Fig. 08 · Robotics Lab · Sixteen servo pulses from two I2C wires

PCA9685 servo arm

Arduino Nano

Loading the figure…

Bench

No reading yet

PWM frequency

No reading yet

Current draw

0 mA

Build and break

1BuildNext up
Strip to the bare PCB
Take every part off the blue board. You will put it back in factory order: the chip, the MOSFET and the small parts first, then the headers, the terminal and the capacitor.
2
Pick and place the SMD parts, then reflow
3
Fit the I2C header and the four servo blocks
4
Fit the V+ terminal
5
Fit C2, 470 µF, the right way round
6
Solder the through-hole parts
7
Wire the Nano
8
Plug the arm's servos into channels 0-2
9
Screw the pack into the V+ terminal
10
Power the bench
11
Upload Centre the arm
12
Upload Sweep the base
13
Upload Pick and place
14
Upload I2C scan
15
Screw the pack in backwardsBreak it
16
Fill the holder with 14500 Li-ion cellsBreak it
17
Power the servos from the Nano's 5 VBreak it
18
Leave out setPWMFreq(50)Break it
19
Bridge the A0 jumperBreak it
20
Fit C2 backwardsBreak it
21
Leave a cold joint on SDABreak it
22
Close the gripper all the way on the cubeBreak it
23
Exploded view

Break it on purpose

Each fault is one that catches real builders. Trigger it, read the symptom on the bench, then fix it.

Pack screwed in backwards
reverse polarity
Why it happens
Symptom: Nothing moves and nothing gets warm; the green LED stays on.
Q1 is a P-MOSFET with its gate on GND. With V+ negative its body diode is reverse-biased and its gate voltage is the wrong way to switch it on, so no current reaches the servos. The LED runs from VCC, the Nano's 5 V, so it says nothing about V+.
Fix: Swap the leads: red to V+, black to GND. The side header's V+ pin has no such protection.
16.8 V on V+
wrong voltage
Why it happens
Symptom: The servos buzz, get hot and stop for good. The chip and the Nano are fine.
An SG90 is made for 4.8 V and takes 6 V at most. Four 14500 Li-ion cells give 16.8 V: the motor and its driver overheat in seconds. Q1 (30 V) passes it on, it is only there for polarity. C2, rated 10 V, is over its rating too.
Fix: Use alkaline AAs (6 V) or NiMH (4.8 V), or a 5 V supply of 2 A or more. Replace the servos, and C2 if it bulges.
Servos on the Nano's 5 V
brown-out
Why it happens
Symptom: The servos twitch, the Nano restarts over and over and the arm never finishes a move.
Moving SG90s take a few hundred mA each, more when they start. Through USB and the Nano's 500 mA fuse that drags 5 V down until the ATmega's brown-out detector (about 2.7 V) resets it. The servos stop, the voltage recovers, the sketch starts again.
Fix: Give the servos their own supply on the V+ terminal. VCC from the Nano is for the chip's logic only.
No setPWMFreq(50)
wrong setting
Why it happens
Symptom: The servos twitch, ignore their commands and get warm.
begin() leaves the chip at about 1000 Hz: a frame just under 1 ms. A 1.5 ms pulse cannot fit in it, so the outputs stay high most of the time and the servos never see a pulse length they understand.
Fix: Call pwm.setPWMFreq(50) after begin(). One frequency serves all sixteen channels.
Board at 0x41
wrong setting
Why it happens
Symptom: The servos stay limp; the sketch runs and prints as if all were well.
The bridged A0 jumper adds 1 to the address. The library writes to 0x40, nobody acknowledges, and the chip at 0x41 never wakes. An I2C scan shows 0x41 and 0x70.
Fix: Remove the bridge, or construct the driver with Adafruit_PWMServoDriver(0x41).
C2 in backwards
part in backwards
Why it happens
Symptom: C2 bulges, hisses and pops the moment V+ comes on.
An electrolytic capacitor holds its charge on a thin oxide layer that only works one way round. Reversed, it conducts, heats up and vents.
Fix: Unplug, fit a new capacitor with the stripe (minus) to the left, away from Q1.
Cold joint on SDA
cold solder joint
Why it happens
Symptom: The arm stutters: some moves happen late, some not at all.
Every command goes over SDA. Through a joint that makes and breaks, some writes never arrive, and the chip keeps sending the old pulse until the next one gets through.
Fix: Reheat the joint with fresh solder until it flows into a shiny cone around the pin.
Gripper stalled on the cube
overcurrent
Why it happens
Symptom: The gripper buzzes, gets warm and the pack current jumps by about 450 mA.
The servo is told 0° but the cube stops the fingers at about 72°. Its potentiometer says it is far from where it should be, so it pushes with everything it has: an SG90 stalls at roughly 360-550 mA. Plastic gears can strip.
Fix: Close only until the fingers touch (70° here). A servo holds best a few degrees past contact, not at the end of its travel.

Meet the bench

build the driver, centre the arm, then pick and place

A servo wants a pulse every 20 ms, 600 to 2400 µs long, and it turns to the angle the length stands for. The Adafruit PCA9685 board makes those pulses for sixteen servos at once: the Nano writes each channel's numbers over two I2C wires, and the chip keeps sending them from its own 25 MHz clock with no help from the Nano. Servo power comes in separately, on the blue V+ terminal, through a MOSFET that blocks a reversed pack. Build the board, wire a Nano, a 4 x AA pack and a small three-servo arm, then make it pick up a foam cube and put it down on the other side. Then break it: the pack backwards, Li-ion cells, servos on the Nano's 5 V, no setPWMFreq(50), the wrong address, the bulk capacitor backwards, a cold SDA joint and a gripper squeezing too hard.

Sources

On the bench

Selected

Sketch uses 6,178 bytes of program storage

Code: Centre the arm

// PCA9685: servo pulses over I2C
#include <Wire.h>
#include <Adafruit_PWMServoDriver.h>

Adafruit_PWMServoDriver pwm;   // the board at its default address, 0x40

void setup() {
  Serial.begin(9600);
  pwm.begin();          // wakes the chip (it powers up asleep, every output LOW)
  pwm.setPWMFreq(50);   // a 20 ms frame: prescale 121, 4096 ticks of 4.88 µs
  for (int ch = 0; ch < 3; ch++) pwm.writeMicroseconds(ch, 1500);   // 1500 µs = 307 ticks: every servo to 90°
  Serial.println("pca9685: 50 Hz, ch0-2 at 1500 us (307 ticks)");
}

void loop() {}   // nothing to do: the chip keeps sending the pulses by itself

Practice tasks

0/2 tasks

Write the code, then run the checks. The run uses a virtual clock and a scripted world, so the same code always gives the same result.

Microseconds are not ticks

Arduino C++

setPWM(ch, on, off) takes ticks, not microseconds: at 50 Hz one tick is 20000 / 4096 µs. The starter sends 1500 as if it were µs, and forgets setPWMFreq(50) too. Write ticksFor() so 1500 µs becomes about 307 ticks, 600 about 123 and 2400 about 492, and set the frame to 50 Hz.

Checks

Not run yetRuns
Not run yetSets the frame to 50 Hz with setPWMFreq(50)
Not run yet1500 µs comes out as 306-308 ticks
Not run yet600 µs comes out as 122-124 ticks
Not run yet2400 µs comes out as 490-493 ticks
Not run yetChannel 0 sends a 1500 µs pulse in a 50 Hz frame

Hints

ticks = us x 4096 / 20000. Multiply first, then divide, or the fraction is lost.
1500 x 4096 is over 6 million, too big for an int on a Nano: cast to uint32_t first.
pwm.setPWMFreq(50); goes right after pwm.begin();

Code for Microseconds are not ticks

Turn the base with a knob

Arduino C++

Add a 10 kΩ potentiometer on A0 and make it steer the base: the knob all the way down is 0°, half way 90°, all the way up 180°. The starter always sends 90°. Read the knob, map 0-1023 to 0-180° and send that.

Checks

Not run yetRuns
Not run yetReads the knob with analogRead(A0)
Not run yetKnob half way: the base at about 90°
Not run yetKnob down: 0° = 600 µs
Not run yetKnob up: 180° = 2400 µs

Hints

int deg = map(analogRead(A0), 0, 1023, 0, 180);
The line after it already turns degrees into 600-2400 µs.
report() prints only when the angle changes, so the Serial monitor stays readable.

Code for Turn the base with a knob

Build and break

1BuildNext up

Strip to the bare PCB

Take every part off the blue board. You will put it back in factory order: the chip, the MOSFET and the small parts first, then the headers, the terminal and the capacitor.

Pick and place the SMD parts, then reflow

Fit the I2C header and the four servo blocks

Fit the V+ terminal

Fit C2, 470 µF, the right way round

Solder the through-hole parts

Wire the Nano

Plug the arm's servos into channels 0-2

Screw the pack into the V+ terminal

Power the bench

Upload Centre the arm

Upload Sweep the base

Upload Pick and place

Upload I2C scan

Screw the pack in backwardsBreak it

Fill the holder with 14500 Li-ion cellsBreak it

Power the servos from the Nano's 5 VBreak it

Leave out setPWMFreq(50)Break it

Bridge the A0 jumperBreak it

Fit C2 backwardsBreak it

Leave a cold joint on SDABreak it

Close the gripper all the way on the cubeBreak it

Exploded view

Meet the bench

build the driver, centre the arm, then pick and place

On the bench

Selected

Sketch uses 6,178 bytes of program storage

Code: Centre the arm

// PCA9685: servo pulses over I2C
#include <Wire.h>
#include <Adafruit_PWMServoDriver.h>

Adafruit_PWMServoDriver pwm;   // the board at its default address, 0x40

void setup() {
  Serial.begin(9600);
  pwm.begin();          // wakes the chip (it powers up asleep, every output LOW)
  pwm.setPWMFreq(50);   // a 20 ms frame: prescale 121, 4096 ticks of 4.88 µs
  for (int ch = 0; ch < 3; ch++) pwm.writeMicroseconds(ch, 1500);   // 1500 µs = 307 ticks: every servo to 90°
  Serial.println("pca9685: 50 Hz, ch0-2 at 1500 us (307 ticks)");
}

void loop() {}   // nothing to do: the chip keeps sending the pulses by itself

PCA9685 servo arm

The bench

Build and break

Strip to the bare PCB

Pick and place the SMD parts, then reflow

Fit the I2C header and the four servo blocks

Fit the V+ terminal

Fit C2, 470 µF, the right way round

Solder the through-hole parts

Wire the Nano

Plug the arm's servos into channels 0-2

Screw the pack into the V+ terminal

Power the bench

Upload Centre the arm

Upload Sweep the base

Upload Pick and place

Upload I2C scan

Screw the pack in backwardsBreak it

Fill the holder with 14500 Li-ion cellsBreak it

Power the servos from the Nano's 5 VBreak it

Leave out setPWMFreq(50)Break it

Bridge the A0 jumperBreak it

Fit C2 backwardsBreak it

Leave a cold joint on SDABreak it

Close the gripper all the way on the cubeBreak it

Exploded view

Break it on purpose

Pack screwed in backwards

16.8 V on V+

Servos on the Nano's 5 V

No setPWMFreq(50)

Board at 0x41

C2 in backwards

Cold joint on SDA

Gripper stalled on the cube

Live status

Current draw

Check yourself

After: Solder the through-hole parts

After: Upload Pick and place

After: Close the gripper all the way on the cube

End of bench quiz

Meet the bench

Sources

On the bench

Practice tasks

Microseconds are not ticks

Checks

Turn the base with a knob

Checks

PCA9685 servo arm

The bench

Build and break

Strip to the bare PCB

Pick and place the SMD parts, then reflow

Fit the I2C header and the four servo blocks

Fit the V+ terminal

Fit C2, 470 µF, the right way round

Solder the through-hole parts

Wire the Nano

Plug the arm's servos into channels 0-2

Screw the pack into the V+ terminal

Power the bench

Upload Centre the arm

Upload Sweep the base

Upload Pick and place

Upload I2C scan

Screw the pack in backwardsBreak it

Fill the holder with 14500 Li-ion cellsBreak it

Power the servos from the Nano's 5 VBreak it

Leave out setPWMFreq(50)Break it

Bridge the A0 jumperBreak it

Fit C2 backwardsBreak it

Leave a cold joint on SDABreak it

Close the gripper all the way on the cubeBreak it

Exploded view

Break it on purpose

Pack screwed in backwards

16.8 V on V+

Servos on the Nano's 5 V