Ethernet following the IEEE 802.3 standard is a technique, which is used in many network environments. At least the lowest level in big network installations is usually realised with CAT5e copper twisted pair lines for signal transmission. This enables for data transmission using the variant 1000BASE-T, which is also known as Gigabit Ethernet. Twisted pair lines of the category CAT5e do not require a metallic shielding. Therefore, these lines are more susceptible against radiated electromagnetic interferences. Especially intentional interferences have been subject of many investigations over the past years . Amongst others, it was shown that interference from high-power electromagnetics (HPEM) can have a dramatic impact on data transmissions over Ethernet networks.

Typically the effects can be perceived on the application layer with regard to their criticality, e.g. when the data transmission rate of a file transfer drops significantly or an interactive network application like voice-over-IP shows big lags . In practise, it is very hard to find out what exactly is causing such a behaviour as the effects on application layer are normally not directly linked to the physical interaction. Figure  illustrates the process of logical communication within modern IP networks and physical interaction on the basis of the TCP/IP reference model . This model can be considered as simpolified breakdown of the more general OSI layer model .

Errors within the communication can be caused by a hardware problem, a bottle-neck in the network infrastructure due to high utilization or even by a intentional electromagnetic interference (IEMI). Observing and detecting the last-mentioned is a real challenge. Reliable detection with field monitors is still subject of ongoing research and in big network infrastructures an effective field monitoring will probably be a high cost factor .

Therefore a new approach shall be discussed which allows for observation of disturbed data transmissions over an Ethernet network. Software-related effects on higher protocol layers, like control algorithms reacting to data errors, have to bypassed. It will be shown that a practical software implementation of a new test method gives a finely granulated time resolution of Ethernet frames and disturbances on the physical layer. This data can be used for further analysis to detect IEMI. Additionally this new method allows for testing network equipment with comparable results independently from user applications and specific software environments.

Many network applications, such as file transfers, require a reliable data transmission. Altering or loss of data is unacceptable. Therefore typical techniques, like the file transfer protocol (FTP), utilize the transmission control protocol (TCP) on the transport layer, as this protocol offers integrated security measures. TCP operates on top of the IP protocol . Whereas IP is only responsible for the logical routing of data packets from one destination to another through the network topology, TCP opens the communication endpoint for an application.

When testing a network for effects due to EMI, traffic has to be generated. This is often done by simply copying a file from one computer to another while the interference is present . This approach presents a problem, as the TCP protocol's integrated security mechanisms will react to occurring packet losses and delays. Therefore, the perceived effects are not linked to the physical processes on the network cables or within the computers. In and it was shown that an observed drop of the transmission speed can be a result of the control algorithms, although the physical degradation of the network is not very strong.

To conclude this, it can be said that investigating an IT network under the influence of EMI with TCP based applications will only give information on this specific network application in a specific network environment. This data is hardly comparable, as the TCP implementation will act as a “black box”. To overcome these issues s new method shall be implemented.

The basic idea for a new measuring method is not to use a “black box” software like a FTP server/client to generate load on the network, which can be observed using network sniffing tools like Wireshark. Rather the data stream itself shall be generated in a well defined way without the influence of software based control algorithms. To achieve this, the user datagram protocol (UDP) will be used. UDP is also a part of the network stack of every modern operating system and operates on top of IP like TCP does .

Compared to this, UDP is a stateless protocol that provides no security and integrity checks on the transport layer. If a data packet gets lost or the data gets corrupted within an UDP connection, there will be no way to detect and handle this on protocol level as opposed to TCP. This makes UDP a very simple protocol with less overhead, but for this reason the application itself has to handle segmenting, flow-control and especially loss or corruption of data. Exactly this behaviour of UDP will be used to detect interferences without having influences of unknown control algorithms.

The new method was used to perform some test of a IT network under the influence of IEMI. A simple topology as shown in Fig.  was set up. Two computers PC1 and PC2 are connected over a line of switches SW1 to SW4. The computers are connected to the switches using CAT5e twisted pair cables. The switches SW2 and SW3 are also connected by twisted pair cables. SW1 and SW2 as well as SW3 and SW4 are linked with an optical fibre. Using an optical connection between these switches ensures that electromagnetic disturbances can not spread out over the whole network when only one component is exposed to interferences.

In this paper the challenges, when assessing the disturbance in a real world Ethernet set-ups, have been discussed. It has to be distinguished between the hardware effects and software-related effects to get comparable results. To overcome this issue, a new test method was developed and implemented as a simple user application using the stateless user datagram protocol. In contrast to observing normal data transfers from e.g. FTP servers/clients, the effects on application layer perceived with this new method are directly linked to the disturbances on the physical layer. The results of the method are presented as time-domain data within a resolution in the range of approximately a few hundred microseconds. This data can be used for further investigations using known techniques of signal analysis to detect intentional interferences.

The implementation of the test-software is a proof of concept and still under development status. The source code can be requested from the corresponding author.Edited by: F. Gronwald Reviewed by: S. Dickmann and one anonymous referee