Abstract: This paper proposes an IP sensor model based on embedded Internet technology, using a simplified IEEE 1451 intelligent sensor information model, and presents a prototype implementation. The network latency problem of the IP sensor is analyzed, and its latency performance is tested in a network environment. Experimental results show that the designed IP sensor can be widely applied to distributed networked sensing, measurement, and control systems. Keywords: Intelligent sensor; IP sensor; Networked sensing; Network latency Introduction The combination of computer network technology and intelligent sensor technology has given rise to the novel concept of networked intelligent sensors. Sensors can function as independent network nodes, transmitting, publishing, and sharing data directly on the network, just like other network devices. Online programming and configuration of field sensors are possible from any node on the network. This combination has greatly promoted the development of sensor technology and the process of informatization. The application of fieldbus technology has facilitated the development of sensors towards intelligence and networking. In the measurement and control stage of automated processes, numerous intelligent sensors are connected together via fieldbus to form a distributed networked measurement and control system. However, due to historical reasons, there is no unified international fieldbus standard. Existing bus standards such as Profibus, FF, Lonworks, HART, and CAN are incompatible, have poor interoperability, and cannot interconnect or be uniformly configured, negatively impacting system expansion and maintenance. Ensuring that designed sensors fully comply with these protocols is difficult, if not impossible, significantly limiting the widespread application of networked smart sensors in industry. The industry urgently needs a widely accepted sensor interface standard with broad application prospects to solve the interconnection problems between sensors and between sensors and networks. 1. Smart Sensors Based on the IEEE 1451 Standard In 1994, IEEE and NIST jointly initiated the development of the "Intelligent Sensor Interface Standard IEEE 1451." After years of effort, the IEEE 1451.2 and IEEE 1451.1 networked smart sensor standards were adopted in 1997 and 1999 respectively. Simultaneously, the P1451.3 and P1451.4 working groups were established to further extend the IEEE 1451.2 standard. The IEEE 1451 standard has received strong support from major companies including Boeing and HP. It uses a common A/D or D/A converter as the I/O interface for the STIM (Sensor Interface Module), converting analog signals from various sensors into data with a standard format. This data, along with the Sensor Electronic Data Sheet (TEDS) and the Network Adapter (NCAP), connects to the STIM, allowing sensor data to access the network according to network protocols and become an independent node with network node configurability and interoperability. The TEDS stores all the information needed to describe a STIM: manufacturer, data format, physical units, serial number, measurement range, and calibration coefficients. This data can be provided to the NCAP or other parts of the system for STIM self-description and calibration. The application of the IEEE 1451 standard has greatly simplified the design of networked smart sensors. 2. IP Sensors Based on Embedded Internet 2.1 The Proposal of IP Sensors The IEEE 1451 standard has greatly promoted the development of networked smart sensor technology. However, since the IEEE 1451.2 standard was published in 1997, it has not received widespread support from sensor manufacturers. Discussions about the standard have remained primarily at the academic research level, with limited practical application. The reasons for this are mainly as follows: a. Difficulty in unifying network protocols. The Network Independent Information Model (NICM) proposed by the IEEE 1451.1 standard theoretically solves the sensor interface problems caused by the incompatibility and inoperability between various bus protocols. However, with each bus technology vendor acting independently to protect its own interests and unwilling to promote its use, the standard has had limited success. b. Sensor-independent interfaces lack broad application prospects. The IEEE 1451.2 standard specifies a digital interface standard (TII) based on a serial peripheral interface protocol. This interface is insufficient for high-speed, high-precision A/D and D/A converters and other high-frequency applications. c. NCAP is too complex and difficult to implement at low cost. The IEEE 1451.1 standard defines a relatively complete general model, but from a practical application perspective, this model is too complex and difficult to implement, lacking a relatively simple intelligent sensor information model. It is noteworthy that, compared to the complex and expensive NCAP, the low-cost STIM has gained popularity among many sensor users and is driving the development of networked smart sensor standards from proprietary bus technology to Ethernet, which has a wider range of applications. This development will inevitably bring about a new "de facto" networked smart sensor standard. Furthermore, the maturity of silicon microelectronics technology has made it possible to implement complex microelectromechanical systems on a single chip, not only solving the technical problem of connecting embedded microcontrollers to the Internet, but also reducing the cost of this connection to an acceptable level for industrial applications. This technological development has led to the emergence of networked smart sensors based on embedded Internet, known as IP sensors. IP sensors refer to a new type of networked smart sensor that organically combines sensors and embedded Internet technology using a modular structure based on the standard TCP/IP protocol. It communicates directly with the computer network as an independent network node, allowing field measurement and control data to be accessed locally and shared in real time within the network's reach. The analog signal output by the sensing element is converted from digital signal to digital signal and processed by an A/D converter, then encapsulated into TCP/IP data packets and transmitted over the network by a network protocol processor. Conversely, the network protocol processor can also receive data and commands from other nodes on the network to operate on its own node. 2.2 IP Sensor Prototype Implementation To simplify design and reduce costs, the IP sensor, based on the IEEE 1451.1 standard, has tailored the smart sensor information model, retaining the STIM structure and functions of the IEEE 1451.2 standard and extending TEDS. It uses the TCP/IP network protocol as a carrier and Ethernet to transmit sensor data. The IP sensor's overall structure mainly consists of two parts: the smart sensor interface module STIM and the network protocol processor module NPPM. The NPPM is primarily used for sending, receiving, and parsing TCP/IP protocol messages. On one hand, it receives data and commands from other network nodes, parses the messages, and then passes the instructions to the STIM for execution; on the other hand, it receives data from the STIM, encapsulates the messages, and transmits them to the designated network node. Here, the network node can be a PC or another IP sensor. Clearly, the IP sensor is essentially an embedded device with Ethernet communication capabilities that integrates the STIM and NPPM. The STIM is used for the sensor interface, and the NPPM is used for the network interface. To coordinate data communication between STIM and NPPM, the IP sensor abandons the difficult-to-implement TII interface based on synchronous serial data transmission protocols. Instead, a dual-port data buffer DPBI is designed based on the ISA standard to ensure reliable data exchange and STIM expansion between the two. The IP sensor, developed based on a general-purpose 8-bit microprocessor, implements a simplified intelligent sensor information model NPPM based on UBICOM's SX52BD. An IEEE 1451.2 standard-compliant STIM structure is completed based on ADI's ADUC812. The ADUC812's built-in flash memory is used for TEDS implementation to support self-identification and self-description of the IP sensor in a distributed network environment. The IP sensor has the following advantages: a. It uses the most popular network communication protocol, TCP/IP, as its carrier, utilizing the inexpensive Internet to transmit sensor data. This means that IP sensors have a wider range of applications. b. The application of the TCP/IP protocol allows technicians to manage and configure IP sensors online through a browser. This makes distributed networked measurement and control based on Ethernet possible. c. The "plug-and-play" nature of IP sensors allows them to be dynamically plugged into existing systems without changing any network structure. This means that the measurement and control system can be dynamically built and reorganized as needed. d. The openness of the IEEE 1451.2 standard allows IP sensors based on this standard to have great flexibility. This means that IP sensor-based systems have good scalability and maintainability. A typical IP sensor-based distributed measurement and control system connects multiple IP sensors, control nodes, and a central control unit via a common network. IP sensors are used to measure parameters and transmit data to other nodes in the network. Control nodes acquire the necessary data from the network as needed and formulate corresponding control methods and execute corresponding control outputs based on this data. In the entire system, each sensor node and control node is independent and autonomous. The number of control nodes and IP sensors depends on the application requirements and can be dynamically increased or decreased as needed. The network can be either an internal enterprise Ethernet or directly on the Internet. 3. IP Sensor Network Latency Analysis IP sensors transmit data via Ethernet using the TCP/IP protocol. Their data transmission performance is inevitably limited by network latency, which directly affects whether IP sensors can achieve widespread practical application. In a networked measurement and control system, IP sensors and control nodes are connected via Ethernet. Different routing choices cause sensor data packets to be transmitted along different lines. Combined with the inherent transmission uncertainties of CSM/ACD, this leads to instability and randomness in IP sensor data transmission and latency. However, with the use of switched cabling and the increase in Ethernet data transmission rates, this problem has been significantly improved. By limiting network load, the probability of data collisions can be greatly reduced. Especially in low-load LAN applications, IP sensors still have broad application prospects. In general, in an Ethernet-based distributed sensor network, if the total network delay of the IP sensors is T_TOT, then: T_TOT = T_c + T_v + T_p where T_c is the communication delay, T_v is the disturbance delay, and T_p is the execution delay. Clearly, only T_c and T_v are affected by network communication, and are the main research content of IP sensor network delay analysis. It is worth mentioning that network delay strongly depends on network load, making it very difficult to construct an accurate mathematical model of network delay. There are no microscopic patterns to follow, and its statistical characteristics can only be studied from a macroscopic perspective. 4. Experiment and Result Analysis In Internet applications, the Control Message Protocol (ICMP) is mainly used to test the network reachability of a destination host. Any host that receives an ICMP echo request will generate an echo reply and return it to the original sender. The returned round-trip time (RTT) is actually the total time it takes for a data packet to travel from the source to the destination and back to the source, which can approximately reflect the changes in the sum of T_c and T_v. This paper tests the network latency performance of an IP sensor based on the ICMP protocol. ICMP echo requests are sent to the IP sensor at equal time intervals via a PC, and the RTT distribution is statistically analyzed. Let t_var = T_c + T_v ≈ RTT, which represents the network latency value in a single test, and T_var represents the corresponding average latency. Considering the representativeness of the test and the possible application modes of the IP sensor, the network latency performance of the IP sensor was tested in two scenarios: a local area network (LAN) (the IP sensor and the PC are located in the same subnet) and a wide area network (WAN) (the IP sensor is located at Shanghai Jiao Tong University, and the PC is located at Huazhong University of Science and Technology) during the same time period (meaning similar network load). The changes in the average latency of the IP sensor under different network data packets were tested. The distribution and statistical results of t[sub]var[/sub] and T[sub]var[/sub] of the IP sensor in LAN and WAN environments are shown in Figure 1, Figure 2, and Table 1. Figure 1 Network latency of IP sensor when data packet is 1024 bytes Figure 2 Distribution of average latency of IP sensor under different network data packets Table 1 Network latency test results of IP sensor As can be seen from the figures, if we do not consider data packet loss and individual abnormal situations, although the network latency of the IP sensor is always random, the amplitude of the network latency and the rate of change of the amplitude are bounded, that is, the network latency t[sub]var[/sub] always varies randomly within a certain range. Data traffic has a certain impact on the network latency of IP sensors. The average latency Tvar shows a subtle increase as the network data packet size increases. Experimental results show that after a certain transmission delay (for a 1024-byte data packet, considering a 90% probability, 23-36ms for LAN and 408-580ms for WAN), IP sensors can reliably exchange data with other network hosts via Ethernet. They can be widely used in networked sensing and measurement systems with non-strict real-time requirements to complete signal acquisition from field devices. 5. Conclusion The 21st century will be the era of embedded Internet. According to relevant experts, embedded devices will greatly increase in the next generation of network equipment, with 70% being embedded devices. If embedded sensor devices can connect to the Internet, information can be transmitted conveniently and inexpensively to any place in the world. It is foreseeable that with the improvement and maturity of network technologies such as Ethernet, IP sensors developed based on embedded Internet technology will be widely used in distributed networked sensing, measurement and control applications, and will bring about new changes in the measurement and control system itself: field sensors will become independent nodes in the network, just like ordinary computers, and sensor information can cross any field that the network can reach without being limited by time and space.