Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Goals
Principal Elements
Process
Path Loss Measurements
Experiment Application Design
Connection Evaluation Steps
NEC Sector Antenna Tilt
Range and Throughput Measurements
Plan
Results
Summary
Conclusions and Next Steps
Authors: Manu Gosain, Tony Michel, Tom Cahill, Harry Mussman
Goals
Validate base station installation and configuration process
Provide comprehensive documentation
Design an experiment to evaluate range and throughput
Document for use by other sites in evaluating their
expected range and throughput
Later: move to OMF/OML environment
Evaluate range and throughput at BBN site
Compare to known calculations, measurements
Document for use by other sites in estimating their
expected range and throughput
Principal Elements
Base station kit (BTS)
Utilizing NEC Profile C IDU and ODU
Rooftop antennas
NEC 120deg sector
Commercial omnidirectional
Anritsu spectrum analyzer, for measuring received power
Linux laptop with Intel 6250 WiMAX modem, acting as a mobile station (MS)
BTS servers, including:
ASN GW with WiMAX RF AggMgr (Case 1b)
Test host
I&M host
Experiment application, running in:
MS (measurement script)
Test host (ping and iperf servers)
I&M host (report script)
Process
1) Conduct power measurements using Anritsu spectrum analyzer
Check for presence of Clearwire signal with Anritsu spectrum analyzer
1) Power Measurements
Power measurements using Anritsu spectrum analyzer
Measured with sector antenna, 6deg mechanical tilt
Near antenna: -34dBm
Point 41, 370ft: -59dBm (good signal)
Point, 520ft: -50dBm (good signal)
Point, 1190-ft: -79dBm (edge of coverage)
Presence of Clearwire signal with Anritsu spectrum analyzer
On roof (line of sight): -60dBm
Point 47: -70dBm
1) NEC sector
2) Omni
GREtunnel
DHCP
NEC Base
Station
(BTS)
WiMAX
modem/antenna:
BTS
ODU/IDU
1) USB-connected Intel
6250/external omni
2) internal Intel 6250/
internal
1) Range/throughput
experiment script
tstats2:
ASNG
W
salamis
2) WiMAX AggMgr
service:
Record location
Scan/connect/chk RSSI
Get IP via DHCP
ping sequence
iperf sequence
Log results
Sponsored by the National Science Foundation
public
Internet
.bbn.dataplane.geni.net
I&M host
black
3) Report script
report:
(manually gather logs
from MS and BTS)
Process logs
Generate location
summaries
Generate run summary
Test host
argos
4) Test
targets:
ping
iperf
(continued)
Step 4) Do a sequence of iperf tests between MS and Test Host
argos
Repeat 3 times
Use TCP
Use -d for double connection, separating DL and UL measurements
Throughput in Mb/s calculated from bytes transmitted within
60sec interval
Print throughput in Mb/s to log
TCP parameters:
use Nagles algorithm
window size and segment size per OS: 16kB
depth read/write buffer in socket, default: 8kB
max segment size: 1408B (MTU size) - 40B = 1368B
Use of TCP gives conservative result, but typical of many
applications
4) Measurements Plan
Focus on line-of-sight, outside only (gives best case)
Points 41 through 48, in a straight line at center of 120deg sector pattern
Optional points 1 through 7, in orthogonal direction (with point 7 obstructed by
building), to verify expected 360deg omni coverage
Keep nominal BTS configuration parameters
Power set to +38dbm, the maximum allowed
Options for base station antenna:
NEC sector base station antenna (at 4deg mechanical down tilt), approx 90ft high
Omni-directional base station antenna, approx 90ft high
Expect sector to work better than omni antenna within 120deg sector pattern,
since has higher gain
Options for Linux laptop mobile station (MS):
External (USB-connected) 6250, with handheld large omni-directional antenna
Internal Intel 6250 WiMAX modem, and internal antenna
Expect large omni antenna to work better than internal antenna
Expect packet loss and throughput to vary from moment-to-moment, due to MS
position and multi-path propagation
10
(continued)
For each option combination:
A) BS sector, MS omni antennas
B) BS sector, MS internal antennas
C) BS omni, MS omni antennas
D) BS omni, MS internal antennas
For each point:
41 48
option for C): 41 48 and 1 7
Plot vs distance (mi) from base station to mobile station:
DL RSSI (db)
UL RSSI (db)
1008byte pings, the % of responses not within window (lost)
DL iperf throughput, min and max over three attempts (Mb/s)
UL iperf throughput, min and max over three attempts (Mb/s)
11
Neighborhood of
BBN Technologies, Cambridge, MA
12
BBN
Base
Station
0
4
1
4
24
4
4
3
4
5
4
6 4
7
4
8
0
(0 mi,
90 ft
up)
BBN
Base
Station
Antenn
a
41
(0.058
mi)
Center
Parking
Lot
42
(0.085
mi)
Social
Security
Entrance
43
(0.090
mi)
NE
FawcettConcord
44
(0.097
mi)
SE
FawcettConcord
45
(0.153
mi)
T (Bus)
Stop
46
(0.200
mi)
S
WheelerConcord
47
West
13
4
3
BBN
Base
Statio
n
0
2
1
0
(0 mi,
90 ft
up)
BBN
Base
Station
Antenn
a
1
(0.040
mi)
Parking
Lot
2
(0.080
mi)
Fawcett
St
3
(0.110
mi)
Fawcett
St
4
(0.140
mi)
Fawcett
St
5
(0.180
mi)
Fawcett
St
6
(0.220
mi)
Fawcett
St
Fawcett
14
15
16
17
18
19
Measurements Summary
RSSIs
DL RSSIs varies from -30db for a strong signal point, down to
-64db for a weak signal point; below that, the connection fails
UL RSSIs remained more constant, often close to -75db for a wide
range of points. Is this due to automatic WiMAX UL transmit
power adjustments?
Ping loss (1008bytes)
Measured delays are relatively constant (80 100ms) until link is
about to fail
For 1008byte pings, the % of responses not within window (lost)
increases quickly as link is about to fail; otherwise 0%
Good measure of overall connection quality
20
(continued)
iperf Throughput
Use of TCP gives conservative result, but is typical of many
applications
Use of TCP results in significant variations over the 3 runs, due to
packet losses and retransmissions; need to consider both min and
max
As link gets poorer, the throughput eventually falls to zero
DL throughput is typically better than UL throughput, following
WiMAX convention
Best case DL throughput is over 10Mb/s
Best case UL throughput is approximately 1 Mb/s
21
(continued)
Range:
Best range (to point 48, 0.254mi) seen with BS sector antenna and
MS handheld large omni antenna
Range is worse, as expected, with BS sector antenna and MS
internal antenna
Worst range (to point 46, 0.2mi) seen with BS omni antenna and
MS handheld large omni antenna
However, range is better with BS omni antenna and MS internal
antenna; why?
Expected packet loss and throughput to vary from moment-tomoment, due to MS position and multi-path propagation, but not
directly verified
Range at points 1 - 7 comparable to range at points 41 47
verifies expected 360deg omni coverage
Signal gone at point 7 obstructed by building
22
23
(continued)
Consider to improve range:
Fix some mistake in BTS parameters
Modify BTS parameters to improve range by forcing reduced rate
Add diversity at BS (requires an extra ODU and an extra antenna)
Use vehicular omni antenna at MS (includes ground plane)
Add diversity at MS?
Tune up TCP and/or WiMAX parameters to improve throughput, e.g.,
reduce iperf buffer length so packets fit within MTU
Turn ON ARQ or HARQ
Utilize for UDP traffic, and accept more lost packets
Can we get to 0.6mi?
Expected to reduce range:
Use of MSs indoors
Leaves on trees starting in spring
24