In reading a scale reflected in a mirror by a telescope
In reading a scale reflected in a mirror by a telescope, the deflection $x$ is measured at a fixed scale distance $s$. How do small errors in the measured deflection $x$ affect the value of the angle $\alpha$?
Explanation of Method
How a scale reading error propagates into the angle — and where the method is weakest.
Step 1: The geometric relation
When the mirror turns through an angle $\alpha$, the reflected ray turns through $2\alpha$ — the standard doubling of the mirror method. The spot on a scale at distance $s$ is therefore displaced by
\[ x = s\tan 2\alpha, \qquad s = \text{const}. \tag{1} \]
Step 2: Differentiate
\[ dx = s\cdot\frac{2}{\cos^{2}2\alpha}\,d\alpha = \frac{2s}{\cos^{2}2\alpha}\,d\alpha. \tag{2} \]
Step 3: Solve for the angular error
\[ d\alpha = \frac{\cos^{2}2\alpha}{2s}\,dx, \qquad \bigl|\Delta\alpha\bigr| \approx \frac{\cos^{2}2\alpha}{2s}\,\bigl|\Delta x\bigr|. \tag{3} \]
Step 4: Reading the answer
Two factors control the error.
The distance $s$. The angular error is inversely proportional to $s$: doubling the distance to the scale halves the error from a given reading error. This is why such instruments are set up with a long optical path.
The angle itself. The factor $\cos^{2}2\alpha$ is largest, namely $1$, at $\alpha = 0$, and decreases as $\alpha$ grows. So the same reading error produces the largest angular error near zero deflection:
\[ \bigl|\Delta\alpha\bigr|\Bigl|_{\alpha=0} \approx \frac{\bigl|\Delta x\bigr|}{2s}, \]
and a smaller one at larger $\alpha$, where the scale is more sensitive because $\tan$ is steepening. The relative error $\bigl|\Delta\alpha\bigr|/\alpha$ is worse still for small angles, since it grows without bound as $\alpha\to0$.
Step 5: Verification
Solving (1) explicitly gives $\alpha = \tfrac12\arctan\dfrac{x}{s}$. Differentiating: \[\frac{d\alpha}{dx} = \frac12\cdot\frac{1}{1+\bigl(x/s\bigr)^{2}}\cdot\frac1s = \frac{s}{2\bigl(s^{2}+x^{2}\bigr)}\]
\[\text{And } \cos^{2}2\alpha = \frac{1}{1+\tan^{2}2\alpha} = \frac{1}{1+(x/s)^{2}} = \frac{s^{2}}{s^{2}+x^{2}}\]
\[\frac{\cos^{2}2\alpha}{2s} = \frac{s^{2}}{s^{2}+x^{2}}\cdot\frac{1}{2s} = \frac{s}{2\bigl(s^{2}+x^{2}\bigr)} \;\;\checkmark\]
Result
Final answer:
\[\boxed{\,d\alpha = \dfrac{\cos^{2}2\alpha}{2s}\,dx\,}\]
Remark
Remark. The form $\dfrac{d\alpha}{dx} = \dfrac{s}{2(s^{2}+x^{2})}$ is the more practical one, since $x$ and $s$ are the measured quantities while $\alpha$ is the unknown. It shows directly that the angular error falls off as the spot moves away from the centre of the scale, and that it is at most $\dfrac{1}{2s}$ per unit of reading error, whatever the deflection.