2017 Exam Flashcards
How a DFE is able to remove noise from ‘post cursor’ symbols (2)
- DFE hardlimits the eq o/p to create noise free est of transmitted symbols
- Passed into feedback filter section which cancels post cursor ISI
Adv/disadv of DFE over LTE (2,2)
Adv
- hardlimited symbols introduce (if correct) no noise into eq o/p symbols
- # taps feedback section = # post cursor taps
- not true for feedforward filter
Disadv
- hardlimited symbols may not be correct, impair eq. when used in the feedback filter to cancel ISI
- Errors can propagate
Adv/disadv of training seq w/ decision directed adaptation (2,2)
Adv
- training seq allows exact calc. of o/p error by providing perfect nowledge of the transmitted symbol
- decision directed calc use imperfect est
Disadv
- No info transmitted, represents an overhead
- allow error calc when actual info is being transmitted
How are training seq and decision directed adaptation best combined? (3)
- Best combo is to use periodic training symbol / seq
- Keep data symbols > training symbols
- Keep loss in useful throughput fairly minimal
- provides accurate error calc to initilalise the eq to a good state before switching to decision directed calc
Benefits of MMSE over ZF (3)
- MMSE considerst the noise power in determining eq. co-effs in order to min. overall error by noise + channel disp.
- ZF only considers the disp + attempts to cancel it completely
- In deep fades, ZF can cause noise ampl.
When do ZF + MMSE perform similarly? When does MMSE work best?
- @ high SNR, (absence of noise), MMSE = ZF
- @ low SNR, noise ampl. of ZF can be severe (MMSE performs better)
Assumptions used in LMS alg. to approx Wiener-Hopf? (2)
- Treats gradients separatly for each dimension (tap) rather than calculate for all dimensions
- LMS uses instantaneous values as approx. of long term av.
Index of Discrimination definition
difference between peak auto and cross-correlation function
How to maximise Index of Discrimination
can be maximised by selection of preferred pairs of spreading codes (or m-sequences)
Near Far Effect definition (2)
For a DS-SS UL w/ all users transmitting @ same pwr level, users near the BS will suppress/saturate far users
Only separated in the code domain
Prevention of Near Far Effect
UL pwr control can be adjusted such that all users are receieved w/ same pwr @ BS receiver
Additional process required for FDD air interface?
As UL + DL have uncorrelated fading in FDD network, closed loop pwr control us used to adjust for freq selective multipath fading
Difference between SDMA w/ BF and Spatial Multiplexing using MIMO? (2)
SDMA BF
applied to separate users provided that they are sufficiently separatd in space
Spatial Multiplexing
exploits the spatial temporal characteristics of the multi-user channel, allowing channel separation even when users are in close proximity
Multiple isolated spatial channels can be associated with …
… a single user thereby incr. the spectrum eff. of the link
Difference between Massive MIMO + sub-6GHz mmWave freq? (3)
- Massive MIMO has an excess of antennas at BS serving fewer users (10-100)
- Sub 6GHz Massive MIMO full digital signal processing is associated w/ each antenna @ BS therefore full spatial multiplex is possible
- mmWave, analogue BF is applied to the wider b/w channel in order to compensate for higher path loss at these freq.