AndersI : I agree with you, ISO says Minimax...
But L2 seems to be more "accurate" than minimax.
The most problem of Chebychev minimization is the start point of the calculation.
You can see it on a curve like this :
The result depends on entered parameters, so it can be a problem.
I made some tests here with a bad part (a poorly crafted part, with big defects that I ask to the machinist to realize), and it seems that L2 is more stable than minimax...
ISO5459 is close to accept least squares on datums, 50 years after the first CMM (Johansson, Zeiss or DEA.. ? I know your answer !), maybe ISO world will accept L2 soon ?
AndersI : I agree with you, ISO says Minimax...
But L2 seems to be more "accurate" than minimax.
The most problem of Chebychev minimization is the start point of the calculation.
You can see it on a curve like this :
The result depends on entered parameters, so it can be a problem.
I made some tests here with a bad part (a poorly crafted part, with big defects that I ask to the machinist to realize), and it seems that L2 is more stable than minimax...
ISO5459 is close to accept least squares on datums, 50 years after the first CMM (Johansson, Zeiss or DEA.. ? I know your answer !), maybe ISO world will accept L2 soon ?
