RobotGuide

Hardware & electrical · topic E5 · 20 hours

Power & thermal management

A robot is a battery-powered mobile device. Voltage sag and temperature rise are the usual root causes of «occasional resets» and «slows down after running a while».

Engineering stage · Intermediate
System software · Aware Embedded & controls · Hands-on Mechanical · Hands-on

Tools
Tools

What to learn

  • Bus voltage selection and current-capacity estimation, line loss and voltage drop
  • Peak vs continuous current, and how capacitors cover the transients
  • Thermal paths, thermal resistance, junction temperature estimation
  • Under-voltage protection and brown-out protection strategies

The lesson · plain words

Power and thermal management is the field nobody notices until something breaks: an unstable regulator and the whole robot reboots at random; insufficient cooling and chips throttle, motor magnets demagnetize. This page teaches you to compute, not to guess.

Thermal circuits: temperature is voltage, power is current

Steady-state junction temperature = ambient temperature + power × thermal resistance Rθ. Drag the power slider below and watch when the steady-state line crosses the red limit — all of thermal design is this one multiplication.

Rθ is a series chain: junction → case → heatsink → air. The Rθ on the heatsink datasheet is the key to saving the design: adding a fan can drop Rθ from 40°C/W to 8°C/W, halving the junction temperature at the same power.

Interactive figure · drag it and see

What this figure shows: junction temperature climbing toward the 105 °C junction limit as you drag the power P (W) slider — the steady-state line is labeled safe or «over the limit — it burns», because heat is an RC circuit where power = current, temperature = voltage, Rθ = R. Takeaway: steady state is just Rθ×P — drag P past 23 W and watch it cross the red line; thermal design is computed, not felt.

Power architecture: partition the tree before picking parts

The robot power tree: battery (12–48V) → motor drivers (feed directly) → DCDC buck to 5V/3.3V → logic. Motors and logic must be fed separately (or isolated with bulk capacitance), or the logic rail collapses the instant a motor stops and the whole machine reboots.

The DCDC/LDO split: DCDC is efficient (>90%) but has ripple — for motors and high power; an LDO has clean output but efficiency = Vout/Vin — for sensors and the ADC reference. The ADC reference must be an LDO or a dedicated reference source — ripple goes straight into the data.

The battery's engineering ledger

Three numbers: capacity (Ah, sets runtime = capacity / average current), discharge rate (C, peak current capability — stall current must stay within the battery's 1C for peace of mind), and the protection board (overcurrent/over-discharge/balancing).

The lithium battery's hidden killers are low temperature (-10°C halves capacity) and over-discharge (below 3V, permanent damage). Set the BMS low-voltage cutoff at 3.2V or higher; do not trust the cell's advertised «cutoff voltage».

The acceptance checklist

① Full-load rail ripple under 100mV (measured with the scope's 20MHz bandwidth limit on); ② 30 minutes continuous full load with key chip surfaces under 70°C (IR gun); ③ logic supply sags less than 10% during motor stall; ④ cold-start succeeds on the first try.

The field check for thermal design: sweep the whole board with an IR camera; any hotspot above 85°C will fail eventually — fixing the cooling early is a hundred times cheaper than a burnt board in the field.

Three questions — do you really get it?

1. Chip power 5W, total thermal resistance 20°C/W, ambient 25°C — what is the steady-state junction temperature?

  • A75°C
  • B125°C
  • C45°C
Show answer

B. Tj = Ta + P×Rθ = 25 + 5×20 = 125°C. Over the limit — time for a heatsink.

2. What should feed the ADC reference voltage?

  • AA DCDC output
  • BAn LDO or dedicated reference source
  • CThe battery directly
Show answer

B. DCDC ripple adds straight onto the sample values. Noise on the reference is noise on the data.

3. What is the biggest risk of motors and logic circuits sharing one supply?

  • AMotor transients collapse the logic rail and cause resets
  • BLow efficiency
  • CHigh cost
Show answer

A. Stall/emergency-stop current spikes couple into the logic rail through shared impedance. Separate rails + isolated regulation is standard.

Keep going deeper

Links without a marker stay on this site; the rest point to the Chinese edition until its translation batch lands.

How it fails in the field

  • Sizing the supply by peak current only: under continuous work, the voltage drop resets the MCU
  • Never computing junction temperature: the case feels cool while the die inside is already at 110°C

You have learned enough when

You can estimate bus power and cooling needs for a 12-joint robot, and explain an incident of «resets as soon as the load spikes».

The Chinese edition carries the full topic page for Power & thermal management, including the complete outline, depth requirements per career track, and further derivations: robotguide.dev/k/E5/.