I have built a difference amplifier and I have a pretty linear characteristic with a coefficient of determination of 0,9999478 (by a linear regression). Is it enough to say it's "pretty linear" (I don't want to make a bold claim but it's not very important in my paper) ?

The difference amplifier is for measurement and control in a DC DC converter. The amplification is approximately 100 times the input value, and the values should be used for stability analysis.

  • 4
    \$\begingroup\$ Well we need more information about your paper. How exactly would the non-linearity affect the conclusion? What values did other researchers from your field accept as linear enough and in what circumstances? \$\endgroup\$ – AndrejaKo Jan 21 '13 at 1:04
  • 1
    \$\begingroup\$ When it becomes a characteristic "line" ;-) \$\endgroup\$ – Oli Glaser Jan 21 '13 at 3:12
  • \$\begingroup\$ It becomes linear when the marketing people get involved. \$\endgroup\$ – Connor Wolf Jan 23 '13 at 4:51

You're doing it backwards. You spec how linear you need it to be, and then design to it.

You're also not measuring linearity the way most folks do. If x is your input, and y is your output, nonlinearity is maximum HORIZONTAL distance from the full scale best fit line as a percentage of full scale.

  • \$\begingroup\$ That is, I deed, how I did it, but I'm just saying "seems ok to me". That's why I wanted to know I there are standards for this. \$\endgroup\$ – Yves Jan 23 '13 at 17:26
  • \$\begingroup\$ Not that I know of. There are standards for how to measure nonlinearity in various device classes, like IEC 60904-10 for photovoltaics, but I don't know of any defining a threshold. \$\endgroup\$ – Scott Seidman Jan 23 '13 at 17:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.