September 1st, 1997 FLYING WITH BREEZECOMS In late August of 1996, the Marketting Manager of BREEZECOM, an Israli company with American outlets contacted our NSF Wireless for Education Project offering the loan of their then new BreezeNET radios for evaluation. We accepted, however decided to purchase several so we could do some extended tests of this interesting - claimed performance, price, size, and features - spread spectrum Part 15 radio. Over the past year we have used it in several configurations, and one upgrade, and are still evaluating it. Or rather 'them', for the minimum configuration of a wireless link requires one Access Point (AP-10) and one Wireless Bridge (WB-10). The BreezeNET Wireless Bridge PRO (WB-10 Pro) and Access Point PRO (AP-10 Pro) comprise a wireless LAN. Both the AP and WP are equipped with 10BaseT ethernet connectors on each radio, 5 watt, 1200 mAmp input transformer, two antenna connectors. The radios operate at 50 mW of output (one twentiethof a watt), in the 2.4Ghz FCC Part 15 frequency range. They employ frequency hopping patterns. They are very small units, being no more than 5 inches long. A number of status lights are on the front of the units, indicating power on, Wireless Link, its relative strength (strong, medium, weak), ethernet connectivity, and data flow. The radios also, although only LAN bridges, with no protocols built in, except conforming to the IEEE 802.11 LAN specifications, nevertheless support SNMP protocols in each radio. So that a remote radio can be monitored or reconfigured by a TCP/IP signal if they are assigned an IP number in the network. They can be configured by attaching, through their own special cable, a PC with a terminal program set to 9,600 baud and pressing a reset button. All the setup parameters are menu driven. Up to 255 WBs can be configured to access one AP. Their rated range is up to 7 miles, line of sight, with a 24dB gain antenna (UNI-24), which has a very narrow beam - 7.5 degrees. Their claimed performance ranges from 3Mbps half-duplex at 1.5 miles, through 2Mbps at 3 miles, and 1Mbps at down to 1Mbps at the 7 mile distance. The list price of an Access Point is $1,295 and a Wireless Bridge, $1,995. The UNI-24 antenna, with 50 feet of RG-8 cable and connectors is $395. For the claimed price, size, and performance, BreezeCOMs promise much. POINT TO POINT FOR A WHILE The first configuration we used it in, after a few cursory tests inside, using only small vertical antennas as an internal LAN substitute - which could be expected to be strong, was between our office and my home-office, using roof corner reflector antennas, with cable lengths up to 30 feet. The buildings were seperated by 1/3d of a mile. In the winter of 1996, with few leaves on the trees, the pair of units performed at about 1Mbps consistently. This was in part because one of the rooftop units were aimed at a planned Access Point site, and not directly back at the corner reflector. It was a satisfactory link for 4 months, until very heavy spring rains started coming to Colorado. The first indication of weakness was after rains after the first strong spring foliage growth. The combination of the leaves blocking the signal, at 2.4Ghz frequencies, and wetness, caused link failures from 3 hours to 24 or more. The drying out would permit the link to be restored. Then in late May, when the leaves had fully flowered, and there was no visible line of sight that 1/3d of a mile, the link failed permanently. It would not have come back until winter, had the radios been left in place. The test of this configuration demonstrated that low power - 50 mW at 2.4Ghz frequencies are very sensitive to line of sight and wet leave issues. Thus the change of the seasons have to be taken into account. A RELIABLE RELAY LINK The next test, still going on, is a much better, and fairer, test of the BreezeCOMs. Since schools were out for the summer, and we could not find school systems where we could fully exercise the radios until fall, we decided to link up our primary Linux server, and the NSF Wireless Field Test NT web server both on a LAN in office, to an upstream Internet provider, Colorado Supernet using the top of office buildings for the antenna sites and the continuous traffic to and from the Linux system and NSF system to stress the reliability. At first, since we had already done some hand-held link tests in the 902-928Mhz range from the 12 story top of the Alamo Plaza office building in downtown Colorado Springs, to the 4 story roof of our office building in Old Colorado City - 2.2 miles away, we tried similar tests with a hand-held BreezeCOM units powered by 5 volt wet gel cell batteries. We got a lesson in the sharp difference between radios operating in the 915Mhz and the 2.4Ghz range. The 902Mhz radios were able to acquire signal with less than a perfect line of sight around the edge of the large bank building that sat between the Alamo Plaza building and the Templeton Building. But the 2.4Ghz radios were NOT able to acquire between the same two points. We would have to install a relay point between the two end points, if the BreezeCOM radios were to work. THE THREE RADIO SOLUTION After examining several possible relay sites, I decided the best one with the cleanest line of sight and equal distance between it and the two end points, was the premises of AA Self Storage business up the hill of south 21st street. It was obviously a popular site, for a giant cell phone tower (perhaps 200 feet high) of US West had leased space there too. A deal was struck with the owner of the business, who, as part of it asked if we could feed him an Internet connection, since he was so dissatisfied with dial up AOL. (this was during the worst of the AOL overload problems). He had a 28.8 modem attached to his MacIntosh and only got marginal performance with the World Wide Web, lots of busy signals, and he had to tie up his office phone when he used AOL to get his e-mail. I was intrigued to see how well his singular computer would perform at the 'relay' point, which was line of sight 1.1 miles from the Alamo Plaza antenna site. The configuration would be as follows: 1. At the Templeton Building (the office building in which Old Colorado City Communications had its offices, servers, and LANs) we would put up, on the roof, with 30 feet of LMR 400 quality RF cable running up to it, a 24Db gain antenna aimed at the relay point. And attach a WB-10 PRO to the cable in the office. 2. At the relay point, we put up a cylindrical white omni antenna tuned to 2.4Ghz on a tripod on the metal one story roof. An AP-10 PRO would be put in a room just below the roof, not more than 10 feet from the antenna to minimize the RF cable length, and connect it to wall power. Though the Access Point did not need to be connected to anything, when configured to act as a relay radio, we knew we could take an Ethernet line off the AP and feed it directly into the MacIntosh in an office 20 feet away. So long as the Mac had its own IP address from our 192.160.122 Class C set, it would act just like any other workstation on the virtual LAN. 3. At the downtown Colorado Supernet POP, their Cisco Routers are in an office on the 12th floor of the Alamo Plaza Corporate Center at 2 South Tejon Street. There were various conduits running to the roof mechanical rooms on the '13th' floor. Above that level, which opened out to the main roof, is a super roof, where Building Management permitted antennas from various enterprises. We calculated that it would take an ethernet run, through conduits of about 75 feet from the Cisco router to a '13th floor' mechanical room where we could put a BreezeCOM WB-10 PRO using wall power, and inside just below the superroof where we would put another 24dB gain directional antenna aimed at the relay point. A LMR 400 cable of 40 feet would have to be used to couple the inside WB-10 radio and the antenna. Such cable operates with about 4 dB of loss per hundred feet of cable. It took many months of negotiation with the building management - which was getting more and more nervous about the proliferation of antennas on its roof - before we were granted approval to install the antenna, set back far enough it could not be seen from the street (aesthetics) and use a tiny corner of the building mechanical room for the radio. And use, and paint, conduits everywhere. In early July of 1997, the installation was completed, with the eager cooperation of Colorado Supernet, a Denver based Internet service that is under contract to the State of Colorado to provide heavily discounted service to state agencies, including all public schools and colleges, which was as eager to learn how the link worked as we were. So Doug Kerry of CSN came down and helped configure the CSN routers. Our BreezeCOM radio then was, in effect, directly connected by ethernet into their Cisco routers in Colorado Springs. No US West phone lines between CSN's POP and our computers 2 miles away. Just wireless, RF cables, and ethernet segments. We could not risk dropping our Old Colorado City Communications 56Kbs frame relay line through US West to Supernet until we had confidence that the radio link would stay up through bad weather as well as good. So CSN gave us a seperate set of Class C numbers to use temporarily, while they provided an Internet feed at ethernet speed to them from their routers. CSNs connection to Denver was DS3. (45Mbps) In order to test the radios without risking our office LAN, we did a rather unusual thing. We connected a Wi-Lan Hopper PLus, 1Mbps 915Mhz radio directly into the WB-10 by means of a 1 foot ethernet wire. Then, on my home office premises, we put another Wi-Lan radio, using a 915Mhz Yagi antenna to reach the office with a corresponding Yagi. This radio was then connected to a hub on my home office LAN, and principally fed into an OS2 Warp system, with a Web Server on it, which has been used in all our preliminary radio tests, and is accessible through the wireless.oldcolo.com Since the Wi-Lan radios operate at half a watt, are in lower frequencies less subject to wet-leaf interference, and the Yagi antennas were directly pointed at each other, these radios worked well to my premises through the leaves. While the BreezeCOMs, with much lower power, at higher frequencies, and with only one antenna directional aimed at the side of the one on our office building, could not stay connected. World of difference between 50Mwatts at 2.4Ghz and 500Mwatts at 915Mhz where obstacles are concerned. In effect we connected the BreezeCOM wireless network to a Wi-Lan wireless network. All operating as a single LAN, with IP protocol. Thus, for over 2 months I was able to access the Internet via this seperate network, and judge its speed and reliability. The whole network worked well and fast. Colorado Supernet measured througput end to end - through the 'relay' point which would cut the speed by half to the far end of the link, at from 360Kbs to 400Kbps consistently. Over 7 times faster than a 56Kbps link. And though John Bickett, owner of AA Self Storage, did not really realize what a good deal he had in operating his Mac over a wireless link to the Internet at no cost to him, he was experiencing nearly 512Kbps of thruput, free. Before we had all the grounding wires connected and conduits in place, a low summer storm passed over, induced a static charge in the antenna dish which blew out the ethernet port of the WB-10 at the Alamo Plaza. The replacement radio, after completing the grounding, has operated through severe storms ever since. LOOK MA, NO WIRES TO AN ISP After several weeks of letting the three radio BreezeCOM network operate with its high rooftop placed exposed antennas, it was clear it was completely reliable. So we cut over from the frame relay telco connection (leaving it in standby mode just in case), put a two ethernet card router (one side using the special IP addresses CSN had given us on the BreezeCOM wireless net side, and our original Class C oldcolo.com IP addresses on the other side) This way we did NOT have to change our oldcolo.com IP addresses at all, and could switch back to frame relay if the radio net went down. It took less than 30 minutes to make the switch over. The radios have not failed since, and lots of heavy weather with much sturm and drang has occured over the buildings, antenna and exposed lines since then. So if you http://wireless.oldcolo.com you will be accessing this NSF Wireless Test Field Web Server Site at about 384Kbps wirelessly, as well as to an ISP site, through a relay site, and three Breeze PRO radios, directly from Cisco servers of CSN, to oldcolo.com. You are reading *about* wireless, *via* wireless. And if you go further to my home-office 'demo' wireless site - (http://192.160.122.3) the OS2 system with web server sitting in my home office 1/3 of a mile from the office office where the BreezeCOMs terminate you will be traversing that distance over Wi-Lan Hopper Plus radios in the 915Mhz range, hanging off the office LAN, just as if an ethernet was extended the distance. But that is another story which presently will be put up on this Case Study section of wireless.oldcolo.com. For it involves a small MAN (metropolitan area network) - a wirelessly connected 'community' - with yet another NSF project at its core - 'Local History by Wireless.' Stay tuned. Dave Hughes Principal Investigator NSF Wireless Field Tests for Education dave@oldcolo.com http://wireless.oldcolo.com