Investigation of Communication Quality Degradation
of 1000BASE-T1 by Pulse Disturbance
Yusuke Yano, Takashi Hisakado, and Osami Wada
Department of Electrical Engineering, Kyoto University, Kyoto Daigaku Katsura, Nishikyo-ku, Kyoto 615-8510, Japan
Email:
[email protected] Abstract—We investigated experimentally an adverse effect of Vdiff as Fig. 2(b) affects adversely on the communication quality.
pulse disturbances on the communication quality of 1000BASE-T1. The cause of this requires further investigation. In the ranges “a”
As a result, it was clarified that the packet loss rate increased most of Fig. 3, it is presumed that the communication quality has
at a specific pulse width which depends on the harness length degraded stochastically. A constant PER was obtained
determining the time between injecting a disturbance and
regardless of t. When t changes, only the time occurring the
returning one reflected at the end of a harness.
cancellation changes, the time that the voltage difference
I. INTRODUCTION affecting the signal does not change.
Ethernet Optical fiber cable
tester
In a previous study [1], for improving the reliability of l = 3.0 or 6.4 m (MT1000A) l = 3.0 or 6.4 m
Ethernet, the adverse effect of pulse disturbances on the
PHY PHY
communication quality of 100BASE-TX was evaluated Module Vdiff
0Ω 300 Ω
experimentally. The quality was most degraded with the same 1000BASE-T1 4.7 nF 4.7 nF
Module
pulse width of disturbance as the 1-bit time of the signal. In this harness & connector Coupling network
(CN)
study, we focus on in-vehicle Ethernet (1000BASE-T1, IEEE Pulse SMA Vcom
generator L-shaped
802.3bv) and investigate the dependency of communication (M8195A) copper plate 50 mm
System GND
quality degradation on the pulse width. Fig. 1. Overview of experimental system.
II. PULSE DISTURBANCE INJECTION
(a)
=
Fig. 1 shows an overview of experimental system to inject
=
t = 50 ns
pulse disturbances into harnesses of 1000BASE-T1. Although
we adopted the common-mode injection method using the
=
coupling network (CN) described in [2], the resistance values
of 0 and 300 Ω, which are different from [2], were used. This
is because we investigate adverse effects on the communication
caused by the disturbance superimposed on the differential (b)
signal. The common-mode voltage injected by using the
-
t = 25 ns
imbalance CN is converted to a differential-mode voltage. The
CN was inserted between the harnesses. Each harness length l
-
was set to 3.0 or 6.4 m. The packet loss rate (PER) was observed
with an Ethernet tester when the pulse width t was changed. The
parameters of the pulse generator are set as the time period of 1 Fig. 2. Injected pulse disturbances (l=3.0 m): (a) t=50 ns and (b) t=25 ns.
s, the voltage source amplitude of 250 mV, and the rise and
Fig. 2(b)
fall times of 0.7 ns. Fig. 2 shows Vcom and Vdiff denoted in Fig.
1 when l was 3.0 m. In Fig. 2(a) (t=50 ns), the reflected wave a 6.4 m
of Vcom is seen at 25 ns. Vdiff is zero at around 50 ns because b
when the reflected wave of Vcom is converted to Vdiff again at the
CN, the phase is reversed with respect to the first Vdiff, and the 3.0 m
amplitude is canceled out. In Fig. 2(b) (t=25 ns), although the
b Fig. 2(a) a
cancellation due to the reflected wave does not occur, the time
change of Vdiff at 50 ns is steep.
III. RESULTS AND DISCUSSION
Fig. 3 shows the relationships between pulse width and Fig. 3. Pulse width and communication quality (PER)
PER. The PER increased with a specific pulse width longer than REFERENCES
the 1-bit transmission time (2.67 ns). In addition, since the [1] S. Matsushima et al., DOI:10.1587/comex.2017XBL0116,
specific pulse width changed according to l, the packet loss was 2017.
occurred due to a factor different from the 1-bit time. In the [2] IEC62228-5 ED1, 47A/1073/CD, 2018.
ranges “b” of Fig. 3, it is considered that the steep change of
