ICCAS2005

June 2-5, KINTEX, Gyeonggi-Do, Korea

ASIC for Ethernet based real_time communication in DCS

Takeshi Nakajima
Development & Engineering Dept. 1 System Div.
Industrial Automation System Business
Yokogawa Electric Corporation , Tokyo, Japan
(Tel:81-422-52-5590;E-mail:T.nakajima@jp.yokogawa.com)
Abstract: We have developed Ethernet based real-time communication systems called Vnet/IP for DCS which is the control
system for process automation. This paper describes the features and the technologies of the ASIC which is utilized in the
communication interface hardware for Vnet/IP.
Vnet/IP has been developed for mission-critical communications. Hence it has real-time feature, high reliability and precise
time synchronization capability. At the same time, it is able to deal with standard protocols without influence on mission-critical
communications.
The communication interface hardware has a host interface and dual redundant network interfaces. The host interface can be
chosen PCI-bus or R-bus which is the proprietary internal bus developed for the high reliable redundant controller. Each network
interface is a RJ45 connection with 1Gbps maximum in compliance with IEEE802.3.
Keywords:Network,Ethernet,Real-time,Reliability

1.

INTRODUCTION

In todays global economy, competition is intense and the

business environment is continuously changing. Efficient
production methods and rapid adaptation to market trends are
thus prerequisites for the survival of any company in the
production industry. Companies therefore require a
production system that can offer 1) integrated information
that enables real-time response to market information, 2)
flexibility to adjust to changes of business climate for
efficient production, 3) reliability to ensure product and
service quality, cost reduction, and on-time delivery, and 4)
the ability to grow in a changing business climate. One key
component of the production system is DCS(Distributed
Control System), and network technologies for DCS systems
need to be improved.
Plant system developers have been developing proprietary
network systems to achieve real-time response and high
reliability required for stable plant operation. To meet the
demand for higher speed and openness, Ethernet-based plant
networks have emerged in recent years. Unfortunately,
introducing such networks has compromised plant stability.
However, Ethernet is already in widespread use and affords
very low-cost communications. In addition, Ethernet is
advancing rapidly as Internet technologies are improved, and
is much more promising than other data communication
systems. With this background, there have been high hopes
for an industrial Ethernet which can be reliably used for
production systems.
We have developed Vnet/IP, an Ethernet-based plant
network system that offers real-time response, high reliability,
and openness.
In this paper, we propose the construction and the feature of
ASIC used with Vnet/IP.
This paper is organized as follows. In Sect.2, we describe
the features of Vnet/IP. In Sect.3, we describe the
construction of Vnet/IP hardware. In Sect.4, we describe the
architecture of ASIC used with Vnet/IP. Finally, we present
conclusions in Sect.5.

2.

FEATURES of Vnet/IP

Through IECs efforts to standardize Real-time Ethernet

(RTE), more than 10 systems have been proposed. However,
most of them are data communication systems for discrete
process, such as a motion control system. No existing RTE
systems fulfill the requirements for communication systems

for continuous process control.

Vnet/IP was developed for DCS to enhance continuous and
batch process control for mid- and large-size plants. It is
optimized to flexibly transfer large amounts of data
distributed within the plant with real-time capability, or at
about 100 ms. However,the performance realized for very
large systems is also effective for stable operation of smalland mid-size plants. An extra capacity is provided to address
the increase in data traffic volumes in order to boost the
productivity of such plants.
Vnet/IP is a plant network based on 1-Gbps Ethernet.
The feature of Vnet/IP is described as follows.
(1) Real-time response
For real-time operation, Vnet/IP adopts technologies for
transmission scheduling, priority control, and high-speed
response.
Transmission scheduling is performed when a large amount
of packets is transmitted by multiple transmitting stations at
one time, in order to prevent delays in the data being
transmitted to receiving stations as well as delays of
transmission retry due to packet loss. Even when a 1-Gbps
collision-free switching hub is used, the internal buffer
overflows if a large amount of packets is transmitted by
multiple ports at one time, eventually resulting in packet loss
and delay. Vnet/IP divides a macrocycle of 100 ms into time
slots of 1 ms, each of which is assigned to different
transmitting stations. Each time slot contains specific types of
data for different categories of communication.
Refer to the next section for priority control and high-speed
response.
(2) High reliability
In the aspect of its fault tolerant, fault avoidance capabilities
and maintainability, Vnet/IP offers high reliability.
The network configuration is dual redundant, with the two
networks independent from each other. When the main
network goes down due to a failure or abnormality, the other
one is immediately activated by network switching for
real-time operation. All routes between stations in the
network are constantly monitored by the exchange of
diagnostic packets. Even an abnormality in routes that are
rarely used can be immediately detected. When switching of
CPU control right is performed for the dual-redundant
controller of the pair-and-spare architecture, destinations of
data being transmitted are automatically changed so that

normal operation can continue. The pair-and-spare is

characterized in that each logical unit has a pair of partner
boards operating at a lock-step which perform the same
computation in every cycle.

(3) Openness
Vnet/IP can simultaneously perform communications using
TCP-based standard protocols, such as FTP and HTTP, and
control communications. The standard-protocol
communications handle engineering data, and maintenance
data. And control communications handle information which
require real-time response and high reliability.
Ethernet-compatible devices of various vendors can be
connected to form a multi-vendor network. In addition to
bandwidth control and distributed communication processing,
cyber security measures have been implemented, not only to
ensure the real-time response and high reliability required for
control communications, but also to enable simultaneous
performance of control communications and open
communications.
In communication stations, the physical layer and the MAC
layer are made dual-redundant for network connection. These
layers are shared by the stack for control communications and
the data link service function for open communications. The
protocol processing function for open communications is
implemented in a CPU other than the one in which the
processing function for control communications is
implemented. The protocol processing is thus separated from
the processing function for control communications, so that
the former does not adversely affect the latter. In addition, the
data link service function for open communications is
implemented with a bandwidth control function that prevents
excessively heavy traffic of open communications from
interfering with control communications.

3.

Construction of Vnet/IP hardware

In order to achieve a real-time response, high reliability, and

openness on Ethernet, we developed ASIC(Application
Specific Integrated Circuit). The ASIC is named VENUS.
Figure 1 shows an example construction of Vnet/IP
hardware.

H
o
s
t
I/F

PHY
VENUS

PHY
RAM

MPU

ROM

Fig.1 example of Vnet/IP hardware

The external interface of VENUS has the host
interface(HostI/F), MPU(Micro Processor Unit) interface,
and two media attachment interface in compliance with
GMII(Giga bit Medium Independent Interface).VENUS is the
integrated DLL(Data Link Layer) corresponding to the layer2
defined by the ISO/OSI seven layer model.
A MPU executes communication process corresponding to
the layer3 and the layer4 defined by the ISO/OSI seven layer
model. It is connected by MPU interface with VENUS.
A PHY is physical layer corresponding to the layer1

defined by the ISO/OSI seven layer model. Two PHYs

achieve the dual redundant network interfaces. Each PHY is
connected to the RJ45 connector with 1Gbps maximum in
compliance with IEEE802.3. It is connected by GMII
interface with VENUS.
A RAM is a shared memory which VENUS and MPU use.
Host I/F is two PCI functions(32bit,33MHz) or R-bus which
is the proprietary internal bus. Those Host I/F can be selected
by the purpose used.
VENUS has two data path for the real-time communication
and standard communication. Fig.2 shows its data flow.
path 1
H
o
s
t
I/F

PHY

VENUS
PHY

path 2

RAM

MPU

ROM

Fig.2 a data flow of Vnet/IP hardware

The path 1 is used for openness of network communication.
By implementing the protocol stack from layer 3 to layer 7
on host CPU which connected with Host I/F , you can use the
application such as its protocol-based maintenance data
gathering.
The path2 is used for the real-time response.
When receiving, VENUS writes the data received from PHY
in external RAM. MPU notifies host CPU if necessary via
Host I/F of VENUS after removing the protocol header of the
received data.
When transmitting, VENUS sends the data on external RAM
by two ways.
One of them is a transmission that MPU sends the response
immediately if it received the data. If host CPU responded, it
cannot take time for processing, and the real-time response
will not be achieved. To achieve the real-time response, MPU
executes UDP/IP protocol stack that optimized to the
real-time response and suppresses the memory copy to
minimum. The other one is a transmission by the notification
from the Host I/F. MPU attaches a header to the data written
in external RAM via Host/IF and rewrites it on the external
RAM, and then VENUS sends the data.
To achieve the high reliability, VENUS is connected with
two PHY via GMII.
MPU transmits the data to one PHY. If the network of its
PHY fails, MPU is switched to the other PHY. The fail
detection time is possible according to a quick response by
150ms or less.

4.

Architecture of VENUS

The scale of VENUS is 300Kgate, and it works based on

62.5MHz clock.
Fig.3 shows the architecture of VENUS.

PCI
Function

Internal
Register

RBUS
Function

Therefore, MPU can achieve the response of higher speed

by using a higher priority buffer.
Fig.5 shows the architecture of Integrated MAC
Transmit
queue 1
Transmit
Selector

Transmit
queue 2

Integrated
MAC1

Receive
queue 1

Priority
Controller

Receive
Selector

Receive
queue 2

Integrated
MAC2

MPU
Interface
Function

Giga
MAC

Fig.5 the architecture of Integrated MAC

Fig.3 the architecture of VENUS

VENUS consists of PCI functions/RBUS function for Host
I/F,MPU interface function, two Integrated MAC(Media
Access Control),Priority Controller and Internal Register.
PCI function has two functions to achieve openness and
real-time response. PCI function for openness performs by
the PCI master function and is connected with each
Integrated MAC. It is the path 1 of Fig 2.
PCI function for real-time response performs by the PCI
slave function, Its PCI function and RBUS function are
connected with MPU Interface Function. It is the path 2 of
Fig 2.
Internal Register memorizes control information of internal
settings and external memories.
MPU interface Function is connected with each Integrated
MAC via Priority Controller besides PCI function/RBUS
function.
Priority Controller equips the priority control mechanism
of transferred data.
When receiving, VENUS checks the Tos (Type of Service:
priority control field defined by IP protocol)field of the data
received from each Integrated MAC, and stores in four
buffers set beforehand according to priority. Those buffers are
mounted on external RAM, and VENUS memorizes their
destination address in Internal Register beforehand. Tos field
is set by MPU when transmitting.

Transmit queue 1 and Receive queue 1 are data transfer

queue for openness. Transmit queue 2 and Receive queue 2
are data transfer queue for the real-time response. Besides
being able to switch two buffers respectively, Transmit
Selector has the function that controls transmission data for
openness. In Transmit Selector, though Transmit queue 2
transmits data immediately, Transmit queue 1 transmits data
only at regular intervals.
The purpose of this is to avoid delaying the data of the
real-time response when open data is crowded, and it is
constituted to achieve real-time response and high reliability.
GigaMAC can set two MAC addresses in addition to the
function of layer2. One of the MAC addresses is for
Openness, and the other one is for real-time response.
To forward received data to Receive queue promptly, Receive
Selector uses only the low 2nd bit of the MAC address of
received data to distinguish the openness from the real-time
response.
Moreover, Vnet/IP hardware uses the low 1st bit to identify
the redundant Integrated MAC.
Therefore, VENUS needs 4 consecutive MAC address. A
MAC address in Fig.3 is as follows.
bit47

bit1 bit0
MAC

From Integrated MAC

Address

for Openness/real-time selection

for Integrated MAC selection

Tos

Received Data
VENUS
RAM

Priority1 Priority2 Priority3 Priority4

Buffer Buffer Buffer Buffer

bit1 bit0
0
0
0
1
1
0
1
1

--- for Integrated MAC1 real-time

--- for Integrated MAC2 real-time
--- for Integrated MAC1 openness
--- for Integrated MAC2 openness

GigaMAC and Receive Selector use the received data of

openness and real-time to identify with hardware
instantaneously.
They are thus one component that achieves the real-time
response.

5.
Fig.4 Receive Priority Control
When transmitting, VENUS controls to transmit data on
external RAM according to priority.
At this time, even if the data of priority that is lower than
the same priority or it is crowded, high priority data is
processable.

Conclusion

In the future, intelligent field devices and wireless

networks of field devices will continue to spread, and the
concept of plant maintenance will change accordingly.
Eventually, when network security technology has improved,
direct access to Internet resources will become possible from
a factory or plant site. We consider that Vnet/IP will be a
major force for the development of production systems.

REFERENCES
[1] Real Time Ethernet Vnet/IP,IEC/PAS 62405/Ed.1,2005