MIMO Measurement Campaign Data and Graduate Project Course in Wireless Implementation

1
MIMO Measurement Campaign Dataand Graduate
Project Course in Wireless Implementation
Per Zetterberg perz_at_s3.kth.se Wireless_at_KTH/Signal
Processing Royal Institute of Technology
2
Multi-Cell MIMO Campaign
BS 2
BS 1
MS
15km of vehicle trajectory!
Simultaneous MIMO measurements on two sites!
3
The Multi-Cell MIMO Measurement Campaign

• Typical European urban environment with mostly
  6-8 stories buildings.
• Outdoor to indoor measurements included.
• Measurements are uplink.
• One 100mW CW per transmitting antenna (offsets
  1kHz).
• Both base-stations receive simultaneously.
• GPS time and position in both BS and MS.

4
Strongest Beam Results
5
F2E5520 - Project Course in Smart Antenna
Implementation
New name in autumn 2005. Start in November
approximately.
6
The Course

• Work in groups of 2 or 3 persons.
• Pre-requisites A good understanding of
  modulation, frequency offset, equalization,
• Working knowledge of programming language C.
• Course starts with lectures and laboratories on
  the test-bed and on the programming environment.
• Main part of course is implementation of a system
  (that the group define themselves) on the MUMS
  test-bed.
• Examination is writing a conference paper on the
  implementation.
• The paper should contain a comparison of theory
  and practice.
• The groups share the available hardware.

7
Multi-User MIMO Test System (MUMS)

• Consists of four nodes having
• Either two radio-transmitters or radio-receivers,
  carrier freq. 1800MHz, Bandwidth 10kHz.
• A EVM6701 DSP-boards.
• Host PCs. Visual C.
• Connected to the other nodes using LAN with AFS.

8
MUMS parts
Projects TX_Algorithm(), RX_Algorithm()
Basic DSP and PC software MUMS_skeleton MUMS_host
Hardware Radio modules, oscillators, antennas,
DSPs, PCs
9
Air-interface common-parameters

• Basic air interface parameters
• Sample rate 48000Hz.
• Symbol rate 9600Hz.
• Pulse shape Root-raised cosine roll-off 0.5.
  Truncated to 16 samples.
• Receiver filter Pulse shape.
• Frame and buffer length 32 symbols 3.3ms.
• Run time 68 frames 200ms (repeated).
• Feedback channel data rate 32bits / frame
  10kbit/s.
• Feedback delay 5 frames 16.5 ms
• Maximum output power 2 x 0.4mW
• Carrier frequency 1766.600MHz
• Receiver noise figure 7dB.

10
Narrow-Band Receiver/Transmitter (14kHz)
11
Single-cell set-up
TX1
Internet
RX1
12
Two-cell set-up
RX1
TX1
TX2
RX2
13
MUMS_skeleton

• Does the DSP side of the start-up procedure.
• Feedback , buffer handling, etc. already
  implemented.
• Starting point for the groups implementations.
• Just change TX_algorithm and RX_algorithm.
• ..
• while ((frame_ixltno_frames_tot)
  (frame_traffic_ixlt(no_frames_traffic2)))
• switch(State)
• case RX1_START ..
• ..
• case RX1_RECEIVING
• DownConvert(
• Filter(.
• RX1_Algorithm()
• modulate_feedback()
• .

14
Just changing TX_Alg Involves

• Air interface.
• Channel estimation.
• Fine synchronization.
• Fine frequency offset estimation/compensation.
• Equalization.
• TX/RX algorithm design e.g. beamforming, Alamouti
  etc.
• Interference rejection ?
• Implementation with memory and clock-cycle
  constrains.

15
Host-control
Define algorithms to run in sequence.
Performance evaluation parameters
Custom parameters
Matlab-script for presentation of results.
Location of files where to store results
16
Performance Presentations Using Matlab
The system automatically estimates an logs
channel and noise based on sinusoid transmission.
17
Groups 2003 edition.

• Spatial Multiplexing (SM).
• Joint Transmitter Receiver Interference rejection
  (RxTxIR).
• (Influence of ADC errors on beam-forming
  performance)
• Alamouti-coding with frequency-domain
  equalization.

• Participants have spent about two man-month on
  the project (but many more calendar months).
• Output 4 conference and 1 journal paper.

18
Spatial Multiplexing (SM)

• Targets high data rate
• High SNR and low mobility
• Two spatially multiplexed data streams using
  Rx-Tx feedback
• Data protection by 64 state TCM code
• Adaptive rate, by constellation size
• Throughput of up to 150 kbps over a 9.6 kHz
  channel
• Decision feedback directed channel and temporal
  equalizer estimation in receiver

19
SM Overview
16 bits /symbol! Two months development time!
TX
Internal RX feedback of channel estimates
RX
Detection Decoding
Post-detection channel estimation
Constellation and beamformer selection
RX to TX feedback
20
SM Mathematical Model
21
SM RX-TX Feedback

• Feedback of unitary precoding matrix V, which is
  parameterized as
• The 32-bit feedback word

22
SM Processing Loads.
SM 70 of C6701. Skeleton (Filtering, feedback
etc) 8.
Relative load of SM sub-tasks
23
SM Performance Result
Achieved SNR (Estimated)
Ideal SNR (without ISI, frequency and
synchronization errors and non-linearities)
24
SM Throughput
Throughput bits/symbol
Ideal SNR
25
2x2 versus 1x1

• Transmitter in end of corridor moving towards
  receiver.

2x2
100kbit/s
1x1
Time (0-45sec)
26
Receiver Transmitter Interference Rejection
(RxTxIR)

• System Setup

Cell 1
Rx1
Tx1
Interference!
Tx2
Cell 2
Rx2
27
Receiver Transmitter Interference Rejection
(RxTxIR)

• Idea
• Perform beam-forming in both transmitters and
  receivers to jointly optimize SNIR.
• Can be seen as pre-whitening in transmitter and
  receiver.
• Requires knowledge and estimation of
  cross-channels.

28
Receiver Transmitter Interference Rejection
(RxTxIR) Experimental Result
Without pre-whitening.
With pre-whitening.
MS1
MS1
MS2
MS2
29
Remember

• A proper course announcement will be made later.
• Give me an indication if you are interested in
  taking the course please give me an indication by
  Tuesday.
• For projects outside single-cell and two-cell
  ask me.
• Course capacity is somewhere in the 2-4 groups
  range.
• If you are interested in the measurement data
  email me.