Download Uncertainty in Measurements: Estimation and Propagation and more Lab Reports Physics in PDF only on Docsity!
APPENDIX A: DEALING WITH UNCERTAINTY
1. OVERVIEW
- An uncertainty is always a positive number δ x > 0.
- If the uncertainty of x is 5%, then δ x = .05x.
- If the uncertainty in x is δx , then the fractional uncertainty in x is δ x/x.
- If you measure x with a device that has a precision of u , then δ x is at least as large as u.
- The uncertainty of x + y or x - y is δ x + δ y.
- If d is data and e is expectation: The difference is Δ = d - e % difference is 100 ( Δ / e) They are compatible IF |d - e | < δ d + δ e
- Fractional uncertainty of z = xy or x / y is: δ z / z = δ x / x + δ y / y
- Fractional uncertainty of f = xp^ yq^ zr^ is: δf / f = p δ x / x + q δ y / y + r δ z/z
- Uncertainty of f(x) is: |f(x+ δ **x) – f(x)|
- ESTIMATING UNCERTAINTIES FOR MEASURED QUANTITIES**
(a) Simple Measurements: The smallest division estimate Suppose we use a meter stick ruled in centimeters and millimeters, and you are asked to
measure the length of a rod and obtain the results (see figure 1a) : L 0 = 5.73 cm. A good
estimate of the uncertainty here is half of the smallest division on the scale, or 0.05 cm. That is, the length of the rod would be specified as: L = 5.73 ± 0.05cm This says that you are very confident that the length of the rod falls in the range 5.73 cm – 0.05 cm to 5.73 cm + 0.05 cm, or the length falls in the range of 5.68 cm to 5.78 cm (see figure 1b).
(b) Manufacturer’s tolerance Suppose I purchase a nominally 100 Ω resistor from a manufacturer. It has a gold band on it which signifies a 5% tolerance. What does this mean? The tolerance means δR/R = 0.05 = 5%, that is, the fractional uncertainty. Thus, δR = R x 0.05 = 5Ω. We write this as R = Rnominal ± δR = 100 ± 5 Ω.
It says that the company certifies that the true resistance R lies between 95 and 105Ω. That is, 95 ≤ R≤ 105 Ω. The company tests all of its resistors and if they fall outside of the tolerance limits the resistors are discarded. If your resistor is measured to be outside of the limits, either (a) the manufacturer made a mistake (b) you made a mistake or (c) the manufacturer shipped the correct value but something happened to the resistor that caused its value to change.
(c) Reading a digital meter. Suppose I measure the voltage across a resistor using a digital multimeter. The display says 7.45 V and doesn’t change as I watch it. The general rule is that the uncertainty is half of the value of the least significant digit. This value is 0.01 V so that half of it is 0.005. Here’s why. The meter can only display two digits to the right of the decimal so it must round off additional digits. So if the true value is between 7.445 V and 7.454 V, the display will get rounded to 7.45 V. Thus the average value and its uncertainty can be written as 7.45 ± 0.005V.
When you record this in your notebook, be sure to write 7.45 V. Not 7.450 V. Writing 7.450 V means that the uncertainty is 0.0005 V.
Note that in this example we assumed that the meter reading is steady. If instead, the meter reading is fluctuating, then the situation is different. Now, you need to estimate the range over which the display is fluctuating, then estimate the average value. If the display is fluctuating between 5.4 and 5.8 V, you would record your reading as 5.6 ± 0.2 V. The uncertainty due to the noisy reading is much larger than your ability to read the last digit on the display, so you record the larger error.
3. USING UNCERTAINTIES IN CALCULATIONS
We need to combine uncertainties so that the error bars almost certainly include the true value.
(a) Adding and Subtracting Let’s look at the most basic case. We measure x and y and want to find the error in z.
If z = x + y
δz = δx + δy
If z = x - y
Our formula for multiplication indicates that multiplying by a perfectly known constant has no effect on the fractional error of a quantity:
F = mg m = 12 ± 1kg g = 9.8 m/s F = 117.6 N
δ F
F
δ m
m
δ g
g
δ m
m
since δg = 0
Then
δ F
F
1 kg 12 kg
and F = 117.6 ± 9.8 N
The uncertainty δF = F
δ m
m
= mg
δ m
m
= gδm.
So, δF is just the constant g times δm.
(c) Multiples
If f = cx + dy + gz where c, d, and g are positive or negative constants then, δ(cx) = |cδx| δ(dy) = |dδy| δ(gz) = |gδz| from the multiplication rule.
From the addition rule,
δf = |cδx| + |dδy| + |gδz|.
(d) Powers
If f = x
p y q z r where p, q, and r are positive or negative constants.
( ) ( ) (^ )
z
z
y
x = p f
f
zr
zr
y
y x
x
f
f q
q p
p
δ δ δ d
δ δ δ δ
(e) General
Suppose we want to calculate f(x), a function of x, which has uncertainty δ x. What is the uncertainty in the calculated value f? We simply calculate f at x, and again at x′ = x + δx, then take the absolute value of the difference:
δf = | f(x′) - f(x) | where x′ = x + δx.
Example: f(x) = sin x x = 30 ± 1 ° δf = | sin (31°) - sin (30°) | = | 0.515 - 0.500 | = 0.
What happens when there is more than one variable? We do the calculation for each variable separately and combine the resulting uncertainties.
δf (x,y) = | f (x + δx,y ) - f(x) | + |f (x,y + δy ) - f (x,y ) |
(f) When are Errors Negligible? Errors are only negligible in comparison to something else and in the context of a particular calculation. So it’s hard to give general rules, but easier for specific cases. Here’s an example of how to think about this question.
You measure a long thin tape (that is, something rectangular). Its length is 10 ± 0.2m, and its width is 2 ± 0.1cm. Which uncertainty is more important? The answer depends on what you want to calculate.
First consider finding the length of the perimeter P of the rectangle formed by the tape.
P = 2 ( L + W ) We apply our addition rule: δ P = 2 ( δL + δW)
now: δL = 0.2m δW = 0.1cm = 0.001m so we can neglect δW since it is much less than δL. We had to put δL and δW into the same units to compare them: 0.2m is much larger that 0.2cm.
Now consider finding the area A = LW of the tape.
The multiplication rule gives:
δ A
A
2.0cm
- 1 cm 10m
0.2m W
W L
L
In this case the uncertainty due to δL is negligible compared to that from δW, the opposite conclusion as for the perimeter calculation! That’s because we are multiplying
and now need to consider not δ L vs δ W , but instead
δ L
L
vs
δ W
W
R δr e δe Δ^ δΔ^ compatible?
5.9cm 0.1cm 6.1cm 0.1cm -0.2cm 0.2cm YES 6.4cm 0.1cm 6.1cm 0.1cm +0.3cm 0.2cm NO 6.2cm 0.2cm 6.1cm 0.1cm +0.1cm 0.3cm YES 6.4cm 0.2cm 6.1cm 0 +0.3cm 0.2cm NO
If only one comparison is to be made, your lab report might contain a sentence like the following: “The measured value was 6.4± 0.2cm while the expected value was 6.1± 0cm, so the difference is +0.3 ± 0.2cm which means that our measurement was close to, but not compatible with what was expected.”
