Interference Effects of Multi-User Ultra-wideband Systems Anup Doshi Carnegie Mellon University July 31, 2003
Outline Intro Models Observations Summary
What is an Ultra-wideband Signal? Short impulses in succession FCC Definition – Bandwidth > 25% of center frequency
Advantages of UWB Low power levels spread over large spectrum Operates below noise floor of narrowband devices Possibility of >500Mbps short range GPS Frequency (Ghz) a -41 dBm/Mhz “Part 15 Limit” UWB Spectrum Source: Intel PCS
Potential Applications are Numerous Personal Area Network Interconnect Computers, Devices, PDAs, Printers Entertainment...TV, Camcorder, DVD Music…MP3, Audio Systems, etc Safety Through-wall Imaging Sensor Network Lots of other exciting applications Broadband UWB LAN/WLAN UWB Image Sources: Intel, AetherWire
Why Only Now? Started as impulse radar, 1960’s Primitive forms, simple communication Studied & used by military New technology enables digital comm., 1990’s Commercial applications seen by several companies Petitioned FCC to review potential uses FCC approves development of conservative applications
Problem… What happens when lots of UWB devices are transmitting in close proximity? Will the combined noise level be too much for a victim narrowband receiver? Existing studies claim minimal effects Done by various agencies and companies Those studies do not examine all cases… This is my job!
VICTIM Constant-Distance Distribution Multiple UWB devices located three meters from a victim
Units turn on and off in a 2-state Markov Process Switching times are Exponential Random Variables Time until on ~ Exponential(λ) => mean 1/λ sec Time until off ~ Exponential(µ) => mean 1/µ sec Rho=ρ= λ/µ Characterizing the Transmitters Unit Off Unit On λ µ
Characterizing the Transmitters Total Number on modeled as a Markov Chain Steady-state probabilities: 012N-1N NλNλ (N-1)λ…λ µ … Nµ2µ
How Does the System Act Over Time? λ =1, µ=2
How Does the System Act Over Time? Total Number of Units Onλ =1, µ=2
Noise Level in Victim Receiver Each UWB signal modeled as White Noise Total Noise= N 0 +M(t)*N 1 Ambient Noise Floor (=kTw) Number of Transmitters On (Markov Chain) Power Received at Victim from UWB Signal
Some Properties of This Model Autocorrelation Spectral Density
Probability of Error in Receiver On Average: (µ=1)
Other Ways to Describe Model Probability of Outage P(outage)=Probability( P err > Pe* ) Pe*=.1,.01 P err = Expected Time of Outage E(T 10 ) = T 1 +a N,1 E(T 20 ) E(T 20 )=T 2 +a N,2 E(T 10 )+b N,2 E(T 30 ) … E(T N0 )=T N +E(T (N-1)0 )
P(outage), Expected Time of Outage (µ=1) (µ=10)
VICTIM Random-Distance Distribution UWB devices distributed uniformly in a circular area around victim
Properties of Random-Distance Model Moved to a computer simulation Experimentally calculated: P(outage), Expected Time of Outage, Max and mean power levels over time Done on Matlab – Monte Carlo simulation
Example Simulation Run
P(outage) & Expected Time of Outage (µ=1)(µ=10)
Max/Mean Power Levels
Observations Multiple transmitters will cause major problems in worst cases Such situations may soon arise in real-life situations Important to consider every possible case in testing Broadband
Future Work Need to consider many more variables Receiver type Frequency, PRR Different distributions Once 802 Standard comes out, incorporate into model Possibly Multi-Band OFDM (TI, Intel) Possibly Dual-Band (Time Domain, Motorola)
Summary Characterized an aggregate of UWB transmitters Realized various methods of measuring effect on victim receiver Concluded that as number of UWB transmitters increase, performance of victim receiver attenuates
Acknowledgments Prof Baum Prof Noneaker, Prof Xu ECE Faculty and Grads NSF