Version 1.0 of the antenna switch didn't work as planned. The beverage antennas acted more like ominis and didn't help during the contests at all. After careful measurements by W4AD and W4RM, we concluded the old relay board layout wasn't providing enough isolation between ports. Two years later - we have version 2.0. The system was installed November 21, 2009 during the ARRL SSB Sweepstakes contest. John, W4NF, reports that the beverages seem to be working well.
The final design is based on relays from COTO (2204) and a circuit board with good RF layout principles. The isolation was measured to be in the order of 65db between the antenna port and radio port. Much better than the 30db we achieved with version 1.0! We added some external coax relays to provide further isolation and to ground the antennas when they aren't in use. (Lightning protection.) Provisions for external splitters were made, which should isolate the receivers when the antennas are shared between stations.
I spent about 80 hours on the second design. Laying out the circuit board (gEDA) took most of the time. After initially laying out a single board, I decided to use a more modular approach. Again, thanks to generous junk boxes (WA0DYJ) and some good finds at Dayton Hamvention, the project cost about $200. Most of that was for the PC Board expenditure ($100), which I would do again - contrary to my claim in the write up for version 1.0.
Version 1 used standard 12V DPDT relays from Allied Control and the board was a kludge of trace cutting and rewiring. We thought it wouldn't have been a big deal at HF frequencies. We were wrong! After we discovered the isolation between antenna and radio ports was in the order of 30db, we disconnected the switch and used the old patch cord system.
I started looking for a better switching element and spent a lot of time looking at PIN diodes, FETs and coaxial relays. PIN diodes are meant for VHF and up. They won't work well at HF. I never found a FET based design I liked. So I concentrated on relays. At first, we thought about using external coax relays, but nixed that idea when we realized it was going to take a 4x8 foot plywood board as a mount and multiple power supplies.
I measured the isolation of several relays. The results are listed here. The Allied Control relay provided the least amount of isolation. (40db at 7MHz) The gold standard appeared to be a Dow Key RF antenna relay, which provided about 65db at 7MHz. After measuring several relays purchased from Mouser and Digikey, I found that it was important to use relay with a coax shield. So the search was on for something better than $4/ea since the project required 30 of them.
Luckily, All Electronics had the right COTO relays for $1 a piece. The COTO 2402 has a coax shield to provide better RF perfomance. I found they were almost as good as the Dow Key relay at HF.
The final isolation measurements of the system are listed here. I measured the isolation betweeen the antenna port and radio ports of a single board with all antennas deselected. All the ports provide better than 60db of isolation. This is a 30db improvment over version 1.0! The results were almost identical when I assembled the final unit. I didn't measure the isolation between the radio ports. (e.g. Radio 1 to Radio 2)
Version 2.0 incoorporates provisions for external RF splitters, which allows us to share antennas. This significantly complicated the switch logic and limited the various antenna/radio combinations. There are four radios (stations 1 through 4) and we have the logic set up so that Stations 1 and 2 can share with station 3 and 4, with only two radios connected at a time. When a second radio wants to share an antenna, an external splitter is switched into line. Although this provides a 3db drop in signal strength, it ensures the receivers don't feed additional noise into the shared receiver.
After a few practical tests, we're not 100% sure this is required. I'll provide an update after the ARRL 160M contest. Note - we were going to build a few splitters based on an Internet article I found, but we decided to buy them from DX Engineering.
I came across an article from SJ2W and he suggests the bandpass filter is all that's required. We need more investigation. (29Nov09)
I used Linux and gEDA to develop the circuit boards. This was a very time consuming process because I had to make several of my own parts/footprints to match the physical layout.
The auto-router function wasn't much help because it doesn't lay out traces according to good RF practices. I decided to use the top of the board for the RF traces and the bottom for the signals. The RF side has a ground plane on the top and bottom. The RF ground is separate from the signal ground.
gEDA provides utilities to produce the gerber plots, including the drill instructions. I didn't modify any of the drill sizes in the file. The PCB board shops either provide a wide variety of sizes or round up as necessary. In some cases, I had to create a new part in the PCB layout package, which required careful measurements with a micrometer. I used the closest standard drill size,leaving a few thousands of an inch for extra clearance.
The circuit boards were made by PCFABEXPRESS. They did an excellent job. The soldermask and silkscreen layers were standard. Next time, I'll take better advantage of the silkscreen layer. Cost was about $120 with shipping, using the 10 business day option. (Doesn't include shipping time.) The boards were 2.5 x 5 inches and used two layers with platted through holes. Minimum clearance is 5 mils, which I took advantage of in some of the areas.
The code is fairly straight forward. The only quirks involve the bit mappings for the port state variables. I had to change the bit order to match the physical layout. Bit 3 is Station 2, for example.
In general, I keep the state of the relays for each board in an array. When the user request and antenna, I check for allowed combinations, set the new new state and then set the switch matrix with a strobe routine. The Micrel latches are addressed like memory chips. That is, I put the data on the parallel 8 bit input bus, which connects to all 6 boards. Then I strobe the right address pin for the desired board. The latches sink the appropriate relays and hold them in that state until a change.
The six circuit boards were mounted in a box made from PC Board material. Note the lower opening for the edge terminal connectors, which are used to connect the external coax relays. There is a missing SMA connector, which I'll add next time I buy something from Digikey or Mouser.
The switch, external relays, splitters and band pass filters are mounted on a sheet of plywood. The plywood is mounted to a wall near the contest station.
Note: some of these pictures reside on the Picasa/Google site.