2014 Exam Flashcards
Fractionally Spaced Equalisation (4)
- used when spacing between the filter taps of an LTE / feed forward taps of DFE are integer fractions of the symbol period
- fractional spacing solves the problem of performance sensitivity to symbol-timing error
- in a symbol spaced device, timing errors result in aliased components cancelling part of the main signal
- w/ fractional eq. the sampling freq. is sufficient to cover the entire modulation signal (incl. excess b/w introduced by bandlimiting filters)
Path Metric
sum of all branch metrics that lead to a given state
Branch Metric
squared error, resulting from a state transition from time ‘k’ to ‘k+1’
Survivor Path (2)
- path w/ lowest path metric leading to a given state
- each state has ONLY one path (survivor) that propagates to the next-time instance
Freq. Domain Equalisation (2)
- any method where the channel is equalised in freq. domain
- unlike convolution in t.d, FDE uses mulitplication in f.d
FDE exploitation in OFDM (3)
- a channel est. is generated for each sub-carrier, resulting in a smapled freq. domain est
- Blocks of received t.d data are transformed into freq domain using FFT
- FDE, each sub-carrier is multiplied by the inverse of channel est.
Why is Guard Time necessary + implementation (3)
- Ensures sub-bands remain orthogonal in a time dispersive channel
- if guard removed, OFDM suffers from interchannel interference
- implemented as a cyclical extension of the OFDM signal
What is CSI + how is it used in Rx (3)
- CSI - a vector of complex samples that represent the magnitude + phase of the channel on each sub-carrier
- Used to generate the FDE weights (either ZF / MMSE)
- FDE vector is then used to correct the channel ampl. + phase distortion on each sub-carrier
DS-SS Process
- Channel encoder
- Modulator
- Pseudo-noise generator
- De-modulator
- Pseudo-noise generator
- Channel Decoder
Why is SDMA beneficial in mmWave access?
high attenuation in mmWave band, SDMA applied to re-use high capacity beams
Angular Spread
spatial domain dual of time dispersion i.e. spreading or smearing of energy in the angular domain
LMS alg, why does the rate of convergence depends on eigenvalues spread of the received auto-correlation matrix?
- Iterative mthod of calc. tap weights relies on calc. est. of the steepest descent down the error gradient
- Eigenvectors of the error surface are the direction that lead to the optimum soln
- Eigenvalues determine the gradient in the direction fo the eigenvector
- Any update vector can be decomposed into a weighted set of eigenvectors
- Only 1 step size in the LMS alg.controls the amount the soln moves in opposition to gradient
- If 1 eigenvalue dominates, then convergence doesn’t progress well in other directions
- larger convergence time
- Measure of eigenvalue domination is eigenvalue spread
What is the Eigenvalue Spread?
ratio fo largest to smallest eigenvalue
What does the symbol represent?
Expectation Process
Mitigation of Near-Far Effect in FDD (3)
- UL + CL pwr control required to balance the Rx pwr of multiple users for UL
- UL pwr control based on DL Rx signal strength is used to adj. for path loss + shadowing in UL Tx
- CL is also used to compensate for freq. sensitive variations in UL + DL by making mobile user adj. Tx pwr
