Good Engineering Practice f or Wireless Netw orks

Transcription

1 Copyright 2004 Tim Pozar 1 Good Engineering Practice f or Wireless Netw orks Tim Pozar Bay Area Wireless Users Group w w w.barw n.org NANOG 32

2 Copyright 2004 Tim Pozar 2 Standards of Good Engineering Practice Was the old set of standards that the governm ent used and evolved into the current FCC s Rules and Regulations. Currently m eant as the proper and standardized way of doing things to insure uptim e and m inim ize interference to yourself and others.

3 Quick review of non-licensed radios, or Why Use ? Biggest reason: Non-licensed Significant ly less regulat ory cont rol No expensive licenses or coordinat ion. The arcane regulatory knowledge needed for licensed radios is not needed. The hardware cost is m agnitudes less than equivalent licensed t echnology. Client cards are from $35 to $150; Access Points are from $50 to $500. Even less than (non) unlicensed "carrier class" radios Exam ple: Western Multiplex/Proxim Tsunam i point-to-point radios are $12,000 per end not including the thousands of $ for antennas, coax, installation, etc. Copyright 2004 Tim Pozar 3

4 Copyright 2004 Tim Pozar 4 Why Use ? (cont.) Speeds range from 11Mb/s to 54Mb/s Another magnitude jum p of low-cost networking hardware from a couple of years ago. (2Mb/s) or even 2 to 3 magnitudes of amateur packet radio (1.2-56Kb/s). Real world speeds about 5Mb/s to 20Mb/s Plug and play Most modern operating systems support i.e. Mac OSX, *BSD, Linux, MS-Win 95 to XP. Depending on implementation, it can be as easy to set up as standard wire or fiber Ethernet and in som e cases easier.

5 Copyright 2004 Tim Pozar 5 Why Not Use ? Biggest reason: Non-licensed Non-existent coordination facilities. Other users on the bands. Folks with large organizations that lobby the FCC (ie. ARRL, SBE/NAB, Public Safet y) Regulations Affecting Deployment : /papers/part15 You have no legal priority or recourse over any other user of the spectrum...

6 FCC R&R Part 15.5 General Conditions of Operation (b) Operat ion of an intent ional, unintentional, or incident al radiator is subject t o t he condit ions t hat no har m f ul interf erence is caused and t hat int erf erence m ust be accept ed that m ay be caused by the operation of an authorized radio station, by anot her intent ional or unint ent ional radiator, by industrial, scient if ic and m edical (ISM) equipm ent, or by an incidental radiator. (c) The operat or of a radio f requency device shall be required to cease operat ing t he device upon notif icat ion by a Com m ission representative t hat the device is causing harm f ul int erf erence. Operat ion shall not resum e unt il the condition causing t he harm f ul interf erence has been correct ed. Copyright 2004 Tim Pozar 6

7 Copyright 2004 Tim Pozar 7 Why Not Use ? (cont.) The protocols have problem s scaling. Com panies like Etherlinks and Cisco have solut ions Modifications to the standards will help m inim ize interference (ie. Autom atic Power Control and frequency select ion) Outdoor deploym ent requires RF engineering knowledge to m inim ize interference issues. Other regulatory issues one m ay have to deal with for outdoor deploym ent. Radio Frequency Radiation exposure standards Local ordinances for antennas

8 Copyright 2004 Tim Pozar 8 Conclusions on issues Building an expensive network on can be risky as you have no rights or priorit ies. Coordinate with other licensed and unlicensed users. A properly designed network will survive longer. Other issues can affect you like FCC Rules changes or pressure to get the FCC to enforce. Be active in watching and changing t he FCC's Rules!

9 From Point A to Point B Proper radio engineering using Good Engineering Practices will only put enough Radio Frequency (RF) energy to get to the reception point. Energy will not be wasted by being distributed in other directions. This also m eans reuse of the spectrum by others. Two prim ary link design considerations are directionality and power (turf). Copyright 2004 Tim Pozar 9

10 Copyright 2004 Tim Pozar 10 Basic Link Design Site Survey Should be done for every deploym ent. Depending on the com plexity of the deploym ent, the engineering study requirem ents will change. It gets down to - Can you get the signal from one antenna to the other so it can be successfully used? Engineering the Link Antenna Requirem ents The signal should only go where it is needed to minimize interference to other networks and to your own. Signal Requirem ents Transmitter power and signal strength needed for the receiver

11 Copyright 2004 Tim Pozar 11 Site Survey Short distances (< 30m eters) can be determ ined by visual inspection. Longer distances will likely need to use visual with m icrowave path engineering software. Exam ples: EDX s of thousands of $ PathLoss ~ $4,000 Radio Mobile - - Free (example later) The m ore you pay the m ore accurate the uptim e and coverage predictions. The more expensive packages have more knobs to turn and are more fine-tuned to what reality is.

12 Copyright 2004 Tim Pozar 12 Site Survey (cont.) Real and potential interference needs to be evaluat ed. Look around. What ant ennas do you see nearby? Objects nearby that will cause m ulti-path? (future slides) Non-intrusive t est ing - sniff around: Use dstum bler or netstum bler to see what SSIDs you can see. Not e channel usage. A spect rum analyzer will reveal non RF. Try it out: Bring m asts and 24 dbi dishes. Not e signal and noise levels.

13 Copyright 2004 Tim Pozar 13 Engineering Software Design - Questions: Can the antennas can see each other? Do objects cut into the Fresnel Zone? How high do we need the antennas to clear? What is the free-space path loss? First cut on what size of antennas and transm itter power output (TPO) needed. What is the predicted up-tim e of the path determ ined from frequency, EIRP, distance, weather, type of terrain, etc. Final cut before field test of antennas and TPO.

14 Copyright 2004 Tim Pozar 14 Loss and Gain in a System Gain: Transm itter Output Power (TPO) in dbm or Watts. dbm = 10*log 10 (power in m illiwatts / 1 m W) 0 dbm /1 m W; 15dBm /30m W; 20dBm /100mW; 30 dbm /1 W Transm itter and receiver am plifiers Transm it and receiving ant ennas. Loss: Coax, connect ors ie. LMR-400: 0.22 db per m eter. Free-space loss db = Log 10 (distance in km) + 20 Log 10 (freq. in GHz) Obstructions and Diffraction (ie. Trees, rain, etc.) At m ospherics (ie. Snow/Rain, Refract ion (ie. Duct ing)

15 Copyright 2004 Tim Pozar 15 Sim plified Path Calculation Schem atic Free- space, et c. loss Direct ional Ant enna (gain) Om ni- direct ional Antenna (som e gain) Connect or and Transm ission Line Loss TX Out put RX Sensit ivit y

16 Copyright 2004 Tim Pozar 16 Signal Level Through the Path Signal St rengt h Transm it t ed Pow er TX TX Coax/ Connect or Loss TX Ant enna Gain Free Space, et c. Loss RX Ant enna Gain RX Coax/ Connect or Loss RX RX

17 Copyright 2004 Tim Pozar 17 Sim ple Path Calculation - /pathcalc Free Space Loss Path Fr equency GHz TPO Wat t s TPO d Bm dbm Transm ission Line Loss db TX Antenna Gain dbi Path Length m iles Free Space Loss db RX Antenna Gain dbi RX Transm ission Line Loss db RX Signal dbm RX t hreshold dbm Fade Margin db

18 Copyright 2004 Tim Pozar 18 Signal Path Loss - Through the aether... Atmospheric Attenuation Rain/Snow Trees (Spring/Sum m er vs. Fall/Wint er) Typical Solution: Just need to have lots of signal Needed fade m argin will increase with distance. Refract ion Therm al Ducting Marine Layers Typical Solut ion: Diversit y Recept ion Fresnel Zone At t enuat ion...

19 Fresnel Schem atic Copyright 2004 Tim Pozar 19

20 Copyright 2004 Tim Pozar 20 Fresnel Zones Calculation: Fr esnel Zones: Distance f r om TX to calc point m iles Path Length m iles Distance f r om RX to calc point m iles Fr equency GHz First Fr esnel Zone Radius f eet Second Fr esnel Zone Radius f eet Third Fr esnel Zone Radius f eet For th Fr esnel Zone Radius f eet

21 Copyright 2004 Tim Pozar 21

22 Copyright 2004 Tim Pozar 22 Signal Path Interference - Other Considerations Multi-path Structures and bodies of water can create a second path that will cause interference. Typical Solutions: Change the polarization of the antenna Change the beam-width of the antenna More gain, tighter beam -width, m ore expense. Add shielding Other users of the band? Typical Solutions: Change frequency. See above

23 Coverage Area Point to Multipoint Questions to ask yourself: How m any clients can see your AP? What sort of signal strength can the clients expect? Where are they located? One area vs. all around you? Propagation Prediction Methods: An elaborate form of ray-t racing. Longley-Rice Terrain-Int egrat ed Rough-Eart h Model (TIREM) Considered a better m ethod Bot h can over-predict 5-17dB. The m ore expensive software m odifies these m ethods for better accuracy. Copyright 2004 Tim Pozar 23

24 Copyright 2004 Tim Pozar 24 Topology Point to Multipoint Traditional Design One AP / Many clients Hidden Xm itter Problem s

25 Copyright 2004 Tim Pozar 25

26 Copyright 2004 Tim Pozar 26

27 Copyright 2004 Tim Pozar 27 Which Antenna for the Job? Multi Point to Multi Point Om nidirectional antennas on both ends Designed is discouraged due to spectrum polution. Point to Multi Point Om nidirectional for the "point" or "center", directional for the endpoints. Typically for last m ile. Point to Point Directional for both ends. Back bone links

28 Different Antennas for the Job Point to Multipoint Di-pole Typical PC card antenna Very low gain, radiates in almost all directions equally Horizontal Polarization Om ni-directional Usually has gain: 0 to 15 dbi Lower gain better for tall mounts. Typically Radiates equally well on the horizontal plane. Vertical Polarization Panel or Sector Used to cover a section of a flat area. Has gain: 3 to 18 dbi Horizontal beam width is typically 10 to 120 degrees. Vertical beam width is typically 10 to 45 degrees Copyright 2004 Tim Pozar 28

29 Copyright 2004 Tim Pozar 29

30 Copyright 2004 Tim Pozar 30 Panel Antennas Great for targeting an area to serve. Typically put around a tower to segm ent coverage of an om ni. Make: SuperPass Model: SPFPGH14S Gain 14 dbi Hort. Beam 60 deg. Vert. Beam 18 deg.

31 Copyright 2004 Tim Pozar 31 Directional Antennas Does not radiate in all directions The m ore focused the antenna, the higher the gain. Will have either vertical, horizontal or circular polarization

32 Copyright 2004 Tim Pozar 32 Why Use a Directional Antenna? Helps your system and be a good neighbour. Raise the effective power to the distant point you are trying to serve. Helps with the fade m argin. Reduce interference to the path. Your receiving antennas will be less sensitive to off-axis signals (ie. Other transm itters or m ulti-path). Reduce interference to others. Transm itting antennas will send less signal to off-axis receivers.

33 Copyright 2004 Tim Pozar 33 Antennas 101 continued... P2P Yagi 5 18 dbi of gain Vertical or horizontal polarit y Parabolic Dish shaped Larger than a Yagi dbi of gain Vertical or horizontal polarit y

34 Copyright 2004 Tim Pozar 34 Helical Antennas Has circular polarization Left or right handed Great for multi-path problems (see next slide) Gain is dbi

35 Copyright 2004 Tim Pozar 35 Antenna Polarity Polarity is a product of the design of the antenna. Each end of the link m ust m atch. It can be used to m inim ize m ulti-path and int erference. Typical Polarit ies: Horizontal Avoids multi-path from vertical objects like buildings Vertical Avoids multi-path from horizontal objects like the ground, bodies of water. Circular It can avoid multi-path from odd-number bounced sources. Left or right handed

36 Frequency Coordination The bands are crowded and the channels 2.4GHz only channels 1, 6 & 11 do not overlap. (next slide.) Need to coordinate internally. Need to coordinate with other band users. 2.4 is used by non-licensed and licensed users Licensed users are: Am ateurs, ENG, Public Safety, etc. Find out who m ay be using the band in your area. Ie. Mountain tops m ay have licensed users. Better would be to use a spectrum analyzer and try to identify the users. Worst case: just fire it up and find out what works. Copyright 2004 Tim Pozar 36

37 Copyright 2004 Tim Pozar 37 Frequencies of b Channels at 2.4 Ghz Channel Bot t om (Ghz) Cent er (Ghz) Top (Ghz)

38 Copyright 2004 Tim Pozar 38 Multiple Aps vs. Frequencies APs should not overlap Channels/ Fr equencies.

39 Copyright 2004 Tim Pozar 39 Access Points Exam ples... Range in costs from $50 to $1000 High end to low: Cisco 1200 Supports b, a & g Space for 2,048 MAC addresses ~ $600 to $800 Cisco 350 Next step in the Aironet line from the 340 series. ~ $500 to $1000 Space for 2,048 MAC addresses

40 Hom e-brew APs - An Exam ple Soekris PCMCIA slots and 1 m ini- PCI for radios (3 radios total) GEODE CPU - like a 486@133MHz 2 100base-T Com pact Flash slot FreeBSD 4.10 in < 32 MB IPFW, NAT, NoCat, Open1X (open1x.org), altq, etc. Copyright 2004 Tim Pozar 40

41 BAWRN Designed Hardware Layer 2/3 switching/routing between radios via and cat-5. Layer 2/3 to access point s Can run fancy routing protocols such as spanning tree or BGP. Can also be its own AP. Copyright 2004 Tim Pozar 41

42 Copyright 2004 Tim Pozar 42 Other Hardware Considerations Am plifiers Great for TX and RX issues if used properly. Best installed next to the antenna to overcom e cable loss. Can cause m ore RF pollution with excessive power. Quality varies by m ake and m odel. Regulation issues Equipm ent Certification. (Soon to be deregulat ed) Cables Bigger the better for longer runs as you will have lower loss. Battery Backup How critical is it if the power goes out? Grounding and Light ing Arrest ors Static and lighting can do som e rather dram atic dam age.

43 Copyright 2004 Tim Pozar 43 Grounding No 100% way to protect from large direct lighting hits. You can m inim ize dam age with static build-up and sm all discharges t hrough proper grounding. Static and lighting will tend to build up on the highest devices (antennas, towers, etc.). Biggest dam age com es from a strike to a single point. One solution: Static-Cat like static dissipation

44 Grounding (cont.) Discharges will go through the m ost direct route to/from ground. Don't design installations where your equipm ent is the best path. Tower/Pole Grounding Transm ission Line Grounding At antenna m ount and base of tower. Gas discharge devices (exam ple next slide) Provide a proper ground and com m on Point. Copyright 2004 Tim Pozar 44

45 Copyright 2004 Tim Pozar 45 Grounding and Lighting Arrestors Gas discharge grounding term ination. Goes in-line with t ransm ission line. Protects center conduct or st rikes. Doesn't conduct until strike.

46 Additional Weather Proofing Seal transm ission line connections. Wrap all connections with rubber tape and then a UV proof tape. Don't reuse connectors. Cheaper in the long run to purchase new. Use outdoor cable (coax and ethernet) Provide drip loops for transm ission lines before entrances to equipm ent and buildings. Use light colours on equipm ent to keep int ernal heat down. Copyright 2004 Tim Pozar 46

47 Copyright 2004 Tim Pozar 47 Future... How can we increase bandwidth while still addressing the physics (turf and spectrum )? Possible st eps: H Spectrum Managed a Autom atic power control Autom atic frequency selection Needed for EU and som ething a/b/g should have. Mesh MeshNetworks, et. al. Phased-Array Antennas Vivato, et. al. Active and passive coginitive radios.

48 Copyright 2004 Tim Pozar 48 Resources Books Building Wireless Com m unity Networks - Rob Flickenger O'Reilly Wireless Networks - Matthew Gast O'Reilly ht t p:// works.org The m eta site for the com m unity groups such as: Bay Area Wireless Users Group - Great m ailing list list. The standards body for 802.(n)(x)

49 Copyright 2004 Tim Pozar 49 Thank You and Q&A Bay Area Research Wireless Network Tim Pozar