Skip to content

Troubleshooting

What each error means on a servo bus, what to look at first, and how to read what the driver kept.

Datasheet sections this page relies on
  • Status packet, error byte and alert bit: ROBOTIS e-Manual, DYNAMIXEL Protocol 2.0 web page, no revision mark, 3 Status Packet; 3.2 Error, p. 1
  • Protocol 1.0 error byte: ROBOTIS e-Manual, DYNAMIXEL Protocol 1.0 web page, no revision mark, 3.1 Error, p. 1

All of them, with the driver API for each: Datasheet map.

First, look at what the driver kept

Every call leaves evidence, with no extra I/O:

Call What it tells you
GetLastRequest() The bytes of the last instruction packet, as put on the line (stuffed, with CRC)
GetLastReply() The decoded error byte of the last status packet: error number, alert flag, protocol
GetLatchedAlert() The first hardware alert seen since ClearLatchedAlert()
GetStats() Transactions, retries, timeouts, CRC errors, skipped garbage bytes, line faults

A growing timeouts with no crc_errors is silence; growing garbage_bytes is noise or a wrong rate; growing crc_errors is damaged packets. Compare GetLastRequest() with what a logic analyser sees: a request that never appears on the wire is the port, one that appears but gets no answer is the servo or the bus.

Ping returns Timeout

Nothing answered, or the answer was not a packet the driver could read. In order of likelihood:

Cause Check
The servo has no power The servo supply: 10.0 V to 14.8 V for the XM430-W350
Wrong ID Ping by ID with ScanIds() (protocol 1.0) or BroadcastPing() (protocol 2.0)
Wrong bit rate The servo's Baud Rate register against Config::baud; X series ship at 57 600, AX at 1 000 000
Wrong protocol Servos of the X330, XL430, XM, XH and XD series can be set to either protocol; AX, RX, DX and MX 1.0 speak 1.0
The line is not turned round The direction pin or half-duplex mode: the reply must be heard (Hardware)
A reply deadline that is too short Raise Config::reply_timeout_us (Configuration); a USB adapter needs UsbAdapterPreset()
TX and RX swapped Cross them (a separate-line transceiver), or check the single-wire stage

Calls return LineFault

The line carried only 00 or only FF for a whole candidate packet, or did not go quiet within 16 flush rounds. The wiring is at fault, not the servo: the data line is stuck low (a short, a servo with no power pulling it down, the transmitter left enabled), stuck high or floating (no pull-up, the transmitter disabled and nothing driving it). Check the direction control and the pull-up, then call Resync().

Calls return EchoMismatch

Config::echo is Loopback and the bytes read back are not the bytes sent: another device is talking at the same time (two masters, a servo that answers a broadcast it should not), or the wiring does not loop back. If TX and RX are not tied, set EchoMode::None.

Calls return Crc, or garbage_bytes grows

Packets arrive damaged. Check, in this order: the bit rate (a rate error of more than 2 to 3 percent damages bytes), the ground between the master and the servos, the length of the cable and its termination at a high rate, and a port that shifts bytes (a Read() that returns more than it was given, a FIFO that overflows). A single Crc is retried for a read; a write is not repeated, so a Crc on a write means the reply was damaged, not that the instruction failed.

Calls return InvalidReply

The bytes looked like a status packet but broke a rule: the length does not fit, the instruction is not a status instruction, the parameter count is wrong for the read, or a protocol 2.0 packet has an unstuffed header. A servo of another protocol on the bus produces this; so does a reply to an older request that arrived late. Call Resync() to empty the line.

Calls return DeviceError

The servo understood the packet and refused it. Read GetLastReply():

Protocol 2.0 error number Meaning Typical cause
1 Result Fail The instruction could not be processed Wrong state, a register not allowed now
2 Instruction Error Undefined instruction, or Action without Reg Write A call the servo's firmware does not have
3 CRC Error The servo's CRC check failed A damaged request
4 Data Range Error The value is outside the register's range A goal beyond a limit
5 Data Length Error The data is shorter than the register A write of the wrong size
6 Data Limit Error The value exceeds the register's limit A goal above the servo's limit
7 Access Error Write to a read-only register, or a protected one Writing the EEPROM area while the torque is on

The most common one: a write to an EEPROM register (ID, baud rate, limits, operating mode) with the torque on. Turn the torque off first (SetTorque(servo, false)).

Calls return DeviceFault

The servo reported a hardware alert (bit 7 of the error byte in protocol 2.0) and Config::alert_policy is Fail. Read Reg::HardwareErrorStatus for the cause: input voltage, overheating, motor encoder, electrical shock or overload. With AlertPolicy::Report (the default) the call succeeds and the alert is latched: poll GetLatchedAlert().

Calls return NotPermitted or Unsupported

Error Cause Fix
NotPermitted A destructive instruction without Config::allow_destructive Set it for that call on purpose (API guide)
Unsupported The instruction does not exist in the configured protocol, the family has no such register, the servo does not answer at the configured status_return, or the port has no SetBaudRate() Check Protocol and Control tables
InvalidArgument A value does not fit the register, an ID is out of range, or the packet does not fit FrameBytes Check the argument; raise FrameBytes for a very long sync read

A sync read returns Ok but a device is not Ok

That is by design: Result says the exchange worked, DeviceReply::result says what happened to each servo. A servo that is Timeout did not answer (wrong ID, no power, or it was shifted out of the packet by an earlier silent one); Crc means its part arrived damaged; DeviceError carries the servo's error byte. A packet that ends early still delivers the parts that arrived whole.

After a Reboot or Factory Reset nothing answers

The servo restarts and may come back with another ID or bit rate: a Factory Reset with ResetOption::All returns the ID to 1 and the rate to the factory one. Use ResetOption::KeepIdAndBaud to keep them, and wait Config::reboot_wait_us before the next call (the driver does).

The servo does not move

Check Why
Torque is on Reg::TorqueEnable is 0 at power-up; SetTorque(servo, true)
The operating mode Position control is the factory mode of most families; a velocity or PWM mode ignores a goal position
The limits A goal outside the min and max position limits is refused or clamped
The supply A servo with a weak supply reboots under load: the voltage in ReadFeedback() sags