Cognitive RF Front-end
Post on 20-Jul-2016
6 Views
Preview:
DESCRIPTION
Transcript
A Probabilistic Performance Metric
forRF Front-ends in Wireless Devices
Eyosias Yoseph Wireless@VT
Presentation Thesis
Traditionally, deterministic metrics are used to quantify the performance of RF front-ends These metrics tell the performance of the RF front-end under
specific scenario They are not convenient to define the reliability of RF front-ends
In our research, we are developing probabilistic metrics These metrics can be used to define the reliability of RF front-
ends
Content
Wireless history: from Marconi to Smartphones The Future of Wireless Probabilistic RF front-end metrics
Content
Wireless history: from Marconi to Smartphones The Future of Wireless Probabilistic approach of modeling RF front-ends
Wireless History: The Birth of Radio 1895: First long
distance radio communication (“Wireless telegraph”)
1902: First wireless trans-Atlantic telegraph
Wireless History: Voice
1912: First radio based voice broadcast 1920: First mobile radios in Detroit police cars 1935: Frequency modulation was invented
Wireless History: Mobile Telephony
1946: Public switching network based Mobile telephony was started 1972: Motorola made the first phone call from portable mobile
telephone
Wireless History: Cellular Systems
Cellular systems were introduced in the 1980’s
Cellular systems divide the geographical areas into smaller cells
Each cell has its own tower Frequency is re-used between
cells Cellular technology presented a
significant increase in network capacity
Wireless History: Frequency Reuse
f1 f2
f3
Wireless History: Frequency Reuse
f1 f2
f3f3
Wireless History: Smartphones Smartphones
Trends in Wireless
Exponential growth Mobile video is the main source of traffic
Trends in Wireless
FCC: The demand for spectrum is not matched by its availability
PCAST: This has huge opportunistic cost on the economy
Content
Wireless history: from Marconi to Smartphones The Future of Wireless Probabilistic RF front-end metric
The Future of Wireless: Spectrum Sharing
Fixed allocation based spectrum management has been used for decades PCAST recommended that the federal government shares its spectrum with
commercial users Example: 3.5 GHz radar bands
The Future of Wireless: Spectrum Sharing
Fixed allocation based spectrum management has been followed for decades PCAST recommended a sharing federal spectrum for commercial use
Example: 3.5 GHz radar bands
The Future of Wireless: Small Cells
Further reducing the size of the cells Example: WiFi router type base-stations
(femtocells) in each home
Millimeter Wave 28 GHz and higher are being considered Easy to obtain 1 GHz of frequency Only for short range, line-of-sight communication Beam forming is crucial (multiple antenna use)
Content
Wireless history: from Marconi to Smartphones The Future of Wireless Probabilistic RF front-end metric
Poorly Selective Receivers
Upcoming wireless technologies likely contain poorly selective receivers
Spectrum Sharing
• Spectrum sharing uses tunable filter• Tunable filters have 10-20% bandwidth• At 1 GHz, this corresponds to 100 -200 MHz bandwidth
mmWave
• mmWave filters have Q 10• At 28 GHz, this corresponds to 2.8 GHz 3-
dB bandwidth
SAW filters can be as selective as 1 MHz at 1 GHz
Receiver Selectivity 101fs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Receiver Selectivity 101fs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Receiver Selectivity 101fs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Receiver Selectivity 101fs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
What if it is impossible or hard to get sufficiently selective pre-selector filter for our receiver?
That is the problem in spectrum sharing based receiversThat is the problem in mmWave receivers
Receiver Selectivity 101fs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Wireless devices of the future will be poorly selective
Selectivity and Reliability
Poor selectivity implies higher rate of “dropped-calls”
Hence, poor selectivity makes a wireless device less reliable
How can we improve the reliability of a poorly selective receiver (wireless device of the future)?
Potential Solutions
Improving the filter technology MEMS SAW filters in mmWave range??
Filter technology did not show fast improvements in the past
Not very prospective
We propose re-defining selectivity using probabilistic performance metrics
We also propose the use of artificial intelligence to control the parameters of the receiver
Receiver Selectivity
Signals outside the pass-band of the filter are rejected --- always
Input
Output Rejection
Receiver Selectivity
Would this type of receiver work?
Input
What if a strong adjacent channel signal occurs only 0.01% of the time?
Level of rejection is not a reasonable reliablity and performance metric
Receiver Selectivity
Would this type of receiver work?
Input
What if a strong adjacent channel signal occurs only 0.01% of the time?
We propose using probability of outage (“drop-call”) as reliablity and performance metric
Practical Measurement
1 2 30
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Number of Active Signals
Prob
abilit
y of O
utage
("dr
op-ca
lls")
Filterless receiverModerately selective receiverHighly selective receiver
Without cognitive engine
Withcognitive engine
Number of Adjacent Channel Interferers
Prob
abilit
y of
Out
age
(“dr
op-c
alls”
)
Strong Received Signals are Rare
Strong adjacent channel signals can be detrimental in a poorly selective receivers
But, probability of receiving a strong adjacent channel signal is very low Wireless@Virginia Tech showed that the probability of received signal
power is exponentially distributed in logarithmic domain
-100 -80 -60 -40 -20 0 200
0.2
0.4
0.6
0.8
1
Power, dBm
CDF
It is rare to receive a signal with power level more than -60 dBm
Using Artificial Intelligencefs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Using Artificial Intelligencefs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Using Artificial Intelligencefs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Using Artificial Intelligencefs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
Using Artificial Intelligencefs/2
fLO fs
Antenna
LNA Mixer Baseband Filter ADC DSPPre-selector
RF frequency
INPUT SEPECTRUM
RF frequency
fLO
LNAOUTPUT
Baseband frequency
DCMIXER OUTPUT
DSP frequency
DCADC OUTPUT
+fs/2-fs/2
By intelligently Controlling the parameters of the receiver, the desired signal can be protected from interference of adjacent channel signals - without using RF filters.
Proposed Receiver Architecture
Receiver Receiver
Towards filter-less receivers
Receiver
Cognitive Engine
Practical Measurement
1 2 30
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Number of Active Signals
Prob
abilit
y of O
utage
("dr
op-ca
lls")
Filterless receiverModerately selective receiverHighly selective receiver
Without cognitive engine
Withcognitive engine
Number of Adjacent Channel Interferers
Prob
abilit
y of
Out
age
(“dr
op-c
alls”
)
Conclusion
Reliability of a wireless device is not necessarily defined by the selectivity of its filter
Using probability of outage (“drop-call”) may be a better reliability metric
This is particularly true in dynamically changing spectrum scenarios in which strong received signals are rare
Adding a cognitive engine can improve the reliability of a poorly selective radio
Thank you, Questions ? Comments?
top related