Anish Laha/Notes: sizing a pan-tilt camera nodeSHEET 12 / 12 ›
DWGAL-12

Notes: sizing a pan-tilt camera node

First-pass numbers for a low-cost outdoor camera node that has to find and follow a small drone. These are my own assumptions, written down so they can be checked and corrected.

Every input here is an assumption I picked to get an order of magnitude. The point is the method and which term dominates, not the exact answer.

1. Why the camera has to move

Take a 0.3 m drone at 500 m and a 1920-pixel-wide camera.

Angular size of target = 0.3 / 500 = 0.6 mrad 60° lens: 1.047 rad / 1920 px = 0.55 mrad/px → about 1 pixel on target 6° lens: 0.105 rad / 1920 px = 0.055 mrad/px → about 11 pixels on target

One pixel isn't detectable. Eleven is workable, but a 6° lens sees almost none of the sky. So a narrow camera has to be steered, and something wider (a wide camera, microphones or RF) has to tell it where to look.

TRY ITDrag the sliders. This is the calculation above, live.

2. It's a positioner, so angle knowledge matters most

A node on a pole doesn't move, so this isn't a stabilised gimbal. The job is to point accurately and to know the angle. Two nodes that each report a bearing can triangulate, and the position error scales with range.

position error ≈ range × angle error = 500 m × 1 mrad = 0.5 m

That argues for an absolute encoder on the output axis, after the gears or belt, so backlash and belt stretch are measured and not guessed.

3. What sizes the motor

Assume a 200 g camera head at a 3 cm radius, accelerating at 20 rad/s², in a 20 m/s wind.

Inertia: J = m r² = 0.2 × 0.03² = 1.8 × 10⁻⁴ kg·m² T = J α = 1.8 × 10⁻⁴ × 20 = 3.6 mN·m Imbalance: T = m g d = 0.2 × 9.81 × 0.005 = 9.8 mN·m (5 mm off-axis) Wind: F = ½ ρ v² Cd A = ½ × 1.225 × 20² × 1.2 × 0.01 = 2.9 N T = F × 0.05 = 147 mN·m (5 cm offset)

Wind is about 40 times the inertial load. So the design priorities are a balanced head, a small frontal area with the centre of pressure on the axis, and a drive that can hold position without power, such as a worm stage.

4. Resolution through a belt

1.8° stepper × 16 microsteps = 3,200 steps/rev 5 : 1 belt reduction = 16,000 steps/rev → 0.0225° per step (0.39 mrad)

Microstep resolution isn't accuracy, which is another reason for the output encoder.

GearsTiming belt
BacklashPresent, shows up on reversalNear zero when tensioned
StiffnessHighStretchy, can resonate
OutdoorsGrit and wearTension drifts with temperature, UV ageing
NeedsPreload or an anti-backlash pairA tensioner

5. Slew rate

Target crossing at 30 m/s: at 100 m → 30 / 100 = 0.3 rad/s (17 °/s) at 20 m → 30 / 20 = 1.5 rad/s (86 °/s)

Open questions

X 0000 Y 0000 HDG 000 ALT 35000 FT