New DSP board
Since I have done many board designs for microcontrollers, a DSP design wouldn't be too difficult. I have the tools to do a board layout and knowlege to make good use of it. This would be something I could contribute to the hobby.

Newer DSP chip
My design would be centered around the ADSP2185M. This is a newer 2.5V core chip that consumes 38mA as opposed to 85mA for the original 2181. This sounds trival, but also consider the 2185 runs at 75MIPS rather than 40MIPS. Twice the speed for half the power. Half the power also means considerable less EMI output.

Simple design
The board is relatively simple with four main parts. The 2185 DSP, a boot EPROM, a CODEC and an serial transceiver. To this I would add a power supply for 2.5VDC and 3.3VDC. Although the power consumption would be less than half a watt, I would add a switching 5V pre-regulator just because the efficiency is less than 25% for linear regulators. I would also add EMI filtering and audio transformers to eliminate ground loops.

Serial remote control
Most of the mess of the interconnections in the orginal design comes from the control functions to RF board. Primarily for the control of the PLL and power. My redesign would add a microcontroller (likely a PIC) to the RF board to handle all of these control functions. Thus a simple easy to filter serial connection could be used to communicate to the DSP.

Serial interfaces
For serial communications, I considered RS562 which is a low voltage interface similar to RS232. The advantage is that voltage levels are +/-5 volts as opposed to +/-12 volts typical of RS232. This would reduce EMI.

Another alternative would be RS485, a bi-directional interface. RS485 allows multiple devices to share a common communications line. Probably the most popular known usage of RS485 is the AppleTalk network. The advantage here would be to allow a network of modular components talk to the PC pver a single port. RS485 is also a low voltage interface which requires only 5VDC or for some ICs 3.3VDC. Another advantage, RS485 uses differntial signaling, unlike RS232 or RS562 which use single ended. This allows better noise immunity, less EMI and longer cable lengths.

From an RF standpoint, it also eliminates another source of EMI. The MAX232 or simliar RS232 drivers generate +/-10VDC through a charge pump. The charge pump is a source of EMI although less than a inductor based design.

I settled on the RS485 interface since it provides many advatages over RS232. With the separate serial ports on the RF and DSP boards, this would require the DSP to talk to the RF board and to the host pc simultaneously. Unfortunately, there is only on serial port on the DSP. While I could do a software serial interface in the DSP, this seems like a waste of resources. Secondly, I like the possibility of running dual RF modules. This would require multiple ports. I could run everything from the host port, but this would require two serial ports on the host PC. With all the modules I would like to add to my station, I decided RS485 would be the more elegant approach.

Networking
AppleTalk networks have been successfully running for many years now. Transfer rate is 253Kbps, much greater than the typical 9600bps of RS232, so controlling multiple components would be no problem.

Some years ago, I installed a network at the local university. This installation spanned several floors and rooms. At one point I reverse engineered the cabling to reduce cost. Appletalk uses a shielded single twisted pair cable with characteristic impedance or 100 ohms. A parallel 100 ohm terminating resistor at each end eliminates SWR related ringing. I should point out that the cabling was a mix of telco 25 pair and ordinary unshield twisted pair. In other words, it worked with almost any old cabling. Nowdays I would use common catagory-5 cable used for twisted pair ethernet.

Each node has a simple 1:1 toroidal transformer with 7uh inductance. I made several of these for the lab to mount permanently in a electrical box on the wall. The transformers serve two purposes. Isolation from ground loops and anti-jabber protection. Since the transformer only passes AC, a node which is stuck at a logic level will not jam the network, just itself. Since any node on the network can become a master, the transformers also help in minimizing problems when collisions occur.

While I am not a big fan of MACs, I do like the simplicity of AppleTalk. I may actually make everything compatible since Linux (and the other OS) has drivers. I have to see how hard this is on a microcontroller. In any event, I will use the hardware concept. I'm leaning toward implementing UDP protocol though since this is more common. TCP/IP protocol would be better, but is more involved.

I also considered USB of IEEE1394, but neither of these are robust enough to allow a wide area network. They would be fine if everything is within a few feet. The final interface I considered was going full blown ethernet 10baseT. The major problem is, the chips involved have high pin counts and usually use dma transfer. This would add a signifcant complexity to the project.

Practical matters
For any design availabilty of parts is aways a concern. The DSP and CODECs are fairly specialized parts and unfortunately all are pretty much single vendor sourced. Hence, availability becomes a problem. A couple of suppliers carry these, but have minimum quantity purchase requirements. Always a nuisance when trying to prototype a design. Fortunately, Analog Devices sells parts direct, although not the best pricing. Unfortunately, they also don't show these parts currently available at their website.

Another problem is the ADSP2185 is packaged in a PQFP100 or BGA. The BGA style would be near impossible for most of to stick down. However, with care and patients the PQFP can be put down even with a small tipped iron and .031 solder. Of course there are much easier ways with solder paste and a hot air gun. Fortunately, I have access to some nice Pace SMT tools, but only on a favor from a friend basis.

I have been doing some research on alternative soldering techniques, which I hope to present on this site as I try them.