Modifying Webgear Aviator 2.4 Wireless Network cards to use an external antenna

I just popped open the antenna flap (the folding out bit) by putting a sharp knife (or chisel works too) into the join starting at the corner nearest the hinge (the weld is weakest there). If you use a wedged blade, the weld should crack a few mm along from where you inserted the blade. Reinsert the blade at the end of the crack. Repeat, if you have the right shaped blade this should be fairly quick, 10 minutes or so. When the weld is cracked all the way around, move on to the join at the back of the hinge (you'll have to shut the antenna flap to do this). Be careful as the coax which goes to the antenna is behind this join. When this is cracked, carefully open up the two halves of the flap, it should pop open with a little effort. You should see the antenna, a printed circuit board with a tinned conductive coating. It has two c shaped cutouts in the coating. These are the two antenna elements. I think that these are to provide two different polarisations so that the card can optimise it's parameters depending on the card orientation and/or polarisation of the incoming signal. The elements are significantly smaller than the wavelength of the microwaves (which are about 12cm).
When I first tried this I just desoldered the mini-coax from the antenna and resoldered a panel mount socket to the end. Actually that's basically what I still do. I'm sure it's radiating like hell from the back of the panel mount, but it seems to work. The panel mount socket (N type, Dick Smith A$6) is mounted in the flap which is still connected to the card by the hinge. The mini-coax is just long enough to do this. I'll post a photo when I get some film for my camera.
I did investigate a little further to find the legendary J2 connector. It's on what I would call the top of the card, but what others refer to as the bottom of the board. This is because the top of the board has been defined as what you see when you pop the metal PCMCIA case apart. You then need to pop the board out to see the J2 and J3 connectors. I didn't rip the case of our cards apart. I carefully plunged the soldering iron into the top of the card to what I guessed was slightly less than the thickness of the plastic case in the position I expected the J2 connector to be, based on the pictures I have seen (I have posted the pictures here). It is about 1cm in and 1cm up from the bottom left corner of the card when the card is in front of you and the label is facing up with the writing the correct way up. Have a look at the picture on Robert Scholten's page, it's easier on the brain. They have connected to J2.
With the benefit of 10/20 hindsight I should have guessed that the plastic case would contain a shield. There is a very fine copper gauze on the inside of the plastic case which you should be very careful to remove to prevent shorting of the components. When you open it with the hot melt soldering iron technique (quick and dirty) it tends to drop down onto the circuit board, so be careful.
I decided that since the manufacturer went to the trouble of attaching mini-coax to the board, I should use that. When I get some spare money i'll probably get a pigtail made up to reduce the stress on the cards. On the card that I opened to see the J2 connector, I checked with the ohm-meter which of the two coax cables was connected to J2 and used that one. On the other card I just guessed. In the absence of any reason to suspect otherwise, I reckon that it won't matter which coax/jumper I use, as the card will probably correct for it. If anyone knows more about this then contact me so I can update this. My assumption is that the card checks the signal quality and optimises the power to the two antenna elements. One element is connected to J2 and one is connected to J3, so pick whichever and let the card do the work.
I can now connect the cards up to the appropriate coax. I used LMR-400 as recommended on the wireless mailing list. If you get the cable assembly made up by rojone it is cheaper than buying the parts unless you have the appropriate crimp tool. I paid A$140 (May2001) for a 20m cable with 2 N connectors including delivery. I actually wanted two shorter cables with an N connector on one end but that's what side cutters are for isn't it?
I initially tried this with BNCs and some 50 ohm ethernet cable I had lying around, but BNCs crap out at about 1Ghz and cheap ethernet can have losses >50dB/10m at 2.4Ghz by my calculations.
We now have one card in Steve's main server (which is now within reach of the antenna) and I have the P90 at my place so I can use it as a wireless router and backup internet server. The system provides a throughput of up to 440Kb/s. This is with almost no leaves on the trees nearby. If anything it seems to work better in the rain, with throughput of up to 550Kb/s. I'll keep this updated with things as the weather and seasons progress. Tribes 2 plays fine over this link.
Update: I'll have to sheepishly admit that we put the last mast up when it was getting dark and I failed to notice that one of the tree trunks between our houses was directly in the line of sight of our antennas. Oops :( . We added a cute little wooden arm to the mast to move the antenna 30 cm to the right and our throughput is up at about 800kb/s. I think that over this distance diffuse reflections are assisting us so hopefully we won't have a problem with wet leaves.
The little bowtie antennas are an unknown quantity, but they sure are small, about 93x107mm and about 30mm thick. The first one was made up by Kerry, a friend of Steve's (thanks for that), and I duplicated this. It would be possible to put these in a jiffy box, and I think I might do this, but currently we just have some clear plastic taped over them to prevent shorting (update: this keeps blowing off so I have left it bare at the moment). If we find that we can't punch through the wet trees come spring, we might try Jason Hecker's helical antennas (Update: the trees do seem to cause problems when it is wet, but the signal still gets through to some degree).

The whole thing is fine for me to play multiplayer games over the internet via Steve's modem. It's a shame that the bandwidth is not quite sufficient for both of us to play, so we have to alternate. Now I'm looking forward to when Steve gets that ADSL connection.... mmmmm... yummy....
Update: This is a separate issue entirely but Telstra refuse to connect me or Steve to ADSL :( . For me it is because I'm on a pair gains line. Technically the exchange has capacity to switch me to all copper twisted pair, but telstra don't do that. My complaint is in but went nowhere. Steve has an ANT-1 device to split his all copper line into a phone and data line. It would seem easy to rip this out and replace it with a similarly functioning device, an ADSL modem, which funnily enough splits your line into a phone line and a data line, but the data line is much faster. Telstra don't do this either. I guess we'll both have to wait for Transact, and that will be a one-way transfer, bye bye Telstra. Customer support, now there's a concept.

History

We initially tried running house to house with the inbuilt antennas, with no real success, despite being only 100m apart. This is within the specs, even for indoor use. Our line of sight was not great and Steve has 15cm aluminium gridded windows (can anyone say "faraday cage"). We built a "doghouse" on stilts and placed one router (fido) outside so that our radio line of sight was sort of OK (there was still a fence in the way). Performance was patchy at best (ping 70-100% packet loss).
We were running a 486 LRP box at one end of our link. At the other Steve ran a cutdown real linux system on a P90.
We added a passive repeater, which was a bowtie antenna connected by high quality coax to a small dipole pickup, which we taped to the card antenna flap. This worked (you can pick up your jaw now). By appropriate orientation we could get 230kbits through in dry weather. The rotation angle was important as well as the direction, so don't forget to try this. It was quite acceptable for the odd game of UT or Diablo II. Parked cars and wet weather presented a problem, so that we could only get 0-60kbits through at times. Then I was convinced to rearrange my house so that the computer room is at the other end with no line of sight. Thus we were forced to upgrade, but this presented the opportunity to get it right, so that (hopefully) the throughput is workable in wet weather. My web access had improved, as even with a patchy 60kbits, then out through Steves modem, the throughput was better than my modem. Due to some problem I have yet to fix (I think it's the lines), I only get a 16.8k connection with my 28.8 modem, and that's with a 40% error rate.
Email Me: mick@faceng.anu.edu.au
Back to Mick's page
Last Modified 5May02