Radio Design for MIMO Systems
with an Emphasis on IEEE
802 11
802.11
Arya Behzad
IEEE DL Series
August 2011
A. Behzad
IEEE DL 2011: MIMO Radios
1
Outline
 Introduction,
I t d ti
Market
M k t Overview
O
i
& Market
M k t Trends
T d
 Communication System Performance in a Multipath Channel:
 The problem with multipath
 Mitigation methods for multipath
m ltipath
 Taking advantage of multipath
 WLAN Evolution & 802.11
 Additions in 802.11n
802 11n PHY and impacts on radio
 802.11n Sensitivity and EVM requirements
 Some optional modes of 802.11n
 Key features of TGac
 Some 802.11 Radio Requirements
 Impacts of some radio impairments on system performance
 Examples of Circuit Techniques and Architectures Used in 802
802.11n
11n Radios
 Evolution of Radios: SISO to MIMO
 Real World Performance
 Conclusion
A. Behzad
IEEE DL 2011: MIMO Radios
2
WLAN in Top Gadgets
 9 of 10 have integrated WiFi
A. Behzad
IEEE DL 2011: MIMO Radios
3
Wireless Connectivity Attach Rates: Up, Up, & Away!
2008
Wi-Fi GPS Bluetooth (BT)
Wi-Fi
BT
Wi Fi
Wi-Fi
BT
2013
Wi-Fi
Bluetooth (BT) GPS
Wi-Fi
Bluetooth (BT) GPS
Wi-Fi
Bluetooth (BT) GPS
Wi Fi
Wi-Fi
Bl t th (BT) NFC
Bluetooth
Wi-Fi
Bluetooth (BT) NFC
NFC
NFC
NFC
Low
Medium
Wi-Fi
High
GPS BT
GPS
BT
Wi-Fi
BT
Wi-Fi BT
GPS
Source: ABI Research, 2010
BT
% WC Attach Rate
A. Behzad
IEEE DL 2011: MIMO Radios
% WC Attach Rate
4
Wi-Fi Growth Driven by CE Adoption and its Utility!
Wi--Fi Chipset Shipments
Wi
2,500
3.8B Units
(over 3 years)
Units (Millionss)
2,000
1 500
1,500
1,000
500
0
2007
2008
2009
2010
2011
2012
2013
2014
2015
Source: ABI Research 2010
Wi-Fi Device Demand Expected to Surpass Cellular in 2014!
A. Behzad
IEEE DL 2011: MIMO Radios
5
A Snapshot of Wi-Fi® Evolution
HDTV
STB
Media Adapter
Media
Blu-ray
Netbook
Printer
Game System
Cable Gateway
Connectivity
Notebook PC
Mobile Phone
Camera
Mobilityy
Kindle
3G Router
802.11 Router
Portable Audio
Player
DSL Gateway
Internet
A. Behzad
Desktop PC
IEEE DL 2011: MIMO Radios
Portable Video
Player
Portable Game
D i
Device
6
Mobility Trend in 2011
Number 1 - Speed
802.11n 2x2 MIMO
- Higher Speed Over 150Mbps
- Connect to 47% of 2.4GHz
MIMO APs
- Perfect for “Tablet Application”
Number 2 - Connection
Convenient P2P Connection
- WiFi Direct
-SoftAP Modem
Number 3 - 2.4GHz Traffic
2.4GHz is getting congested
- Smart Coexistence Required
- 2.4GHz band for Internet Access
- 5GHz band off-load for P2P
WLAN, Bluetooth,
Zigbee, Cordless
phone, etc
2.4GHz
A. Behzad
IEEE DL 2011: MIMO Radios
WLAN Only
5 GHz
7
Outline
 Introduction,
I t d ti
Market
M k t Overview
O
i
& Market
M k t Trends
T d
 Communication System Performance in a Multipath Channel:
 The problem with multipath
 Mitigation methods for multipath
m ltipath
 Taking advantage of multipath
 WLAN Evolution & 802.11
 Additions in 802.11n
802 11n PHY and impacts on radio
 802.11n Sensitivity and EVM requirements
 Some optional modes of 802.11n and 802.11ac
 A brief introduction to 802
802.11ac
11ac
 Some 802.11 Radio Requirements
 Impacts of some radio impairments on system performance
 Examples of Circuit Techniques and Architectures Used in 802
802.11n
11n Radios
 Evolution of Radios: SISO to MIMO
 Real World Performance & Multi-Vendor Interoperability
 Conclusion
A. Behzad
IEEE DL 2011: MIMO Radios
8
Multipath Channels
 Multipath (Rayleigh fading):
 Is a phenomena caused by the multiple arrivals of the transmitted
signal to the receiver due to reflections off of “scatterers”
 For traditional wireless systems, more problematic if a direct line-ofsight (LOS) path does not exist between the transmitter and the receiver
(Rayleigh Fading)
 Example and vector-magnitude representation shown below
Access Point
Client
A. Behzad
Fi after
Fig.
ft reff [1]
IEEE DL 2011: MIMO Radios
9
Multipath Channels
 Received signal power as a function of receiver-to-transmitter
distance for a multi-GHz transmission in a multi-path indoor
environment is shown below
 Received signal power can vary quite significantly with a slight change
in distance
Fig. after ref [3]
A. Behzad
IEEE DL 2011: MIMO Radios
10
Performance in a Multipath Channels
 In an AWGN channel, a “reasonable” SNR is sufficient for obtaining a low
probability of error
 The only factor that can impact an error is large additive noise
 In a faded environment, a very large SNR would be required for a low
probability of error
 Not a significant difference between a coherent or non-coherent scheme
 Probabilistically, in the high SNR region, errors often occur due to the deep fade
present in the channel
Fig. after ref [4]
A. Behzad
IEEE DL 2011: MIMO Radios
11
Diversity
 Without the use of some kind of diversity
diversity, performance in a faded
environment is quite poor
 Error probabilities decay quite slowly with an increase in SNR
 The fundamental reason is that the reliability of the link is dependent on
a single net “path” (i.e. no “diversity”)
• High probability that this single net path will be in a deep fade
 The technique used to improve the situation by providing multiple
independently faded signal paths is “diversity”
 As long as at least one of the signal paths has a reasonable SNR,
reliable communication can be achieved
A. Behzad
IEEE DL 2011: MIMO Radios
12
Diversity
 One or more dimensions ((“degrees
g
of freedom”)) can be
exploited in a faded wireless system to enable diversity
 Time
• Interleaving of coded symbols
 Frequency
• Can be applied when bandwidth of the modulated signal is wider
th the
than
th coherence
h
bandwidth
b d idth off th
the channel
h
l
• Frequency response of the channel is not flat over the
bandwidth of the signal
• Can be implemented in the form of:
– Inter-symbol equalization
– Spread-spectrum techniques
– OFDM
 Space
• Use of multiple Rx and/or Tx antennas
– Selection diversity
– Space-time coding
A. Behzad
IEEE DL 2011: MIMO Radios
13
Diversity
 Di
Diversity
it can be
b utilized
tili d to
t improve
i
the
th reliability
li bilit off th
the lilink
k
 Repetition coding is the simplest form of diversity
• Same information is communicated over multiple signal paths
• Can
C b
be applied
li d tto titime, ffrequency or space di
diversity
it
• Is quite wasteful of the degrees of freedom of the channel
 Using more sophisticated coding techniques allows to obtain
diversity gain as well as a “coding
coding gain”
gain
A. Behzad
IEEE DL 2011: MIMO Radios
14
OFDM Subdivision of Wideband Modulation
 The received signal
g
in a multi-path
p
environment will suffer “fades”
 For wideband channels (often required for high data rate
communications), the fade is often frequency-selective
 OFDM coding is a very effective method of combating frequencyfrequency
selective fades
-8.125
8.125 MHz
Multipath channel
response
Broadband channel with no OFDM subdivision
-8.125
-26
26
A. Behzad
-.312
.312
8.125 MHz
-1
1 +1
+26
S b
Subcarrier
i IIndex
d
Broadband channel with many OFDM (narrowband) subcarriers
IEEE DL 2011: MIMO Radios
15
OFDM Coding
 802.11a/g
802 11 / standards
t d d are b
based
d on O
Orthogonal
th
lF
Frequency Di
Division
i i
Multiplexing (OFDM)
 The OFDM concept has been around for a long time
 OFDM provides some immunity to multi
multi-path
path
 Orthogonality:
 Obtained through the use of multiples of a sub-carrier frequency over an integer
cycle (property of transform used DFT/IDFT)
 Sub-carrier over-lap is allowed
• Better spectral efficiency can be achieved
 Is deteriorated by phase noise, distortion, frequency inaccuracy, IQ imbalance, …
• Cause inter-subcarrier interference
f
A. Behzad
IEEE DL 2011: MIMO Radios
f
16
Multi-Antenna Systems
 The use of multiple
multiple-antennas
antennas is a very popular technique and the
subject of very active research
 Multiple antennas can provide:
 Power gain
• Can be achieved with SIMO, MISO, or MIMO
• Rx Maximum ratio combining
• Tx beam forming
g
 Diversity gain
• Can be achieved with SIMO, MISO, or MIMO
• Selection diversity
• Space-time coding
 “Degree of freedom” gain
• Can only be achieved with MIMO
– Requires far-apart transmitter antennas OR far-apart receiver antennas in a non
multi-path environment
– Reasonably closely spaced Rx and Tx antennas can be used as long as
scatterers and reflectors are placed far apart in the environment
» We
W are taking
t ki advantage
d
t
off the
th multi-path
lti th environment!!
i
t!!
• Allows use of spatial multiplexing to increase the capacity of the channel
A. Behzad
IEEE DL 2011: MIMO Radios
17
Multi-Antenna Systems: Spatial Diversity
 Antenna
A t
Diversity
Di
it iis also
l referred
f
d tto as S
Spatial
ti l Di
Diversity
it
 Can be achieved by using multiple antennas at the transmitter or the
receiver
 Antennas are required to be placed “sufficiently” far apart in order to
create (reasonably) independent fading channels
 Antenna separation is a function of the carrier wave length and the
scattering environment
• For an indoor environment, an antenna separation of greater than ½ carrier
wavelength is typically required
 Systems
S t
with
ith multiple
lti l ttransmit
it and
d multiple
lti l receive
i antennas
t
offer
ff
more than just diversity.
 They offer degrees of freedom which can be exploited for increasing the
channel capacity
For an excellent detailed discussion of the communication theory aspects of
multi-antenna systems see ref. [4]
A. Behzad
IEEE DL 2011: MIMO Radios
18
Multi-Antenna Systems: Spatial Diversity
 Multi-Antenna systems can be implemented with Rx and/or Tx
di
diversity
it
 If each Rx antenna has a complete analog path to digital, Rx Diversity is
often referred to as single-in, multi-out (SIMO)
 If each
hT
Tx antenna
t
h
has a complete
l t analog
l path
th ffrom di
digital,
it l T
Tx Di
Diversity
it iis
often referred to as multi-in, single-out (MISO)
 If each Rx and Tx antenna has a complete analog path to and from digital,
Rx and Tx Diversity is often referred to as multi
multi-in,
in multi
multi-out
out (MIMO)
Ant1 Tx
Ant1.Tx
Ant1.Tx
1
Ant 1
Ant.1
Rx
Ant.1
Rx
Ant.Tx
Ant.1
CHANNEL
CHANNEL
Rx
CHANNEL
Ant2 Tx
Ant2.Tx
A t2 T
Ant2.Tx
Ant 2
Ant.2
Rx
Ant.2
Rx
Rx Diversity
A. Behzad
Tx Diversity
IEEE DL 2011: MIMO Radios
Rx and Tx Diversity
19
Selection Diversity
 In a simple
p Rx selection-diversityy system:
y
 Received power at each antenna is examined one after another
(during preamble processing, for example)
• Often a “diversity
diversity switch”
switch is used to multiplex the antennas to the
common receiver block
 The antenna path with the largest signal strength is selected
Ant.Tx
CHANNEL
Tx
A. Behzad
L > λ/2
Ant.1
Ant.2
Ant. Div.
S it h
Switch
Radio
IEEE DL 2011: MIMO Radios
ADC
PHY + MAC
20
Analog Maximum Ratio Combining (MRC)
 MRC is a form of Rx diversity
y that is more sophisticated
than Rx selection diversity
 MRC works by simultaneously enabling multiple receive
antennas,, equalizing
q
g the g
gain on each p
path,, and aligning
g g the
delays of the received signals before (coherently) combining
them
 If g
gain is not equalized
q
((only
y delays
y are aligned)
g
) system
y
is often
referred to as a “single weight combiner” or SWC.
t
t
Ant.1
CHANNEL
Tx
L > λ/2
Ant.Tx
t
Var Delay
t
Ant.2
t
+
Var Delay
A. Behzad
IEEE DL 2011: MIMO Radios
21
Analog Maximum Ratio Combining (MRC)
 In an MRC system,
system signals on each path are added coherently
coherently,
whereas noise for each path is added incoherently
 MRC allows for an SNR improvement of up to [10 log(m)] over a Rx
selection diversity scheme
scheme, where m is the number of Rx antennas
• 3dB SNR improvement for 2 antenna system
 MRC can be performed in the digital domain also (with full
independent Rx paths to baseband)
 More on this later
A. Behzad
IEEE DL 2011: MIMO Radios
22
Analog Maximum Ratio Combining (MRC)
 When the bandwidth of the modulated signal is narrow as compared
to the carrier frequency, a variable delay block can be replaced with
a programmable phase shifter which is much simpler to implement
 IIn this
thi case th
the MRC system
t
iis also
l referred
f
d tto as a ““phased
h
d array””
 MRC/SWC Phased array can be implemented in the LO path (e.g. ref.
[12])
 MRC/SWC Phased array can be implemented in the signal path (e
(e.g.
g
ref. [13])
Ant 1
Ant.1
Ant.1
Rx
Rx
Phase Shift
Ant.2
Rx
LO
Phase Shift
Ant.2
LO
Rx
Phase Shift
A. Behzad
IEEE DL 2011: MIMO Radios
23
MIMO
 A n X m MIMO system can offer a diversity gain of up to n X m
 A multi-antenna
lti t
system
t
with
ith a ffullll analog
l path
th ffor each
h antenna
t
allows for maximum system performance and flexibility
 Phase shifting, combining, beam-forming, etc. can be performed easily
at digital
g
baseband
• On a per subcarrier basis in OFDM systems, for example
 Allows for “spatial multiplexing”
 Consumes more power and area than analog-only multi-antenna
solutions
Ant1.Tx
Ant.1
Rx
ADC
DAC
CHANNEL
Tx MIMO
PHY + MAC
A t2 T
Ant2.Tx
A t2
Ant.2
Rx
ADC
DAC
A. Behzad
Rx MIMO
PHY + MAC
IEEE DL 2011: MIMO Radios
24
MIMO
 So
Some
eo
of tthis
sd
diversity
e s ty ga
gain ((i.e.
e reliability)
e ab ty) ca
can be ttraded
aded o
off for
o “spatial
spat a
multiplexing” (i.e. a more efficient packing of bits)
 Spatial multiplexing can allow for an increase in the channel capacity as
compared
p
to a SISO system
y
 SISO capacity is given by Shannon’s famous relation:
CSISO  BW  log2 (1  SNR)
• Capacity increases linearly with BW which is the channel bandwidth and is a
very expensive and precious commodity
• Capacity increases logarithmically with SNR. Increasing channel capacity by
trying to increase SNR is futile.
 MIMO capacity (in a high SNR environment) is given by
CMIMO  min(n, m)  BW  log2 (1  SNR)
• n is the number of transmitter antennas and m is the number of receiver
antennas
• A huge increase in capacity can be obtained because of spatial multiplexing
A. Behzad
IEEE DL 2011: MIMO Radios
25
Why MIMO OFDM?
 OFDM has been used for legacy 802.11a/g to enable reasonably high data rates
at the
h presence off multi-path
li
h ffading
di present iin typical
i l iindoor
d
environments.
i
• OFDM combats the multi-path fading by dividing the wide bandwidth required for high
data rate communications to multiple narrow subcarriers such that the modulated
bandwidth of each subcarrier is << than 1/(rms delay) of the channel
 Utilizing MIMO OFDM, in general, a “rich scattering” environment can be used.
The capacity increases with min(M, N), i.e. the minimum of the number of
transmitter and receiver antennas.
 Independent data
streams are sent on
each transmit
antenna at the same
time and on the
same frequency
band
 Need multiple
antennas on both
sides
A. Behzad
IEEE DL 2011: MIMO Radios
26
MIMO-OFDM
 In
I OFDM,
OFDM the
th
channel is broken
into L (in this case,
53) parallel flatfading channels
channels,
each representable
by a single
coefficient.
bits in
 In MIMO OFDM,,
there is an NxM
matrix of channel
coefficients per
subcarrier, where M
is the number of
transmitter antennas
and N is the number
of receiver antennas.
H-24(0,0)
MIMO
TX
H-24(0,1)
bits out
H-24(1,1)
(1 1)
 Y 24 0   H  24 0,0  H  24 0,1  X  24 0   N  24 0 
÷

÷
÷
÷
 
×
 
÷
÷
÷
÷










 Y 24 1   H  24 1,0 H  24 1,1   X  24 1   N  24 1 
-26 -25 -24
OFDM Subcarrier Index
A. Behzad
MIMO
RX
H-24(1,0)
-1
0
1
25 26
Carrier Frequency (f)
(fc)
IEEE DL 2011: MIMO Radios
27
Space Division Multiplexing (SDM)



H-24(0,0)
One can transmit an
independent data stream
on each transmit antenna
provided the receiver has
at least two antennas.
In this 2x2 SDM case, the
data may be recovered
perfectly on any subcarrier
if its 2x2 channel matrix is
invertible (2 equations, 2
unknowns) and SNR is
high enough.
The simplest linear
receiver inverts the
channel matrix to recover
transmitted symbols and is
referred to as “Zero
ZeroForcing”.
bits in
MIMO
TX
H-24(0,1)
bits out
H-24(1,1)
1
 Xˆ  24 0   H  24 0,0  H  24 0,1  Y 24 0
÷ 

÷ 
÷
 Xˆ 1÷   H 1,0  H 1,1÷ × Y 1÷
 24
   24 
  24    24

"Zero-Forcing" Receiver
-26 -25 -24
OFDM Subcarrier Index
A. Behzad
MIMO
RX
H-24(1,0)
-1
0
1
25 26
Carrier Frequency (f c)
IEEE DL 2011: MIMO Radios
28
Outline
 Introduction,
I t d ti
Market
M k t Overview
O
i
& Market
M k t Trends
T d
 Communication System Performance in a Multipath Channel:
 The problem with multipath
 Mitigation methods for multipath
m ltipath
 Taking advantage of multipath
 WLAN Evolution & 802.11
 Additions in 802.11n
802 11n PHY and impacts on radio
 802.11n Sensitivity and EVM requirements
 Some optional modes of 802.11n and 802.11ac
 A brief introduction to 802
802.11ac
11ac
 Some 802.11 Radio Requirements
 Impacts of some radio impairments on system performance
 Examples of Circuit Techniques and Architectures Used in 802
802.11n
11n Radios
 Evolution of Radios: SISO to MIMO
 Real World Performance & Multi-Vendor Interoperability
 Conclusion
A. Behzad
IEEE DL 2011: MIMO Radios
29
WLAN Evolution
 .11b
 2.4 GHz band
 Direct sequence spread spectrum (DSSS): 1 and 2 Mbps
 Complimentary code keying (CCK): 5.5 and 11 Mbps
 .11a
11
 5 GHz band
 Orthogonal frequency division multiplexing (OFDM)
• 6,9,
6,9,12,18,24,36,48,54
, 8, ,36, 8,5 Mbps
bps
 Signal BW = 20 MHz
 .11g
 2.4 GHz band
 .11b + .11a rates: 12 rates
 .11n
 2.4 and 5 GHz bands
 .11g
11g rates + 77 Modulation and coding schemes (MCSs)
 Top PHY rate of 600 Mbps. Rate increase over .11a/g achieved by:
• Wider Signal BW of 40 MHz
• Multiple input multiple output (MIMO)-OFDM: Exploit spatial richness of the channel
• Higher
Hi h coding
di rates
A. Behzad
IEEE DL 2011: MIMO Radios
30
WLAN Evolution
 .11ac
 5 GHz band only
 Backward compatible with 11a/b/g/n
 Top PHY rate of 6.933Gbps with Nss=8, 160MHz channel, R=5/6, Short GI. Rate
increase over .11n
11n achieved by:
• Wider Signal BW of 80 and 160MHz with channel bonding
• 256-QAM modulation
A. Behzad
IEEE DL 2011: MIMO Radios
31
What was added in .11n PHY & MAC?
 Spatial Division Multiplexing (SDM) through MIMO-OFDM
 Bandwidth Expansion
 Higher Rate Binary Convolutional Code (BCC)
 New Frame Formats
 Reduced Interframe Spacing
p
g ((RIFS))
 Short Guard Interval (GI)
 Space-Time Block Code (STBC)
 Transmit Beam Forming (TxBF)
 Low Density Parity Check Code (LDPC)
 New Modulation and Coding Schemes (MCSs)
 Aggregation Techniques
A. Behzad
IEEE DL 2011: MIMO Radios
32
Impact of SDM on Radio: Area and power consumption
 A MIMO SDM system requires multiple Rx/Tx chains and AFEs
 Compact design becomes much more important
 Low
L
power d
design
i b
becomes much
h more iimportant
t t
 To increase robustness, even more Rx or Tx chains may be desired
DSP
Bit
Split
Bits
TX
A. Behzad
DSP
Radio
Radio
Radio
Radio
CHANNEL
IEEE DL 2011: MIMO Radios
DSP
DSP
Bit
Merge
Bits
RX
33
Impact of SDM on Radio: SNR and EVM requirements
 As compared
p
to non SDM systems
y
((SNR ~ 22dB for legacy
g y a/g),
g), a higher
g
SNR is required to obtain the highest data rates promised by MIMO SDM
systems (SNR ~ 30dB)
After ref [5]
A. Behzad
IEEE DL 2011: MIMO Radios
34
Impact of SDM on Radio Design: Crosstalk
 Linear
ea (additive)
(add t e) and
a d Nonlinear
o
ea (multiplicative)
( u t p cat e) crosstalk
c ossta have
a e very
ey
different impacts on MIMO system performance
TX1
I1/Q1
Shared LO
(see below)
OFDM
spectrum
g
Signal-1
TX2
Desired
Signal-2
I2/Q2
Linear (additive) crosstalk
Nonlinear crosstalk
Figs. & Discussions of crosstalk largely after ref. [6]
A. Behzad
IEEE DL 2011: MIMO Radios
35
Impact of SDM on Radio Design: Crosstalk
0
PER
10
-1
10
-2
2
10
22
Isolation
Infinite
20dB
15dB
10dB
24
26
28
30
SNR [dB]
32
34
• MIMO processing is inherently trying to cancel additive crosstalk
(channel + radio)
• 20dB isolation is adequate
pp y usual isolation tricks in design/layout
g
y
• Apply
A. Behzad
IEEE DL 2011: MIMO Radios
36
Impact of SDM on Radio Design: Other Requirements
 Tx or Rx chains are not required to have absolute phase matching
 Any mismatch or variations of Rx or Tx chains is tracked as part of the
overall MIMO channel
 Channel is “learned”
learned on the receiver DSP
DSP-side
side based on the received
preamble
 Frequency and phase of the multiple spatially multiplexed streams is
tracked jointly
 A single common reference source (crystal) should be used for all
streams
 All else
l b
being
i th
the same, and
d assuming
i that
th t the
th phase
h
noise
i ffrom th
the
multiple streams are correlated, the phase noise requirements for an
SDM system are no more stringent than for a SISO system. In reality a
better PN is highly desirable,
desirable however
however, since:
• In reality, the correlation factor is not exactly 1
• Because of other reasons, such as higher coding rate required in 11n
• …
A. Behzad
IEEE DL 2011: MIMO Radios
37
One Good Design Choice: Shared LO
 Avoids duplication
p
of synthesizer/VCO
y
p
power dissipation
p
 Better area and power consumption
 Provides two estimates of same phase noise-- Better phase noise tracking
 Better effective EVM
 Avoids pulling between multiple VCOs
 Reduced nonlinear cross-talk
 If nott possible
ibl to
t use same PLL,
PLL att least
l
t share
h
the
th same reference
f
crystal
t l oscillator
ill t
Uncorrelated
components
reduced
Pilot tones

+

A. Behzad
=
Uncorrelated
components
IEEE DL 2011: MIMO Radios

Fig. after ref. [6]
38
Bandwidth Extension & Spectral Mask: (legacy A/G)
PSD
0dBr
-20dBr
-28dBr
-40dBr
-30
-20
-11 -9
9 11
20
30
Freq [MHz]
Subcarrier Index
-26
+26
 For legacy A/G, the spectral mask floor is set to -40dBr.
 For legacy A/G, there are a total of 52 subcarriers (# -26 through #26
with
ith #0 excluded)
l d d)
A. Behzad
IEEE DL 2011: MIMO Radios
39
Bandwidth Extension & Spectral Mask: 802.11n 20MHz Mode
PSD
0dBr
-20dBr
-28dBr
-45dBr
-30
-20
-11 -9
9 11
20
30
Freq [MHz]
Subcarrier Index
-28
+28
 For 802.11n
802 11n 20MHz mode
mode, the spectral mask floor is set to -45dBr.
45dBr
 For 802.11n 20MHz mode, there are a total of 56 subcarriers (# -28
through #28 with #0 excluded)
 8% increase in PHY rate relative to legacy A/G
A. Behzad
IEEE DL 2011: MIMO Radios
40
Bandwidth Extension & Spectral Mask: 802.11n 40MHz Mode
PSD
0dBr
-20dBr
-28dBr
-45dBr
-60
-40
-21
20MHz-mode
20MH
d ttones moved
d up
and down by 20MHz, the DC
tones filled and additional tones
are added in the middle (added
tones shown in blue)
-19
19
21
40
60
Freq [MHz]
Subcarrier Index
-57
+57
 For 802.11n 40MHz mode, the spectral mask floor is set to -45dBr.
 For 802.11n 40MHz mode, there are a total of 114 subcarriers (# -57 through
# +57 with #0 excluded))
 Use of 108 data subcarriers increases PHY rate by 2.25x relative to legacy A/G
A. Behzad
IEEE DL 2011: MIMO Radios
41
Bandwidth Extension & Spectral Mask
Implications on Radio (20MHz mode)
 Additional subcarriers in 20MHz mode has:
 No impact
p
on Rx or Tx RF sections
 No impact on PLL
 On baseband Rx and Tx sections:
•
Flatter frequency response across the desired passband
– Less amplitude ripple
– Less group delay variation
• Sharper filter roll-off for same ACRR and ALCR
 Lower spectral mask floor (-45dBr) requires:
 More linear transmitter
 Lower noise transmitter
 Lower noise VCO and LOGEN chain
A. Behzad
IEEE DL 2011: MIMO Radios
42
Bandwidth Extension & Spectral Mask
Implications on Radio (40MHz mode)
 Wide-band 40MHz mode has:
 No or minimal impact
p
on Rx or Tx RF sections
 Phase noise of PLL needs to be considered up to +/-20MHz
• Particularly for fractional-N implementations
 On baseband Rx and Tx sections:
• High speed baseband blocks required
– e.g. opamps with >1GHz GBW product necessary for opamp-RC style filters and
VGAs
•
Wide band and flat frequency response across the desired passband
Wide-band
– Low amplitude ripple
– Low group delay variation
• Sharper filter roll-off for reasonable ACRR and ALCR
 Higher speed ADCs and DACs needed
A. Behzad
IEEE DL 2011: MIMO Radios
43
Bandwidth Extension & Spectral Mask
Implications on Radio (40MHz mode)
 Low spectral mask floor (-45dBr) requires:
 Linear and wide-band transmitter
• Implications and complications with respect to digital predistortion
 Low noise transmitter
 Low noise VCO and LOGEN chain
A. Behzad
IEEE DL 2011: MIMO Radios
44
Higher Rate Binary Convolutional Codes
Implications on Radio
 Legacy code rates of 1/2
1/2, 2/3
2/3, and
¾ are carried over.
 With the addition of the 4
subcarriers in the 20MHz mode, a
data rate of 58.5Mbps is achieved
with a BCC rate of ¾.
 A mandatory code rate of 5/6 has
been added for MCS7
 An additional increase of 11% in
data rate
 A higher coding rate reduces the
redundancy and therefore
robustness of the code
 A better EVM is required on Tx
side
 A worse sensitivity is achievable
on the Rx side (with all else being
the same)
A. Behzad
Index
Modulation
Code
Rate
Data Rate
(Mbps)
0
BPSK
½
6.5
1
QPSK
½
13
2
QPSK
¾
19.5
3
16-QAM
½
26
4
16-QAM
¾
39
5
64-QAM
2/3
52
6
64-QAM
¾
58.5
7
64-QAM
5/6
65
IEEE DL 2011: MIMO Radios
45
Space Time Block Codes
 S
Space Time
Ti
Bl
Block
kC
Code
d (STBC)
 Increases rate at range for scenarios with more transmit chains than
receive chains
 Useful
f especially for
f transmitting to single antenna devices
 Does not require closed-loop operations
 Same data sent in different timeslots and on different antennas
 If channel paths to the RX are statistically independent, then probability
of simultaneous fading becomes very small
 STBC is an optional mode of 802.11n
 Requirements on radio:
 If sent to a single antenna receiver, no particular requirement on the
receiver
 If combined with SDM, the requirements are the same as outlined in
SDM discussion
A. Behzad
IEEE DL 2011: MIMO Radios
46
Beamforming / MRC
 Uses
U
multiple
lti l ttransmit
it and/or
d/ receive
i radios
di tto fform coherent
h
t signals
i
l
 Receive beamforming / combining boosts reception of signals
D
S
P
Radio
Bits
Radio
TX
Channel
Radio
Bits
RX
 Phased
Ph
d array transmit
i b
beamforming
f
i to ffocus energy to each
h receiver
i
and provides “array gain”
D
S
P
Bits
TX
A. Behzad
Radio
Radio
Channel
IEEE DL 2011: MIMO Radios
Radio
RX
Bits
47
Rx MRC
 MRC is performed on a per subcarrier basis to help reduce multipath
deep nulls
 Note difference with analog MRC
 The system will pick the higher SNR of each subcarrier between the
multiple MRC channels
Fig. after ref. [7]
A. Behzad
IEEE DL 2011: MIMO Radios
48
MRC
 In
I order
d to
t maximize
i i benefits
b
fit off beamforming,
b
f
i
the
th radio
di mustt
preserve high SNR on all subcarriers
 In particular the SNR of the inner tones which are subject to DC and low
frequency effects must be preserved
• For example HPFs utilized for DC cancellation can adversely impact
the SNR of the inner tones for all beamformed channels, with no
recovery of those subcarriers possible
A. Behzad
IEEE DL 2011: MIMO Radios
49
Transmit Beamforming
 TX beamforming implies channel knowledge at the TX
 Direction of the beam is dependent on the relative phase of the
signals at the Tx antennas
 Requires some form of feedback loop
• Explicit feedback: the receiver estimates the channel state
information (CSI) and sends back to transmitter through a special
message
• Implicit feedback: The transmitter estimates the channel based on
the packets received from the intended receiver and the assumption
of channel reciprocity
 The required beamforming phase accuracy is dependent on the
width of the beam
 The width of the beam is dependent on the number of transmit
antennas
A. Behzad
IEEE DL 2011: MIMO Radios
50
LDPC
 Optional
p
Low Density
y Parity
y
Check (LDPC) code provides
higher performance than BCC
code
 12 different
diff
t codes
d based
b
d on
same code structure
 Same code rates as BCC code
• ½ , 2/3,
2/3 ¾
¾, 5/6
 Three block lengths
• 648, 1296, 1944
 The higher
g
p
performance of
LDPC codes as compared to
BCC codes would ideally
somewhat relieve the
requirements on the radio Rx
and Tx EVM
Fig. after ref. [9]
 They are optional and we cant
bank on them!
A. Behzad
IEEE DL 2011: MIMO Radios
51
Modulation Coding Schemes & Implications on Radio
 The modulation coding schemes (MCS) allow for the various
mandatory and optional modes of 802.11n. These include:
 Use of multiple streams
 Use of different modulation types
 Use of different binary convolutional codes
 Use of the different bandwidths (20MHz and 40MHz)
 Use of robust duplicate mechanism in 40MHz mode
 Use of asymmetric MCSs for transmit beam forming and STBC
 …
 The implications on radio implementation has been individually
di
discussed
d under
d each
hh
heading
di
A. Behzad
IEEE DL 2011: MIMO Radios
52
Modulation and Coding Scheme (MCS)
default OFDM
guard
d iinterval
t
l
NSD
NES
NCBPS
Bits 0-6
in HTSIG1
(MCS
index)
Number of
spatial
streams
Modulation
Coding
rate
0
1
BPSK
½
1
1
52
108
52
1
1
QPSK
½
1
1
52
108
2
1
QPSK
¾
1
1
52
3
1
16 QAM
16-QAM
½
1
1
4
1
16-QAM
¾
1
5
1
64-QAM
6
1
64-QAM
7
1
8
GI = 800ns
20
40
20MHz
40MHz
GI = 400ns
Rate in
Rate in
Rate in
Rate in
20MHz
40MHz
20MHz
40MHz
108
6.5
13.5
7 2/9
15
104
216
13
27
14 4/9
30
108
104
216
19.5
40.5
21 2/3
45
52
108
208
432
26
54
28 8/9
60
1
52
108
208
432
39
81
43 1/3
90
1
1
52
108
312
648
52
108
57 7/9
120
¾
1
1
52
108
312
648
58.5
121.5
65
135
64-QAM
5/6
1
1
52
108
312
648
65
135
72 2/9
150
2
BPSK
½
1
1
52
108
104
216
13
27
14 4/9
30
9
2
QPSK
½
1
1
52
108
208
432
26
54
28 8/9
60
10
2
QPSK
¾
1
1
52
108
208
432
39
81
43 1/3
90
11
2
16-QAM
½
1
1
52
108
416
864
52
108
57 7/9
120
12
2
16-QAM
¾
1
1
52
108
416
864
78
162
86 2/3
180
13
2
64-QAM
1
1
52
108
624
1296
104
216
115 5/9
240
14
2
64-QAM
¾
1
1
52
108
624
1296
117
243
130
270
15
2
64-QAM
5/6
1
1
52
108
624
1296
130
270
144 4/9
300
A. Behzad
⅔
⅔
20
40
IEEE DL 2011: MIMO Radios
53
Modulation and Coding Scheme (MCS)
NSD
NES
NCBPS
GI = 800ns
Bits 0-6
in HTSIG1
(MCS
index)
Number
of spatial
streams
Modulation
Coding
rate
16
3
BPSK
½
2
2
52
108
156
17
3
QPSK
½
2
2
52
108
18
3
QPSK
¾
2
2
52
19
3
16-QAM
½
2
2
20
3
16-QAM
¾
2
21
3
64-QAM
22
3
64-QAM
23
3
24
20
40
20MH
z
40MH
z
GI = 400ns
Rate in
Rate in
Rate in
Rate in
20MHz
40MHz
20MHz
40MHz
324
19.50
40.50
21.67
45.00
312
648
39.00
81.00
43.33
90.00
108
312
648
58.50
121.50
65.00
135.00
52
108
624
1296
78.00
162.00
86.67
180.00
2
52
108
624
1296
117.00
243.00
130.00
270.00
2
2
52
108
936
1944
156.00
324.00
173.33
360.00
¾
2
2
52
108
936
1944
175.50
364.50
195.00
405.00
64-QAM
5/6
2
2
52
108
936
1944
195.00
405.00
216.67
450.00
4
BPSK
½
2
2
52
108
208
432
26.00
54.00
28.89
60.00
25
4
QPSK
½
2
2
52
108
416
864
52.00
108.00
57.78
120.00
26
4
QPSK
¾
2
2
52
108
416
864
78.00
162.00
86.67
180.00
27
4
16-QAM
½
2
2
52
108
832
1728
104.00
216.00
115.56
240.00
28
4
16-QAM
¾
2
2
52
108
832
1728
156.00
324.00
173.33
360.00
29
4
64-QAM
2
2
52
108
1248
2592
208.00
432.00
231.11
480.00
30
4
64-QAM
¾
2
2
52
108
1248
2592
234.00
486.00
260.00
540.00
31
4
64-QAM
5/6
2
2
52
108
1248
2592
260.00
540.00
288.89
600.00
32
1
BPSK
½
1
1
⅔
⅔
20
40
48
48
6
6.67
Further MCS entries (33-76) are defined to support unequal
modulation for tx beamforming using SVD.
A. Behzad
IEEE DL 2011: MIMO Radios
54
TGac Maximum PHY Data Rate Examples
• In TGac significantly higher PHY rates can be achieved
• Maximum Nss is set to 8
• Maximum constellation is 256-QAM
256 QAM
• Maximum channel bandwidth is 160MHz (contiguous or non-contiguous)
• With a code rate of 5/6 and a short GI
A. Behzad
BW = 80 MHz
BW = 160 MHz*
Nss = 1
433 Mbps
867 Mbps
Nss = 2
867 Mbps
1.7 Gbps
Nss = 4
1.7 Gbps
3.5 Gbps
Nss = 8
3.5 Gbps
6.9 Gbps
IEEE DL 2011: MIMO Radios
MU-MIMO Highlights
 S
Significant
g ca t ga
gain in agg
aggregation
egat o possible
poss b e (no
( o maximum
a
u PHY rate
ate
increase)
 OFDMA or Pre-coding used to achieve orthogonality amongst users
2 streams
STA 1
AP
2 streams
STA 2
IEEE DL 2011: MIMO Radios
A. Behzad
B
d
P
i t
&
5 GHz Channelization (US and EU)
Regulatory Power Limit (dBm)
Weather Radars
30
23
16
52
36
40
44
56
60
100 104 108 112 116 120 124 128 132 136 140
64
149 153 157 161 165
48
Frequency
5180 MHz
US
5320 MHz
5500 MHz
DFS
EU
US and EU
A. Behzad
IEEE DL 2011: MIMO Radios
5700 MHz
5825 MHz
Outline
 Introduction,
I t d ti
Market
M k t Overview
O
i
& Market
M k t Trends
T d
 Communication System Performance in a Multipath Channel:
 The problem with multipath
 Mitigation methods for multipath
m ltipath
 Taking advantage of multipath
 WLAN Evolution & 802.11
 Additions in 802.11n
802 11n PHY and impacts on radio
 802.11n Sensitivity and EVM requirements
 Some optional modes of 802.11n and 802.11ac
 A brief introduction to 802
802.11ac
11ac
 Some 802.11 Radio Requirements
 Impacts of some radio impairments on system performance
 Examples of Circuit Techniques and Architectures Used in 802
802.11n
11n Radios
 Evolution of Radios: SISO to MIMO
 Real World Performance & Multi-Vendor Interoperability
 Conclusion
A. Behzad
IEEE DL 2011: MIMO Radios
58
802.11n Sensitivity Requirements
 802
802.11n
11n sensitivity requirements:
 At the levels stated below, PER < 10% at 4096 byte packets is required
 Power levels are average levels on each Rx antenna
A. Behzad
IEEE DL 2011: MIMO Radios
59
802.11n EVM Requirements
 EVM requirements
equ e e ts a
are
e stated be
below
o
 Stated in RMS
 Averaged over subcarriers, OFDM frames and spatial streams
 Requires to 20MHz and 40MHz channels
 Linear cross-talk between the spatial streams are corrected for
• Be aware of this fact when measuring EVM of a MIMO transmitter with a
single-channel
single
channel VSA
A. Behzad
IEEE DL 2011: MIMO Radios
60
Practical Advantages of a Better EVM
 Although the 11n standard requires an Tx EVM specification of only -28dB
at the highest rate, a better radio EVM can provide many advantages:




Improved sensitivity level for an AWGN channel at the receiver
More EVM budget for the PA
Even more aggregation and longer packets due to lower probability of error
Possibility of higher order proprietary modulation schemes for higher data rates
 As a tradeoff,
tradeoff a better overall EVM (radio Tx + PA) requires:
 A better transmitter (PN, IQ balance, linearity, etc.)
 More back off on the PA transmit power
• Efficiency impact
• Range impact
A. Behzad
IEEE DL 2011: MIMO Radios
61
What are the Special Requirements of 802.11n radios?
 So
Some
e of
o the
t e specific
spec c requirements
equ e e ts (on
(o isolation
so at o o
of tthe
e cchains,
a s, for
o
example) have already been discussed. Beyond that, in general,
 A “better” radio (transmitter, receiver and PLL) are required for
satisfying
y g the 802.11n requirements
q
 Not all radio characteristics have to be better for an 802.11n radio
• For example, a better NF is not required to meet the bare minimum standard
requirements on sensitivity
• But from an customer point of view, a better performance than the legacy
products is expected
 In particular, to meet the better EVM requirements radio impairments
have to be reduced
reduced. In particular
particular,
• Better PN is required
• Better quadrature balance is required
 As an example,
example impact of quadrature imbalance on image rejection and
the subsequent impact of image rejection on EVM is shown in the next
two slides
A. Behzad
IEEE DL 2011: MIMO Radios
62
Impact of Quadrature Imbalance on Image Rejection
 Quad
Quadrature
atu e gain
ga and
a d phase
p ase e
errors:
os
-10
Amplitude Error (%)
-20
Imag
ge Rejection (dB
B)
10%
-30
5%
2%
-40
1%

 2  G 2
IR 
0.5%
-50
4
0%
60
-60
Ref: Lee, pgs. 558-559
-70
0.1
1
10
Phase Error (deg)
A. Behzad
IEEE DL 2011: MIMO Radios
63
Impact of Image Rejection on EVM
EVM vs Image Rejection (measured)
(dotted line is extrapolated)
-20
-22
-24
EVM [dB]
-26
26
-28
-30
EVM
-32
-34
-36
-38
Fig. after ref. [10]
-40
-50
-45
-40
-35
-30
-25
-20
Image Rejection [dBc]
A. Behzad
IEEE DL 2011: MIMO Radios
64
How to Improve Radio Performance to satisfy 802.11n/ac
 Ge
General
e a ccircuit
cu t pe
performance
o a ce improvement
p o e e t tec
techniques
ques sshould
ou d be
utilized
 Variety of techniques available depending on the block. A couple of
examples:
p
• Retiming the VCO dividers with the VCO output helps reduce the
contribution of the dividers to the PLL noise
• Utilizing linearization techniques on the transmitter (and receiver) allows for
a smaller contribution of nonlinearities to the EVM
 Attention to details
 For example, how do the following factors contribute to the PLL noise?
•
•
•
•
•
Supply
S
l utilized
tili d ffor various
i
PLL bl
blocks
k
Bias currents utilized for various PLL blocks
The LO generation and distribution blocks
Are the varactor models accurate?
Are all factors contributing to the loaded tank Q of the VCO taken into
account?
 Calibrate, calibrate, calibrate!!
A. Behzad
IEEE DL 2011: MIMO Radios
65
Calibration Techniques
 As much as calibrations were “nice to have” for legacy 802.11 systems, they are
significantly more important for 802.11n and even more so for 11ac
 Chip level and system level auto-calibration is required for overcoming difficulties
of integrated radio design, increasing chip and system yield and improving
performance
performance.
 Some examples:
 VCO tuning
 Tx and Rx IQ Calibration
 R-Calibration on bandgap blocks
 RC time constant calibration
 Integrated power detector
 Integrated temperature sensor
 Transmit LO feedthrough cancellation
 AFC
 Process sensing
 TSSI
A. Behzad
IEEE DL 2011: MIMO Radios
66
TGac: A Few Additonal Challenges
 Radio
 256-QAM requires higher EVM (< -32dB ) requirements
 Ability to bond two dis-contiguous 80Mhz channels into one logical
160MHz channel
 Ability
Abilit tto h
handle
dl llarger signal
i
l BW
BWs up tto 160Mh
160Mhz on b
both
th ttransmit
it and
d
receive chains
 Mixed-signal
Mixed signal
 Need 4x higher sampling ADC/DACs from current generation 11n
 PHY
 ML (or AML) for handling 256-QAM rates
 VHT preamble processing
 Synchronization challenges and demod accuracy to support 256-QAM
 Multi-user MIMO
 RTL and MAC
 Higher (4x) clocking and ability to handle much higher TPUTs
A. Behzad
IEEE DL 2011: MIMO Radios
Outline
 Introduction,
I t d ti
Market
M k t Overview
O
i
& Market
M k t Trends
T d
 Communication System Performance in a Multipath Channel:
 The problem with multipath
 Mitigation methods for multipath
m ltipath
 Taking advantage of multipath
 WLAN Evolution & 802.11
 Additions in 802.11n
802 11n PHY and impacts on radio
 802.11n Sensitivity and EVM requirements
 Some optional modes of 802.11n and 802.11ac
 A brief introduction to 802
802.11ac
11ac
 Some 802.11 Radio Requirements
 Impacts of some radio impairments on system performance
 Examples of Circuit Techniques and Architectures Used in 802
802.11n
11n Radios
 Evolution of Radios: SISO to MIMO and Single Function to Multi-Function
 Real World Performance & Multi-Vendor Interoperability
 Conclusion
A. Behzad
IEEE DL 2011: MIMO Radios
68
TX Architecture
Ref. G. Chien, et. Al., RFIC 2006
TX_I±
TX2_OUT
PPA2
VGA
2
5X
2
TX_Q±
TX5_OUT
PPA5
VGA
2

Used in a 2x3 multi-band SiGe super-heterodyne implementation

Sideband rejection TX architecture to suppress the lower-sideband mixing component for 2.4GHz
operation

0.5dB/step VGA used to compensate for gain variation

Fully-Differential signal path with on-chip balun at TX outputs
A. Behzad
IEEE DL 2011: MIMO Radios
69
IF Current Amplifier
A. Behzad
in
n-
in+
+
out-
out+
Ref. G. Chien, et. Al., RFIC 2006

IF Current Amplifier bases on a (Nx)
current multiplier (Q1/2, Q3/4)

An auxiliary amplifier (Q5/Q6) is used to
enhance HF beta performance

No additional current is used for the
aux. amp.
IEEE DL 2011: MIMO Radios
70
Adaptively Biased PA
Ref. D. Rahn, et. Al., JSSC Aug. 2005
A. Behzad

Used in a 2x2 multi-band SiGe supery implementation
p
heterodyne

Adaptive biasing used to reduce
power consumption and increase
PAE

Main PA device is Q1

Adaptive bias applied through R1
and generated using Q2, Q3 and Q4

Allows for lower quiescent power
consumption and higher linearity for
large input signals

Achieves 11/13.5 dBm with 4% EVM
for 2.4/5GHz band
IEEE DL 2011: MIMO Radios
71
PA driver Transconductance Linearization
Ref. A. Behzad, et. Al., ISSCC 2007
VDD
R1
R2
L1
L2
To Output Balun
b3
M11
M12
M7
M8
M3
M4
M9
M10
M5
M6
M1
M2
Stage3
b3
b2
Stage2
b2
b1
Stage1
b1

U d iin a 2x2
Used
2 2 multi-band
lti b d CMOS direct-conversion
di t
i iimplementation
l
t ti

Offset bias voltages b1, b2, b3 allow for the linearization of the transconductance
 Lower quiescent power consumption
 Higher linearity under large input signals
A. Behzad
IEEE DL 2011: MIMO Radios
72
PA Driver Transconductance Linearization

Simulated each stage Gm and effective total Gm under various process corners
A. Behzad
IEEE DL 2011: MIMO Radios
73
Transmitter Architecture & Tx Calibration Path


Used in a 2x2 multi-band CMOS direct-conversion implementation
LOI
LOFT and Tx IQ Calibration path shown
To ext PA or
Antenna
BB_I
DAC
Gm
LPF
Transconductance
Stage I
BB_Q
DAC
Gm
LPF
RFPGA
PAD
Transconductance
Stage Q
LOQ
CM
REFERENCE
OUT
VARIABLE
GAIN
R1
500 KHz RC
HIGH PASS
FILTER
Ref. C. P. Lee, et. Al., ISSCC 2006
A. Behzad
ENVELOPE
DETECTOR
+
LPF
R2
IEEE DL 2011: MIMO Radios
74
Tx Mixer Variable-Gain Transconductor Stage

Highly linear mixer Gm

Offset based LOFT scales with gain
control
RF_LOFT_IP
To a Gilbert cell mixer
Quad
VDD
R3
R4
R5
RF_LOFT_IN
R6
VCASC
M5 1/gm5,6
M1
in
R1
M5
R7
R2
M3
R8
iout
M2
ip
1/gm5,6
1/gds7
R10
iin
BB_LOFT_IP
M6
R11
Gain
Control
1/gm7
M4
R9
M7
M6
R12

BB_LOFT_IN
Mechanisms for cancelling offset
based and coupling based LOFT
Ref. C. P. Lee, et. Al., ISSCC 2006
GND
A. Behzad
IEEE DL 2011: MIMO Radios
75
PA Nonlinearity
 High order modulations are quite sensitive to phase nonlinearities
nonlinearities, in
addition to amplitude nonlinearities
Pout
φout
Pin
AM-AM
Pin
AM-PM
 3 different schemes of Tx/PA linearization will be presented in the
next several slides
A. Behzad
IEEE DL 2011: MIMO Radios
76
Power Amplifier
Ref. Y. Palaskas, et. Al., ISSCC 2006
• Class-AB
to matching
network
t
k
• Thick gate devices for
good efficiency and
reliability
• P1dB≈20dBm,
PAVG
≈12dBm
12dBm (54Mb/s, EVM
EVM=25dB)
from mixer
• Apply linearization to increase PAVG and efficiency
A. Behzad
IEEE DL 2011: MIMO Radios
77
Digitally-Assisted AMPM Linearization
Ref. Y. Palaskas, et. Al., ISSCC 2006
filter
I
D/A
Q
D/A
PA
φtank (deg)
(d )
I2  Q 2
Lookup
Table
Δφ
D/A
fin
f (GHz)
(GH )
 5GHz implementation
signal dependent
 Varactor introduces phase shift to cancel signal-dependent
phase shift of PA
 Varactor controlled by amplitude of digital IQ data
EVM=-25dB
25dB (w/o linearazation 12dBm)
 PAVG = 16dBm @ EVM
 2.8x efficiency improvement
A. Behzad
IEEE DL 2011: MIMO Radios
78
AMAM Analog PA Linearization Loop
M. Terrovitis, et. Al., ESSCIRC 2009
 2.4GHz
G implementation
p e e tat o
 Achieves 13.6dBm -28dB EVM without linearization and
18.4dBm with at the chip output
g
but balun is external
 TR switch is also integrated
A. Behzad
IEEE DL 2011: MIMO Radios
79
Analog + Digital Tx Linearization
Top Level Block Diagram:
Top-Level
3-stage
g PA with
On-chip balun
To an
ntenna
Feedback path
used for PA
pre-distortion
3.3v
3.3v
PA
Attn.
PAD
1.2v
VGA
LO
Ref. A. Afsahi, et. Al., RFIC 2009
A. Behzad
IEEE DL 2011: MIMO Radios
I
LPF
Q
LOI
LOQ
G
LPF
LPF
DSP
VGA and PAD Circuit Implementation
VDD
• Multiple
Multiple-branching
branching
for gain control
L
+ Vout -
• Utilizing gain boost
Vc
• Thick-gate
Thick gate device for
cascode in PAD
M3
-
M5
M6
+
Vin
M1
Vb1
A. Behzad
Vc
M4
Vin
M2
M7
Vb1
IEEE DL 2011: MIMO Radios
Vb2
40
Ga
ain(dB)
Gain (dB)
35
30
25
20
Max_gain (5.9G)
15
Max_gain (4.9G)
10dB Gain backoff (5.9G)
12
17
Pout (dBm)
22
Gain vs. Pout for
G
f two
different gain codes
A. Behzad
45
120
110
40
100
35
90
Gain w/o GB
Gain w GB
80
Phase w/o GB
Phase w GB
25
10
7
130
30
10dB Gain backoff (4.9G)
2
50
27
70
0
10
20
Pout(dBm)
30
5.5GHz
5
5GH AM/AM and
d AM/PM
with and without gain boost
IEEE DL 2011: MIMO Radios
Phas
se(deg)
Simulated Gain Control and Gain Boost
PA Circuit Implementation
Vout
•
•
Utilizing gm-linearization
technique
Thick-gate device for
cascode
•
On-chip balun
•
Capacitor divider for
feedback path
+
T feedback
To
f db k
mixer
-
Vin+
M3
M4
C2
C1
Vc
Vin- Vin+
Aux
Vb2 (Class B) Vb2
A. Behzad
Cbyp
VDD
IEEE DL 2011: MIMO Radios
Vb1
M5
M6
M1
M2
Main
Vc
Vin-
Vb1
0.5
Main_gm
Aux gm
Aux_gm
Total_gm
G a in (d B )
0.4
Gm(S)
G
0.3
0.2
0.1
0
-2
-1
•
•
A. Behzad
0
Vin(v)
1
2
16
14
12
10
8
6
4
2
0
80
70
60
50
40
30
20
10
0
G i w// Gm-Lin
Gain
G Li
Gain w/o Gm-Lin
Phase w/ Gm-Lin
Phase w/o Gm-Lin
0
5
10
15
20
25
Pout(dBm)
P1dB is improved from 21
21.7dBm
7dBm to 26
26.6dBm
6dBm by using GmGm
linearization
Lower gain due to class-B operation of Aux section
IEEE DL 2011: MIMO Radios
30
P h a s e (d e g )
Gm Linearization
Die Photo
5GHz
5GH
TX1
2.4GHz
2
4GH
TX1
5GHz
5GH
TX2
2.4GHz
2
4GH
TX2
VGA
VGA
PAD
PA
A. Behzad
PAD
PA
VGA
PAD
PA
2.4GHz
0.074mm2
0.132mm2
0.31mm2
5GHz
0.085mm
0
085mm2
0.092mm
0
092mm2
0.27mm
0
27mm2
IEEE DL 2011: MIMO Radios
Gain and Drain Efficiency
45
40
2.4G Gain (Sim)
11
2.4G Drain Eff (Meas)
20
15
10
2.4G
G Drain
a Eff (Sim)
(S )
9
7
5
0
5
10
A. Behzad
15
20
Pout(dBm)
25
30
25
5.5G Gain (Meas)
11
5.5G Gain (Sim)
20
5.5G Drain Eff (Meas)
5.5G Drain Eff (Sim)
9
15
10
7
5
5
30
0
10
15
20
Pout(dBm)
PA gain
Psat
Peak Drain
Eff
2.4G
14dB
28.3dBm
35.30%
5.5G
12dB
26.7dBm
25.30%
IEEE DL 2011: MIMO Radios
25
30
Dra
ain Eff(%)
2.4G Gain (Meas)
13
Gain(dB)
G
Gain(dB)
G
13
35
30
25
35
Drrain Eff(%)
15
EVM vs. Pout
-15
2.442G With Lin
2.442G W/O Lin
5 18G With Lin
5.18G
Li
5.5G With Lin
EVM(dB)
-20
-25
5.805G With Lin
-30
-35
-40
0
Pout @
-25dB EVM
Pout @
-28dB EVM
A. Behzad
5
10
15
Pout(dBm)
20
25
2442MHz
5180MHz
5500MHz
5805MHz
22.4dBm
21.1dBm
20.5dBm
19.7dBm
21.1dBm
19.7dBm
19.5dBm
18.3dBm
IEEE DL 2011: MIMO Radios
Spectral Mask
•
A. Behzad
22dBm Po at 2.442GHz
IEEE DL 2011: MIMO Radios
Comparison with published Dual-Band WLAN PAs
ref[a]
ref[b]
A Afsahi [RFIC 2009]
Psat (2.4G)
25dBm
28.3dBm
Psat(5.5G)
23.5dbm
26.7dBm
Power/Eff @ -28dB
EVM (2.4G)
15.5dBm/19%
17.5dBm/16%
21.1dBm/18%
14.5dBm/12%
17dBm/12.5%
19.5dBm/12.5%
Power/Eff @ -28dB
EVM (5.5G)
Max CW Eff (2.4G)
50%
35.3%
Max CW Eff (5.5G)
30%
25.3%
Linearization
DPD
No
DPD & Offset-Gm
On chip balun
No
N/A
Yes
Supply
3.3v
3.3v
3.3v
Technology
90nm CMOS
SiGe HBT
65nm CMOS
• Ref. [a]: O. Degani et al., “A 1x2 MIMO Multi-Band CMOS Tranceiver with an Integrated Front-End in 90nm
CMOS for 802.11a/g/n WLAN Application” ISSCC 2008.
• Ref. [b] H.H. Liao et al., “A Fully Integrated 2x2 Power Amplifier for Dual Band MIMO 802.11n WLAN
Applications using SiGe HBT Technology”, RFIC 2008
A. Behzad
IEEE DL 2011: MIMO Radios
Outline
 Introduction,
I t d ti
Market
M k t Overview
O
i
& Market
M k t Trends
T d
 Communication System Performance in a Multipath Channel:
 The problem with multipath
 Mitigation methods for multipath
m ltipath
 Taking advantage of multipath
 WLAN Evolution & 802.11
 Additions in 802.11n
802 11n PHY and impacts on radio
 802.11n Sensitivity and EVM requirements
 Some optional modes of 802.11n and 802.11ac
 A brief introduction to 802
802.11ac
11ac
 Some 802.11 Radio Requirements
 Impacts of some radio impairments on system performance
 Examples of Circuit Techniques and Architectures Used in 802
802.11n
11n Radios
 Evolution of Radios: SISO to MIMO and Single Function to Multi-Function
 Real World Performance
 Conclusion
A. Behzad
IEEE DL 2011: MIMO Radios
90
Evolution of Radios: SISO to MIMO
TX
802.11a radio
(ISSCC 2003)
2x2
2
2 802
802.11a/b/g/n
11 /b/ / radio
di
(ISSCC 2007)
A. Behzad
3x3
3
3 802
802.11a/b/g/n
11 /b/ / S
SoC
C
(ISSCC 2011)
IEEE DL 2011: MIMO Radios
91
Evolution of Radios: SISO to MIMO-An Example
802.11a
2x2 802.11n
3x3 802.11n
Behzad,
ISSCC 2003
Behzad,
ISSCC 2007
A-Alibeik
ISSCC 2011
Radio/SoC
Radio-only
Radio-only
SoC
Process
0.18um CMOS
0.18um CMOS
65nm CMOS
Area (mmsq)
11.7
18
10.4 (radio)
In-band PN
~ -100
~ -108
~ -102
~ -34
~ -40
~ -36
Rx NF (dB)
4
4
4
Chariot
Throughput
24
> 200
> 300
Author/Conf
(5GHz; dBc/Hz)
Tx EVM Floor
(5GHz; dB)
A. Behzad
IEEE DL 2011: MIMO Radios
92
Evolution of Radios: Multi-Function Radios
802.11a/b/g/ssn + FM/BT SoC
(ISSCC 2010)
A. Behzad
IEEE DL 2011: MIMO Radios
93
Real-World Performance/Throughput
 Chariot throughput test is a very high level test and can be limited by
many factors
f






Radio performance
Digital PHY performance
MAC performance
Processor performance
Network switch performance
…
 Many setups to measure throughput/performance
 Cable-connected AWGN channel or controlled channel emulator setup
• Results are for the most part repeatable and easily comparable to other
similar measurements
• Direct comparisons of subcomponents still not possible due to the high-level
nature of the Chariot test. Comparisons of the overall systems, however is
quite valid
 Over-the-air measurements (rate vs. range over the air)
• Requires the different solutions to be tested in the exact same setup for fair
comparisons to be made
• However, general conclusions can still be made and very high level
comparisons may still be possible
A. Behzad
IEEE DL 2011: MIMO Radios
94
Real-World Performance/Throughput
 Cable test
 2x2 system
 Throughput >
Ref. A. Behzad,et.
, al.,, ISSCC 2007
200Mbps
 PHY Rate = 270Mbps
 2.442GHz
2 442GHz channel
 Close-range
Close range (10
(10-ft.)
ft )
Ref. A. Behzad,et. al., ISSCC 2007
A. Behzad
over the air test at
5.24 GHz
 2x2 system
 Max throughput: 198
Mbps, Average
throughput > 193
Mbps
IEEE DL 2011: MIMO Radios
95
Real-World Performance/Throughput
 TCP throughput over range data
 Channel 6 (2437MHz)
 Mix LOS/NLOS Environment
 3x3
Ref. P. Petrus, et. al., ISSCC 2007
A. Behzad
IEEE DL 2011: MIMO Radios
96
Conclusions
 MIMO
O OFDM
O
systems
syste s a
are
e a relatively
e at e y new
e top
topic
co
of interest
te est a
and
d we
e
have briefly touched on various aspects of this very rich topic and in
particular the requirements for the radio
 Integrated
g
radios,, especially
p
y CMOS ones,, have come a very
y long
g
ways in a very short period of time
 Improvements in performance
 Reductions in size
 Higher degrees of integration
 Higher yields
 Multi-function
Multi function capability and co
co-existence
existence
 Radios will have to become even cheaper and even better with more
iimprovements
t iin co-existence
i t
by
b utilizing
tili i innovative
i
ti systems
t
and
d
circuits techniques as well as taking further advantage of the power
of DSP for self-calibration in order to accommodate the complex
needs of future communication standards
A. Behzad
IEEE DL 2011: MIMO Radios
97
Acknowledgements
With many thanks to the following individuals who have contributed to the
sslides
des a
and/or
d/o helped
e ped review
e e tthe
e sslides
des
Rohit Gaikwad
George Chien
David Su
Jason Trachewsky
Iason Vassiliou
Yorgos
g Palaskas
Bill McFarland
Won-Joon Choi
Vinko Erceg
Michael Hurlston
A. Behzad
IEEE DL 2011: MIMO Radios
98
References
[1] A. Behzad, “The Implementation of a High Speed Experimental Transceiver Module with an Emphasis on CDMA
A li ti
Applications”,
” El
Electronic
t i Research
R
h Labs,
L b U
U.C.
C B
Berkeley,
k l
1994
1994.
http://bwrc.eecs.berkeley.edu/Publications/1995/theses/Imp_hi-spd_ex_trnscvr_mod_emp_CDMA_ap/index.htm
[2] S. Sheng, “Wideband Digital Portable Communications: A System Design”, Electronic Research Labs, U.C.
Berkeley, 1991.
[3] T. S. Rappaport, Wireless Communications - Principles & Practice, IEEE Press, 1996.
[4] D. Tse, et. al. Fundamentals of Wireless Communications, Cambridge U. Press, 2005.
[5] B. McFarland, “Advanced Wireless CMOS Receivers”, GIRAFE Forum Presentation, ISSCC 2006
[6] Y. Palaskas, et. Al. ISSCC 2006
[7] W. Choi, et. al., “MIMO Technology for Advanced WLAN”, DAC, 2005
[8] W
W. Choi
Choi, et
et. al
al, “Circuit
Circuit Implications of MIMO Technology for Wdvanced WLAN”
WLAN , RFIC Symposium
Symposium, 2005
[9] D. Browne, “Experiments with an 802.11n Radio Testbed”, UCLA/802.11n committee, July 2005.
[10] A. Behzad, “WLAN Radio Design” ISSCC Tutorial, 2004.
[11] I. Vassiliou, “A System-Level Approach for RF CMOS: A Digitally Calibrated 802.11a/b/g Direct Conversion
Transceiver”, MEAD Course, 2006.
[12] H
H. H
Hashemi,
h i et. al,l “A 24
24-GHz
GH SiG
SiGe phased-array
h
d
receiver-LO
i
LO phase-shifting
h
hif i approach”,
h” IEEE T
Transactions
i
on
Microwave Theory and Techniques, Feb. 2005.
[13] J. Paramesh, et. al. “A four-antenna receiver in 90-nm CMOS for beamforming and spatial diversity”, JSSC Dec.
2005
[14] 802.11n draft: http://www.ieee802.org/
A. Behzad
IEEE DL 2011: MIMO Radios
99
General Radio/WLAN Design Bibliography
For the interested reader and to find out more about Radio design for WLAN
systems, the following references offer valuable information:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
http://www.ieee802.org/
http://grouper.ieee.org/groups/802/11/QuickGuide_IEEE_802_WG_and_Activities.htm
A. Behzad, “WLAN Radio Design” ISSCC Tutorial, 2004
A Montalvo, “Highly integrated RF & Wireless Transceivers”, ISSCC 2003 Tutorial
A. Behzad, “Wireless LAN Radios: System Definition to Transistor Design”, IEEE Press, 2007.
T Tuttle, “Introduction to Wireless Receiver Design”, ISSCC 2002 Tutorial
D. Shoemaker, et. al., “Wireless LAN: Architecture and Design”, ISSCC 2003 Tutorial
B. Cutler, “Effects of Physical Layer Impairments on OFDM Systems”, RF Design, May 2002
R. Van Nee, R. Prasad, “OFDM For Wireless Mulimedia Communications”, Artech House Publishers, 2000.
I. Bouras, et. al., “A Digitally Calibrated 5.15GHz - 5.825GHz Transceiver for 802.11a Wireless LANs in 0.18um
CMOS”, ISSCC 2003
M. Zargari, “Challenges in the Design of a CMOS RF Transceiver for IEEE 802.11a Wireless LAN”, VLSI
Symposium Short Course, 2003
D. Su, et. al., “A 5GHz CMOS Transceiver for 802.11a Wireless LAN”, ISSCC 2002
A. Behzad, et. al., “A Direct-Conversion CMOS Transceiver with Automatic Frequency Control for IEEE 802.11a
Wireless LAN”, ISSCC 2003
T. Schwanenberger, et. al., “A Multi Standard Single-Chip Transceiver Covering 5.15 to 5.85GHz”, ISSCC 2003
B. Cutler, “Effects of Physical Layer Impairments on OFDM Systems”, RF Design, May 2002
J. Troychak, “The Design and Verification of IEEE 802.11a 5GHz Wireless LAN System”, Agilent-Eesof Technical
Note
“Making 802.11g Transmitter Measurements”, Agilent Application Note 1380-4
“RF Testing of Wireless LAN Products”, Agilent Application Note 1380-1
B C
B.
Come, “I
“Impactt off Front-End
F tE dN
Non-Idealities
Id liti on Bit E
Error R
Rate
t performance
f
off WLAN
WLAN-OFDM
OFDM T
Transceivers”.
i
”
A. Behzad
IEEE DL 2011: MIMO Radios
100
General Radio/WLAN Design Bibliography (2)
18
18.
A Abidi,
A.
Abidi “Direct
Direct Conversion Radio Transceivers for Digital Communication
Communication,” JSSC,
JSSC Dec
Dec. 1995.
1995
19.
A. Abidi, “CMOS Wireless Transceivers: The New Wave,” IEEE Communications Magazine, Aug. 1999.
20.
F. Behbahani et al., “CMOS Mixers and Polyphase Filters for Large Image Rejection,” IEEE J. of Solid State
Circuits, Jun. 2001.
21
21.
R Carsello
R.
Carsello, “IMT-2000
“IMT 2000 Standards: Radio Aspects
Aspects,”” IEEE Personal Communications Conference,
Conference Aug
Aug. 1997
1997.
22.
J. Crols et al., “Low-IF Topologies for High-Performance Analog Front Ends of Fully Integrated Receivers,” IEEE
Transactions on Circuits and Systems-II, Mar. 1998.
23.
T. H. Lee, “The Design of CMOS RF ICs,” Cambridge University Press, Jan. 1998.
24
24.
B R
B.
Razavi,i “RF microelectronics,”
i
l t i ” Prentice
P ti Hall
H ll, Nov.
N
1999
1999.
25.
R. Adler, “A Study of Locking Phenomena in Oscillators,” Proc. Institute of Radio Engineers, Vol 34., pp. 351-,
June 1946
26.
M. Steyaert et al., “A 2V CMOS cellular transceiver front-end,” IEEE J. of Solid State Circuits, Dec. 2000.
27
27.
J F
J.
Fenk,
k “Hi
“Highly
hl IIntegrated
t
t d RF IC
ICs ffor GSM and
d DECT S
Systems
t
– A Status
St t Review,”
R i ” JSSC,
JSSC Dec.
D
1997.
1997
28.
J. Strange et al., “A Direct Conversion Transceiver for Multi-Band GSM Application,” IEEE RFIC Symposium,
2000.
29.
A. Behzad, et. al., “A 5-GHz Direct-Conversion CMOS Transceiver Utilizing Automatic Frequency Control for
IEEE 802.11a Wireless LAN Standard”,
Standard , JSSC, December 2003
D.H. Morais, K. Feher, “The Effects of Filtering and Limiting on the Performance of QPSK, Offset QPSK, and
MSK Signals,” IEEE Transactions on Communications, Vol. 28, pp. 1999-2006, December 1980.
J. F. Sevic, J. Staudinger, “Simulations of Adjacent Channel Power for Digital Wireless Communication
Systems,” Microwave Journal, pp. 66-80. October 1996.
A Hajimiri
A.
Hajimiri, T
T. Lee
Lee, “ Design Issues in CMOS Differential LC Oscillators,
Oscillators ” JSSC,
JSSC pp.717-724,
pp 717 724 May 1999
1999.
P. Gray, R. Meyer, “ Analysis and Design of Analog Integrated Circuits,” Third Ed., Wiley, 1993.
30.
31.
32
32.
33.
A. Behzad
IEEE DL 2011: MIMO Radios
101
Talk Abstract
 An Introduction
t oduct o to 802.11a/b/g/n/ac
80
a/b/g/ /ac Radio
ad o Design
es g : From
o
Systems to Transistors
 A short lecture on the evolution of the 802.11 standard with an
emphasis
p
on the radio design
g will be p
presented. After a brief discussion
of the market trends, the performance of the 802.11’s OFDM-based
modulations as well as multi-in multi-out (MIMO) design in an multi-path
environment will be outlined. The general evolution of the 802.11 PHY
from b to a to g to n to ac is presented and the impact of these PHYs on
the radio design is discussed. Further, the impacts of various radio
impairments (noise, quadrature inaccuracy, nonlinearity, etc.) on the
performance of the various
ario s PHYs will
ill also be presented
presented. Some specific
examples of circuit techniques and architectures used in various
802.11n radios will be discussed. The evolution of 802.11 radio
performance metrics throughout the past several years will be outlined
outlined.
The lecture will wrap up with some real-world throughput measurement
results and a brief discussion of the future trends of radio design.
A. Behzad
IEEE DL 2011: MIMO Radios
102
Author Bio
 Arya
y Behzad obtained his BSEE and MSEE from ASU and UC Berkeley
y in
1991 and 1994 respectively. After working for UTC, Microunity and Maxim,
he joined Broadcom in 1988 where he is currently a Senior Director of
Engineering working on radios and SoCs for current and future generation
wireless products
products. He was designated as a Broadcom Distinguished
Engineer in 2007 and as a Broadcom Fellow in 2009 for the design and
productization contributions to CMOS RF transceivers and power amplifiers.
He has published numerous papers and is an inventor on well over 100
issued patents in the areas of precision analog circuits, gigabit Ethernet,
set-top boxes and wireless networking. He has taught courses and
presented technical seminars at various conferences and at several
universities He is a retired member of the ISSCC Wireless Technical
universities.
Committee as well as a retired Guest and Associate Editor of the JSSC.
He has authored the book, “Wireless LAN Radios”, IEEE Press/Wiley. He is
an IEEE Distinguished Lecturer as well as an IEEE Fellow.
A. Behzad
IEEE DL 2011: MIMO Radios
103