Determining and Characterizing Frequency Hopping Interferers
This article discusses the challenges of determining and characterizing the number of frequency hopping interferers in wireless communication systems. Frequency hopping spread spectrum devices offer resistance to narrow band interference,
- Uploaded on | 0 Views
About Determining and Characterizing Frequency Hopping Interferers
PowerPoint presentation about 'Determining and Characterizing Frequency Hopping Interferers'. This presentation describes the topic on This article discusses the challenges of determining and characterizing the number of frequency hopping interferers in wireless communication systems. Frequency hopping spread spectrum devices offer resistance to narrow band interference,. The key topics included in this slideshow are . Download this presentation absolutely free.
Slide1Determining and Characterizing theNumber of Frequency Hopping Interferers using Time and Frequency Offset Estimation Alican Gök Prof. Danijela Cabric
Slide203/12/20092 FREQUENCY HOPPING SPREAD SPECTRUM DEVICES • Resistance to Narrow Band interference • It's hard sniff the signal w/o knowing the hopping sequence • Can share frequency band with conventional conventional signals – At 2.4GHz FHSS devices have co-existence issues with variety of other wireless devices – Example: Bluetooth (79 1 MHz channels interfering with Wi-Fi networks)
Slide3PROBLEM• Cognitive Radio Perspective – Need to detect transmitters & classify transmission parameters • System Security Perspective – Need to recognize transmitters & identify whether they belong to the network • Motivations : – Each device has an inherent and unique center frequency offset/imperfection due to manufacturing • This will be used to distinguish between devices & characterize them – The starting time of transmissions for different transmitters are random across time • It is unlikely that any two transmitters begin transmission at the same time. (within ±1 µs) • Thus, temporarily unique “timing offsets” can be utilized
Slide4ASSUMPTIONS1. Each transmitter performs frequency hopping over a known set of M subchannels (79, 1MHz channels) 2. The transmission frame period is identical for each transmitter and is known a priori (625 µs) 3. Transmissions are assumed to have a duty cycle of 100% and are also active every frame period during the observation interval 4. Transmission start times are assumed to be uniformly distributed over one frame period 5. The maximum number of transmitters is much less than the number of subchannels 6. The signal-to-noise ratio (SNR) for detected signals are high (fixed at 20 dB for the simulations)
Slide5PROCEDURE1. Estimation of Number of transmitters: – Detection of transmission start times – Construction of the Time Offset Histogram – Threshold the Histogram to estimate # TX 2. Estimating the frequency of bursts – A two stage algorithm is proposed. – Different methods (Sliding Correlation, Recursive Methods) are explored. 3. Post-processing to find the final carrier frequency offsets for the number of transmitters
Slide6SYSTEM BLOCK DIAGRAM
Slide7ESTIMATION OF NUMBER OF TRANSMITTERS• Using 5 µs-long observation intervals with 100 MHz sampling, get snapshots of the full 79-MHz spectrum. • Detect the subchannels in use by a simple energy detection algorithm based on thresholding • The FFT window is shifted by 1 µs successively (STFT) • By comparing the channels in use at consecutive snapshots, the “new” transmitting subchannels are detected Start Time of a Burst
Slide8•By associating each burst start with the time modulo 625 µs, and accumulating over T 0 seconds, a histogram is obtained. • A threshold on the Histogram is applied to find the number of transmitters, and their time offsets. ESTIMATION OF NUMBER OF TRANSMITTERS
Slide9FREQUENCY OFFSET ESTIMATION• Using the same subchannel detection algorithm as in the previous stage, the burst starts are detected. • If the time of detection matches a time offset found in the previous stage, frequency offset estimation is conducted. • Over T 1 seconds, for all transmitters, multiple bursts are acquired, which improves the frequency estimates.
Slide10FREQUENCY OFFSET DETECT ALGORITHM• For each burst, mix the signal to baseband by using the subchannel frequency. • Low Pass Filter and Decimate (by 100 to get 1 MHz sampling) • Use the FFT of the baseband signal to run the Sliding Correlation Algorithm – With 625 µs pulses, can have 625 points at 1 MHz sampling. If 512 pt FFT is used frequency resolution is ~2 kHz Idea : If we wanted to obtain 2 kHz freq. res in a single stage method without mixing and decimating, then we would need ~50.000 pt. FFTs at 100 MHz sampling. Not practical to implement 03/12/2009 10
Slide11SLIDING CORRELATION• Inspired by the Matched Filter detection • Use the average PSD of CPM pulses with given parameters (Bluetooth – CPM) as a template for the correlation. (Shown in RED) • Smooth the FFT of the received pulse (GREEN) with a smoothing parameter (BLUE) • Slide the template over a range to find the max. correlation with the blue FFT. • Estimated Freq Offset = Frequency at which the maximum correlation occurs! – Can do the correlation using abs(FFT), abs(FFT)^2 or log(abs(FFT)) 03/12/2009 11
Slide12DEALING WITH COLLISIONS• Collision Superposition of pulses • Decrease the performance of frequency offset estimation. • Solution: Track end times of the bursts to cancel out bursts which are affected by collisions.
Slide13RESULTS• A summary of the most significant results will be shown. • Simulink was used to simulate Bluetooth devices. • Simulates desired number of FHSS transmitters, with configurable: – Frequency offsets, starting time offsets – SNR, channel noise – CPM characteristics – Data rate – Hopping sequences, duty cycle – Burst time, frame periods • Reception of signals, time and frequency offset estimation is conducted using embedded Matlab codes in Simulink.
Slide14DETECTION OF THE BURSTS
Slide15ESTIMATION OF NUMBER OFTRANSMITTERS
Slide16FREQUENCY OFFSET ESTIMATION(Characterization of the SC Algorithm)
Slide17FREQUENCY OFFSET ESTIMATION(Without Collision Detection)
Slide18FREQUENCY OFFSET ESTIMATION(With Collision Detection)
Slide19FUTURE WORK• Theoretical Analysis • Gain robustness against spillage, by improving subchannel detection • Remove certain assumptions to enable the system to work with different devices, such as Software Defined Radios in the 2.4 GHz band. • Real-world implementation, using USRP2 ( Universal Software Radio Peripheral) Boards, BEE2 or other hardware.
Slide21WHY USE CARRIER FREQUENCY OFFSETS? • We would like to characterize transmitters, to – Help spectrum sensing in Cognitive Radio – Find out sources of interference to 802.11b/g networks. – Create fingerprints for devices for future use. In trying characterize bursts coming from different transmitters: • Hard to find protocol-specific pulse characteristics: – Protocol type (Bluetooth, 802.11, generic SDR), modulation type (FSK, OFDM, QPSK), alphabet size • Some device-specific characteristics are hard to estimate, unreliable or does not help us in distinguishing btw transmitters: – Pulse power, pulse start-time, pulse periods, ramp times • Need to use carrier frequency offsets – Inherent in and Unique to any transmitter 03/12/2009 21
Slide22FREQUENCY RESOLUTION ISSUES• Pulse duration: t, Sampling time: f s, FFT size: N=t*f s • FFT spans the [ –f s /2 : f s /2 ] range • Resolution = Distance btw two consecutive FFT bins = f s /N = f s / (t* f s ) = 1/t • For a duration of pulse t Get 1/t freq. resolution – Zero padding, interpolation, etc. does not give us any additional statistical data. • Called “Heisenberg-Gabor inequality” • Vital to capture the whole pulse to get a better freq. resolution, thus a better freq. offset estimate 03/12/2009 22
Slide23System Performance - SNR• Single Transmitter / Single Burst • Sliding Correlation Method • Most of the uncertainty in the estimation of offsets is due to the sampled spectrum. Significant variance exists even at very high SNR values.
Slide24DETERMINATION OF THE THRESHOLDFOR THE HISTOGRAM (ESTIMATION OF NUMBER OF TRANSMITTERS)