This is a purely fictional example: suppose I want to find the voltage drop across a resistor $R$ where a current $I$ flows. My resistance supplier established a tolerance of 5% for $R$, so I know that the true value of my resistance is somewhere between $[R-0.05R;R+0.05R]$ (let's call $\Delta R=0.05R$). Also the measurement device for the current only gives me values until the second decimal place, that is, the value of current that I measure is somewhere between $[I-0.01;I+0.01]$ ($\Delta I=0.01$). That being said, what is the error/uncertainty in $V=IR$?

One could argue that: $$ \Delta V=\bigg|\frac{\partial V}{\partial I}\bigg|\Delta I+\bigg|\frac{\partial V}{\partial R}\bigg|\Delta R $$

which makes sense, because it gives me a rectangle of possible values for my new variable $V$. But I have also seen (and used extensively) the uncertainty propagation formula:

$$ \Delta V=\sqrt{\bigg(\frac{\partial V}{\partial I}\Delta I\bigg)^2+\bigg(\frac{\partial V}{\partial R}\Delta R\bigg)^2} $$

This also makes some sense, in a way that it gives me an ellipse of possible values for $V$, but which one is more appropriate and in which cases? What is the difference between these two formulas?

The goal is to present the result as $V=(V\pm \Delta V)$.

  • $\begingroup$ Sum of normally distributed random variables. $\endgroup$
    – Farcher
    Commented Nov 17, 2019 at 22:52
  • $\begingroup$ Yes, that helps but only understanding the second expression. In the sense that it gives me standard deviation of my variable V, that is, I will get $V\pm \Delta V$ 68% of the times. But what about the first one. How does it differ? Does the first assume a uniform distribution? $\endgroup$
    – Bidon
    Commented Nov 18, 2019 at 16:41

1 Answer 1


Your choice of an example considering resistance complicates matters which I will mention later.

If the error is the maximum error then using the first equation you are finding an error based on the assumption that both values have a maximum error simultaneously which is not a very likely event.
This means that the error found this way is an over estimate.

For resistor the complication is that a supplier quotes a tolerance as a percentage and usually produces resistors with a series of tolerances, eg $1\%,\,2\%, \,5\%,\,10\%,$ etc.

So if you have some $5\%$ $100\Omega$ resistors it is possible/probable that your resistors are all have values in the range $95\Omega$ to $98\Omega$ and $102\Omega$ to $105\Omega$ because all the resistors with a smaller tolerance have been removed from the batch for sale at a higher price.
It is also possible that your batch of $5\%$ $100\Omega$ resistors contain only resistors in the range $95\Omega$ to $98\Omega$ because all resistor above the nominal value have been "trimmed" during processing to a value closer to the nominal value.

  • $\begingroup$ could we say that the second equation assumes a normal distribution of the values of $I$ and $R$ and so it gives me standard deviation of my variable V, that is, I will get V±ΔV 68% of the times, whereas the first one assumes a uniform distribution for both variables with some maximum deviation? $\endgroup$
    – Bidon
    Commented Nov 18, 2019 at 17:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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