DNP3
Distributed Network Protocol
Last updated
Distributed Network Protocol
Last updated
This is an excerpt from the dnp3.org website of the DNP3 organization. See dnp3.org
The development of DNP3 was a comprehensive effort to achieve open, standards-based Interoperability between substation computers, RTUs, IEDs (Intelligent Electronic Devices) and master stations (except inter-master station communications) for the electric utility industry. Also important was the time frame; the need for a solution to meet today's requirements. As ambitious an undertaking as this was, we reached this objective. And since the inception of DNP, the protocol has also become widely utilized in adjacent industries such as water / waste water, transportation and the oil and gas industry.
DNP3 is based on the standards of the International Electrotechnical Commission (IEC) Technical Committee 57, Working Group 03 who have been working on an OSI 3 layer "Enhanced Performance Architecture" (EPA) protocol standard for telecontrol applications. DNP3 has been designed to be as close to compliant as possible to the standards as they existed at time of development with the addition of functionality not identified in Europe but needed for current and future North American applications (e.g. limited transport layer functions to support 2K block transfers for IEDs, RF and fiber support). DNP3 has been selected as a Recommended Practice by the IEEE C.2 Task Force; RTU to IED Communications Protocol.
DNP3 was developed by Harris, Distributed Automation Products. In November 1993, responsibility for defining further DNP3 specifications and ownership of the DNP3 specifications was turned over to the DNP3 Users Group, a group composed of utilities and vendors who are utilizing the protocol.
DNP3 is an open and public protocol. In order to ensure interoperability, longevity and upgradeability of the protocol, the DNP3 Users Group has taken ownership of the protocol and assumes responsibility for its evolution. The DNP3 Users Group Technical Committee evaluates suggested modifications or additions to the protocol and then amends the protocol description as directed by the Users Group members.
Complete, comprehensive documentation of the protocol is available to the public. The Document Library contains the protocol specifications, as well as details of what is required at the different sublevels, how to implement secure authentication, how to create XML device profiles, and conformance test procedures.
DNP3 was designed to optimize the transmission of data acquisition information and control commands from one computer to another. It is not a general purpose protocol like those found on the Internet for transmitting email, hypertext documents, SQL queries, multimedia and huge files. It is intended for SCADA (Supervisory Control and Data Acquisition) applications.
Figure 1 shows the master-outstation relationship and gives a simplistic view of the databases and software processes involved. The master is on the left side of figure 1, and the outstation is on the right side.
A series of square blocks at the top of the outstation depict data stored in its database and output devices. The various data types are conceptually organized as arrays.
An array of binary input values represents states of physical or logical boolean devices.
Values in the analog input array represent input quantities that the outstation measured or computed.
Counter Input represents count values, such as kilowatt hours, that are ever increasing (until they reach a maximum and then roll over to zero and start counting again.)
Control Outputs are organized into an array representing physical or logical on-off, raise-lower and trip-close points.
Lastly, the array of analog outputs represents physical or logical analog quantities such as those used for setpoints.
The elements of the arrays are labeled 0 through N - 1 where N is the number of blocks shown for the respective data type. In DNP3 terminology, the element numbers are called the point indexes. Indexes are zero-based in DNP3, that is, the lowest element is always identified as zero. Notice that the DNP3 master also has a similar database for the input data types (binary, analog and counter.) The master uses values in its database for the specific purposes of displaying system states, closed-loop control, alarm notification, billing, and much, much more. An objective of the master is to keep its database updated. It accomplishes this by sending requests to the outstation asking it to return the values in the outstation’s database. This is termed polling. The outstation responds to the master’s request by transmitting the contents of its database. Arrows are drawn at the bottom of Figure 1 showing the direction of the requests (toward the outstation) and the direction of the responses (toward the master). Later we will discuss systems whereby the outstations transmit responses without being asked.
The master and the outstation shown in Figure 1 each have two software layers. The top layer is the DNP3 User Code. In the master, it is the software that interacts with the database and initiates the requests for the outstation’s data. In the outstation, it is the software that fetches the requested data from the outstation’s database for responding to master requests. It is interesting to note, that if no physical separation of the master and outstation existed, eliminating the DNP3 might be possible by connecting these two upper layers together. However, since physical, or possibly logical separation of the master and outstation exists, DNP3 software is placed at a lower level. The DNP3 user’s code uses the DNP3 software for transmission of requests or responses to the matching DNP3 user’s code at the other end.
More will be said about data types and software layers later, but first we want to examine a few typical system architectures where DNP3 is used.
Figure 2 shows common system architectures in use today. At the top is a simple one-on-one system having one master station and one outstation. The physical connection between the two is typically a dedicated or dial-up telephone line.
The second type of system is known as a multi-drop design. One master station communicates with multiple outstation devices.
Conversations are typically between the master and one outstation at a time.
The master requests data from the first outstation, then moves onto the next outstation for its data, and continually interrogates each outstation in a round robin order. The communication media is a multi-dropped telephone line, fiber optic cable, or radio. Each outstation can hear messages from the master and is only permitted to respond to messages addressed to itself. Outstations may or may not be able to hear each other.
In some multi-drop forms, communications are peer-to-peer. A station may operate as a master for gathering information or sending commands to the outstation in another station. And then, it may change roles to become an outstation to another station.
The middle row in Figure 2 shows a hierarchical type system where the device in the middle is an outstation to the master at the left and is a master with respect to the outstation on the right. The middle device is often termed a sub-master.
Both lines at the bottom of Figure 2 show data concentrator applications and protocol converters. A device may gather data from multiple outstations on the right side of the figure and store this data in its database where it is retrievable by a master station on the left side of the figure. This design is often seen in substations where the data concentrator collects information from local intelligent devices for transmission to the master station.
Many vendors offer products that operate using TCP/IP to transport DNP3 messages in lieu of the media discussed above. Link layer frames, which we have not talked about yet, are embedded into TCP/IP packets. This approach has enabled DNP3 to take advantage of Internet technology and permitted economical data collection and control between widely separated devices.
Communication circuits between the devices are often imperfect. They are susceptible to noise and signal distortion. DNP3 software is layered to provide reliable data transmission and to effect an organized approach to the transmission of data and commands. Figure 3 shows the layering that was not shown in Figure 1.
The link layer has the responsibility of making the physical link reliable. It does this by providing error detection and duplicate frame detection. The link layer sends and receives packets, which in DNP3 terminology, are called frames. Sometimes transmission of more than one frame is necessary to transport all of the information from one device to another. A DNP3 frame consists of a header and data section. The header specifies the frame size, contains data link control information and identifies the DNP3 source and destination device addresses. The data section is commonly called the payload and contains data passed down from the layers above.
Every frame begins with two sync bytes that help the receiver determine where the frame begins. The length specifies the number of octets in the remainder of the frame, not including CRC check octets. The link control octet is used for the sending and receiving link layers to coordinate their activities.
The destination address specifies which DNP3 device should process the data, and the source address identifies which DNP3 device sent the message. Having both destination and source addresses satisfies at least one requirement for peer-to-peer communications because the receiver knows where to direct its responses. 65520 individual addresses are available. Every DNP3 device must have a unique address within the collection of devices sending and receiving messages to and from each other. Three destination addresses are reserved by DNP3 to denote an all-call message; that is, the frame should be processed by all receiving DNP3 devices. One address is a universal address, the details of which are not given here, and twelve addresses are reserved for special needs in the future.
The data payload in the link frame contains a pair of CRC octets for every 16 data octets. This provides a high degree of assurance that communication errors can be detected. The maximum number of octets in the data payload is 250, not including CRC octets. The maximum length link layer frame is 292 octets if all the CRC and header octets are counted.
One often hears the term “link layer confirmation” when DNP3 is discussed. A feature of DNP3's link layer is the ability for the transmitter of the frame to request the receiver to confirm that the frame arrived. Using this feature is optional, and it is often not employed because there are other methods for confirming receipt of data. It provides an extra degree of assurance of reliable communications. If a confirmation is not received, the link layer may retry the transmission. Some disadvantages to using link layer confirmation are the extra time required for confirmation messages and waiting for multiple timeouts when retries are configured.
The transport layer has the responsibility of breaking long application layer messages into smaller packets sized for the link layer to transmit, and, when receiving, to reassemble frames into longer application layer messages. In DNP3 the transport layer is incorporated into the application layer. The transport layer requires only a single octet overhead to do its job. Therefore, since the link layer can handle only 250 data octets, and one of those is used for the transport function, each link layer frame can hold as many as 249 application layer octets.
Application layer messages are broken into fragments. Maximum fragment size is determined by the size of the receiving device’s buffer. The normal range is 2048 to 4096 bytes. A message that is larger than a one fragment requires multiple fragments. Fragmenting messages is the responsibility of the application layer. Note that an application layer fragment of size 2048 must be broken into 9 frames by the transport layer, and a fragment size of 4096 needs 17 frames. Interestingly, it has been learned by experience that communications are sometimes more successful for systems operating in high noise environments if the fragment size is significantly reduced.
The application layer works together with the transport and link layers to enable reliable communications. It provides standardized functions and data formatting with which the user layer above can interact. Before functions, data groups and variations can be discussed, the terms static, events and classes need to be covered. In DNP3, the term static is used with data and refers to the present value. Thus static binary input data refers to the present on or off state of a bi-state device. Static analog input data contains the value of an analog at the instant it is transmitted. One possibility DNP3 allows is requesting some or all of the static data in an outstation device. DNP3 events are associated with something significant happening. Examples are state changes, values exceeding some threshold, snapshots of varying data, transient data and newly available information. An event occurs when a binary input changes from an on to an off state or when an analog value changes by more than its configured deadband limit. DNP3 provides the ability to report events with and without time stamps so that if desired, the master will have the information to generate a time sequence report. The master’s user layer can direct DNP3 to request events. Usually, a master is updated more rapidly if it spends most of its time polling for events from the outstation and only occasionally asks for static data as an integrity measure. The reason updates are faster is because the number of events generated between outstation interrogations is small and, therefore, less data must be returned to the master. DNP3 goes a step further by classifying events into three classes. When DNP3 was conceived, class 1 events were considered as having higher priority than class 2 events, and class 2 were higher than class 3 events. While that scheme can be still be configured, some DNP3 users have developed other strategies more favorable to their operation for assigning events into the classes. The user layer can request the application layer to poll for class 1, 2 or 3 events or any combination of them.
DNP3 has provisions for representing data in different formats. Examination of analog data formats is helpful to understand the flexibility of DNP3. Static, present value, analog data can be represented by variation numbers as follows:
A 32-bit integer value with flag
A 16-bit integer value with flag
A 32-bit integer value
A 16-bit integer value
A 32-bit floating point value with flag
A 64-bit floating point value with flag
The flag referred to is a single octet with bit fields indicating whether the source is on-line, the data source restarted, communications are lost with a downstream source, the data is forced and the value is over range.
Not all DNP3 devices can transmit or interpret all six variations. Later, DNP3 levels are discussed, but for now, suffice it to say that DNP3 devices must be able to transmit the simplest variations so that any receiver can interpret the contents. Event analog data can be represented by these variations:
A 32-bit integer value with flag
A 16-bit integer value with flag
A 32-bit integer value with flag and event time
A 16-bit integer value with flag and event time
A 32-bit floating point value with flag
A 64-bit floating point value with flag
A 32-bit floating point value with flag and event time
A 64-bit floating point value with flag and event time
The flag has the same bit fields as for the static variations.
It appears by looking at the above variations that variation 1 and 2 analog events cannot be differentiated from variation 1 and 2 static analog values. DNP3 solves this predicament by assigning group numbers. Static analog values are assigned as group 30, and event analog values are assigned as group 32. Static analog values, group 30, can be formatted in one of 6 variations, and event analog values, group 32, can be formatted in one of 8 variations.
When a DNP3 outstation transmits a message containing response data, the message identifies the group number and variation of every value within the message. Group and variation numbers are also assigned for counters, binary inputs, controls and analog outputs.
In fact, all valid data types and formats in DNP3 are identified by group and variation numbers. Defining the allowable groups and variations helps DNP3 assure interoperability between devices. DNP3's basic documentation contains a library of valid groups and their variations.
When data from an index is transmitted across the wire, the sender must suitably encode the information to enable a receiving device to parse and properly interpret this data. The bits and bytes for each index appearing in the message are called an object. That is, objects in the message are the encoded representation of the data from a point, or other structure, and the object format depends upon which group and variation number are chosen.
The master’s user layer formulates its request for data from the outstation by telling the application layer what function to perform, such as reading, and by specifying the data types it wants from the outstation. The request can specify how many objects it wants or it can specify specific objects or a range of objects from index number X through index number Y. The application layer then passes the request down through the transport layer to the link layer that, in turn, sends the message to the outstation. The link layer at the outstation checks the frames for errors and passes them up to the transport layer where the complete message is assembled in the outstation’s application layer. The application layer then tells its user layer which groups and variations were requested. Responses work similarly, in that, the outstation’s user layer fetches the desired data and presents it to the application layer that, in turn, uses the group and variation numbers to format user layer data into objects. Data is then passed downward, across the communication channel and upward to the master’s application layer. Here the data objects are then presented to the master’s user layer.
Reading data was briefly described in the above two paragraphs, but DNP3 software is designed to handle other functions. For one, the master can set the time in the outstation. The master can transmit freeze accumulator requests, and it can transmit requests for control operations and setting of analog output values using select-before-operate or direct-operate sequences.
One area that has not been covered yet is transmission of unsolicited messages. This is a mode of operating where the outstation spontaneously transmits a response without having received a specific request for the data. Not all outstations have this capability. This mode is useful when the system has many outstations and the master requires notification as soon as possible after a change occurs. Rather than waiting for a master station polling cycle to get around to it, the outstation simply transmits the change. Before configuring a system for unsolicited messages, a few basics need to be considered. First, spontaneous transmissions should generally occur infrequently, otherwise, too much contention can occur, and controlling media access via master station polling would be better. The second basic issue is that the outstation should have some way of knowing whether it can transmit without stepping on another outstation’s message. DNP3 leaves specification of algorithms to the system implementer.
One last area of discussion involves implementation levels. The DNP3 organization recognizes that supporting every feature of DNP3 is not necessary for every device. Some devices are limited in memory and speed and do not need specific features, while other devices must have the more advanced features to accomplish their task. DNP3 organizes complexity into three levels. At the lowest level, level 1, only very basic functions must be provided and all others are optional. Level 2 handles more functions, groups and variations, and level 3 is even more sophisticated. Within each level only certain combinations of request formats and response formats are required. This was done to limit software code in masters and outstations while still assuring interoperability.
It should be apparent by now that DNP3 is a protocol that fits well into the data acquisition world. It transports data as generic values, it has a rich set of functions, and it was designed to work in a wide area communications network. The standardized approach of groups and variations, and link, transport and application layers, plus public availability makes DNP3 a protocol to be regarded.
