wi-fi and lte 2.3 ghz co-location field testing pdf, 4.5 mb

Copyright © 2014 7signal Solutions, Inc.

Wi-Fi and LTE 2.3 GHz co-location field testing

25.9.2014

Final report

Version 1.0

Tuomas Aaltonen

Veli-Pekka Ketonen

1


Table of content

 General Test setup Log of changes Mechanisms of possible LTE impact Ethernet performance trending Spectrum analysis SLA analysis of tests during LTE test days Day internal (24h) analysis during LTE test days

–  Day 10.9–  Day 11.9

 Day to day comparison, LTE on/off–  Retransmissions between clients and APs–  Retransmissions between sensors and APs –  Data rates between clients and APs–  Retransmissions between sensors and APs

 Challenges Observations Conclusions

2


General This is the final report of 2.3 GHz LTE – Wi-Fi co-location tests at Victoria station Project was a collaborative effort between Ofcom, BSkyB and 7signal Purpose of the project was to determine how much 2.3 GHz LTE transmitter impacts to nearby Wi-Fi operation

 Project duration was three weeks –  1st week with LTE, one-two sensor, 2.4 GHz only

–  2nd week with LTE, four sensors, 2.4 GHz only

–  3rd week without LTE, four sensors, 2.4 GHz and 5GHz

 Data collection started with one Eye sensor on 2.9. Three more eye sensors were added later to the profile; One sensor on 3.9 and two sensors on 5.9.

 5 GHz radios were tested with sensor 8 starting from 5.9 (AP03 and AP06). Rest of the 5G radios were added to the test profile on 15.9 This report includes data from all the weeks e.g. LTE testing results, 2.4 GHz and 5 GHz Wi-Fi performance comparison and analysis and conclusions of the results Conclusions are presented at the end of the report

3


Test setup

4


Test setup

 A LTE test transmitter and Ruckus WLAN access point were co-located 7signal Sapphire Wi-Fi service assurance and performance optimization product was used to analyze performance variations during the times when LTE transmitter was on and off Two types of 7signal Wi-Fi sensors were used

–  Three 1000 series Eyes with beam steering and one smaller Micro Eye sensor were used

 7signal measurement engine (Carat) was cloud based. Sensors were connected to cloud server through Ethernet. All collected data was stored to cloud

 During active tests 7signal sensors measure against a test end point. This test end point was also in cloud server During the test period, Victoria station Wi-Fi access point were connected to internet through 16/1Mbit/s ADSL connection. Normally a faster multi-ADSL line has been available, but due to failure of aggregating device, only one ADSL line was available for all Wi-Fi traffic in the station

5

Copyright © 2014 7signal Solutions, Inc.Copyright © 2014 7signal Solutions, Inc.

Elements of 7signal Sapphire


Diagram of test setup elements

 Eyes performance active and passive tests Active tests operate over the LAN and WLAN separately Passive tests capture radio packets and calculate metrics from them


AP and Eye locations

8


LTE test transmitter

9

 LTE test signal was 20 MHz wide QPSK & 16QAM modulated synthesized TDD LTE traffic pattern Test signal center frequency was 2380 MHz Radiated output power levels were varied during the test

–  30, 36, 49 dBm mean EIRP values were tested


Eye 3 and LTE BTS

10


Log of changes

11


Test log from week 1

12



13



14


Mechanism of possible impact

15


Key aspects

 The key challenge with Wi-Fi and mobile base station co-location is lack of RF filtering in receivers. Wi-Fi access points and end user radios commonly do not have any or only very limited RF filtering preventing out-of-Wi-Fi-band signals entering Wi-Fi receiver RF parts.  Effective radiated signal levels from mobile network base stations may be very high (1000’s RF watts) compared signal levels used in Wi-Fi (typically below 0.1 RF watts) Unfiltered RF entering Wi-Fi receivers may degrade Wi-Fi performance in different ways Currently studied LTE band in UK is 2350-2390MHz. Likely deployment scenario is one 20 MHz carrier/base station in one site. This is also the test scenario in this study.

16


Main scenarios

 Scenario 1:–  WLAN receiver is de-senzitized due to high level RF entering the receiver–  Receiver gain control circuitry may be operating incorrectly under high pout of band

RF signal influence

 Scenario 2:–  High level RF signals create intermodulation products within WLAN receiver which

lands on the Wi-Fi frequency band. Typically the 3rd order harmonics are most likely causes

–  The main third order intermodulation products are generated between two signal as follows

•  Carrier 1: Frequency 1•  Carrier 2: Frequency 2•  Intermodulation products:

–  Frequency 2 + (Frequency 1-Frerquency 2)–  Frequency 1 – (Frequency 2-Frequency 1)

–  Intermodulation product may be caused of signal at different bands or from two signals in nearby band

 Scenario 3:–  Wide band noise leakage from nearby RF carrier to Wi-Fi band

17


Wi-Fi and main nearby mobile network frequency bands

18


3rd order interference generation scenarios: Case 1

19



20



21


Test scenario in this study:LTE 2.3 GHz and Wi-Fi 2.4 GHz

22


Ethernet performance

Sensor Ethernet port towards the cloud based test end point

23


Eye Ethernet tests, downlink TP(Against cloud test end point)

24

 ADSL line connecting all access points to internet is a remarkable bottleneck


Eye Ethernet tests, downlink TP(Against cloud test end point)

25

 ADSL line connecting all access points to internet is a remarkable bottleneck Wired network throughput towards the internet dropped regularly to 100-200 kbit/s range during busy traffic


Results and conclusions

 Used 16/1 ADSL is not sufficient enough to provide well performing WLAN to Victoria station end users

 This bottleneck also impacts test results in this project. Results from active end-to-end tests are mostly not limited by Wi-Fi radio connection but by the internet connection speed

 This limitation caused project to focus more on radio metrics than end-to-end performance

26


2.4 GHz spectrum analysis

Sensors performed spectrum analysis

27


2.4GHz spectrum, the first week

28

 No visible impact from LTE transmitter in spectrum results AP channel changes are visible in the spectrum over time


2.4GHz spectrum, the second week

29

 No visible impact from LTE transmitter in spectrum results Some other strong non- Wi-Fi interference is visible near sensor 1 (“Standard Eye 1”)


2.4GHz spectrum, the third week

30

  No visible impact from LTE transmitter in spectrum results  Some other non- Wi-Fi interference is visible near sensor 1 (“Standard Eye 1”)  This overlaps significantly with Wi-Fi channels and negatively impacts to Wi-Fi operation


Spectrum results, 11.9 LTE on/off

31


Spectrum results and conclusions

 No clear impact is visible in the 2.4 GHz band spectrum during the LTE is transmissions. This is the case even when the Tx power is set to maximum (49 dBm) Significant non Wi-Fi interference is present near to Eye 1. Interferer is continuous, non pulsed signal type signal. This causes continuous spikes is spectrum which overlap with Wi-Fi channels and degrade Wi-Fi performance Spectrum results shows that some APs are changing the channel often and they are using all available channels. Some of the APs change their channel more often than others. This seems to be the case especially near to Eye 1 where the interference is high. Ruckus AP Channel Fly tries to avoid the interference, but since its across the band, that’s not helping

32

 Spectrum results shows also that there is also other interference. For example Bluetooth signal can be seen in spectrum as narrow 1 or 2 MHz spaced spikes across the whole 2.4 GHz band


SLA table analysisLTE on / LTE OFF, 9-13.9

Network performance is compared to pre-defined thresholds

33


SLA thresholds used in this report

34


9.9 2.4 GHz SLA table, averaged data (All APs)

35

 General performance level is modest


10.9 2.4GHz SLA table, averaged data (All APs)

36

 => No clear impact from LTE


11.9 2.4G SLA table, averaged data (All APs)

37

 => No clear impact from LTE


12.9 2.4G SLA table, averaged data (All APs)

38

 General performance level is modest


Single AP SLAs, LTE on times only, 9-13.9

39

• • • 

• • • 

 => No clear difference between LTE ON days and LTE OFF days


Single AP SLAs, LTE on times only, 9-13.9

40

 => No clear difference between LTE ON days and LTE OFF days


24h LTE on/off testing results, 10.9

2.4GHz band AP graphs, 24h history with 10min average

41


# of clients on AP BSSID

42


AP signal strenght at Eye

43


Beacon availability

44


Radio attach success

45


Latency

46


TCP download throughput

47


TCP upload throughput

48


Eye retry rates in TCP test

49


AP retry rates in TCP test

50


AP retries

51


Client retries

52


WLAN utilization

53


Probe density from AP

54


24h LTE on/off testing results, 11.9

55


# of clients

56


AP signal strenght at Eye

57


Beacon availability

58


Radio attach success

59


Latency

60


TCP download throughput

61


TCP upload throughput

62


Eye retry rates in TCP test

63


AP retry rates in TCP test

64


AP retries

65


Client retries

66


WLAN utilization

67


Probe density from AP

68


LTE testing results, comparing day to day

69


Retransmission in traffic between clients and APs

70


AP retries towards clients,7.00-9.00, LTE at 36dBm/OFF

71


Client retries towards APs,7.00-9.00, LTE at 36dBm/OFF

72



73



74



75



76


Retransmission between Eye and APs during active tests

77


AP retries towards Eye,7.00-9.00, LTE at 36dBm/OFF

78


Eye retries towards APs,7.00-9.00, LTE at 36dBm/OFF

79



80



81



82



83


Data rates/MCSs between clients and APs

84


AP04, Downlink data rates,AP traffic towards clients, LTE +36dBm

85


AP04, Uplink data rates,Client traffic towards APs, LTE +36dBm

86


AP04, Downlink data rates,AP traffic towards clients, LTE +49dBm

87


AP04, Uplink data rates,Client traffic towards APs, LTE +49dBm

88


Data rates/MCSs between Eye and APs during active tests

89


Downlink data rates, active test, AP traffic towards Eye, LTE +36dBm

90


Uplink data rates, active testEye traffic towards APs, LTE +36dBm

91


Downlink data rates, active test, AP traffic towards Eye, LTE +49dBm

92


Uplink data rates, active testEye traffic towards APs, LTE +49dBm

93


Challenges

94

 Certain compromises were made due to very rapid deployment schedule Slow ADSL from Victoria station to internet limited maximum throughputs to very low values (100’s kbit/s during load). Instead of looking at the end-to-end throughputs, more focus was put to observing radio behavior during the active and passive tests when LTE transmitter was turned on and off. This was after all the goal of the project  High packet loss in internet connection caused sensors drop off from test profile at times. This caused some delays and at times gaps to data collection, but did not impact accuracy of the gathered data when tests was made No local test end point was available, only cloud. Planned location for Victoria station onsite active test end point (Sonar) did not offer enough space for a planned laptop device Not all planned sensors were online due to placement and available PoE power limitations. Also, some sensors at times rebooted due to voltage drops on power feed Sensor placement was not optimal. Sensors had to get PoE power from access points and they were co-located with access points. Sensors were also positioned less than optimally due to temporary nature of the setup Access point band steering feature delayed access to 2.4 GHz radio measurements. Sensor has capability for 5 GHz band and band steering delayed responses at 2.4 GHz. This slowed down test cycle speed but did not impact accuracy of the relevant test data when tests were performed. Only beacon availability, attach success rate and attach time metrics were negatively impacted Despite the challenges, a lot of valid data was collected and valuable conclusions have been made


Observations

 General Wi-Fi performance level is modest, the main reason being slow ADSL connection from the Victoria station to internet at the time of testing due to ADSL equipment failure before the project. The wired connection alone drops throughputs below 1 Mbit/s in downlink and to 100-200kbit/s in uplink. Wi-Fi is capable of 10’s Mbit/s to even some 100 Mbit/s throughputs in good conditions  RF operation is also clearly less than optimal, including very high retry rates and high utilization at 2.4 GHz band and some strong sources of non-Wi-Fi interference Spectrum analysis measurements did not show noticeable impact from LTE traffic Day internal active and passive testing did not show any meaningful correlation with degraded performance and LTE transmitter presence

95


Conclusion

 Tested scenario does not show 2.3 GHz LTE transmitter impacting negatively on Wi-Fi performance

96

wi-fi and lte 2.3 ghz co-location field testing pdf, 4.5 mb

Documents