assembly for PIC16 family ? I need the POW routine > > >with a precision of 32 bits. > > > > > >Thanks. > > > > > >[]4s > > >Jeferson Martin > > > > > >-- > > >http://www.piclist.com hint: The PICList is archived three different > > >ways. See http://www.piclist.com/#archives for details. > > > > Larry Bradley > > Orleans (Ottawa), Ontario, CANADA > > > > -- > > http://www.piclist.com hint: The PICList is archived three different > > ways. See http://www.piclist.com/#archives for details. > > > > > >-- > >-- >http://www.piclist.com hint: The PICList is archived three different >ways. See http://www.piclist.com/#archives for details. Larry Bradley Orleans (Ottawa), Ontario, CANADA -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Hello Larry, it4s my case. I need to calculate the follow eq: (p/p1)^0.190255, p and p1 are variables. Can you explain how i can do it ? I need to use PIC16 due to PIC17 be very large. I have already downloaded the stadard FP routines (+,-,...) from Microchip site, but i don4t know how i can use it to calculate the pow function. Thanks a lot! []4s Jeferson Martin Larry Bradley <lhbradleyKILLspamIGS.NET> escreveu: {Quote hidden}> There are two appnotes. One is the basic add, subtract stuff, the other > (AN660) is for the power, log, etc. It is a lot of code - the power > function requires some of the others, as I recall. > > I needed a power function for the system I'm working on - The exponent in > my case is a constant (y=x^0.386), so I fit a 3rd degree polynomial to the > curve, and used that and the standard FP routines. If the exponent is a > variable, of course, you can't do that. > > Larry > > At 10:31 AM 1/9/2003 +0000, you wrote: > >Yeah, but not a POW floating point routine for PIC16, only > >for PIC17. I really need for PIC16 family. I had not seen > >any POW floating point routine on the Microchip web site > >for PIC16... > > > >Thanks.... > > > >[]4s > >Jeferson Martin > > > > > >Larry Bradley <.....lhbradleyKILLspam.....IGS.NET> escreveu: > > > > > There are floating point routines on the Microchip web site. > > > > > > At 09:52 AM 1/9/2003 +0000, you wrote: > > > >Hello, > > > > > > > >Please, does anyone know where i can find a floating point > > > >routine in assembly for PIC16 family ? I need the POW routine > > > >with a precision of 32 bits. > > > > > > > >Thanks. > > > > > > > >[]4s > > > >Jeferson Martin > > > > > > > >-- > > > >http://www.piclist.com hint: The PICList is archived three different > > > >ways. See http://www.piclist.com/#archives for details. > > > > > > Larry Bradley > > > Orleans (Ottawa), Ontario, CANADA > > > > > > -- > > > http://www.piclist.com hint: The PICList is archived three different > > > ways. See http://www.piclist.com/#archives for details. > > > > > > > > > > >-- > > > >-- > >http://www.piclist.com hint: The PICList is archived three different > >ways. See http://www.piclist.com/#archives for details. > > Larry Bradley > Orleans (Ottawa), Ontario, CANADA > > -- > http://www.piclist.com hint: The PICList is archived three different > ways. See http://www.piclist.com/#archives for details. > -- -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Scott, i4m trying to do an altimeter based on pressure altitude. So, the eq is: H = 44330,769[1 - (P/P1)^0,190255] Where H is the altitude, P is the air pressure at altitude H at a given and variable sea level pressure P1. Thanks for your help ! []4s Jeferson Martin Scott Dattalo <scottspam_OUTDATTALO.COM> escreveu: {Quote hidden}> On Thu, 9 Jan 2003 @spam@jmartinKILLspamCOMP.UFSCAR.BR wrote: > > > Hello Larry, it4s my case. I need to calculate the follow eq: > > (p/p1)^0.190255, p and p1 are variables. Can you explain how i can > > do it ? I need to use PIC16 due to PIC17 be very large. I have > > already downloaded the stadard FP routines (+,-,...) from Microchip > > site, but i don4t know how i can use it to calculate the pow function. > > What's the dynamic range of p and p1? (i.e. are they 8-bit signed, 32-bit > unsigned?) > > Scott > > -- > http://www.piclist.com hint: The PICList is archived three different > ways. See http://www.piclist.com/#archives for details. > -- -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

On Thu, 9 Jan 2003 KILLspamjmartinKILLspamCOMP.UFSCAR.BR wrote: > Scott, i4m trying to do an altimeter based on pressure altitude. > > So, the eq is: > > H = 44330,769[1 - (P/P1)^0,190255] > Where H is the altitude, P is the air pressure at altitude H at > a given and variable sea level pressure P1. That doesn't answer the question. Again, what are dynamic ranges of P and P1? Do these come from an 8-bit A/D converter or what? You originally asked for 32-bit precision, but I doubt you need it. It'd probably make sense in a PIC to re-write this equation like: H = A * ( 1 + P^B * (1/P1)^B) Use a lookup table to compute f(x) = x^B and then perform the simple arithmetic computation that follows. Of course, you'll need to scale the 1/P1 in a range suitable for the table. But I can't tell what the scale factors should be without first knowing the dynamic range of your variables. Scott -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Scott: p1 and p ranges: 100.000 hPa to 1200.000 hPa ok ?? Tanks!! []4s Jeferson Martin Scott Dattalo <RemoveMEscottTakeThisOuTDATTALO.COM> escreveu: {Quote hidden}> On Thu, 9 Jan 2003 spamBeGonejmartinspamBeGoneCOMP.UFSCAR.BR wrote: > > > Scott, i4m trying to do an altimeter based on pressure altitude. > > > > So, the eq is: > > > > H = 44330,769[1 - (P/P1)^0,190255] > > Where H is the altitude, P is the air pressure at altitude H at > > a given and variable sea level pressure P1. > > That doesn't answer the question. Again, what are dynamic ranges of P and > P1? Do these come from an 8-bit A/D converter or what? You originally > asked for 32-bit precision, but I doubt you need it. > > It'd probably make sense in a PIC to re-write this equation like: > > H = A * ( 1 + P^B * (1/P1)^B) > > Use a lookup table to compute f(x) = x^B and then perform the simple > arithmetic computation that follows. Of course, you'll need to scale the > 1/P1 in a range suitable for the table. But I can't tell what the scale > factors should be without first knowing the dynamic range of your > variables. > > Scott > > -- > http://www.piclist.com hint: The PICList is archived three different > ways. See http://www.piclist.com/#archives for details. > -- -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

I doubt if the sea level pressure (P1?) varies that much - 800 to 1200, perhaps. Even 800 suggests a hurricane!!! What I did with my system was to use the equation (in my case y=x^0.386 or whatever) to generate a set of values of y vs x over the expected range of x (in my case, x was from 1 to 100). Then I used a program I have to fit a polynomial to the data. I ended up having to break the data up into two ranges, and fit a 3rd degree polynomial to each range (1<=x<=12, and x>12). I got a fit to about 0.1% accuracy. I then used the floating point package in my program to evaluate the polynomial as I read X from the A/D converter. I'll use your equation and try to generate a polynomial or two for you. Read the AN660 app note - you might be able to use it if you have a PIC with enough program memory. Larry At 03:01 PM 1/9/2003 +0000, you wrote: {Quote hidden}>Scott: > >p1 and p ranges: 100.000 hPa to 1200.000 hPa > >ok ?? > >Tanks!! > >[]4s >Jeferson Martin Larry Bradley Orleans (Ottawa), Ontario, CANADA -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

Hello Mike, Look at the top of page 7 of Microchip AN660: "Currently, this function (Pow) is only supported on the PIC17CXXX". []4s Jeferson Martin "Jon M (Mike) Jones" <TakeThisOuTpmjones3EraseMEspam_OUTATTBI.COM> escreveu: {Quote hidden}> I may be missing something but the Appendix (A&B) of Microchip AN660 claims > to have floating point code for both the 17C and the 16C both with the POW24 > and POW32 routines. I can't comment on how good either of these routines are > but after much research they are the only ones I have found. > > I plan on attempting to convert the 17C code for use on the 18F452 soon. I > hope I'm not biting of more than I can chew. > > (New to the list) > Mike Jones > > {Original Message removed}

OK, here are two 3rd degree polynomials to approximate y=x^0.190255, where x is P/P1 in your case. I have assumed that 100<= P<= 1200. P1 is always < P, thus the lowest value for x will be 100/1200 = 0.0833. I generated a table of y=x^.190255 for 0.08<=x<=1.0. As in my case, it was necessary to split the data into two ranges to get a good fit. Thus we get For 0.08<=x<=0.3: y=0.48093+2.12405x-5.51251x^2+6.42179x^3 and for 0.3<=x<=1: y = .612857+.763474x-.569529x^2+.193612x^3 Thus select a poly based on the pressure ratio P/P1 and evaluate it, and then evaluate the rest of the function to find the altitude. Larry At 12:13 PM 1/9/2003 +0000, you wrote: {Quote hidden}>Hello Larry, it4s my case. I need to calculate the follow eq: >(p/p1)^0.190255, p and p1 are variables. Can you explain how i can >do it ? I need to use PIC16 due to PIC17 be very large. I have >already downloaded the stadard FP routines (+,-,...) from Microchip >site, but i don4t know how i can use it to calculate the pow function. > >Thanks a lot! > >[]4s >Jeferson Martin > >Larry Bradley <RemoveMElhbradleyTakeThisOuTIGS.NET> escreveu: > > > There are two appnotes. One is the basic add, subtract stuff, the other > > (AN660) is for the power, log, etc. It is a lot of code - the power > > function requires some of the others, as I recall. > > > > I needed a power function for the system I'm working on - The exponent in > > my case is a constant (y=x^0.386), so I fit a 3rd degree polynomial to the > > curve, and used that and the standard FP routines. If the exponent is a > > variable, of course, you can't do that. > > > > Larry > > > > At 10:31 AM 1/9/2003 +0000, you wrote: > > >Yeah, but not a POW floating point routine for PIC16, only > > >for PIC17. I really need for PIC16 family. I had not seen > > >any POW floating point routine on the Microchip web site > > >for PIC16... > > > > > >Thanks.... > > > > > >[]4s > > >Jeferson Martin > > > > > > > > >Larry Bradley <lhbradleyEraseME.....IGS.NET> escreveu: > > > > > > > There are floating point routines on the Microchip web site. > > > > > > > > At 09:52 AM 1/9/2003 +0000, you wrote: > > > > >Hello, > > > > > > > > > >Please, does anyone know where i can find a floating point > > > > >routine in assembly for PIC16 family ? I need the POW routine > > > > >with a precision of 32 bits. > > > > > > > > > >Thanks. > > > > > > > > > >[]4s > > > > >Jeferson Martin > > > > > > > > > >-- > > > > >http://www.piclist.com hint: The PICList is archived three different > > > > >ways. See http://www.piclist.com/#archives for details. > > > > > > > > Larry Bradley > > > > Orleans (Ottawa), Ontario, CANADA > > > > > > > > -- > > > > http://www.piclist.com hint: The PICList is archived three different > > > > ways. See http://www.piclist.com/#archives for details. > > > > > > > > > > > > > > > >-- > > > > > >-- > > >http://www.piclist.com hint: The PICList is archived three different > > >ways. See http://www.piclist.com/#archives for details. > > > > Larry Bradley > > Orleans (Ottawa), Ontario, CANADA > > > > -- > > http://www.piclist.com hint: The PICList is archived three different > > ways. See http://www.piclist.com/#archives for details. > > > > > >-- > >-- >http://www.piclist.com hint: The PICList is archived three different >ways. See http://www.piclist.com/#archives for details. Larry Bradley Orleans (Ottawa), Ontario, CANADA -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

On Thu, 9 Jan 2003 EraseMEjmartinCOMP.UFSCAR.BR wrote: > Scott: > > p1 and p ranges: 100.000 hPa to 1200.000 hPa Let me try one more time: http://www.its.bldrdoc.gov/fs-1037/dir-013/_1853.htm When we say "dynamic range" in a system such as this we're typically referring to the number of bits in the quantized signal. If you say that the signal is quantized with an 8-bit A/D then I can infer the dynamic range in dB as 20log(256) = 48dB. However, most people will just say the dynamic range is "8-bits". (Assuming all 8-bits are used). So in your system, you're probably obtaining p1 over some range of pressures; e.g. 0 to 100 hPa (this seems a little strange). Also, this will be quantize: p1 = A*d1 + B p1 is the pressure d1 is the digitized number representing pressure A and B are constants. For example, if you're using an 8-bit A/D, and when the A/D gives you 42 that corresponds to p1=0 and when it gives you 168 when p1=100, then you can write: p1 - 0 d1 - 42 -------- = -------- 100 - 0 168 - 42 p1 = 100/126*(d1 - 42) = 50/63*d1 - 100/3 This equation simply tells you how d1, numbers from the A/D, are converted into pressure. In computing the altitude, you're probably going to want to work with either d1 values directly or perhaps with a scaled version. The answer depends on many factors such as how much time do you wish to spend (e.g. you could convert everything to floating point and let the computation take forever to compute) or how precise should the answer be (it doesn't make sense computing altitude to the nearest nanometer!). Scott -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

> p1 and p ranges: 100.000 hPa to 1200.000 hPa You get floating point pressure values out of your pressure sensor? (I guess it's possible, especially if you're reading it from something like a GPS "thing"...) The point is that usually a pressure sensor is connected to an integer A-D converter, and that it may make a lot more sense (performance-wise) to stay in that integer domain as long as possible while you're doing your math, especially since your end result is probably an integer as well (feet or meters of altitude, right?) Why convert your integer pressure reading to a floating point number, plug it into a complex floating point equation, and finally convert back to an interger readout when the PIC would rather not do floating point? That means manipulting the formula quite a lot, but makes things a lot easier for the PIC. BillW -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.