Downloaded from https://iranpaper.

ir
https://www.tarjomano.com/order

2017 IEEE 23rd International Symposium for Design and Technology in Electronic Packaging (SIITME)

Equalization techniques for high data rates

D. Bucur, E. Goicea, N. Militaru

Telecommunications Department
Faculty of Electronics, Telecommunications and Information Technology, University POLITEHNICA of Bucharest
Bucharest, Romania
[email protected]

Abstract— Signaling specification at multi gigabit-per-second bounce, ISI. In order to avoid reception errors the channel
clock rate have to be met independently of loss characteristics of equalization algorithms are used. Equalization is the inverse of
the dielectric or conductive material, length or topology of distortion caused by a signal transmitted through a channel and
interconnect. The lossy line should be equalized by different filter is used to achieve the frequency response. When a channel has
schemes to achieve the best performance of the route and to been equalized, the input signal is reproduced at the output.
obtain a proper opening of the eye diagram. This paper emphasis
the performance of an equalization filter variant considering the In general, a communication system is composed of a
interconnect as a transmission line starting with synthesis of the transmitter, Tx, a receiver, Rx, and a transmission medium
filter and up to the fabrication of the test board. Applying on through which information is propagated, Fig. 1.
both electric and optics signaling, equalization is needed to avoid
amplitude and phase distortion of the received pulse.

Keywords— equalization, signal integrity, high speed design,

eye diagram, ISI

I. INTRODUCTION
Fig. 1 Communication system block diagram
For high data-rates specific rules must be applied to
properly model interconnects and their behavior in both time II. EQUALIZATION CONCEPT
and frequency domains. Each domain allows different types of The losses at high frequency over the transmission line are
analysis that can easily highlight different aspects that appears determinate by the propagation constant. The transmission line
when the trace is represented through microstrip or stripline with the linear parameters RLLLCLGL has a similar behavior
technologies. The transmission line has to be modelled for with a low pass filter so the equalizer must be designed like a
dielectric and conductive losses considering their frequency high-pass filter in order to compensate the effects caused by the
dependent behavior [1]. High frequency signal losses and transmission line losses.
crosstalk can have a significant impact on signal integrity. With
these aspects in mind, a transmission line might distort in both There are two ways of compensation; either the high
amplitude and phase the pulses during propagation as losses frequency signals are amplified or the low frequency signals
are proportional to the length of the line. The higher the rates are attenuated to have a balance through low and high
and the longer the line will result in closed eye diagrams and frequencies regions.
ISI, inter-symbol interference, at the receiver end.
This paper is focused to provide the reader with an
overview of the equalization in the high speed digital circuits.
The goal of equalization is to reduce the ISI in order to recover
the transmitted symbols. The main objective is to analyze and
develop efficient topology circuits for compensating high
frequency losses by using equalization techniques to improve
the quality of the signal. Using these equalization techniques,
the real waveform and the frequency of the transmitted signal
are kept. The equalization filters should cancel the group and Fig. 2 Transfer function of the trace, the equalizer and the compensated
phase delays between different frequency components. response
High transmission rate over long distance, a bandwidth Having the transfer function of the line (orange in Fig. 2),
efficiently used and a correct data reception that represents a ( ), with the aspect of a low-pass filter, the equalization
copy of the emitted signal are the most important performance filter should compensate the line’s transfer function via a high-
factors in today’s system design aspects. In reality, there are pass filter (green in Fig. 2), ( ), and the relation between
many sources that have a negative effect to the transmitted them will be given by [2]:
signal so the reception errors occur. Some of the problems are
caused by: mismatched impedances, reflections, ground

978-1-5386-1626-0/17/$31.00 ©2017 IEEE  26-29 Oct 2017, Constanța, Romania

 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2017 IEEE 23rd International Symposium for Design and Technology in Electronic Packaging (SIITME)
( )
( )=
( ) Input signal EQ TL Output signal
where ( ) is the compensated transfer function with the
lossy line and the equalization filter included. Transmitter Channel Receiver

A. Data rate x Post-emphasis: the filter placed after the transmission

line
The Fig. 3 represents a series of rectangular pulses. The
binary sequence is represented by a line code such as NRZ,
non-return to zero, in order to ensure integrity of the Input signal TL EQ Output signal

transmitted signal. Receiver

Transmitter Channel

NRZ x Combined pre- and post-emphasis: the filter placed

1U before and after the transmission line.
I
2UI
Input signal EQ TL EQ Output signal
Fig. 3 Ideal model of pulses 1010101 NRZ
Double polarity data transmission is preferred because it Transmitter Channel Receiver
assures a better noise protection. Also, by using this kind of
coding the exactness of the received data is ensured. The The signal of a transmission line with losses is
width of pulse is the bit time and represents the propagation represented on the eye diagram by a closure. An opened eye
time often referred to as unit interval (UI). indicates a set of signal date that can be correctly received at
The data rate is defined as the number of bits transmitted the output while the closed eye indicates an incorrect
per interval of time and has units of bits per second [2]: interpretation of the data signal.
The data gathered from the simulations showed that
1 in order to ensure a high system performance, the most
= = effective technique is the combined one in which there is a
RLC filter placed before and after the transmission line. The
The period of the signal, T=2UI determines de simulations were made in Microwave Office from AWR [3]
fundamental frequency: and Advanced Design System [4].
1 1
= =
2
One bit is transmitted at each UI, so the frequency is half of
the data rate.
B. Eye diagram
Each bit can be analyzed in detail, but it is difficult to
establish which one has an incorrect amplitude or time margin.
The data eye diagram is constructed from a digital waveform
Fig. 4 Microstrip line, substrate definition and the passive RLC filter
by folding the parts of the waveform corresponding to each
The substrate used to realize the test board is a Roger
individual bit into a single graph. The eye diagram shows the
RO3003 having a high stability of its dielectric constant and
errors of the digital signal such as jitter which is caused by the
loss tangent with the frequency variation. The substrate as it’s
fast or slow transition of the bit. The eye opening corresponds
illustrated in Fig. 4 was considered at the simulation point for
to one bit period and it is typically called the UI width of the
a 8.89cm microstrip line length and 0.6mm wide that was
eye diagram.
matched for 10GHz frequency prior to equalization process.
As the speed of the high frequency digital signals
reach multi-Gbps, using the eye diagram the signal parameters 0
Return loss magnitude Eq off

can be precisely measured. These parameters indicate the

system performance. -20

-40 DB(|S(1,1)|)
III. PRE-EMPHASIS AND POST-EMPHASIS SIMULATIONS Eq S parameters

The equalization is realized by using a passive filter. -60

Depending on the filter position referenced to the transmission -80

line there are three techniques by which the equalization can
be realized. -100
0 20 40 60 80 100
x Pre-emphasis: the filter placed before the Frequency (GHz)

transmission line Fig. 5 Return loss magnitude

978-1-5386-1626-0/17/$31.00 ©2017 IEEE  26-29 Oct 2017, Constanța, Romania

 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2017 IEEE 23rd International Symposium for Design and Technology in Electronic Packaging (SIITME)
The return and insertion loss are presented in Fig. 5
and Fig. 6 where it is obvious that the termination is properly
selected having the best return loss at 10GHz and smaller
when the frequency increase and also is shifted from the
multiples of the working frequency due to loss effect. The
synthesized equalization filter corresponding to the ( ) of
the line is represented in Fig. 6. The effect of the equalization
filter is analyzed in three variants, as mentioned earlier, by
placing it before and after the line, and both in the same
design.
Insertion loss magnitude Eq off
0

DB(|S(2,1)|)
Eq S parameters Fig.7 Eye diagram of the transmission line with losses without equalization
-1

The closed shape of the eye diagram can be explained

-2
easier by analysing the transmitted and the received signal
waveforms. As shown in Fig. 8 the pulses at the receiver are
-3
reduced in amplitude, have rounded edges, and has sloping top
-4
and bottom portions. Clearly, a transmission line does not
0 20 40 60
Frequency (GHz)
80 100
preserve the proper amplitude. Because line losses are
DB(|S(4,3)|)
proportional to line length, the distortions will grow with the
0
Filter transfer function RLC Eq S parameters
transmission line’s length.

-0.5

-1

-1.5
0 20 40 60 80 100
Frequency (GHz)

Fig. 6 Insertion loss magnitude and RLC transfer function

The synthesis of this filter starts with the expression

of insertion loss of the line [5]:

⎡ ⎤
−1
( ) = 10 log ⎢1 + ⎥
⎢ ⎥
1+
⎣ ⎦
where K determine the filter’s insertion loss at the cut-off Fig. 8 Waveforms at Tx and Rx without equalization
frequency, f , and the flatness of the transfer function
associated. From the above equation the filter’s insertion loss After applying the combined equalization technique
is K=1.13. Using this parameter the filter’s components were it can be clearly seen a significant improvement of the
determined from the below equations, where and are de received data waveform. This improvement is represented by
inductance and capacitance at the cut-off frequency [5]: the opening of the eye diagram in Fig. 9.
√ √
= = ; ; = ( − 1); = .
The final values of the filter’s components are
computed based on the above relations:
= 6,440 ; = 0,0462 ; = 16,728Ω;
= 904,411Ω
Eye diagrams and bits sequence are represented for a
better understanding. Fig. 7 indicates the received data. It can
be observed the closed eye that indicates a high level of
distortion. This is the result of the transmission line losses and
high data rate signal transmitted on it.

Fig. 9 Eye diagram of the transmission line with losses with equalization

978-1-5386-1626-0/17/$31.00 ©2017 IEEE  26-29 Oct 2017, Constanța, Romania

 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2017 IEEE 23rd International Symposium for Design and Technology in Electronic Packaging (SIITME)
The effect of the filters placed before and after the
transmission line can also be seen if we analyze the data
waveform from the transmitted and received signal, Fig. 10.

Fig. 10 Waveforms at Tx and Rx with equalization

The amplitude of the logic signals was restored to the initial

values and the noise was eliminated leading to the opening of
the eye.
IV. MEASUREMENTS RESULTS AND CONCLUSIONS
A PNA-X N5245A Network Analyzer from Agilent was
used to make measurements of the S parameters on the test
board, Fig. 11.

Fig. 12 Test board

ACKNOWLEDGMENT
This article was supported by University Politehnica of
Fig. 11 Test board
The magnitudes of reflection parameters on both Bucharest, through the ”Excellence Research Grants”
Program, UPB-GEX, identifier: UPB-EXCELENTA-2016,
ends, , and also the insertion loss magnitude, , are
research project title ”Innovative microwave filtering
illustrated in Fig. 12 for the line without any equalization
structures for telecommunication and radar applications”,
scheme, with pre- and post-equalization filter RLC, and
Contract number 102/26.09.2016.
combined variant. As it can be seen from the measurements
the magnitude of for the combined pre- and post-
equalization scheme gives the best results, -23dB at 10GHz REFERENCES
comparing to -10dB in pre- or post- equalization or -3dB at the [1] S. Hall, H. Heck, “Advanced signal integrity for high-speed digital
line with no equalization. The overlapping waveforms of , designs”, John Wiley & Sons Inc., 2009
demonstrates a good fabrication process. [2] Stephen Thierauf, “High-speed circuit board signal integrity”, Artech
Other adaptive approaches of equalization schemes House Inc., USA, 2004;
can be further found and applied on a specific design and also [3] AWR Corp., v.13 - MWO, AXIEM, Analyst, http://www.awrcorp.com
the presented circuit can be improved from a cost perspective [4] ADS, Keysight Technologies, http://www.keysight.com
view of the circuit by replacing the lumped elements of the [5] F. E. Terman, “Radio Engineers’ Handbook”, McGraw-Hill Book
filter with microstrip lines. Company, USA, 1943.

978-1-5386-1626-0/17/$31.00 ©2017 IEEE  26-29 Oct 2017, Constanța, Romania