| View previous topic :: View next topic |
| Author |
Message |
Cogitum
Joined: 22 Mar 2012
Posts: 70
Location: France (Paris)
|
| 18F2480 and interpolation ? |
 Posted: Tue Jun 05, 2012 9:09 am |
 |
|
Hi
Looking to translate the code below for CCS compiler Ver 4.132.
| Code: |
int32 Regime_Sp[10]={500,1000,1500,1750,2000,2500,3000,3500,4000,4500};//Tr/mn
int32 Couple_Sp[10]={60,125,225,240,235,227,215,210,185,160};//N/m
public static double interp1(double X, double[] Xref, double[] Table)
{
int longXref = Xref.Length;
int j = 0; // Index de position en X dans Table
while ((j < longXref - 1) & (X > Xref[j]))
{
j++;
}
double x1, x2, y1, y2, t, resultat;
// Calcul de x1 et y1
x1 = Xref[j];
y1 = Table[j];
// Calcul de y2
if ((j < longXref - 1) & (j > 0))
{
x2 = Xref[j + 1];
y2 = Table[j + 1];
}
else
{
y2 = y1;
x2 = x1;
}
if (x2 == x1)
{
t = 0;
}
else
{
t = (X - x1) / (x2 - x1);
}
resultat = y1 + (y2 - y1) * t;
return resultat;
}
|
Thanks in advance for your help  |
|
 |
FvM
Joined: 27 Aug 2008
Posts: 2337
Location: Germany
|
|
| Posted: Tue Jun 05, 2012 10:05 am |
|
|
If the code is intended for PCH compiler (PIC18), double (64 bit floating point) isn't supported. PIC24 PCD compiler however does.
Xref.Length seems to be a private non-standard C syntax extension. Referring to standard C, it would be necessary to provide the array size in a separate function parameter.
You should better show a complete example application including a function call. Which problems did you observe? |
|
|
Cogitum
Joined: 22 Mar 2012
Posts: 70
Location: France (Paris)
|
| 18F2480 and interpolation next |
| Posted: Tue Jun 05, 2012 11:57 pm |
|
|
HI,
Thanks for your quick answer.
This code is made with C #. 64 Bits are not necessary for my application.
I'm just looking to translate this code for CCS compiler or looking for CCS example using interpolation calculation.
Regards |
|
|
RF_Developer
Joined: 07 Feb 2011
Posts: 839
|
|
| Posted: Wed Jun 06, 2012 1:25 am |
|
|
OK then, go ahead and translate it. Its simple; just linear interpolation: the sort of thing I've done over and over again.
RF Developer |
|
|
FvM
Joined: 27 Aug 2008
Posts: 2337
Location: Germany
|
|
| Posted: Wed Jun 06, 2012 4:00 am |
|
|
| Quote: | This code is made with C #. 64 Bits are not necessary for my application.
I'm just looking to translate this code for CCS compiler or looking for CCS example using interpolation calculation. |
Ah, C#. I'm not familiar with this modern stuff. At first sight, you would write float for double, presumed you want to use floating point arithmetic. And supply length as an additional function parameter. Hopefully, that's it. |
|
|
Cogitum
Joined: 22 Mar 2012
Posts: 70
Location: France (Paris)
|
| 18F2480 and interpolation |
| Posted: Thu Jun 07, 2012 10:41 am |
|
|
Hi,
This code RUN but the result is wrong !
| Code: |
int32 interpolation(int32 X,int32 Xref[10],int32 Table[10])
{
int32 x1, x2, y1, y2, t,t1,t2, resultat;
int longXref = 10;
int j = 0;
while ((j < longXref - 1) && (X > Xref[j]))
{
j++;
}
// Calcul de x1 et y1
x1 = Xref[j];
y1 = Table[j];
// Calcul de y2
if ((j < longXref - 1) && (j > 0))
{
x2 = Xref[j + 1];
y2 = Table[j + 1];
}
else
{
y2 = y1;
x2 = x1;
}
if (x2 == x1)
{
t = 0;
}
else
{
t = (X -x1)/(x2 - x1); // normal > -1,2
}
resultat = y1 + (y2 - y1) * t; // Normal 207
return resultat;
}
|
note: X= 1200, x1= 1500, X2= 1750, Y1= 225, Y2= 240
resultat = 10693 !! manual calculation is 207.
How to solve this problem with CCS ?
Thanks in advance |
|
|
gaugeguy
Joined: 05 Apr 2011
Posts: 303
|
|
| Posted: Thu Jun 07, 2012 12:01 pm |
|
|
| An unsigned int can only hold positive integers. No negatives and no fractions. |
|
|
FvM
Joined: 27 Aug 2008
Posts: 2337
Location: Germany
|
|
| Posted: Thu Jun 07, 2012 2:34 pm |
|
|
| I previously suggested to use float data type for the calculations. With integer, you have the problem, that t is truncated to an integer value, which can't give a correct result. Another problem, already mentioned by gaugeguy, in contrast to the C standard, CCS C is understanding int as unsigned int, you need to write signed int explicitely if you mean it. |
|
|
Mike Walne
Joined: 19 Feb 2004
Posts: 1785
Location: Boston Spa UK
|
|
| Posted: Thu Jun 07, 2012 4:23 pm |
|
|
Three issues:-
(1) You're extrapolating not interpolating. You example X is outside the range of x1 to x2.
(2) As already noted CCS treats ints as unsigned. This is causing most of your error. X is outside the range x1 to x2 on the negative side. Therefore (X -x1) wraps round to a huge positive value instead of a small negative one.
(3) There's no need to move to floats. Simply change the order in which you perform the calculations. Again as already noted you are incurring a massive rounding off error with your calculation of t. Perform the divisions last. As it stands you're multiplying your rounding off error by (y2-y1).
Mike |
|
|
FvM
Joined: 27 Aug 2008
Posts: 2337
Location: Germany
|
|
| Posted: Thu Jun 07, 2012 10:51 pm |
|
|
| Quote: | | You're extrapolating not interpolating. |
Yes. The algorithm should work for both. The problem is caused because the code selects an index j with x < x1 and x <x 2,
but it should select for x1 < x < x2.
| Quote: | | There's no need to move to floats. |
If not float, fixed point calculation has to be used. Although the signed/unsigned problem causes
most of the error in your example, the code as shown is only meaningful for float. |
|
|
|
FAQ 
Search
Register
Profile 
Log in to check your private messages 
Log in
