Dynamixel Protocol

Introduction

A Dynamixel is an off-the-shelf servo with integrated control that are controlled by commands over a digital interface. There are three protocols for such an interface: Modbus RTU, Dynamixel v1, and Dynamixel v2. We will only consider Dynamixel v2, which will be simply referred to as the Dynamixel protocol. However, before we do so, we must look at the lower-level electrical system.

This page provides an overview of the Dynamixel communication. It is not an exhaustive list of what is already in the Robotis manual but gives further insight into things left out of the manual, such as the low-level electrical side and some peculiarities. The Robotis manual should be read after the reader has read this page.

The target-audience is somewhat broad and can be anywhere between

• a team-member who is experienced with the Dynamixel system but wants to cross-reference something,
• a new electrical, computer, or mechatronics engineering student who has had some experience with some of these concepts such as UART in ELEC2720 but may want to refresh some of the basics and needs to learn some new things such as RS485 and the Dynamixel protocol,
• and a computer-science student who has some experience with telecommunications such as Ethernet but may have no experience at all with UART and RS485.

Universal Asynchronous Receiver-Transmitter

A universal asynchronous receiver-transmitter (UART) is an electrical device that sends data at an agreed upon bitrate or baudrate. Unlike Serial Peripheral Interface (SPI) and Inter-Integrated Circuit (I2C), there is no signal for the clock, and both devices on either end have to know the common baudrate beforehand. It is common with UART to have to set a baudrate at the very beginning externally either by rewriting firmware or configuring a peripheral.

A UART has two pins, an input called RX, and an output called TX.

The term UART can also refer to the way how data is framed. If one will, it can almost be informally thought of as a very low-level and very basic protocol. Here, UART can refer to the special bits that are needed to frame data, all of which need to be agreed upon between both devices on either end.

The receiver looks for the start-bit's falling edge. When it has found the edge, it begins to time and to sample each bit thereforth based on the agreed upon bitrate. Most setups are written by three symbols, e.g. 8N1 means that there are eight data-bits, no parity bits, and one stop-bit. Both Dynamixel protocols and many other serial ones are 8N1. Henceforth in this guide, this is the assumed framing.

However, pure UART often only refers to abstract bits or logic levels, not voltages. Signalling techniques such as RS485, RS232, and TTL are what are further needed for a complete system to define the electrical behaviour as well as maybe flow-control (e.g. RS232).

Duplexity

Firstly, one must understand the concept of duplexity since the half duplexity of the Dynamixel protocol is not always intuitive for beginners. There are three kinds of digital communication, namely simplex, full duplex, and half duplex.

Simplexity is when data goes in only one direction, from Device-A to Device-B, where only the former transmits and only the latter receives. There is only one channel, and there may often be only one electrical signal. An analogy would be Alice giving instructions to Bob through a loud-speaker very far away.

Full duplexity is when data can go in both directions at the same time, between Device-A and Device-B. There are two channels, one for each direction. This is like having two simplex channels. There is no clash when both Device-A and Device-B send data at the same time since the two channels are separate. An analogy would be Alice and Bob writing letters or texting.

A pure UART will have two signals, RX and TX. RX is used for receiving signals and TX is used for transmitting signals. This is an example of full duplexity.

Half duplexity is when data can go in both directions, but not at the same time. There seem to be two channels but they are really combined into one. An analogy would be Alice and Bob talking to each other by walkie-talkies.

The reader will soon see that the reason why the protocol is half duplex is the fewer number of wires - especially for RS485 - and the fact that servos only respond to instructions.

Often, a Dynamixel device's microcontroller has a UART to communicate on the bus. This UART has two pins, RX and TX. Therefore, in theory, it could be used for full duplexity. However, RX and TX get switched between by a transceiver, which is another chip on the board. The transceiver can switch between receiving and transmitting on the same bus-line. This transceiver has another input pin to set the direction, sometimes called the TX-enable or the direction.

Signal-levels

Transistor-Transistor-Logic

Transistor-Transistor-Logic (TTL) is a family of logical interfaces mainly for logical circuits. Its name refers to the transistors inside the chip used to control the voltage-level. It also refers to signal-levels for many UARTs and protocols. There are only three wires for a TTL setup for a servo, namely ground, power, and data. The NC7WZ241 chip is an example of a TTL transceiver that is found on the OpenCR.

RS485

Transceiver

RS485 is an electrical standard that defines the signal-levels for many serial devices; it does not define a protocol but is instead used for many protocols. Unlike TTL, this has two wires for the data, here called positive and negative, but sometimes under different names such as either plus and minus or A and B. These two wires make up a differential signal. The THVD1550 chip is an example of an RS485 transceiver that is found on NUSense.

Differential Signal

In order to appreciate the power of such a differential signal, one must understand the challenges of noise and interference. For very long cables, the wires often behave like antennae and can thus pick up nearby interference from other electrical devices, especially from servos which tend to emit switching noise. As such, a big enough interference can flip a bit and lead to an error.

However, RS485 and other such techniques have an answer for this. Given a message $m(t)$, the transmitter sends a positive copy $p(t) = m(t)$ and a negative copy $n(t) = -m(t)$ together as a pair. As long as the two physical wires are close together, any interference $w(t)$ is both added such that the received positive signal is $p'(t) = m(t) + w(t)$, and the received negative signal is $n'(t) = -m(t) + w(t)$. The receiver then subtracts the two, often with a difference amplifier, such that the interference is cancelled out, and the received message is $m'(t) = p'(t) - n'(t) = m(t) + w(t) - ( -m(t) + w(t) ) = 2 m(t)$.

Indeed, this is a simple example as there is some more complexity in the way how the RS485 transceiver drives the signals, but this should give the reader the broad idea of how it works and to understand its advantage. Furthermore, this kind of interference is known as common-mode noise. As an aside, a differential pair cannot handle differential mode noise, which is however less common in our case. The reader can learn more about the two here.

Twisted Pair

The robustness to interference is better when the two wires are twisted. This method is very common not only for RS485 but throughout telecommunications. When a pair of wires is untwisted, the area between the wires, although very small and thin, may make up a big enough loop that can pick up magnetic fields going through it. However, if the pair is twisted, then the big loop is now split up into smaller loops which are alternating in polarity such that the current induced across one small loop is cancelled out by that across the next small loop. Furthermore, a twisted pair is easier to manage and keeps the two wires close at all times.

Terminating Resistors

Further to RS485 are terminating resistors which are often important to minimise reflections and to maximise the received power. RS485 can only have two terminating resistors, best at either end of a bus. However, RS485 doesn't always need two terminating resistors in order to work, but it is best to have them for the best performance. Nearly always, the terminating resistor is 120 Ω. The reader can learn more here.

v2 Protocol

Asymmetry

The Dynamixel v2 protocol is a basic serial protocol much like Modbus. The most important thing to understand is that the Dynamixel protocol is fundamentally asymmetric and works with two classes, a controller and a device, analoguous to SPI, I2C, and Modbus for example. The controller only sends an instruction or a request, only to which a device responds with a status or a response. Robotis specifically use the terminology of 'instruction' and 'status' in their documentation, but the reader may also see 'request' and 'response' in other places. Only the controller can iniate an instruction-status exchange. There can be many devices - at most 253 -, but to avoid collisions, there can only be one controller on a bus at a time. One must remember this when connecting the Dynamixel wizard externally to a robot while it is running.

Packets

An instruction-packet is made up such:

A status-packet is made up such:

Cyclic Redundancy Check

The cyclic redundancy check (CRC) is a value that is appended at the end which helps tell whether there has been any corruption in the received packet. Firstly, the sender computes the CRC of the whole packet soon to be sent - excluding the CRC of course - and appends it on the end. When the receiver has received this packet, it computes its own version of the CRC and compares it to the appended CRC. If the two CRCs are different, then there has been an error. One can learn more here about how the CRC works mathematically.

There are different ways to calculate the CRC depending on the polynomial. The Dynamixel protocol uses CRC-16-IBM which has $x^{16} + x^{15} + x^{2} + 1$ as the polynomial, otherwise wrtten as 0x8005. An algorithm for the specific CRC that the Dynamixel protocol uses as well as a look-up-table can be found here.

Error

One difference between an instruction-packet and a status-packet is the extra error-field in the latter as well as the fixed instruction-field of 0x55. Firstly, if the most significant bit of the error-field is 1, then there has been a hardware problem; reading the hardware-error-status in the device can give more details. The rest of the bits can tell the following.

Control-Table

Before we go further, some context may be needed for the protocol and how it is used as a means to an end. The Dynamixel servos have their own integrated control with a closed-loop, feedforward gains, etc. Therefore, one controls the servo by giving a target, often the angular position of the shaft. Such data can be written to and read from a unified control-table which is a block of memory that is accessable by the Dynamixel protocol. For example, the two most common registers to access are the goal-position and the present position; the former is written as a target for the servo; the latter is read as the actual position. One can look at the control-table for the XH540-W270 as an example.

Instructions

The controller can send a variety of instructions to either write to and read from these registers as well as to control the servo as a whole. The Dynamixel v2 manual has a full overview of the different instructions. However here, we will only look at the most common ones that are used at NUbots.