VLSI Design EE213 VLSI DesignStephen Daniels 2003 VLSI Design EE213 Dr. Stephen Daniels.
VLSI Egypt Magazine Issue#2
-
Upload
vlsi-egypt -
Category
Documents
-
view
224 -
download
2
description
Transcript of VLSI Egypt Magazine Issue#2
Diamond is the newSilicon
60 GHz Operation FPGA911 Gigabit Home Networks
60 GHz TranceiversExamples
Best of CES2012
FirstYear-Issue2-April2012
Visit VLSI Egypt New Websitewww.vlsiegypt.com
Improve and Share Your Knowledgeeven more
Stay Connected
Get tuned to our recent updates andactivities by following us at
VLSI Egypt Grouphttp://www.facebook.com/groups/164586813554814/
VLSI Egypt Pagehttp://www.facebook.com/pages/VLSI-Egypt/214146611943790
VLSI Egypt Official Twitter Pagehttp://twitter.com/#!/VLSI_Egypt
A Quarter -AnnualMagazine by VLSI Egypt
Haytham Ashour
Vice President
3
9
Editorial WordWelcome to VLSI Egypt magazine
Best of CES 2012Gigabit Home Networkusing plastic optical fiber
Japan EDA Insight (1)T-Engine
Diamond is the new siliconFPGA-911
1 0
1 4
1 7
1 8
Editorial Team
Hussein Ali
Ahmad Sadek
Mohamed Atef
Website
www.vlsiegypt.com
2 ISSUE 2
The realization of an RF PLL at frequencies from58.32 to 64.8 GHz is very challenging, especiallyfor the VCO and the first frequency divider stage.
60 GHz TransceiverExamples22
VLSI-Egypt is a non-profit, service oriented,community based NGO
Contacts
60 GHz OperationTo Enable Gbps Wireless Communication
4
Ahmad Ibrahim
Editor in Chief
Khaled Khalaf
Ahmad Abou El-Saoud
Muhammad Zanaty
Khaled Kabil
Mahmoud Nassar
60 GHz TransceiverExamples22
Editorial Word
3ISSUE 2
1 0
1 4
1 8
After the introduction of the first issue of the magazine we were eager to
know impression about it. We opened many channels to collect feedback
and monitor the response and see the impact of it. We found that the first
issue of the magazine was successful enough to give us the sign that we
should continue and make the second issue. Work actually didn’t stop since
then and we formulated an editorial group of volunteers who collaborated
together to make this issue.
First of all, VLSI-Egypt moderation committee would like to say that we are
proud that VLSI-Egypt Magazine is started now to run as an independent
project within VLSI-Egypt Organization by a dedicated editorial group who
we believe are able to make this issue better than the first one in many as-
pects.
We would like to express our great gratitude to VLSI Egypt Magazine edit-
orial group for working on this issue. We know that they spent a lot of time
planning, organizing, discussing, preparing and reviewing all work needed
for this issue to see the light. They did all of that for nothing but feeling the
responsibility toward the VLSI community in Egypt and the wellness to
share information and experience with all ofus.
We would also like to thank all the authors from outside the editorial group
who had participated in this issue by their valuable articles that we hope it
would have a great impact on the VLSI community in Egypt. Also, we en-
courage all those interested in writing articles, whether business or technic-
al, to contact the editorial group and we would be more than happy to
answer all your questions.
This issue of VLSI-Egypt Magazine covers the aspects of Multi-Gigabit cir-
cuits, systems and networks in three different articles. We also provide a
quick glimpse on the world of ubiquitous computing in Japan as well as the
use of Diamond for tranistors. We also start a new series about the basics of
FPGA design.
We hope you like this issue and you find it up to your expectations. We
would also like to ask you for all your feedback as your opinions had been
the main benchmark we had been using to asses our success in achieving
our goals. Finally, we would like to thank you for keeping VLSI-Egypt such a
vibrant community.
Welcome to VLSI Egypt magazine
Technical
60 GHz operation to enableGbps wireless communicationBy: Khaled Khalaf
o one can imagine carrying a laptop or any other portable device which is notconnected to the internet, or even to a local network, from which you’retransmitting and receiving information. These can be ranging from simple
text information to streaming video data that requires large data rates of fewgigabits per second. Wireless communication at 60 GHz is claimed to be the mostappropriate way to achieve the huge demands for the next generation consumerapplications.
DefinitionThe unlicensed band centered at 60GHz lies in the extremely
high frequency (EHF) band, which is the highest radio fre-
quency band according to the International Telecommunica-
tion Union (ITU), running the range between 30 and 300GHz
[1 ] . This frequency range is equivalent to wavelengths
between 10mm to 1mm in free space. That is why it is also
called the millimeter-wave (mm-wave) band. The Institute of
Electrical and Electronics Engineers (IEEE) has another fre-
quency band nomenclature that assigns 60GHz to the V band.
The V band includes frequencies ranging from 40 to 75GHz
[2] .
MotivationThe increasing demands of society for technology driven ap-
pliances are pushing the trend to shift operation to higher fre-
quencies, and the advancement in silicon technology makes
this shift feasible. Data transmission is the current example of
our choice. Communication rates of few gigabits per second
are already available in consumer applications, which is a
challenge to be supported by wireless channels. An example is
shown in Figure 1 . According to the current HDTV standard,
up to 5.6 Gbps is required for TV display with a resolution of
1920x1080, 90 Hz frame rate and 30 bits per channel per pixel.
Even higher data rates are expected to be available in the fu-
4 ISSUE 2
ture. If such data rates are to be con-
sidered for wireless transmission, large
bandwidths are required, which are
available at higher frequencies.
60 GHz band allocationsSeveral regulatory bodies in different
countries started from 2000 to allocate
unlicensed frequency bands around 60
GHz for the use of commercial
products. As shown in Figure 2, typic-
ally 7 GHz (from 5 to 9 GHz) are avail-
able for wireless communication around
60 GHz. Following band allocations,
several efforts to standardize commu-
nication over the available band were
done by standardization organizations,
resulting in standards such as
IEEE802.15.3c [4] , ECMA 387 [5] ,
WirelessHD [6] , WiGig [7] and
IEEE802.11ad [8] . These standards split
the 60 GHz band into four channels,
with around 2 GHz each. Several digital
modulations for different modes of op-
eration are also defined for different ap-
plications, reaching data rates up to
around 25 Gbps and transmission dis-
tances up to around 10 meters.
BeamformingNatural spatial isolation caused by
propagation loss due to free space path
loss (FSPL) and oxygen absorption
makes communication in this frequency
band only viable over short ranges (till
10 meters). FSPL can be calculated in its
simple form using the following equa-
tion:
FSPL(dB) = 20log(distance) + 20log
(frequency) – 147.55
This shows around 88 dB loss for 60
GHz signal at 10 m. Figure 3 illustrates
the oxygen absorption peak in the
60GHz region.
One advantage of the implicit attenu-
ation and limited range of operation at
60GHz is the possibility of bandwidth
reuse in two close offices. Directional
propagation is used to enhance signal
transmission and reception. In the
transmitter, radiated power is concen-
Fig.1 Data rate requirements for different resolutions, frame rates, and numbersof bits per channel per pixel for HDTV standard [3] .
Fig.2: Worldwide spectrumavailability in the 60 GHz band [7] .
Fig.3: Atmospheric propagation attenuation versus frequency [9] .
5ISSUE 2
trated towards the receiver instead of
being wasted in unwanted directions.
Similarly, gain is boosted in one direc-
tion and unwanted interferers can be
spatially attenuated in the receiver. This
suggests using multiple antennas at the
transmitter, to direct and enhance signal
transmission, and at the receiver, to im-
prove the sensitivity and reduce inter-
ference. The size of an antenna is
inversely proportional to the operating
frequency. For example, 60GHz opera-
tion allows the use of 16-element an-
tenna array that occupies the same area
as a dipole antenna at 5GHz [10] .
Beamforming is a signal processing
technique used in sensor arrays for dir-
ectional signal transmission or recep-
tion [11 ] . The term beamforming is
derived from spatial filters that were de-
signed to form pencil beams (Figure 4)
[12] . As shown in Figure 5, an array of
antennas with variable gain and phase
shifting (or time delay elements) can
form different antenna patterns, one of
which is a beam with a specific radi-
ation angle θ [10] . Time delays among
different antenna paths need to be com-
pensated by true time delay elements for
coherent signal combination [13] . As-
suming no channel bonding, signal
bandwidth is around 2GHz, which is
very small compared to the 60GHz car-
rier frequency. In narrowband systems,
phase shift blocks can be used instead of
true time delay elements as an approx-
imation for the multi-path signal to add
constructively [13] .
Beamforming implementationchallengesPhase shifting in a receiver can be im-
plemented in four different ways: at RF
after the LNA, at baseband after the
mixer, in the LO path or using signal
processing in the digital domain [15] .
Signal combination in the digital do-
main uses the most hardware and is the
most power hungry because the whole
front-end is copied as many times as the
number of antenna paths. Phase shifting
at RF places lossy elements directly after
the LNA which degrades the system
gain, noise figure and bandwidth. Sys-
tem gain and noise figure are less sens-
itive to amplitude variations in the large
LO signal. Thus, phase shifting at LO
provides the lowest effect on signal
quality. Phase shifting after the mixer
Fig.4: Radiation pattern of a beamformer [14] .
Fig.5: Beamforming system with antenna arrays and transceivers [10] .
6 ISSUE 2
causes insignificant deterioration of
gain and noise figure (as compared to
phase shifting at RF). Signal combina-
tion is performed at baseband in both
LO and baseband phase shifting. In
both cases, in order to avoid using mul-
tiple PLLs, LO signal should be distrib-
uted to different antenna paths. This
includes other problems, such as cross-
talk and low LO power levels.
System architectureOne antenna path of a direct conversion
60 GHz receiver is shown in the block
diagram of Figure 6. After the receiving
antenna, the receiver includes I/Q
QVCO, LNA and I/Q mixers in the
font-end. Phase shifting and signal-
combination are performed at baseband
together with LPFs, VGAs and ADCs.
In a full system, the digital output signal
is then processed in the DSP. The trans-
mitter architecture uses similar system
blocks, with the LNA and ADC are re-
placed by the PA and DAC, respectively.
Enabling technologyIn a mixed-signal chip that includes
analog and digital circuits, CMOS tech-
nology is preferred over bipolar for large
volume applications, where the digital
part dominates. Moore’s law states that
on-chip density of transistors doubles
every two years. This doubling is due to
the fabrication of transistors with smal-
ler minimum length. Smaller size tran-
sistors enable operation at higher
frequencies. That’s the reason for which
operation at mm-wave became possible
nowadays after it was just a dream years
ago.
Scaling also has drawbacks. Smaller
transistors usually require lower supply
voltage due to the lower gate oxide. For
example, the breakdown voltage is 1 .8V
for 0.18µm devices and 1 .2V for 0.13µm
ones [16] . This reduces the available
voltage headroom, and thus, decreases
voltage swings. The reduced supply
voltage also limits the number of
stacked transistors between supply and
ground terminals. Smaller size transist-
ors have more mismatch. This is be-
cause transistor mismatch is inversely
proportional to the square root of its
area according to the following equa-
tion:
where AVT is a technology constant (for
example, AVT = 5 mVµm for NMOS
device in 65nm technology).
Although devices are smaller in size, al-
lowing for higher frequency operation,
interconnects are not scalable as a con-
sequence. Taking 60GHz as an example,
wavelength in free space is 5mm. As-
suming that the effective dielectric con-
stant of a microstrip line is 4, the
on-chip wavelength at 60GHz becomes
2.5mm. This means that a track length
of more than 2500µm carrying a signal
with frequency components of 60GHz
will cause a considerable difference in
signal characteristics. This suggests the
use of electromagnetic wave simulators,
such as Agilent ADS [18] or Ansoft
HFSS [19] , to model relatively long in-
terconnects with the help of a parasitic
extraction tools, such as Mentor Graph-
Fig.7: 60GHz potential applications [7] .
Fig.6: 60GHz receiver architecture.
7ISSUE 2
ics Caliber [20] or Cadence Assura [21 ] ,
for medium and short interconnects.
Current and future applications
The large bandwidth allocated for the
60GHz frequency band could be used to
transfer tens of gigabytes of data in few
seconds. Short range indoor applica-
tions like broadband internet access and
high speed point-to-point wireless com-
munication could utilize this capability.
The only wide-spread available product
in the market that utilizes the 60 GHz
frequency band is a one port kit to re-
place the HDMI cable in HDTV con-
nections. Other products including 60
GHz Home Cinema projectors are com-
ing into market. The coming effective
target application is to embed 60 GHz
operation in WiFi. 802.11 wireless
routers will have additional great capab-
ilities when using the large 60 GHz band
together with the 2.4 and 5 GHz ones.
802.11ad wireless standard is expected
to be finalized in 2012, with the expect-
ations to be employed in commercial
wireless routers.
Figure 7 gathers the main 60 GHz usage
models, ranging from peer-to-peer high
data rate communications to a complete
home wireless high definition network
between electronic devices.
References[1 ] International Telecommunication
Union. [Online] . http://www.itu.int/
[2] IEEE Std 521-2002, IEEE Standard
Letter Designations for Radar-Fre-
quency Bands, 2003.
[3] Su-Khiong (SK) Yong, Pengfei Xia
and Alberto Valdes Garcia, 60 GHz
Technology for Gbps WLAN and
WPAN: From Theory to Practice, John
Wiley & Sons, 2011 .
[4] IEEE Std 802.15.3c-2009, Wireless
MAC and PHY Specifications for High
Rate WPANs, Oct. 2009.
[5] "Standard ECMA-387 – High Rate
60GHz PHY, MAC and HDMI PAL,"
ECMA International, Dec. 2008.
[6] “WirelessHD Specification Version
1 .1 Overview”, WirelessHD, May 2010.
[7] “Defining the Future of Multi-Gig-
abit Wireless Communications”, WiGig,
July 2010.
[8] Status of project IEEE802.11ad. ht-
tp://www.ieee802.org/11/Re-
ports/tgad_update.htm
[9] Eino Kivisaari, 60 GHz MMW Ap-
plications, Helsinki University of Tech-
nology, Telecommunications and
Multimedia Laboratory, Apr. 9, 2003.
http://www.tml.tkk.fi/Opinnot/T-
109.551/2003/kalvot/60GHz.ppt
[10] Ali M. Niknejad and Hossein
Hashemi, mm-Wave Silicon Techno-
logy: 60 GHz and Beyond, Springer,
2008.
[11 ] Wikipedia, Beamforming. ht-
tp://en.wikipedia.org/wiki/Beamform-
ing
[12] B.D. Van Veen and K.M. Buckley,
"Beamforming: A Versatile Approach to
Spatial Filtering," IEEE ASSP Magazine,
vol.5, no.2, pp.4-24, Apr. 1988.
[13] Dixian Zhao, 60 GHz Beamform-
ing Transmitter Design for Pulse Dop-
pler Radar, M.Sc. thesis, Delft
University ofTechnology, Feb. 2009.
[14] J. Bass, E. Rodriguez, J. Finnigan,
C. McPheeters, Beamforming Basics,
Connexions Web site, Jul. 27, 2005. ht-
tp://cnx.org/content/m12563/latest/
[15] Karen Scheir, CMOS Building
Blocks for 60 GHz Phased-Array Re-
ceivers, Ph.D. thesis, Vrije Universiteit
Brussel, Dec. 2009.
[16] John R. Long, RF Integrated Circuit
Design, TU-Delft MSc. course lecture
notes, 2009.
[17] M.J.M. Pelgrom, A.C.J. Duinmaijer
and A.P.G. Welbers, "Matching Proper-
ties of MOS Transistors," IEEE J. Solid-
State Circuits, vol.24, no.5, pp.1433-
1440, Oct. 1989.
[18] Advanced Design System (ADS),
Agilent Technologies.
http://www.agilent.com/find/eesof-ads
[19] HFSS, Ansoft.
http://www.an-
soft.com/products/hf/hfss/
[20] Software for IC Design and Circuit
Design Verification, Mentor Graphics.
http://www.mentor.com/products/ic_n-
anometer_design/
[21 ] Assura Physical Verification.
http://www.ca-
dence.com/products/mfg/apv/
Khaled Khalaf
Millimeter-wave researcher at imec
PhD student at VUB
8 ISSUE 2
Editors' ChoiceBest of CES 2012Sonyericcsson SmartWatchCheck out Facebook™ posts. Catch up on tweets, texts and
Email. See who's calling. Or escape with some tunes.
Discover SmartWatch. An Android™ watch with a scratch
& splash-proof multi-touch display. Clip it on. Wear it
with any 20mm wristband. Enjoy your favourite
SmartWatch apps. It's up to you.
LightPadConnected with a smartphone, The LightPad™ provides
virtually all functions of a notebook computer for an
individual. This is mainly thanks to a rebirth of rear
projection using a pico-projector. The high gain and high
contrast rear projection screen provides for a high quality
display.
ATC Mini Action CameraThe ATC Mini Action Camera is a super lightweight,
compact, waterproof and rugged action camera suitable
for the toughest of conditions. The camera records at 720P
HD video and can plunge to depths of 20 meters making it
suitable for all your on-land, off-land and even
underwater activities.
VLSI Egypt Corner
9ISSUE 2
Toward Gigabit Home NetworkUsing Plastic optical Fiber
he increasing demand for broadband services raises the need for a highbandwidth link, which should extend from the terminals to the customer’spremises. In-building networks presently are using a wide range of
transmission media: coaxial copper cables, twisted copper pair cables, free-spaceinfrared links, wireless local area network (LAN) links, etc. Each of these networks isoptimized for a particular set of services; this complicates the introduction of newservices and the creation of links between services (such as between video and dataservices). A single broadband multi-services network could provide an efficientsolution to host and connect all existing and upcoming services together.
T
The target for data rate (DR) delivered
in home could be up to 1Gbit/s in case
of fiber to the home (FTTH) or up to
120Mbit/s in case of very high bit rate
digital subscriber line (VDSL2)
technology. The home network must
not represent a bottleneck for the
expected evolution for services such as
the introduction of high definition
quality internet protocol television
(IPTV), multi-room/multi-vision
configuration and high quality video
communication via the TV set. The
home network can be used, for example,
to share multimedia contents not
necessarily delivered in real time by
access network, this content can be
stored in a device inside the house and
use it afterwards [11 ] .
At present, twisted pair and coaxial
cables are used as the physical medium
to deliver telecom services within the
customer’s premises. These two
transmission mediums suffer from
serious shortcomings when they are
considered to serve the increasing
demand for broad-band services. For
instance, twisted pair has a limited
bandwidth and it is susceptible to
electromagnetic interference (EMI).
Coaxial cable offers a large bandwidth,
but it poses practical problems due to its
thickness and the effort required to
make a reliable connection. Moreover,
the coaxial cable is not immune to EMI.
Optical fiber is extensively used for
long-distance data transmission and it
represents an alternative for
transmission at the customer premises
10 ISSUE 2
By: Mohamed Atef
as well optical fiber connections offer
complete immunity to EMI [2] . Glass
optical fibers (GOF), however, are not
suitable for use within the customer
premises because of the requirement of
precise handling, and thus, the high
costs involved. On the other hand, it is
important to have very simple and low-
cost solutions. Also the enormous
capacity of the single-mode GOF is
never necessary in this short distance
application. The Poly-Methyl-
Methacrylate Plastic Optical Fiber
(PMMA-POF) is an excellent candidate
for implementing such a short distance
network.
Certain users find that POF systems
provide benefits compared to GOF and
copper wire, which include: simpler
and less expensive components,
operation in the visible range (the
transmission windows are 530nm,
570nm, and 650nm), greater flexibility
and resilience to bending, shock and
vibration, ease in handling and
connecting (standard step-index POF
core diameters are 1mm compared with
8–100µm for glass), use of simple and
inexpensive test equipment. Finally,
POF transceivers require less power
than copper transceivers. These
advantages make POF very attractive for
use within in-building networks,
industrial control and automotive fields
as depicted in Figure 1 [3] .
The main disadvantage of the PMMA-
POF is its high transmission loss
(150dB/km at 650nm and less than
90dB/km at 530nm and 570nm) which
limits the use of PMMA plastic fibers
for transmitting light to less than 100m.
Most cheap commercial POFs have a
uniform, or step index of refraction that
is the same across the width of the fiber,
and step-index fibers (SI-POF) have the
lowest bandwidth among multimode
fibers [3] . This small bandwidth limits
the maximum data rate (DR) which can
be transmitted through POF.
To overcome the problem of POF high
transmission loss, very sensitive
receivers must be used to increase the
transmitted length over PMMA POF.
The SI-POF limited bandwidth problem
can be decreased by using multilevel
signaling like Multilevel Pulse
Amplitude Modulation (M-PAM), see
Figure 2, and Qudrature Amplitude
Modulation (M-QAM). Also the DR
can be increased more with the limited
POF bandwidth by using spectral
efficient modulation techniques like
Discrete Multi Tone (DMT) [4,5,6,7] .
Fig.1 Main POF applications
Figure 2 Multilevel pulse amplitude modulation (M-PAM)
11ISSUE 2
Pre-equalization for the light source and
post-equalization techniques can be
used to equalize for the SI-POF small
bandwidth [8,9,10] . Pre-equalizing of
light source (peaking) lowers the light
source modulation depth; this reduces
the actual power per pulse compared to
rectangular pulses without peaking.
This is at the expense of system power
budget. Post equalizer introduces
additional noise and a higher optical
power is needed to achieve the required
sensitivity. It follows that the use of pre-
or post-equalization methods are of
particular interest in systems that have
adequate power reserves. Also, for a
fixed equalization if the frequency
response changes, as a result of different
lengths of the POF or a bend in the
fiber, the result will be too much or too
little compensation and the bit error
rate (BER) will increase.
Today on the market several suppliers
offer PMMA POF media converter
solutions at 100 Mbit/s. With such
performance SI-POF may be used in the
home to interconnect all devices usually
communicating through Fast Ethernet
interfaces; for example the link between
the home gateway and the Set Top
Boxes (STB). The standard
requirements for 100 Mbit/s and 1Gbit/s
PMMA SI-POF systems introduced by
The European Telecommunications
Standards Institute (ETSI) are illustrated
in Table 1 .
For multilevel modulation, there is a
need for an integrated optical receiver
with a good performance. The optical
receiver used with multilevel signaling
must deliver a linear response over a
large input optical power range. This
leads to a more sophisticated design of
the automatic gain control (AGC)
compared to the conventional binary
receivers, where a significant part of the
large dynamic range is achieved by a
simple limiting amplifier [11 ,12] .
To receive the multilevel optical signal
we need a high linearity optical receiver
with multilevel signal, the use of the
conventional optical receivers (with
limiting amplifier) will not be useful.
The use of conventional optical
receivers with limiting amplifiers will
result in a distorted output signal with
unequal voltage levels. This makes the
signal decoding very hard or even
impossible. In the design of the
multilevel signaling optical receiver a
linear optical receiver (no limiting
amplifiers) will be considered to have
equally spaced output signal voltage
levels. Also a linear AGC is needed to
have a constant output voltage over
wide input optical power. This equally
spaced output signal voltage levels will
ease the decision levels selection for
signal decoding from multilevel signal
to the original binary signal.
For many applications it is desirable to
integrate the photodiode (PD) with a
transimpedance amplifier (TIA) into the
same chip. Placing the TIA adjacent to
the PD improves the performance by
reducing lead capacitance and
sensitivity to interference, thereby
giving higher speed and lower noise. A
further advantage of the integration of
PD with TIA is the reduction in the
external circuitry required. Hence
overall cost and PCB board size can be
reduced.
Figure 3 Media converter [13]
Table 1 ETSI for 100 Mbit/s and 1Gbit/s PMMA POF systems
12 ISSUE 2
Figure 3 shows the general block
diagram of the media converter. Both
laser diode and fully integrated optical
receiver were mounted in a plastic
optical clamp. The laser diode is driven
with a commercial IC. The integrated
optical receiver (A3PICs) consists of an
integrated transimpedance amplifier
with an integrated 400μm diameter
photodiode, limiting amplifier and line
driver on a single chip. The use of high
performance fully integrated optical
receiver enables the media converter to
reach data rate of 1Gbit/s over 50m SI-
POF.
References
[1 ] “ETSI TS 105 175-1 V1 .1 .1 (2010-
01),”
http://www.etsi.org/WebSite/homepage.
aspx, January 2010.
[2] I. Mollers, , D. Jager, R. Gaudino, A.
Nocivelli, H. Kragl, O. Ziemann, N.
Weber, T. Koonen, C. Lezzi, A.
Bluschke, and S. Randel, “Plastic Optical
Fiber Technology for Reliable Home
Networking: Overview and Results of
the EU Project POF-ALL,” IEEE
Communications Magazine, vol. 47 no.
8, pp. 58–68, 2009.
[3]P. Polishuk, “Plastic Optical Fibers
Branch Out,” IEEE Communications
Magazine, vol. Volume: 44, Issue: 9, pp.
140–148, September 2006.
[4]R. Gaudino, E. Capello, G. Perrone,
G. Perrone, M. Chiaberge, P. Francia,
and G. Botto, “Advanced Modulation
Format for High Speed Transmission
over Standard SI-POF Using DSP/FPGA
Platforms,” POF Conference 2004,
Nuerberg, pp. 98–105, September 2004.
[5] F. Breyer, S. Lee, S. Randel, and N.
Hanik, “PAM-4 Signalling for Gigabit
Transmission over Standard Step-Index
Plastic Optical Fibre Using Light
Emitting Diodes,” 34th European
Conference and Exhibition on Optical
Communication (ECOC 2008),
Brussels, Belgium, vol. 3, pp. 81–82,
September 2008.
[6] S. C. J. Lee, F. Breyer, D. Cardenas, S.
Randel, and A. M. J. Koonen, “Real-
Time Gigabit DMT Transmission over
Plastic Optical Fibre,” Electronics
Letters, vol. 45 ,no. 25, pp. 1342–1343,
2009.
[7] S. C. J. Lee, F. Breyer, S. Randel,
R.Gaudino, G. Bosco, A. Bluschke, M.
Matthews, P. Rietzsch, H. P. A. van den
Boom, and A. M. J. Koonen, “Discrete
Multitone Modulation for Maximizing
Transmission Rate in Step-Index Plastic
Optical Fibers,” Journal of Lightwave
Technology, vol. 27, no. 11 , pp.
1503–1513, June 2009.
[8]O. Ziemann, L. Giehmann, P. E.
Zamzow, H. Steinberg, and D. Tu,
“Potential of PMMA Based SI-POF for
Gbps Transmission in Automotive
Applications,” the 9th International POF
Conference, Cambridge, pp. 44–48,
October 2000.
[9] F. Breyer, N. Hanik, S. Randel, and B.
Spinnler, “Investigations on Electronic
Equalization for Step-Index Polymer
Optical Fiber Systems,” Proceedings
Symposium IEEE/LEOS Benelux
Chapter, Eindhoven, pp. 149–152,
November/December 2006.
[10]F. Breyer, S. Lee, S. Randel, and N.
Hanik, “500-Mbit/s Transmission over
50 m Standard 1-mm Step-Index
Polymer Optical Fiber using PAM4-
Modulation and Simple Equalization
Schemes,” Proc. ePhoton One Summer
School , Brest, France, July 2007.
[11 ]M. Atef, R. Swoboda, H.
Zimmermann, An Integrated Optical
Receiver for 2.5Gbit/s Using 4-PAM
Signaling, The 22nd International
Conference on Microelectronic
(ICM2010), Cairo, Egypt, 2010.
[12]M.Atef, R.Swoboda,
H.Zimmermann, Optical receiver front-
end for multilevel signalling, Electronics
Letters Journal, January 2009.
[13] Olef Zieman, POF-Plus Handbook,
Handbook of the European POF-PLUS
Project 2008 - 2011 , 2011 .
Dr. Mohamed Atef
Assistant Prof. Electrical Eng. Dept. ,
Assiut Univerity, Egypt.
13ISSUE 2
JapanEDAInsight (1)
T-Engine – Steering the UbiquitousComputing World in JapanThroughout my experience studyingand living in Japan for over than fiveyears, I had the opportunity to meetwith many colleagues working in thefield of EDA whether in famous com-panies or in Universities, and togetherwith my humble Japanese language, thathelped me a lot to break the culture bar-rier/shock that any middle-east foreign-er pass through when he lives in afar-east country like China, Korea or Ja-pan with these amazing 10,000 Kanjicharacters. I managed to build a solidknowledge base of many ideas and tech-nologies influencing and driving the Ja-
panese community that I would like toshare with you in a series of articleswhich I write from the perspective of anEDA engineer. So I chose the name “Ja-pan EDA” as a title, and because thesearticles actually open the door for morereading and analysis, I complementedthe title with the word “insight” meaninga piece of Information. The topics wouldvary from the Educational System inUniversities, Research Organizations,Design Tools, Hardware Platforms, andmany other EDA trends, which I trulyhope that VLSI-Egypt magazine readersfind useful and interesting.
By: Muhammad AbdelSalam
State of the Art
14 ISSUE 2
Prof. Ken Sakamura is a Japanese pro-
fessor in Information science at the Uni-
versity of Tokyo. He is the creator of the
Real-Time Operating System architec-
ture TRON. As of 2006, Prof. Sakamura
started the ubiquitous networking
laboratory (UNL), located in Gotanda,
Tokyo as well as the T-Engine forum for
consumer electronics which contains
many Japanese companies and I was ac-
tually a member of representing my
University. The joint goal of Ubiquitous
Networking specification and the T-En-
gine forum, is to enable any everyday
device to broadcast and receive inform-
ation as shown in the use model of Fig-
ure 1 . In this article, I would present the
T-Engine development kit and in the
following article, I would present more
about its applications that I experienced
in Ubiquitous-Japan.
T-Engine Development Kit is an open
platform for embedded system develop-
ment offered as a total complete package
with H/W, S/W, and a development en-
vironment. T-Engine consists of three
boards: CPU, LCD, and debug boards,
as shown in Figure 2 and internal archi-
tecture is shown in Figure 3. The CPU
board mounts an SH7727 SoC
(SH3/DSP core) from Renesas Techno-
logy (There are actually different ver-
sions/series of T-Engine if you explore
more at http://www.t-engine.org/), in-
ternal clock 96 MHz/external clock 48
MHz, 8 MB flash memory, 32 MB
SDRAM, CF card interface, USB host,
PCMCIA interface, sound generator
chip, 2-channel serial interfaces to
which one channel is connected to the
host PC for debugging and download-
ing the application S/W, power supply
controller, and a Real Time clock (RTC).
The LCD board mounts a touch panel
interface, infrared remote control photo
acceptance unit, and a Key SW {SW1,
SW2, SW3}. The LCD panel is 240(H) x
320(V), 262,144 colors and it is con-
trolled by an SH7727 on-chip LCD con-
troller (LCDC). The CPU board has a
bus expansion slot to which a debugger
board is connected. The debugger board
is used mainly in our experiment to cal-
ibrate the embedded S/W running on T-
Engine using a JTAG (ICE) debugger for
(SH3/DSP) core connected via a JTAG
connector.
JTAG (Joint Test Action Group, IEEE
Standard 1149.1 and IEEE Standard Test
Access Port and Boundary Scan Archi-
tecture) was originally designed for
boundary scan. It allows read/write in-
ternal registers via limited number of
Fig.1 Ubiquitous Computing Applications
15ISSUE 2
pins (typically 5 pins). It is now becom-
ing popular to use JTAG interface and
use DCU (Debug control Unit) inside
the CPU. JTAG ICE provide very basic
debug feature such as run-control,
memory/register view/edit, download,
etc.
When developing applications on this
platform, we typically write it in C/C++
and link it with the T-Kernel RTOS,
other middle ware, device drivers, and
GUI widget components and use Part-
ner-J/SH, a JTAG ICE from Kyoto Mi-
cro-Computer, together with H-UDI
(Hitachi user-debugging interface) sup-
ported by the SH7727, to perform on-
chip debugging. The H-UDI is a serial
interface which is compatible with JTAG
specifications. The JTAG debugger is
accompanied with S/W debugger pro-
gram to monitor program execution,
and perform real-time trace.
Dr. Muhammad Abdel-Salam
Technical Lead,
Mentor Graphics Emulation Division
(MED)
Fig.3 T-Engine SH7727 Architecture
Fig.2 T-Engine Development Kit
Fig.4 Developing Applications on T-Engine
16 ISSUE 2
For almost 60 years now, Silicon has
been the crowned king for semicon-
ductors industry. It all started in 1954
when the first Transistor was invented
by Texas Instruments [1 ] .It enabled the
avalanche of a great revolution in the
electronics industry with minichips, mi-
crochips and IC’s.
Year by year scientists managed to pro-
duce smaller, cheaper, more functional-
ities and increased capacity of
transistors by unit area, now as we ap-
proach the final barriers of progress
where heat dissipation is becoming an
imminent barrier , a major change is
needed [2] .
It seem that diamond the hardest
known material on earth will soon be-
comes the most used element in elec-
tronic devices. Diamond can be used as
a semiconductor, which is something
not globally known.
It was by 1952 when scientists noticed
that blue diamond shows semiconduct-
ors P-type conductivity behavior, due to
the natural impurities inside it. Since
this time it was proven that diamond
could be used in electronic circuits
manufacturing but the main barrier was
the problem of manufacturing the dia-
mond itself [3] .
Naturally made diamond is produced
when carbon buried in High Pres-
sure/High Temperature (HP/HT) medi-
um for hundreds of years but with the
current technologies, such effects can be
simulated to produce synthetic dia-
monds in the matter ofmonths.
With the current advances and the us-
age of enhanced methods such as
“Chemical Vapour Deposition”
(CVD)[4] , the manufacturing process of
artificial diamonds is becoming much
feasible, which in return opens the way
for different diamond usage that wasn’t
acceptable before due to the cost and
uniqueness of each diamond piece.
As the manufacturing process of artifi-
cial diamonds becomes more economic
and much faster, the financial barrier is
becoming smaller on using diamonds in
electronic applications.
Diamond has few properties that could
make it very lucrative option such as:
1 - Its ability to disperse heat and con-
duct it 5 times faster than normal con-
ductive/semi conductive materials.
2- Has the highest density per atoms,
which allows 1 cm of diamond to with-
stand 10 million EV.
3- Extreme resistance against shock,
pressure, radiation and even acid.
All of this could play an important role
in high processing devices “i.e server
farms”. A big problem arises with the in-
crease demand on processing power is
the heat barrier and the need for more
power-hungry cooling systems however,
with diamond this problem is going to
be minimized.
For long time the major diamond deal-
ers considered the synthetic diamond as
an enemy to their market share but now
with the current technology advances,
they decided to work by the motto say-
ing, “if you can't beat them, join them.”.
Recently, De Beers a huge multinational
company in the field ofnatural diamond
has decided to extend it’s business do-
main into the huge industry of com-
puter chips by investing millions of
dollars in starting a patner company
named Element Six , that would spe-
cialize in manufacturing synthetic dia-
monds and exploring it’s potential usage
in power microwave electronics indus-
tries [5] .
Currently, research groups around the
world are racing to produce the first
diamond based commercial electronic
devices “i.e. Diodes, Transistors and
switches” [6] [7] .
For some time now, diamond have been
used in high duty applications such as
drilling heads in petroleum industry or
in laser optics application and now is
the time for making use of its near per-
fect electrical characteristics for another
revolution in the electronics industry.
References
[1 ]http://en.wikipedia.org/wiki/Transist
or
[2] .http://www.pbs.org/wgbh/nova/tech
/digital-diamonds-au.html
[3]http://www.semi1source.com/notes/v
iewfile.asp?Which=85
[4]http://en.wikipedia.org/wiki/Chemic
al_vapor_deposition
[5]http://www.investmentu.com/2011/D
ecember/synthetic-diamonds-to-
replace-silicon-in-microchips.html
[6]http://www.youtube.com/watch?v=h
mhxX4JI3Uw
[7]http://www.aist.go.jp/aist_e/latest_res
earch/2005/20050615/20050615.html
Ahmed Sadek
Research Assistant at Nile University
Information Security Department
Diamond is thenew SiliconBy: Ahmad Sadek
State of the Art
17ISSUE 2
FPGA 9-1-1By: Muhammad AbdelGhany
FPGAs or field programmablegate arrays are an attractive solu-tion for many design challenges.The adoption of FPGA in newerdesigns is increasing rapidly andit's significantly compared tonumber of successful new ASICdesigns and the total growth insemiconductors market.
In this series we will try to get youthrough successful design withFPGAs while will highlight differ-ent design methodologies andguidelines to help you efficientlyuse FPGA in your next design.
Technical
18 ISSUE 2
Market for Standard Cell ASICs, ASSPs, and Programmable LogicIBS, Inc., “System IC Market Trends” (Q2/2010 Global System IC
Industry Service), August 2010
ProgrammabilityFPGA or field programmable gate array
is a programmable device just like tradi-
tional microcontrollers and processors.
A designer just needs to translate his
design requirements into chunks of
code segments, integrate them, simulate,
debug, compile and then download to
the device to test or run the design.
The difference between a MCU and FP-
GA is "what is programmable?" MCU is
a fixed structure that can execute a fixed
set of pre-defined operations (instruc-
tions). Designer needs to translate his
design into a list of those operations ex-
ecuted in a way to implement the re-
quired behavior of the design. It's the
order of those operations and the flow
of data that differentiate one design
from another and do the required job.
Unlike MCU, FPGA has a program-
mable structure. The idea is to be able to
alter what the hardware can do not what
the hardware should do.
MCU built to be as generic as possible
to cover a range of applications possibil-
ities. One can split all MCU available in
the market into different categories ac-
cording to the applications that those
controllers are best fit. A vendor of
MCU cannot meet the requirements of
all market segments in a single device.
Some of market segments require a su-
per performance while others need
MCU with minimum power consump-
tion or the cheapest MCU. All or some
of those requirements can conflict with
each other and this is why it's im-
possible to have a single MCU that can
do everything and everywhere. In the
market a MCU that targeting industrial
applications can totally differ than a
MCU that targeting telecommunication
industry. The difference can reach the
whole architecture not only a set of
peripherals.
The main drawback of this is that de-
signers need to learn many and different
design environments, processors archi-
tectures, debugging and verification
tools, software libraries, operating sys-
tems, etc. Even if designer is not shifting
his or her job in different market seg-
ments, he or she may need to learn new
tools and processor architecture for a
new design in the same market seg-
ment. This leads to reducing ROI and
increasing time to market.
FPGA made to solve such issues and
many other issues in embedded system
designs using regular MCU.
MethodologiesDesign with FPGA is basically a kind of
digital logic design. The designer needs
to translate his architecture into a mix of
state machines, logic circuits, memories,
registers, etc. which requires good un-
derstand of underlying architecture of
FPGA.
Some of FPGA design methodologies
adopted higher level of abstraction in
which the designer can model the sys-
tem using high level languages like
Matlab or C or even using a graphical
model. Then a specialized compiler can
translate this model to a synthesizable
RTL code. This methodology was first
defined as Electronic System Level
(ESL) design by Gartner Dataquest, a
EDA-industry-analysis firm, on Febru-
ary 1 , 2001 .
ESL mainly depends on a set of pre-
designed library of primitive models
like RAMs, ROMs, shift registers, pro-
cessors, state machines, encoders, de-
coders, math functions, etc. The
designer should utilize those primitives
in his design with limited configuration
possibilities. The final results mainly
depend on the capabilities of the com-
piler to merge down this model to the
minimum set of FPGA primitive slices.
ESL methodology still not adopted in
wide scale in the industry as it not giv-
ing the optimum results of profession-
ally hand-written code.
Even if you are happy with the results of
ESL based tools, you still need to take
the decision of which FPGA to use in a
certain design. Selecting a device for a
design depends on the understanding of
your design requirements, possible fu-Fig.1 ASICs, Microcontrollers, and FPGAs
Functional Comparison
19ISSUE 2
ture updates and features, and how this
will translate to the infrastructure of a
FPGA device. Again you need to master
the device architecture and learn how
specific device architecture can fit with
your design requirements.
The good news that newer generations
of FPGA are mostly built on the same
technology but differ in terms of per-
formance, cost, pin count, and resources
count. An example of this is the 7th
series of Xilinx FPGAs. This will reduce
the time selecting a device for every new
design.
StructureFPGA have a programmable structure.
In the heart of this structure is pro-
grammable interconnection matrix and
configurable logic slices. This program-
mable structure is surrounded by pro-
grammable input/output blocks.
Understanding this structure and its
capabilities will help you to effectively
translate your design to a synthesizable
logic on this structure. Unlike ASIC
design, in FPGA you are limited to the
capabilities of the configurable logic
cells and the interconnection between
them. You cannot pack a bigger logic
circuit in a physical location than the
configurable blocks can handle.
This is one cause of performance de-
gradation in FPGA designs over similar
ASIC designs. For instance a simple de-
coder circuit can run much faster in
ASIC implementation than FPGA im-
plementation, why?
AnswerInside configurable logic blocks there
are programmable Look-Up Tables or
LUT. LUT is the primary method to
implement Boolean functions in FPGA.
We can consider LUT as single bit
ROM. The depth of that ROM depends
on number of possible inputs to LUT.
Today's FPGAs are based on 4 or 6 in-
put LUT technology. 4-input LUT can
implement any 4 input Boolean func-
tion. See figure 5 for example.
So what is the case if we need 5-input
Boolean function?
Fig.2 Example ofESL design flow using commercial design package fromBinachip, Inc.
Fig.3 FPGA Structure.
Fig.4 Example ofConfigurable LogicCell.
20 ISSUE 2
As we can see in Figure 6, increasing
the width of input enforces the synthes-
izer to split the function into multiple
LUTs and then combine those using ad-
ditional LUTs or multiplexers.
Every LUT has an intrinsic constant
delay regardless the complexity of the
function it implements. Splitting larger
functions in many LUTs creates multiple
logic levels and the total delay increases
as the number of logic level increases
too. Additional to delays in the logic
cells, there are delays in the routing
between them and this is the killing
factor in FPGA if you are seeking a
higher performance. Routing delays are
very significant especially in large logic
functions when the design must spread
over large areas of the same FPGA die.
In ASIC the situation is bit different as
wider Boolean functions can be imple-
mented using custom logic cells and
minimizing routing delays by physically
locate those cells very close on the die.
The solution to such issues is design
dependent. Some designs can adopt
pipelining in logic functions imple-
mentation breaking down the total
delay into multiple clock cycles (the
same concept of instructions pipelining
in microprocessors). In other designs we
can manually place and route logic im-
plementation in tight areas inside FP-
GA. In either of both solutions we must
first try to optimize the functions for
best implementation on FPGA.
NextIn next episodes of this series we will
introduce different FPGA features and
discuss the effective utilization of them.
We will provide valuable design tips for
those planning to use FPGA in a real
design. Also we will introduce design
methodologies in more details. We
would like to listen to your feedback
and suggestions regarding those topics
that you think we should cover in this
series.
Muhammad Abdel-Ghany
Founder & Chairman of Luxor Instru-
ments.
Fig.5 4-input Boolean function implementation in Xilinx Spartan-3E device.
Fig.6 5-input Boolean function implementation in Xilinx Spartan-3E device.
21ISSUE 2
60 GHz TransceiversExamples
Current Trends in RF and Microwave Integrated Circuits Research (2)
RFIC and microwave-IC research can bedivided into many areas such as ultralow power RF frontends, wide bandcircuits and cognitive radio design, theaim of this research it to build auniversal radio frontend, self healing RFcircuits with Built In Self Test (BIST),and millimeter-wave circuits and
systems from 30-300 GHz frequencyrange. Within this article currenttrends in millimeter-wave research areaddressed from devices performancemetrics to highly integrated radiofrontends. I t also provides some designaspects and precautions for such highfrequency circuits and systems.
By: Muhammad Elkholy
State of the Art
22 ISSUE 2
RFIC and microwave-IC research can bedivided into many areas such as ultralow power RF frontends, wide bandcircuits and cognitive radio design, theaim of this research it to build auniversal radio frontend, self healing RFcircuits with Built In Self Test (BIST),and millimeter-wave circuits and
systems from 30-300 GHz frequencyrange. Within this article currenttrends in millimeter-wave research areaddressed from devices performancemetrics to highly integrated radiofrontends. I t also provides some designaspects and precautions for such highfrequency circuits and systems.
FETs vs. HBTsBefore analyzing the two devices using
small signal models, it is important to
compare their fundamental physical
constructions shown in Fig. 1 . In field
effect transistors, the controlling charge
resides on the gate, is of opposite sign,
and is physically separated from the
controlled charge which travels through
the channel between the source and
drain. Since only one type of carriers
(electrons or holes) contributes to cur-
rent flow in the active mode of opera-
tion typically employed in HF circuits,
FETs are also described as unipolar
devices. On the contrary, in HBTs, the
controlling charge (holes in npns, elec-
trons in pnps) is collocated in the base
with the controlled charge (electrons in
npns, and holes in pnps). This explains
the bipolar nature of these devices. The
second and third main differences
between the two structures reflect the
direction of current flow, and the tech-
nological control of the minimum fea-
ture size, gate length, L, and vertical
distance between emitter and collector,
for FETs and HBTs respectively. Intrins-
ic FET speed is therefore driven by
lithography whereas HBT speed is de-
termined by the precision with which
we can grow thin semiconductor layers
vertically, for example by atomic layer
deposition techniques. Historically, ver-
tical control of semiconductor layers has
been several generations ahead of litho-
graphic resolution and less costly to
realize. However, as devices are scaled to
smaller dimensions, 3D parasitics be-
come dominant and limit real device
performance. Therefore, the most ad-
vanced HBTs and FETs require scaling
in both vertical and lateral dimensions.
All the features discussed above are
summarized in Table I.
Nanoscale MOSFETs show many bi-
polar-like features, such as gate leakage
current not unlike the base current of
the HBTs, exponential sub threshold
behavior, similar output characteristics,
and almost identical in form (although
physically different) small signal and
noise equivalent circuits. For the high
frequency circuit designer, designing
with either FETs or HBTs should be al-
most transparent.
The high frequency small signal equi-
valent circuits of both nMOS and HBT
devices is shown in Fig. 2. In the HBT at
high frequency, Cπ is dominating the
input impedance of the device and rπ
can be omitted. And this reduces the
equivalent circuit of the HBT to the
equivalent circuit of the MOS device
and one general small signal model can
be used as depicted in Fig. 3. This small
signal circuit is a good approximationFig. 1 Charge control principle in FETs and HBTs.
Table I. Comparison ofFETs and HBTs
Fig.3 general small signal equivalent circuit of both HBT and nMOS atmillimeter-wave Frequencies.
Fig.2 Small signal equivalent circuit of both MOS and HBT in high frequencies.
23ISSUE 2
for first order hand calculation and
design. Then more accurate simulation
with either BSIM 4 for CMOS or VIBIC
& HICUM models for HBTs should be
carried out for final design steps.
60 GHz Transceivers Examples60 GHz Communication Link Budget
Calculations.
Prior to presenting highly integrated 60
GHz transceivers, 60 GHz communica-
tion system link budget calculation
should be carried out. This will demon-
strate why beamforming and phased ar-
ray systems are of great interest in
millimeter-wave communication sys-
tems.
Assuming that 1 dB compression point
of the transmitter is 10 dBm and the
noise figure of the receiver is 10 dB.
Those are reasonable numbers for cur-
rent silicon technologies (90 nm
CMOS). The bandwidth is 1 GHz, this
leads to – 74 dBm noise floor of the re-
ceiver. The path loss of 60 GHz signal is
68 dB for 1 meter distance. For simple
modulation such as FSK or QPSK the
required SNR at the demodulator is
from 10 to 14 dB. Using those numbers
one can see that there is no link margin
for fading or shadowing.
One solution for this problem is using
highly directive antennas; this will en-
hance the link margin by the antenna
gain in both sides TX and RX. Highly
directive antennas are suitable for con-
sumer electronics market specially
WLAN and WPAN. Here arises the
need to develop phased-array transmit-
ters and receivers working at mm-wave
frequencies to provide high link gain
without sacrificing angular coverage.
The main advantage of the phased-ar-
rays is that electronics beamforming
and steering can be achieved. In trans-
mitters, phased-arrays are used to in-
crease the effective isotropic radiated
power (EIRP), while in receivers, they
are used to increase the signal to inter-
ference-noise ratio (SINR). Higher EI-
RP and SINR are translated into higher
bit rate and longer distance.
Phased array system can be defined as:
Multiple antenna system, electronically
change the direction of the beam, by in-
troducing a phase shift and amplitude
control for each element as shown in
Fig. 4. Assuming that the transmitter
will has 4 TX elements and the receiver
has 4 receive channels. The link budget
will increase by 18 dB (12 from TX and
6 from the RX). As a result the system
will be more robust and high distance
could be achieved.
Single Element Transceiver
Recently fully integrated 60 GHz trans-
mitter and receiver have been presented
in the literature. In this section an ex-
ample of 60 GHz transmitter and re-
ceiver will be presented. Both
transmitter and receiver are based on
sliding IF architecture, in which one
PLL or frequency synthesizer is used to
drive both RF and IF mixers. This will
eliminate the need of two local oscillat-
ors hence reducing the complexity of
the chips. Fig. 5 shows the block dia-
gram ofboth TX and RX.
In transmitter, the 48 GHz output of the
PLL is divided by 4 to generate 12 GHz
I/Q LO signal to drive the IQ modulat-
or. The differential baseband input are
fed to the modulator through baseband
VGAs. The output signal of the modu-
lator is up-converted with the Gilbert
cell mixer. The 60 GHz generated signal
is then filtered using on chip integrated
passive filter and fed to the power amp-
lifier. The purpose of the filter is to at-
tenuate both the image at 36 GHz and
the VCO feed-through at 48 GHz. A
strong feedthrough signal would affect
the linearity of the receiver frontend,
because the transmitter and receiver
antennas are placed in close proximity.
The Q-factor of integrated inductors for
the 60 GHz range is low (typically 15 to
20), resulting in high insertion loss and
limited selectivity. For a compact
design, a lumped element filter type was
chosen. The measured insertion loss is
3.3 dB at 60 GHz. The image rejection
at 36 GHz is 27.7 dB. The VCO feed-
through at 48 GHz is attenuated by 12
dB.
The saturated output power of the
power amplifier is around 20 dBm de-Fig.4 A generic beam forming system employs arrays of antennas and transceivers
to boost gain, power, and sensitivity.
24 ISSUE 2
livered to differential 100 Ω antennas. It
consists of three stages of cascode amp-
lifiers offering high gain around 30 dB
@ 60 GHz. The PA layout is drawn sym-
metrically utilizing the ac ground for the
differential signal at the symmetry line
of the layout. The matching topology
between stages is an L-C structure. The
inductances were realized as lines in the
top metal layer. They were bent in the
layout in order to end at the symmetry
axis to utilize the ac ground. The modu-
lation scheme used for data transmis-
sion is OFDM which is sensitive to
nonlinear distortion. This means that
the PA has to be optimized for high out-
put P1dB rather than for saturated out-
put power. To achieve a high P1dB, a
class A PA was chosen. The PA draws
190 mA from a 3.7 V supply. Full de-
scription and schematic of the PA is
presented in [1 ] .
The receiver exhibits the same architec-
ture; a differential architecture is adap-
ted from antenna to baseband because
of its robustness with respect to bond
wires and its common-mode rejection
ability. The quadrature LO signals for
the second down-conversion mixers are
generated by a divide-by-four circuit in
the PLL chain, which gives perfect IQ
signals because both I and Q signals re-
spond only to the rising edge of the 48
GHz VCO output.
The receiver front-end consists of a low-
noise amplifier (LNA), a mixer, a PLL
and an IF demodulator. The LNA is a
three stage common emitter amplifier
with 18 dB gain and 22 GHz bandwidth.
Its main design and implementation is-
sues have been shown in [2] . Minor
modifications have been made to the
LNA for technology and process migra-
tion. However, only the simulated over-
all noise figure was given in [2] due to
the lack of noise measurement. The 12
GHz demodulator is designed to have a
conversion gain of 50 dB with more
than 30 dB gain control range. An SPI is
introduced for the gain and IQ VGA
mismatch control, thereby reducing the
number of bond pads. The complete re-
ceiver analogue frontend has a 78 dB
conversion gain. The RX consumes 980
mW from two different supplies: 3.3
and 2.5 V.
PLL Synthesizer
The realization of an RF PLL at fre-
quencies from 58.32 to 64.8 GHz is very
challenging, especially for the VCO and
the first frequency divider stage. Using a
sliding-IF topology, the PLL frequency
is reduced to 80 percent of these values.
As a result, the following frequencies
must be generated: 46.656 GHz, 48.384
GHz, 50.112 GHz and 51 .84 GHz. If two
VCOs selectable by a digital command
are used, a VCO tuning range of
2.16×0.8=1 .728 GHz plus safety margin
is required. The feasibility of such a
solution was first shown in [3] , where
an integrated 48 GHz PLL in SiGe-BiC-
MOS with 2.4 GHz tuning range has
been demonstrated. The PLL, unlike
other TX and RX blocks, features both
bipolar and CMOS transistors. The
VCO and static dividers are realized
with bipolar transistors, while the PFD
and charge pumps employ CMOS tran-
sistors. The 48 GHz-band PLL is used to
down-convert the 60 GHz RF signals to
the 12 GHz-band. A sliding IF of about
12 GHz is generated from the 48 GHz
VCO using a 1 :4 frequency divider,
which is used for the downconversion to
baseband with an I/Q demodulator. Fig.
6 shows the basic PLL architecture sug-
gested in [5] . The VCO output fre-
quency is divided by four to generate
quadrature signals at IF of about 12
GHz. This signal is divided by the pro-
grammable divider consisting of a 1 :45
bipolar divider and a programmable
low-speed CMOS divider. The phase-
Fig.5 60 GHz TX and RX architecture
25ISSUE 2
frequency detector (PFD) compares the
9.6 MHz signal of the divided crystal
frequency with the divided VCO fre-
quency. The PFD output is connected to
a high-current charge pump (CP1) for
VCO fine tuning and a low-current
charge pump (CP2) for coarse tuning.
The two parallel tuning loops allow a
low phase noise, a low level of reference
spurs and a large tuning range to be
achieved simultaneously. The voltage di-
vider at the output of CP1 keeps the dc
voltage at the VCO fine tuning input
roughly constant [4] , which makes the
loop bandwidth fairly independent of
device parameter variations with pro-
cess, supply voltage and temperature
(PVT). The total tuning range is defined
by the slow coarse tuning loop, in which
the noise of CP2 is almost eliminated by
a large external capacitor. In this trans-
ceiver, a 48 GHz PLL as described in [2]
contains only one VCO. An IEEE com-
patible PLL using an array of two VCOs
has been tested successfully and will be
included in a future design. The meas-
ured PLL phase noise at 1 MHz offset
was -98 dBc/Hz [3] . This corresponds to
an integrated RMS phase error after
baseband filtering below 1 .5o, which is
more than sufficient for 16-QAM OF-
DM transmission [5] .
ConclusionIn this article, we presented an intro-
duction to the current research in milli-
meter wave integrated circuits. It also
illustrates the new applications in which
millimeter wave circuits and systems
can be applied. Limitation and chal-
lenges for millimeter wave design in
current silicon technologies are ad-
dressed. Finally a simple description of
one of the state of the art 60 GHz trans-
mitter and receiver is provided.
In the next articles we will focus on mil-
limeter wave phased array systems from
architecture to circuit implementation
in silicon technologies.
References[1 ] Glisic, S.; Scheytt, J.C.: A 13.5-to-17
dBm P1dB, Selective High-Gain Power
Amplifier for 60 GHz Applications in
SiGe, Bipolar/BiCMOS Circuits and
Technology Meeting (BCTM), Oct.
2008.
[2] Sun, Y.; Borngräber, J.; Herzel, F.;
Winkler, W.: A Fully Integrated 60 GHz
LNA in SiGe:C BiCMOS Technology,
Bipolar/BiCMOS Circuits and Techno-
logy Meeting (BCTM), Santa Barbara,
USA, Oct. 2005, pp. 14-17.
[3] Herzel, F.; Glisic, S.; Winkler, W.: In-
tegrated frequency synthesizer in SiGe
BiCMOS technology for 60 and 24 GHz
wireless applications, Electronics Let-
ters, 43 (2007), 154–156.
[4] Herzel, F.; Osmany, S.A.; Scheytt,
J.C.: Analytical phase-noise modeling
and charge pump optimization for frac-
tional-N PLLs. IEEE Trans. Circuits
Syst. I, Regular Papers, 57 (2010),
1914–1924
[5] Herzel, F.; Choi, C.-S.; Grass, E.:
Frequency synthesis for 60-GHz OFDM
transceivers, European Conference on
Wireless Technology (EuWiT2008),
Amsterdam, 2008, pp. 77-80.
Muhammad Aly El-Kholy
Microwave, milli-meter researcher , IHP
M.Sc Electronics and communications
Fig.6 Schematic of frequency synthesizer.
26 ISSUE 2
Sponsors Corner
Reserve your area now inthe Sponsors Corner
Deliver Your message with us
For more details, please contact us at:email: [email protected]: +20122 4100 229