Balanced equations
Title seems to be bad ... but look...
We have already dealt with microprocessors and microcontrollers assembler programming, and we have found some situations hard to solve, because there were few avalibale programming tools. Nowadays the microcontrollers are programming in languages like “C”, Basic, that already has important tools of programming.
We remember the implementation of floating point math routines (in Assembler) for microcontrollers. The objective of these has been to obtainn an appropiated precision in some calculations reported by equipment, with no need of matrix calculations or complex series. At last, there always is a way to solve complex equations by simple methods of : - Approach, - Limitations, - Ranges, - Tendencies, - Error Control In order to mention one excellent characteristics of our company (we have solved many things with few software tools) is to interpret the objective of each one of the variables locked up in a function correcting, limiting or excluding the same ones if it were the case. Here we have an interesting example resolved:
For meter calibration is very important to have the explession that define to KFactot. That is to say, which Kfactor has the meter in low flow and high flow. In order to obtain each one of them, 3 samples will be taken in each flow, to calculate the averages of them and thus to define the Kfactor in low flow and high flow. 6 measurements (runnings) will be made, and the final result will be a straight line. However, if instead of having a Kfactor in low and high flow we obtain a Kfactor in low, middle and high flow, the final result will be a curve, and not a straight line. This result is closer than to reality, and therefore is more precise. But now, would we have to make 9 samples?, 3 in low 3 in middle and 3 in high flow ? … No! We can generate a curve of second degree, by means of the minimum squares, in which, all the points involved in the creation of the curve, have the same "weight", that is to say, that a sample in low flow will influence in the position by where it passes in high flow. Briefly, all the samples are useful, and the uncertainty does not change. Anyway, always it will be better to have many samples than few ones
Returning to the subject, applying a little reasoning on the problematic one, if with the same amount of data an expression of second degree can be obtained, why to use a straight line? … In this and other cases, as much within the programming of a microcontroller, like within the preparation of a list “Excel”, the processes to verify why of each variable, to limit it, to exclude it or to replace them, is essential to be able to obtain an equation, that not although being the ideal one, Are Balanced.