Computer Chaos in Kerkennah.

Together with Manfred DK1BT I have been responsible for the networking for the TS7N expedition which took place from November 18^th to December 2^nd 2003. This is a short rundown on the problems we encountered, and the solutions we were able to provide.

How it started

Somewhere in June 2003 I got an email from Andy, DJ7IK. He was preparing an international dxpedition to Kerkennah Island, AF-073. It would include the CQWW-CW contest, and the team was in need of CW contest operators. I looked at the pictures from the previous expedition in the year 2000, and decided I wanted to join. The next question was, what could I bring to the team? I said CW, Contesting and maybe I could help with the network, as I had a wireless LAN running under Linux at PA0R. They had decided that to reduce weight, this time networking would be done by means of WLAN.

The Challenge

There would be two sites which were 300 meters apart, and 3 shacks. And by the way, we should not forget the internet connection to the DX cluster… We would use CTWin 9.92 for the contest, which supposedly had all the bugs fixed. To reduce weight, we would use laptops for logging. A quick roundup by email showed that several laptops would be available. Three with Windows XP, one with Windows 2000, three with Windows 98 and some with W95. I planned to take my own Linux laptop to do the network configuration and maintenance. It looked like we had more than enough computing power to run the 5 concurrent stations during the dxpedition.

The WLAN preparation

For the 300 meters distance we wanted to use a point-to-point connection with two access points from Symbol, which had 1 MBit capability, and more important, a connection for a separate antenna. The German WIMO Company sponsored two 13 cm antennas. We hoped that would be sufficient. A bunch of 802.11b AirLancer PCMCIA cards was bought cheaply. They were compatible to the standard, but the speed was only 2Mbit (the full 802.11b standard covers up to 11 MBit). It took me a week of trial and terror to get these AirLancer cards going under Windows XP. There was a fault in the driver. After I had told them to do a (non-compatible) Wavelan protocol instead of 802.11b they worked. I tested them with my two 802.11b access points, and all looked fine. The system was able to relay the broadcast packets of CTWin via the WLAN from the shack to the house over a distance of 60 meters. Manfred, DK1BT was able to do the same test for Windows 98, and got it going after using the same workaround. We have not been able to get the cards going under W95, but we had enough laptops to do the job.

On-site, compatibility chaos prevails…

Right after the first antennas had been built up we started to have a look at the laptops which the dxpedition members had brought along. At first we installed Manfred’s laptop in shack nr. 2, so the first station could start making qsos logging with CT for DOS. This was fixed in 15 minutes. The first qsos came rolling in. We used Andy (DJ7IK)’s laptop for the internet connection, which worked immediately. Happy faces all around.

We needed 5 laptops with similar capabilities. One serial port for Rig Control, a parallel port for CW and PTT out, and a PCMCIA slot for either an Ethernet connection (for qth2) or a WLAN card (for qth1). One laptop would be connected to the internet to feed cluster data into the WLAN network and for updating the website and the on-line log (http://ts7n.tk).

Frustration started. We had French laptops, German laptops and Swiss laptops. With either XP, W2k or W98 on them. On the French one there was no COM port, only USB and printer port. It had an AZERTY keyboard. On one IBM Thinkpad we could not get the parallel port going. On two Compaqs the PCMCIA port did not have the right drivers. On three laptops could not get the Rig Control going via the serial port. Whenever we wanted to change something in the configuration (even a simple IP number) Windows asked for the original CD-ROM, which in most cases was not available. Moreover, where it was available the owner of the laptop would not let us reinstall the OS as he had not had time to back up the company data. After considerable effort we got the COM-port capability going on the French laptop with a USB to COM converter. Until somebody disconnected the USB to do a quick upload of the photos in his camera. We never got it going again. After a few days of work on the computers we decided that if we wanted a minimum capability of 4 laptops during the contest we had to ask HB9CRV, who would join us for the second week, to bring two more laptops. Which he did. Of these two Compaq machines we got one working with CTWin. I.e. we did not manage to let it control an IC746, so we decided it had to serve a listening station for multipliers. It also gave us some nice views on the characteristic blue screen which has made the Windows OS so famous.

During the contest we could use 2 laptops for the run stations with full capability. Two hours before the start of the contest, when I wanted to proudly demonstrate the CT capability to jump to a spot frequency from the Packet map, it turned out that the new CT 9.94 we downloaded did not have this capability. I just did not work. On the positive side, we did not experience a single computer crash during the contest. We made over 10k qsos in 48 hours.

And what about the WLAN network?

After a first look at the site we were happy to see that there was line-of-sight between beach qth2 and the hotel with qth1. We roped the 13 cm antenna to a suitable tree, wrapped the Symbol access point in plastic bags and connected it with 50 Ohm coax to the 2 laptops in the shack. The other Symbol access point was tied to a lamp post on the roof of the hotel, together with the 802.11b access point, with just a reversed UTP cable between them. The connection did not work. Until we noticed that I had brought the wrong blue cable, the non-reversed type. Measurements showed that the walls of the hotel, 30 cm thick with lots of steel reinforcement presented a major problem for the 2.4 GHz signal. This stuff was completely different from the standard Dutch home wall!! We had to move the internet gateway from the hotel managers’ office into the lobby. Close enough to the front door to receive the signal from our access point. It must have been a reflection from the restaurant at the opposite side of the street. This also made it necessary to provide a guard for the night when the hotel lobby was empty…. Fortunately we were able to buy a telephone line extension cable in the village. We used WintelnetX from K1TTT to feed cluster data into the WLAN network. I mounted the second 802.11b access point in shack 1 and everything worked immediately. For about 30 minutes. Thereafter the AirLancer cards refused to receive the IP broadcast packets with address 255.255.255.255. No problem with TCP/IP. Ping worked, ftp worked, even telnet worked. And the CT packets were sent across the network. But the AirLancer drivers refused to receive them!! We are still in the dark why. When I tested it at home it worked fine under Linux. Fortunately Andy had arranged two spare 802.11b standard cards which did not have the problem. So in the end we had a working system.

System performance was excellent. During nice weather, which was 90% of the time, the link quality was very good. During the one day that it stormed and rained, we had 2% packets lost on the link. This was more than enough quality for the CT network. The only problem was that the access points on the hotel roof were powered from the main building, which had irregular power brownouts. As the Symbol link required 2 minutes to resynchronize we had some data losses on qth2 because of that.

What have we learnt?

WLAN is a very good alternative for networking.

Don’t use Windows if you can avoid it.

You cannot use DOS either, because it does not support WLAN…

If you cannot avoid using Windows, use one version (XP).

If you have to use W98 better be prepared to reinstall it a few times.

Use one type and brand of laptop. A Chinese no-name is better than your QRL laptop with corporate software.

Don’t use the latest version of the software (CT bugs in 9.94); use the one you tested at home.

Test the complete system while you are at home. This includes network components, computers and rigs.

Although the weather was nice, we would have preferred to make qsos during the time we were debugging.

Even under circumstances which are not ideal you can make 54k qsos.

Next time our mistakes will be different.

Rein, PA0R

(PA0R@EUDXF.ORG)

Photo’s:

1.: WLAN Access point roped to a tree at qth2.

2.: Line-of –sight to qth2 from the hotel roof.

3.: WLAN access points on the hotel roof.

4.: The TS7N networking team at work: DK1BT and PA0R.