Building a Decent RF Network

8/13/2019 Building a Decent RF Network

1/4

Building a Decent RF Networkthe ARRL's HSMM (High Speed Multi-Media)project buildsupsteam, Isensethatwe'll be building more RF networks in thenearfuture. Networksarenothingnew,but it's sur-

prising to methat many peoplewho haven't beeninvolved in the digital side of amateur radio for along time try to reinvent the wheel when it comesto buildinga network. Iwrote about thistopic near-ly ten years aao, and while much of what wroteMesh Star Double Star L~near

back then stiliapplies, perhaps it's time to take afresh look at building a decent RF network. Fig. I Commonnetwork topologies.A mesh net-A bief word about HSMM: M~~~hams make workish&h/yreliable, due toredundancy, andhasthemistake ofequating HSMMwith 802.11under the greatest capacity, but costs the most A starpart 97, ut it's muchmore than that. HSMM is a network covers wide areas inexpensively, but aunified approach to making data compatible failure can cut off service and link capacity can beacross many different platforms. While a decent quick&saturated. The linearnetwork covers longnetwork can be built with 802. equipment a distances inexpensively, but with lowered relia-good portionf the HSMM working groupIs focus bility. Costsavingsforstarandlinearcanbeusedis on more effi ientnd robust HF communica to harden the network to improve reliabilitytions as well. somewhat, and carefully chosen alternative linksBeforewe really get started, you should under- Can be cost-effective.Most are astand that good networkdesign has nothing to do combination of all three types.with the kindof networkina software you use. Anysoftware will perform poofiy on a poor network. lnaddition, understand that this topic can fill a thickbook, so what I'm covering here is necessarilybrief and limited.Network Facility lassesTo examine the two networkfunctions morecare-fully, we can break down each function as eithera userportor a backbone.Each of these types offacility has unique requirements and needs.A user access facility is shared by a number ofstations, all accessing a single User Port. Theseindividualstations may or may not beable to hearone another, andthey all have highlyvariable sig-nal qualities, parameter settings, and datathroughput requirements.A backbone, in contrast, is a point-to-point inkwith only two stationsonthechannel. Thetwo sta-tions trade data only with one another, with goodsignal quality, and link parametersare set to opti-mize link performance. Subsets of the backboneare the wormhole, a link that passes through anon-amateur service (such as the internet), andthe gateway, which provides direct access toanother type of network, usually non-amateur.With manygateways, non-amateuraccessispos-sible, so access control must be considered.Now we'll take a look at user ports and back-bones, and how to best design them.User PortsFirst, a little background. In the RF world, whentwo stations try to transmit at the same time, they*P.O.Box 114, Park Ridge, NJ 7656e-mail:

66 . Q October 2 5

will interfere with one another if they are closeenough in frequency and geography. This inter-ference, known as a collision, distorts the signalenoughto make hevoiceordataunusable. Whentwo stations do transmit at the same time, thereis a data collision, which renders both transmis-sions unusable. It should be clear that data colli-sions are bad, since they waste channeltime, andmust be avoided.In hedigitalworld, wesimply listenonthechan-nel for data. If we detect data, we wait until thechannel is quiet to transmit. This is called DataCarrier Detection, or DCD, and it's a key part ofCSMA (Carrier Sense Multiple Access), a com-mon channel-sharing scheme usedby manydatanetworks, including AX.25 Packet and 802.1 1.Some other sharing methods include CDMA(CodeDivisionMultipleAccess) andTDMA (TimeDivision Multiple Access), used by some digitalcellular telephones.When someone puts upa user port, he (or she)may be thinking like a repeater owner, trying tomaximize the port's coverage area. Installing auser portatopa tall buildingwith a 100-watt rans-mitter may offer great coverage, and possiblyboost the sysop's ego a bit, but this is poor net-working practice.The first problem encountered is the ExposedTransmitter Syndrome (ETS). This is where theuser port hears so much RFactivity that the DCDnever turns off, effectively shutting it down. Thechannel is simply so saturatedwith data (at leastfrom the port's perspective) that nothingworks atall. One option for such a site is to install an atten-uator inthe receiveside, sothat fewer stationsareheard, but this creates another serious problem,the Hidden Transmitter Syndrome (HTS)

Visit Our Web Site


2/4

Isolation:Access point nam e:

The SSID:Channel:

Basic rates (Mbits):Supported rates (Mbits):

Transmission rate (Mbits s):Preamble type:

Disable O Enable

Wireless6 v (US/FCC: 1-11, Europe/ETSI: 1-13, Ja pa n/M W 1-14)0 1 0 2 0 5 . 5 0 1 1 (R ates for m anagem ent pa cke ts)0 1 0 2 0 5 . 5 0 1 1 ( Rate for data packets)Automatic v

|Long vjLong = Universal Compatibility (e g , ORINOCO cards)Short Highest Performance (5.2 to 5.5 Mbps)Both = Not fully supported by Intersil

AP Visibility When Invisibility is selected , the AP is protected against APdiscovery by NetStumbler and ApSnrff ndall wireless clientsmust explicitly use and know the SSID .visibilitystatus: Visible O Invisible

WEP configuration For 64 bit keys you must enter 10 hex digits into the key box.For 128 bit keys you must enter 26 hex digits into the key box. Ifyou a key box blank then this means a key of llzeros.WEP enabled:

WEP key lengths: 64 bit v (This length applies to all keys)JSCSHHWEP key 1: |i28bit IWE P k e y :WEP key 3:WEP key 4: ( |

WEP key to use: Ke y v (This is the key to use for transmitted data)| Save | ("Cancel

Fig. 2- The main configuration screen for my 802.11b Access Point (AP). Notethat WEP encryption is not intended to limit access; instead it serves to preventeavesdropping, not useful for amateur radio applications. A simple way to prevent casual access is to set the AP Visibility to Invisible, but then you need toworry about station identification (normally done by setting the SSID to your callsign) if you re operating under Part 97.

Hidden Transmitter Syndrome is theabsolute worst thing that can occur onany shared data channel. In order for aCSMA channel to work properly, everyone on the channe l must be able to heareveryone else. If this is not the case, astation m ight transmit while another station (not he ard by the first station) is alsotransmitt ing, meaning the second station is a hidden transmitter to the firststation. A CSMA channel that has hidden transmitters on i t wi l l always perform poo rly, since a very large percentage of channel t ime is wasted oncollisions. It should be painfully obviousthat ETS and HTS are very bad thingsand must be avoided, even at considerable cost.

What can you do? Instal l ing a ful l-duplex data repeater wil l solve HTSproblems, but ETS can sti l l be a problem where more than few hams can hit

the repeaternot to ment ion theexpen se of a repeater. There's also theissue of using up two chann els, plus theimpractical i ty of a repeater for real lywideband modes such as 8 0 2 . 1 1 . Thesolution is real ly much less expensive,though , and de-cept ively s impleCellular User Ports.The Cellular SolutionInstead of creating a small number ofwide-coverage userports,the better wayis to create a large number of relativelysmall cel ls, each supporting perhaps 10to 20 users (of which only two or threemay be a ctive at a given time). Cells areplaced near the users and designed sothat adjacent cel ls cannot hear oneanother. Data is then transported on abackbone channel, where there are nousers. Cellular phone companies usethis scheme quite successfully.

Mk6 0 shopping days'til Christmas.^Jilv i s i t w w w . k j i e l e c t r o n i c s . c o m or the KJ I S tore 9 7 3 - 3 6 4 - 1 9 3 0

LicensedBefore 1981?Q C W A i n v it e s y o u t o j o i n w i t ht h o s e d i s t i n g u i s h e d a m a t e u r sl i censed 25 o r more yea rs ago .R e q u e s t a n a p p l i c a t i o n f r o m :

QCWA, Inc . , Dept . CPO Box 3247

Fra mingh a m, M A 01705 -3247w w w . q c w a . o r g

W4RTIecifOtvcs>

P rovenP er forma nceYAESU F I L T R S , coMFT-817 FT-857 FT-897 IC-703 IC-718

C W : $ 1 1 5 S S B $ 1 3 0I W 4 R T c a n i n s t a ll fo r y o u , a s kA n t e n n a B O S S & A n t e n n a B O S S I IThe BEST Motor ized Antenna Control lers

Antenna BOSS II interfaces directlywith ICOM IC-706 series of radiosfor automatic tuning of motorizedantennas from Hi-Q Antennas, HighSierra Antennas, & Tarheel Antennas& others that draw < 750 mA whilerunning. Tuning is activated bypressing the IC-706 Tune button.The BOSSH can be interfaced with many other radios from ICOM,Yaesu. and Kenwood. The BOSS II provides coverage capability from6-160 m, even for Yaesu radios. Interfacing with non-ICOM radiosrequires the use of an appropriate One-Touch Tune module, BOSS IIfeatures dynamic motor braking, motor stall sensing & braking, userselectable stall current sensing, no special sensors in antenna,programming or computer are required. The Antenna BOSS makesessentially any motorized antenna "look like" the Yaesu ATAS-100/120to the FT-100/D. FT-847, FT-857. and FT-897 radios. Use the radio'sown TUNE button for automatic tuning the way it was intendedArguably the Most Impor tan t Accessory fo r YourFT-817 , TS-50 , IC-706 , and MOR E

The bhi, Ltd. DSP M odule providesGREAT noise cancellation technology.Has 4 levels of noise cancellation, singlebutton operation, & low-distortion of audiosignals. Dynamically-adaptive neural-S i 2 9 DIY network technology provides trulyW4RTinssa vaabe astounding improvement.O R D E R O N - L I N E a t WWW W4RT COIT1or ask your favorite Dealer

Phone Orders Only: 866.S3S.4442 -GigaParts Shtppir lAddrtwwi

w w w . c q - a m a t e u r - r a d i o . c o m O c t o b e r 2 0 0 5 C Q 6 7
http://www.kjielectronics.com/http://www.qcwa.org/http://www.cq-amateur-radio.com/http://www.cq-amateur-radio.com/http://www.qcwa.org/http://www.kjielectronics.com/


3/4

One cardinal rule of networks is thatyou never carry traffic on a user accessfrequency (something 802.11 l ikes todo, in i ts native version). When you doso , the channel capacity is immediately cut in half: half the time is use d for theuser to get the data to the user port, andthe other half is used by the port in sending the data somewhere else. Sure, i tcosts a l i ttle mo re, but are you buildinga high-perform ance network here or justfool ing around?BackbonesThe s ingle most important thing youcan do for backbone l inks is to insistthat each one is a dedicated point-to-point l ink (DPPL). That means only tworadios on any given frequency (hopeful ly not on a band where user channels can exist), wi th high-gain di rectional antenna s pointing at one a nother.As soon as you introd uce a third stationto a backbone channel , throughputdrops l ike a rock.

I real ly can't emphasize this enough.The reason behind this throughput lossis in the nature of the da ta transfer protocol. Packets of data (not only AX.25uses packetized data) are sent, and abrief packet is returned to acknowledgethe data packet, al lowing the next groupof packets to be sent. Because of theway the timing is set up, a third stationcauses additional waiting time between

data and acknowledgem ent. That wai ting t ime is wasted performance. Foruser channels i t isn't as cri t ical, butlarge-volume users ( large volume isabout 10% of the chann el capa ci ty)should be encouraged to set up theirown DPPL into the network backbones.It is far better to set up a lower speedDPP Lthan a higher speed shared backbone. When loading is very heavy

which is whe n capacity becomes importantthe DPPLs wi l l outperform theshared zoo chann el . Aga in, i t doescost a little more to do things right, butthe results are worth it.Given the realit ies of the typical am ateur radio budget, i t can be unreasonable to expect a large initial outlay ofresources. However, the goal shouldalways be to migrate towards the ideal,even if i t is presently unattainable. Ifyour present network already suffers ashared backbone channel , careful lymonitor the data activity between eachsite. Often two s i tes pass more databetween them than b etween any othertwo sites. Upgrade that l ink to a DPPLwhen the resources become available.There are tools available to monitorwireless network performance, andalthough I haven't found on e that's free,many offer free trial versions. Most802.11 access points have some buil t-in functionali ty, at least for signalstrength,while tools are general ly av ailable for other RF gear. As an alterna

t ive,the w ired side of a l ink can bemonitored,or test files can be sent across al ink to gauge performance.That brings up another point:Intel l igently derived backbone speeds.If a multi-port network hub servicesthree user ports and concentrates thatdata load onto one ma jor backbon e, thebackbone must be able to handle theload without s lowing down anyone. Thisseems ob vious, but some people don' tf igure i t into their plans. The data loading on each backbone l ink should bemeasured orest imated ,and l inkspee dsadjusted accordingly.

Once you have your network's performance opt imized (given the resources avai lable), you should startthinking about redundancy. Draw thenetwork on a piece of paper and lookfor links that, if theyfail,cause a part ofthe network to become unreachable.Then look for a place to instal l anotherbackbone l ink, even a lower speed one,to provide a backup. It is reasonable toset up bunch of backup l inks on ashare d cha nne l, as long as i t is real ly abackup and not a primary l ink. Thissaves money while preserving performance, since only two stations shouldbe affected by a broken l ink. Only thosetwo stations wil l be using the sharedchannel, with the others just l istening,effectively a DPPL.It should go w ithout say ing that tenuous or weak RF paths are unacceptable.Add power, or better yet, antenna

gain to the link until it func tions wel l ;perform the basic l ink calculat ions anddetermine the actual fade margin.Antenna gain is usually cheaper than RFpower, narrows the be am w idth to al lowmore reuse of channels, and helps yourreceiver, too. Breaking up a long pathinto two segments can also make a bigdifference, or you can mo ve down infrequency for better propagation.Avoid confusing radio networks withwire-based networks. There are farmore wire-based network designers outthere, but thei r knowledge and experi

ence often doesn' t apply to the RFworld. Be wary of experts who don' thave any experience with radios.

Securi tyOne commonly encountered issuewhen deploying a network using commercial ly available gear (such as802.11) is the need to prevent non -am ateurs from passing da ta over l inks operating under Part 97. Note that encrypt ion such as WEP does nothing toprevent access to your network. It onlyprevents viewing of the data being car-

u. i l . l * filtering: DMAC address 1:MAC addicts 2:MACadd iess 3 :MAC addte s 4:

MAC addte s S:MAC address 6:MAC address 7:MAC address8:MAC address9 :

MA; address 1 :

Fig. 3- On this screen you can explicitly set which network interface cards will berecognized by the Access Point. If the MAC address isn t on this screen, you willbe unable to access the network. While hams w ill have to register w ith you for access,knowing who they are may help drum up support for network improvemen ts.

6 8 C Q . O c t o b e r 2005 Vis i t Our W e b Site


4/4

Why D ig ipea t ing Won' t WorkBefore networks there were digipeaters. By hopping down a chain of other hams' TNCs(terminal node controllers, for the newbies among us), you could go farther than yourradio's range. It sounds okay, until reality is considered. The problem is that digis aredumb; they just take what they hear and repeat it. The digi doesn't acknowledge yourpackets; that is done only by the destination station. Thus, each digi just passes yourpacket down the line, and passes the distant station's acknowledgement packet back upthe line toyou.Inthe real world, links between digis are not 100% efficient; not all packets make it through without getting lost or corrupted. Note that an 802.11 network extender is nothing but a digipeater.Ifyouuse only onedigi,you have to get across four links before you can send the nextpacket-two to getthere,and two for the acknowledgement to get back. If we assume thatnine out of ten packets m ake it on each hop (actually qu ite good), then that one -digi pathhas an efficiency of only (0.9 x 0.9 x 0.9 x 0.9) = 0.65, or 65%. With two digis, it dropsto 53%.With only half of your packets making it through on any given try, it is very difficult, ifnot impossible, to pass any reasonable amount of data. If other users are on frequency,the hops get worse than90%,easily down to the range of50%perhop.Do the math andyou'll see why for longer-distance communications digipeating just won't work.

ried (see fig. 2). Instead what you needis access control.Virtually all Access Points allow foraccess control based on the MACaddress (a 12-cha racter unique serialnumber assigne d to every networkinterface card) of the wireless NetworkInterface Cards (NICs) trying to accessthe network. For userports,the users willhave to register their wireless LAN cardswith you so they can be put into theaccess table. Backbone l inks should beset to recognize only one another, ofcourse (see fig. 3).The dow nside is thatany hams in the area wil l have to contact the n etwork sysop to get into the network. This has advantages, however:Once you know who is using the network, it is easier to get them to supportthe network w ith time , talent, or treasu re.You can also plan the network better,considering loading and HTS (HiddenTransmitter Syndrome). It definitelykeeps casual intruders out. AlthoughMAC addresses can be spoofed, you'dneed to know what's on the list first. Itisn't perfect security, but it should bemore than adequate for our purposes.Conc lus ionThere are many resources on the internet that will help you design a great network. M uch of it is com mo n sen se, suchas performing l ink RF budg et calculat ions, and some of i t is a bit more subt le. There wi l l a lways be compromisesbased on l imited resources, but thereare some things on which you can compromise, and some things on which youshould not.Don' t compromise, for example, onnetwork topology. Running backbonedata on user channels wil l bring performance to i ts knees . Don't wait for high-tech solutions when low-tech solutionswil l work soone r. Network design is a l ittle bit logic, a little bitskill,a little bitpolitics,and a l i tt le bit luck. A w ell-designednetwork wil l always outperform a poorly designed one, often at lower cost.There are people out there who havespent their entire careers doing networkdesign, so don't expect to become anexpert overnight, but a little effort willpay off handsomely. There aren't manybooks out there that deal with radio-based networking design, but the engineering l ibrary at a local col lege shouldhave som ething usefu l, even if i t's onlya single chapter.Well, that's al l the time we have thismonth. I hope you found this columninteresting and useful, especial ly nowas we find more and more uses for anon-internet network solut ion. As our

RF-network performance improves tomeet and exceedT and cable m odemlevels, we'l l be seeing more and morehamsespecial ly those wi th dial -upconnect ionspreferr ing to get thei rham radio information over radio in

stead of wires. Packet radio is almostdead, not because of slow data butbecause there was l imited useful content. HSMM wi l l have the content,something I am looking forward to.Unt il next t ime . . . 73, Don, N2IRZ

The Most BeautifulKey Ever |Builtin out

more at :S T R A I G H T B R A S S . C O M

E N H A N C E D X R E C E P T I O NAS AP -2 Antenna S wi tch & P reamp Use WithHFTransceivers1.8-30MHzFreq.Coverage Select of 4 Receive Ant. or XmitAntenna Select-20, 0, +20 dB Gain5dBNF,+30 dBm 3 r dOIPUses Rig Key Line for T/RGreat forN S E WBeverages, ShieldedLoops, WhipsASAP-1 Receive Only/SWL model

ASAP-2 @ $149.50ASAP-1 @ $119.50PS-1 12V AC Adapter @$13.95Check, M.O. Buy-On-Line (PayPal)Post Paid - Priority Mail (U.S.)850)936-7100 www.j-tecradio.com J T E C

E N H A N C E D X R E C E P T I O NAS AP -2 Antenna S wi tch & P reamp Use WithHFTransceivers1.8-30MHzFreq.Coverage Select of 4 Receive Ant. or XmitAntenna Select-20, 0, +20 dB Gain5dBNF,+30 dBm 3 r dOIPUses Rig Key Line for T/RGreat forN S E WBeverages, ShieldedLoops, WhipsASAP-1 Receive Only/SWL model

ASAP-2 @ $149.50ASAP-1 @ $119.50PS-1 12V AC Adapter @$13.95Check, M.O. Buy-On-Line (PayPal)Post Paid - Priority Mail (U.S.)850)936-7100 www.j-tecradio.com J T E C

www.cq-amateur-radio.com October 2005 CQ 69
http://straightbrass.com/http://www.j-tecradio.com/http://www.j-tecradio.com/http://www.cq-amateur-radio.com/http://www.cq-amateur-radio.com/http://www.j-tecradio.com/http://www.j-tecradio.com/http://straightbrass.com/

Documents

Building a Decent RF Network