Page 1
I
Design Matched Filter for Digital Transmission
“Ethernet”
Eman Salem Electrical Engineering Department
Benha Faculty of Engineering Benha University - Egypt
[email protected]
Hossam Labeb Electrical Engineering Department
Benha Faculty of Engineering
Benha University - Egypt
@bhit.bu.edu.eg
Abdelhalim Zekry Electronics and Communications
Department
Faculty of Engineering Ain Shams University - Egypt
[email protected]
ABSTRACT
Digital transmission makes out the major
part of the digital communication
networks. The core of the communication
networks is based on digital carriers.
Local area networks exchange their
information on digital carriers called
Ethernet. Unfortunately, the signal is
contaminated by thermal noise. These
noise signals can be partly removed by the
matched filter.
Ethernet is the most ubiquitous
networking technology. It has grown from
its roots in enterprise networks, and now
addresses other markets such as data
centers, storage, metro, wide area, and
carrier networks. The IEEE 802.3
Ethernet Working Group develops
Ethernet’s physical layer standards and
distinguishes each of these links by its
port type or port name. In this paper, we
show simulation results of matched filter
in fast Ethernet system which supports
100Mbps data rate and 1 Gigabit Ethernet
which supports 1000Mbps data rate.
Keywords
Ethernet, fast Ethernet, Gigabit Ethernet,
matched filter, Simulation, BER.
1. INTRODUCTION
Ethernet is the most common type of
connection computers in a local area
network (LAN). The original Ethernet
was created in 1976 at Xerox’s Palo Alto
Research Center (PARC). It has gone
through four generations (standard
Ethernet (traditional), fast Ethernet,
1Gbps Ethernet and 10Gbps). Ethernet
technologies are still in constant evolution
since its inception in 1976, thus increasing
the ability to expand and accommodate
the Permanent largest possible number of
devices that are connected with the
possibility of securing transport at high
speeds during small times.
Fast Ethernet began to be widely deployed
in the mid-1990s. Fast Ethernet supports
a maximum data rate of 100 Mbps. It is
named because original Ethernet
technology supported only 10 Mbps.
Ethernet networks use a variety of cable
types (such as fiber optics and twisted pair
cable). Gigabit Ethernet is the version of
Page 2
2
Ethernet. Gigabit Ethernet offers higher
performance 1000Mbps (1Gpbs) that is
one hundred times faster than the original
Ethernet.
2. ETHERNET OVERVIEW
Ethernet is the most widely deployed
Local Area Network (LAN) protocol
and has been extended to Metropolitan
Area Networks (MAN) and Wide Area
Networks (WAN). The major advantages
that characterize Ethernet can be stated as
its cost efficiency, bit rate increase (from
10 Mbps to 10 Gbps) and simplicity. ). It
has gone through four generations
(standard Ethernet (traditional), fast
Ethernet, 1Gbps Ethernet and 10Gbps).
Standard Ethernet
The Standard Ethernet defines several
physical layer implementations; four of
the most common, are shown in Figure
(1).[1]
Figure 1:Categories of Standard Ethernet
10Base5: Thick Ethernet
The first implementation is called
10Base5, thick Ethernet, or Thicknet. The
nick name derives from the size of the
cable, which is roughly the size of a
garden hose and too stiff to bend with
your hands. 10Base5 was the first
Ethernet specification to use a bus
topology [1].
10Base2: Thin Ethernet
The second implementation is called
10Base2, thin Ethernet, or Cheaper net.
10Base2 also uses a bus topology, but the
cable is much thinner and more flexible.
The cable can be bent to pass very close to
the stations [1].
10Base-T: Twisted Pair Ethernet
The third implementation is called
10Base-T or twisted pair Ethernet.
10Base-T uses a physical star topology .
The stations are connected to a hub via
two pairs of twisted cable [1].
10Base-F: Fiber Ethernet
Although there are several types of optical
fiber 10Mbps Ethernet, the most common
is called10Base-F. 10Base-F uses a star
topology to connect stations to a hub. The
stations are connected to the hub using
two fiber-optic cables [1].
Encoding and Decoding
All standard implementations use digital
signaling (baseband) at 10Mbps .At the
sender, data are converted to a digital
signal using the Manchester scheme; at
the receiver, the received signal is
interpreted as Manchester and decoded
into data. Figure (2) shows the encoding
scheme for Standard Ethernet [1].
Page 3
3
Figure 2: Encoding in a Standard Ethernet
implementation.
Fast Ethernet
Fast Ethernet supports a maximum data
rate of 100 Mbps. It is so named because
original Ethernet technology supported
only 10 Mbps. Fast Ethernet began to be
widely deployed in the mid-1990s as the
need for greater LAN performance
became critical to universities and
businesses. IEEE created Fast Ethernet
under the name 802.3u. Fast Ethernet is
backward compatible with Standard
Ethernet, but it can transmit data 10 times
faster at a rate of 100Mbps.The goals of
Fast Ethernet can be summarized as
follows [1]:
1. Upgrade the data rate to 100Mbps.
2. Make it compatible with Standard
Ethernet.
3. Keep the same 48-bit address.
4. Keep the same frame format.
5. Keep the same minimum and maximum
frame lengths.
The physical layer in Fast Ethernet is
more complicated than the one in
Standard Ethernet. We briefly discuss
some features of this layer [1].
Fast Ethernet implementation at the
physical layer can be categorized as
shown in Figure (3).
Figure 3: Fast Ethernet implementations.
100Base-TX
Uses two pairs of twisted pair cable
(either category5 UTP or STP). For this
implementation, the MLT-3 scheme was
selected since it has good bandwidth
performance However, since MLT-3 is
not a self-synchronous line coding
scheme, 4B/5B block coding is used to
provide bit synchronization by preventing
the occurrence of a long sequence of 0s
and 1s .This creates a data rate Of
125Mbps, which is fed into MLT-3 for
encoding [1].
100Base-FX
Uses two pairs of fiber optic cables.
Optical fiber can easily handle high
Bandwidth requirements by using simple
encoding schemes. NRZ-I scheme was
selected for this implementation.
However, NRZ-I has a bit synchronization
problem for long sequences of 0s (or 1s,
based on the encoding).To overcome this
problem, the designers used 4B/5B block
encoding as we described for 100Base-
TX. The block encoding increases the bit
Page 4
4
rate from 100 to 125Mbps, which can
easily be handled by fiber optic cable [1].
Table 1: Summary of Fast Ethernet
implementations
Encoding
Manchester encoding needs a 200-Mbaud
bandwidth for a data rate of 100Mbps,
which makes it unsuitable for a medium
such as twisted-pair cable. For this reason,
the Fast Ethernet designers sought some
alternative encoding/decoding scheme.
However, it was found that one scheme
would not perform equally well for all
three implementations.
Therefore, three different encoding
schemes were chosen (see Figure 4) [1].
Figure 4: Encoding for Fast Ethernet
implementation.
Gigabit Ethernet
Gigabit Ethernet is the version of
Ethernet. It offers 1000Mbps (1 Gbps)
bandwidth, that is 100 times faster than
the original Ethernet, yet is compatible
with existing Ethernets [2].
Gigabit Ethernet can be categorized as
either a two wire or a four wire
implementation as shown in figure (5).
Figure 5: Gigabit Ethernet
implementations.
Table 2: Summary of Gigabit Ethernet
implementations.
Encoding
Figure (6) shows the encoding/decoding
schemes for the four implementations.
Page 5
5
Figure 6: Encoding in Gigabit Ethernet
implementations.
Ten Gigabit Ethernet
As advances in hardware continue to
provide faster transmissions across
networks, Ethernet implementations have
improved in order to capitalize on the
faster speeds. Fast Ethernet increased the
speed of traditional Ethernet from 10
megabits per second (Mbps) to 100 Mbps.
This was further augmented to 1000 Mbps
in June of 1998, when the IEEE defined
the standard for Gigabit Ethernet (IEEE
802.3z). Finally, in 2005, IEEE created
the 802.3ae standard introduced 10
Gigabit Ethernet, also referred to as
10GbE. 10GbE provides transmission
speeds of 10 gigabits per second (Gbps),
or 10000 Mbps, 10 times the speed of
Gigabit Ethernet [3].
Physical Layer
The physical layer in Ten Gigabit
Ethernet is designed for using fiber optic
cable over long distances. Three
implementations are the most common:
10GBase-S, 10GBase-L, and 10GBase-E.
Table (3) shows a summary of the Ten-
Gigabit Ethernet implementations [1].
Table 3: Summary of Ten-Gigabit
Ethernet implementations.
3. Fast Ethernet design
Figure (7) illustrates the main building
blocks of fast Ethernet systems
(100basefx)
Figure 7: Block Diagram of fast Ethernet
system.
The main component is block coding
(4B/5B) which converts each 4-bit of
information into a 5-bit code resulting in
an effective bit rate of 125 Mbps
according to the table (4) which shows the
corresponding pairs used in 4B/5B
encoding.
Table 4: 4B/5B mapping codes [4].
Page 6
6
Then, we used scrambler to the purpose of
scrambling is to reduce the length of
strings of 0s or 1s in a transmitted signal,
since a long string of 0s or 1s may cause
transmission synchronization problems.
the basic system of scrambler transmitter
is shown in figure (8).
Figure 8: The basic system of the
scrambler transmitter [5].
A circuit in Figure (9) show scrambler
which we used in design 100base-fx Its
characteristic polynomial is
1+ x 9+ x
11 because the taps are
connected at the output of registers 9 and
11, which repeats its sequence after 2 N =
2047 bits [6].
Figure 9: scrambler with polynomial
1 + x 9 + x
11 [6]
The (scrambled) bit-stream is encoded
with a NRZI encoding. NRZI is a method
of mapping a binary signal to a physical
signal for transmission over some
transmission media. The two level NRZI
signal has a transition at a clock boundary
if the bit being transmitted is a logical 1,
and does not have a transition if the bit
being transmitted is a logical 0 figure (10)
shows example of NRZI coding.
Figure 10: Example NRZI encoding [7].
Before a signal is transmitted over a
channel, the bits of information are coded
into symbols using quadrature Amplitude
Modulation (QAM). For this modulation
scheme, a symbol is encoded into discrete
signal levels. The amplitude of each pulse
is proportional to the amplitude of the
message signal at the time of sampling.
The Raised Cosine Transmit Filter up
samples and pulse shaping of the input
signal using a square root raised cosine
FIR filter, Figure (11) shows Impulse
response of pulse shaping filter RRC at
Group delay = 10, N samples = 5 and roll
off factor = 0.001 which we used in our
design.
Figure 11: impulse response of RRC
Page 7
7
The AWGN Channel adds white Gaussian
noise to transmitted signal. We used
AWGN channel with SNR= 12dB.
The signal has now been transmitted over
the channel and it needs to be recovered.
The steps to recover the original signal are
as follows:
1. Recover the signal from the RRC
(root raised cosine filter).
2. Demodulate the signal.
3. Decoding
4. Matlab model
Figure (12) illustrates the constructed
Simulink model.
Figure 12: Matlab model for fast Ethernet
(100basefx).
5. Simulation results
The simulation results at each step are
shown below. The results are displayed in
the form of snapshots of scope signals.
Signals at Transmitter
By using Bernoulli Binary Generator
block, we generated binary data stream of
100Mbps data rate. The serial data
stream is converted into 4-bit parallel.
Each 4-bit of information are converted
into a 5-bit code resulting in an effective
bit rate of 125 Mbps over the transmission
media by 4B5B encoder shown in the
figure (13)
Figure 13: 5 bit after 4b/5b encoder.
Then, we used scrambler to reduce the
length of strings of 0s or 1s in a
transmitted signal, since a long string of
0s or1s may cause transmission
synchronization problem .The signal after
scrambler is shown in figure (14).
Figure 14: Scrambled signal.
The (scrambled) bit-stream is encoded
with a NRZI encoding to convert digital
data to digital signal to be suitable for
Page 8
8
transmission over some transmission
media as shown in figure (15).
Figure 15: Signal after NRZI.
Before a signal is transmitted over a
channel, the bits of information are coded
into symbols using (QAM) modulation
figure (16) illustrates the signal after
QAM modulation.
Figure 16: Modulated Signal.
Then, we used square root raised cosine
filter (pulse shaping filter) the signal after
pulse shaping filter is shown in figure
(17).
Figure 17: Signal after pulse shaping
filter.
Then, adds white Gaussian noise to signal as
shown in figure (18).
Figure 18: Signal after AWGN.
Signal at Receiver
The first step is to recover the signal
from the RRC. Figure (19) illustrates
signal after matched filter.
Figure 19: signal after matched filter
After filtering the signal with the RRC,
we'll demodulate the signal using QAM
as shown in figure (20).
Figure 20: demodulated Signal
Then, we used NRZI decoder to
convert digital signal to binary signal as
shown in figure (21).
Figure 21: Digital data after line
decoding.
Page 9
9
Then, descrambles input signal we used
the same scrambler polynomial figure (22)
show signal after descrambler.
Figure 22: Signal after descrambler.
After descrambler we recover 5 bits which
enter to 5b4b decoder to obtain 4 bits
which was transmitted the figure (23)
shows Signal after 4B5B encoder (delayed by
10 samples) and after descrambler.
Figure 23: signal after 4B/5B encoder and
descrambler
Figure (24) shows Signal after scrambler
(delayed by 40 samples) and signal before
descrambler.
Figure 24: signal after scrambler and
before descrambler
Finally we obtain the recovered signal,
figure (25) shows transmitted signal
(delayed by 4 samples) and received signal.
Figure 25: Transmitted and Received
Signals
Page 10
10
6. BER performance
The BER plot showed the different
responses of the model corresponding to
the different values of SNR. .The BER is
supposed to be decreasing with the
increase in SNR. To investigate the
modified model performance, we
compared its BER to the theoretical one.
Figure (26) shows theoretical QAM.
Figure 26: BER of theoretical QAM.
Figure (27) shows BER comparison
between theoretical QAM and simulation
results of model at different values of
rolloff factors R of square root raised
cosine filters.
Figure 27: BER comparison between
theoretical QAM and simulation results of
model at different values of R.
Figure (28) shows BER comparison
between theoretical QAM and simulation
results of model at R=0.01 and R =0.001.
Figure 28: shows BER comparison
between theoretical QAM and simulation
results of model at R=0.01 and R =0.001.
Page 11
11
7. Gigabit Ethernet design
Figure (29) illustrates the main building
blocks of 1Gigabit Ethernet system over
fiber optic.
Figure 29: Block Diagram of 1Gigabit
Ethernet system.
The main component is block coding
(8B/10B) which converts each 8-bit of
information into a 10-bit code resulting in
an effective bit rate of 1.25 Gbps. The
8B/l0B block coding is actually a
combination of 5B/6B and 3B/4B
encoding, as shown in Figure (30).
Figure 30: 8B/l0B block encoding [8].
So, we design 5B\6B encoder, 3B\4B
encoder and disparity which keep track of
excess 0s over 1s (or 1s over 0s).
8. Matlab model
Figure (31) illustrates the constructed
Simulink model of Gigabit Ethernet.
Figure 31: Matlab model for Gigabit
Ethernet over fiber optic.
8B/10B Encoder
The serial data stream is converted into 8-
bit parallel. Each 8-bit of information are
converted into a 10-bit code resulting in
an effective bit rate of 1.25 Gbps over the
transmission media by 8B/10B encoder.
This coding scheme is used for high-speed
serial data transmission.
Figure 32:8B/10B coding scheme.
The coding scheme breaks the original 8-
bit data into two blocks, 3 least significant
bits (y) and 5 most significant bits (x).
Page 12
12
From the least significant bit to the most
significant bit, they are named as H, G, F
and E, D, C, B, A. The 3-bit block is
encoded into 4 bits named j, h, g, f. The 5-
bit block is encoded into 6 bits named i, e,
d, c, b, a. As see in Figure (32), the 4-bit
and 6-bit blocks are then combined into a
10-bit encoded value [9].
We design 5b/6b encoder and 3b/4b
encoder by logic gates according to table
(5) and table (6).
Table 5: 4b/5b code.
Table 6: 3b/4b code
Disparity
A DC-balanced serial data stream means
that it has the same number of 0’s and 1’s
for a given length of data stream. In order
to create a DC-balanced data stream, the
concept of disparity is employed to
balance the number of 0’s and 1’s. The
disparity of a block is calculated by the
number of 1’s minus the number of 0’s.
The value of a block that has a zero
disparity is called disparity neutral.
Running Disparity
The transmitter assumes a negative
Running Disparity (RD-) at start up.
When an 8-bit data is encoding, the
encoder will use the RD- column for
encoding. If the 10-bit data been encoded
is disparity neutral, the Running Disparity
will not be changed and the RD- column
will still be used. Otherwise, the Running
Disparity will be changed and the RD+
column will be used instead. Similarly, if
the current Running Disparity is positive
(RD+) and a disparity neutral 10-bit data
is encoded, the Running Disparity will
still be RD+. Otherwise, it will be
changed from RD+ back to RD- and the
RD- column will be used again. The state
diagram in Figure (33) describes how the
current Running Disparity is calculated
[9].
Figure 33:Running disparity state
machine.
Page 13
13
Disparity design in transmitter
We use MATLAB-SIMULINK toolboxes
to simulate disparity as shown in figure
(34).
Figure 34: Disparity design at transmitter.
Scrambler
Figure 35: The basic system of the
scrambler transmitter.
The purpose of scrambling is to reduce the
length of strings of 0s or 1s in a
transmitted signal, since a long string of
0s or1s may cause transmission
synchronization problems the basic
system of scrambler in transmitter is
shown in figure (35).
We used scrambler 16 bits with
characteristic polynomial is
1+ x 11
+ x 13
+ x 14
+ x 16.
Scramble polynomial :A polynomial that
defines the connections in the scrambler
[1 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 1].
Non Return to Zero Invert (NRZI)
Encoder
We design NRZI encoder by matlab as
shown in figure (36), we used XOR gate
and D flip flop.
Figure 36: Non Return To Zero Invert
(NRZI) Encoder.
Raised Cosine Transmit Filter
The Raised Cosine Transmit Filter
upsamples and pulse shaping of the input
signal using a square root raised cosine
FIR filter. Figure (37) shows Impulse
response of pulse shaping filter RRC at
Group delay = 25, N samples = 10 and roll
off factor = 0.0025 which we used in our
design
Figure 37 : impulse response of RRC.
The AWGN Channel adds white Gaussian
noise to transmitted signal. We used
AWGN channel with SNR= 12dB.
Page 14
14
The signal has now been transmitted over
the channel and it needs to be recovered.
The steps to recover the original signal are
as follows:
1. Recover the signal from the RRC
(root raised cosine filter).
2. Demodulate the signal.
3. Decoding
9. Simulation results
Signals at Transmitter
By using Bernoulli Binary Generator
block, we generated binary data stream of
1Gbps data rate. The serial data stream is
converted into 8-bit parallel. Each 8-bit of
information are converted into a 10-bit
code resulting in an effective bit rate of
1.25 Gbps over the transmission media by
8B/10B encoder shown in the figure (38).
Figure 38: 5 bit after 8b/10b encoder.
Then, we used scrambler 16 bits to reduce
the length of strings of 0s or 1s in a
transmitted signal, since a long string of
0s or1s may cause transmission
synchronization problem. The signal after
scrambler is shown in figure (39).
Figure 39: Scrambled signal.
The (scrambled) bit-stream is encoded
with a NRZI encoding to convert digital
data to digital signal to be suitable for
transmission over some transmission
media as shown in figure (40).
Figure 40: Signal after line coding
Before a signal is transmitted over a
channel, the bits of information are coded
into symbols using (QAM) modulation
figure (41) illustrates the signal after
QAM modulation.
Figure 41: Modulated Signal
Then, we used square root raised cosine
filter (pulse shaping filter) the signal after
pulse shaping filter is shown in figure
(42).
Figure 42: Signal after pulse shaping
filter.
Page 15
15
Then, adds white Gaussian noise to signal as
shown in figure (43).
Figure 43: Signal after AWGN.
Signal at Receiver
The first step is to recover the signal from
the RRC. Figure (44) illustrates signal
after matched filter.
Figure 44: Signal after matched filter.
After filtering the signal with the RRC,
we'll demodulate the signal using QAM
as shown in figure (45).
Figure 45:Signal after demodulation
Then, we used NRZI decoder to convert
digital signal to binary signal as shown in
figure (46).
Figure 46: Digital data after line decoding.
Then, descrambles input signal we used
the same scrambler polynomial figure (47)
show signal after descrambler.
Figure 47: Signal after descrambler
before 10b/8b decoder we recover 10 bits
which enter to 10b/8b decoder to obtain 8
bits which was transmitted the figure (48)
shows Signal after 8B/10B encoder
(delayed by 20 samples) and before
10b/8b decoder.
Figure 48: signal after 8B/10B encoder
and before decoder.
Page 16
16
Finally we obtain the recovered signal.
Figure (49) shows transmitted signal
(delayed by 8 samples) and received signal.
Figure 49: Transmitted and Received
Signals
10. BER performance
The BER plot showed the different
responses of the model corresponding to
the different values of SNR. .The BER is
supposed to be decreasing with the
increase in SNR. To investigate the
modified model performance, we
compared its BER to the theoretical one.
This is done using bertool. BERTool is a
bit error rate analysis application for
analyzing communication systems bit
error rate (BER) performance. Figure (50)
shows theoretical QAM.
Figure 50: BER of theoretical QAM.
Figure (51) shows BER comparison
between theoretical QAM and simulation
results of model at different values of
rolloff factors R of square root raised
cosine filters.
Figure 51: BER comparison between
theoretical QAM and simulation results of
model at different values of R.
Figure (52) shows BER comparison
between theoretical QAM and simulation
results of model at R=0.025 and R
=0.0025.
.
Figure 52: shows BER comparison
between theoretical QAM and simulation
results of model at R=0.025 and R
=0.0025.
Page 17
17
The BER results indicate that the system
response changes with the change of the
values of roll off factor R of square root
raised cosine filters. The BER
performance at R=0.0025 is better than
the BER performance at R=0.025.
11. CONCLUSION
Ethernet is the most widely used local
area network (LAN) technology. The
original version of Ethernet supports a
data transmission rate of 10 Mb/s. Newer
versions of Ethernet called "Fast Ethernet"
and "Gigabit Ethernet" support data rates
of 100 Mb/s ,1 Gb/s and 10Gbps. An
Ethernet LAN may use coaxial cable or
fiber optic cable. "Bus" and "Star" wiring
configurations are supported.
There are three types of Fast Ethernet:
100BASE-TX for use with UTP cable,
100BASE-FX for use with fiber-optic
cable, and 100BASE-T4 for use with UTP
cable. We design fast Ethernet (100base
Fx). We designed 100BaseFX which use
fiber optic cable.
Gigabit Ethernet is the version of
Ethernet. It offers 1000Mbps (1 Gbps )
bandwidth, that is 100 times faster than
the original Ethernet.
At 100 Mbps, a technique known as 4B/
5B is used to provide extra symbols for
encoding .Different techniques for line
encoding are used depending whether
copper or fiber is used as the physical
layer. The line encoding varies depending
on the physical layer used. In our design
we used NRZI line coding which use in
fiber optic.
In design 1 Gigabit Ethernet over fiber
optic system we used 8B/10B technique
which converts 8 bits to 10 bits. And line
coding NRZI which is suitable for fiber
optic.
we developed a MATLAB model of
matched filter for fast ethernet
(100baseFx) which support 100 Mbps and
1 Gigabit ethernet over fiber optic.
Finally, a complete systems was
designed and tested.
12. REFERENCES
[1] B. A. Forouzan,” Data
Communications and Networking,”
McGraw-Hill Companies, Inc, ISBN-13
978-0-07-296775-3 - ISBN-to 0-07-
296775-7, Fourth Edition, 2007.
[2] V. Moorthy, ” Gigabit Ethernet,” Aug
14, 1997.
[3] L. Parziale, D.T. Britt, C.Davis, J.
Forrester and W. Liu,” TCP/IP Tutorial
and Technical Overview,” International
Business Machines Corporation IBM
Corp, Eighth Edition, 2006
[4] N. Vlajic, “Digital Transmission of
Digital Data: Line and Block Coding,
Digital Transmission Modes,” York
University, Computer Science, CSE 321,
2010.
[5] M. P. Spratt,” The Use of Scramblers
with an Anti-Locking Circuit,” Hewlett-
Packard Laboratories, December 1992.
Page 18
18
[6] V. A. Pedroni,” Digital Electronics
and Design with VHDL,” Morgan
Kaufmann, 2008.
[7] C. Yao,” Line Coding in Digital
Communication,” Fiber Optic Training &
Tutorials FAQ, Tips & News, November
2011.
[8] A. Balchunas, ” Ethernet Technologies
,” v2.01, 2012.
[9]”8b/10b Encoder/Decoder”, Lattice
Semiconductor Corp, February 2012.
