On 12/6/2013 4:27 AM, Helmut Wabnig wrote:
On Fri, 06 Dec 2013 03:57:40 -0600, John S
wrote:

On 12/5/2013 4:38 PM, Helmut Wabnig wrote:
On Thu, 05 Dec 2013 14:42:02 -0600, John S
wrote:

On 12/5/2013 1:49 PM, John S wrote:

Well I did it using an old dirty trick:
adding one fake data point at the end to bend the curve.

http://img560.imageshack.us/img560/1845/yi2.gif

You must then not use the values
which are above the last valid data point.

You may program this formula into an Arduino .-)

w.

Yes, the trend line is very close. But, please perform a calculation on
the 498R value using the equation and let me know if it agrees. If yours
does, then something is wrong here.

Thanks,
John


Helmut -

Please see this. I think Excel is screwed up.

https://imageshack.com/i/mrg919p


Yes, you are right.
Although EXCEL calculated the curve correctly in its own diagram,
it outputs false (rounded) parameters for the curve fitting polynom.
Compare with the Graphmatica plot, they are ident.
Have to find out tomorrow how to get the polynom factors without
rounding errors, if possible.
http://img834.imageshack.us/img834/8177/altv.gif

w.


I found these coefficients at a site called ZunZun dot com:

y = a + bx + cx2 + dx3 + fx4 + gx5

Fitting target of lowest sum of squared absolute error =
7.0645972802275931E-06

a = 3.3214807109861584E-02
b = 7.2539508790925885E-04
c = -3.2830626216867792E-06
d = 7.9094355769986563E-09
f = -9.4574857953126017E-12
g = 4.3679252922923517E-15

Worst case error is .202% at 50.

John


Yes, that's better

neither EXCEL nor OPEN OFFICE can do it correctly.
Excel generates the correct formula, but outputs only truncated or
rounded coefficients.
I do not know how to access the internal correct coefficients
in Excel.

Then I found this site:
http://www.xuru.org/rt/PR.asp#CopyPaste
Which outputs the following:

http://img43.imageshack.us/img43/6194/jmph.jpg


Which gives the correct curve when inserted into EXCEL
with some editing.
=4,368089718E-15*(B4)^5-9,457445532E-12*(B4)^4+0,000000007909410217*(B4)^3-0,000003283056669*(B4)^2+0,0007253920186*(B4)+0,03 321499593
http://img547.imageshack.us/img547/1012/o8io.gif

To improve the curve fitting I suggest to take additional measurements
in the upper range.

w.


Together, I think we got a reasonable answer. Thanks for your
information. I did not know about adding an additional data point to
help the curve. Nice work.