Saved

Press the star next to a tool's name to save it.

All my tools

Calibration Curve Calculator

Use this calibration curve calculator to fit a least-squares line to your standards, get R² and the scatter, and turn the signal of an unknown sample into a concentration.

Restored your last values ·

Enter the concentration and the instrument signal of each standard, one pair per line. The calculator fits the least-squares line, gives R² and the scatter, and turns the signal of an unknown sample into a concentration.

Everything is worked out in your browser. Nothing you type is sent anywhere.

–

–

Enter your standards, then press Calculate.

A calibration curve calculator fits a straight line through the signals measured for standards of known concentration, then uses that line to find the concentration of an unknown sample from its signal. Enter your standards, add the signal of the unknown, and it gives the slope, the intercept, R², the scatter of the points and the concentration with its standard error.

How to use the calibration curve calculator

  1. Enter each standard on its own line: concentration first, then the signal.
  2. Optionally enter the signal of the unknown and the number of replicates.
  3. Press Calculate.

What a calibration curve is

A calibration curve is a method for determining the value of an unknown from known values by drawing a line, and it is widely used in analytical chemistry [1]. The page writes the line as y = mx + y₀, where the slope m is the sensitivity of the method [1]. You measure the signal of several standards, fit the line, then read the unknown off it. The page gives the interpolation as a graphical step, and the inverse formula x = (y − b) ÷ m is our own rearrangement of the line.

Least-squares slope and intercept

The calculator uses the ordinary least-squares line. The page on simple linear regression gives the slope as the sum of (x − x̄)(y − ȳ) divided by the sum of (x − x̄)², and the intercept as ȳ − slope × x̄ [2]. These choices make the sum of the squared vertical distances as small as possible. The page writes the line as α + βx, and the calibration page as mx + b, but the formulas are the same.

R² and the scatter

The coefficient of determination equals the square of the correlation coefficient when the model is linear with one independent variable [2]. A value near 1 means that the points lie close to the line. The scatter of the standards is the standard deviation of the residuals, s_y, the square root of the sum of squared residuals divided by n − 2 [1]. This needs at least three standards, because two points always fit a line exactly.

The unknown and its error

Given the signal y of the unknown, the concentration is x = (y − b) ÷ m. The page gives the standard error of that result as s_x = (s_y ÷ |m|) × the square root of 1/n + 1/k + (y − ȳ)² ÷ (m² Σ(xᵢ − x̄)²), where k is the number of replicate measurements of the unknown [1]. The calculator reports it, so you can see how uncertain the reading is. The error is smallest when the signal of the unknown lies near the mean of the standards.

A worked example

Take the standards 0 → 0.1, 1 → 1.0, 2 → 2.1, 3 → 2.9 and 4 → 4.1. The line is y = 0.99 x + 0.06, with an R² of about 0.998. A sample with a signal of 2 has a concentration of (2 − 0.06) ÷ 0.99 = 1.96. For a perfect line, such as the points 0 → 1, 1 → 3, 2 → 5 and 3 → 7, the slope is 2, the intercept 1 and R² is exactly 1, and a signal of 7 gives 3. These are our own calculations.

Keeping to the linear range

The Beer–Lambert law, A = εcl, is the classic reason why a plot of absorbance against concentration is a straight line [3]. The page says that the law tends to break down at very high concentrations, and it gives 0.2 to 0.5 as an ideal range of absorbance for linearity [3]. Choose standards that cover the range of your unknowns and stay inside the linear region. Do not use a point from outside the range of the standards.

Good practice

Use at least five standards spread evenly over the range, include a blank, and measure the unknowns within that range. Plot the residuals to look for a curve, since a high R² can hide a systematic bend. If the standards are curved, fit a smaller range or use a different model. Measure the standards and unknowns under the same conditions on the same day.

Planning a calibration

Prepare the standards by serial dilution from a stock solution, so that errors in the stock affect every point in the same way and the line is shifted rather than bent. Run the standards in random order to avoid mistaking drift in the instrument for a trend in concentration. Include a blank and a standard in the middle of the range as a check, and repeat the run if the check is off by more than your method allows.

Reading the output

The slope tells you the sensitivity: the change in signal per unit of concentration. A larger slope means that small differences in concentration are easier to see. The intercept is the signal at zero concentration and should be close to the blank. If it is large, look for contamination or a baseline offset. The standard error of the unknown grows as the signal moves away from the middle of the standards, which is why it pays to bracket your samples with standards on both sides.

When the line curves

If the residuals show a pattern, such as positive in the middle and negative at the ends, the response is curved. Dilute the samples into the linear range, drop the highest standards or use a curved model. A straight line forced through curved data gives biased results for the unknown.

Common mistakes

  • Extrapolating beyond the standards. Results outside the range are unreliable.
  • Trusting R² alone. Look at the scatter and at the residuals.
  • Putting the signal first. Each line is concentration, then signal.
  • Using only two standards. The scatter cannot be estimated.

What next

Use this often? Add Calibration to your Home screen.

Frequently asked questions

What is a calibration curve calculator?

A tool that fits a line through the signals of known standards and uses it to find the concentration of an unknown from its signal.

How do I find the concentration of an unknown?

Fit y = mx + b to the standards, then use x = (y − b) ÷ m with the signal of the unknown.

What does R² tell me?

How close the points are to the line. A value near 1 means a good linear fit, but check the residuals too.

How many standards do I need?

At least three to estimate the scatter, and preferably five or more spread over the range of your samples.

Can I use the line outside the standards?

No. Results outside the range of the standards are extrapolations and are unreliable.

What went wrong?

We only receive the reason you pick, never what you typed into the tool.

Thanks for your feedback.

More Physics & Science Tools