Wirelessly Connecting two FPGA development boards (Celoxica RC100 boards)

Can you explain a bit more? Are you planning on looking to replace the Celoxica boards with something else, or do you want to mate the Celoxica boards to some optical or wireless transmission system? If so, how would you want to transfer data to/from the optical or wireless interface boards?

analog

board

board.

Is

and

encoding).

Thank for replying to my post. To answer your first question I want to mate an optical or wireless communication interface to the Celoxica RC100 boards. Set up so far is as follows:

Camera -> Celoxica Board -> Ribbon Cable -> Celoxica Board -> Monitor

The Ribbon cable is connected to the Celoxica boards using the expansion header on the celoxica boards. This expansion header allows digital communition in and out of the FPGA on the Celoxica board. The data on the ribbon cable ha a bus width of 32 (i.e. is 32 bits wide) and and the data changes every 100 ns (10 MHz). All I want to do is remove the ribbon cable and replace it with an optical or wireless communication system (preferably a wireless system). So system would be:

Camera -> Celoxica Board -> Wireless transmitter -> Receiver -> Celoxica board -> Monitor

One board and the transmitter would be at one end of a room, the other board and receiver at other end of room. The transmission range is to be small e.g. max of 5-8 meters. The data rate is fairly high so not sure if a wireless system would be up to the task. Hope this has clarified my situation.

I'm not familiar with what's available in the radio area.

You can get fiber transducers that will do the job. Look for gigabit ethernet parts and browse around when you find an interesting vendor. The interface is usually differential PECL.

Beware of parts for telco usage. They will be high power and very sensitive to get long range. Expensive.

You have to do a parallel to serial conversion on the transmit end. The receiver end has to do clock recovery, serial to parallel, and also lock onto the byte/word boundaries.

There is actually another level of complication. The clock recovery needs enough transitions so you probably need something like 8b/10b encoding and decoding.

You might find a PHY type chip set that will do all that. But maybe that stuff is all included in the big PCI chip now.

If you can't find any PHY chips, consider using several fibers and doing your own Manchester encoding/decoding. Decoding is pretty easy (at least after you see it) if you have a clock that's running at 8x the bit rate. You can drop that to 4x for all but the first FF if you process 2 bits in parallel. Just use enough fibers so your receiver can do the decoding at a sane clock speed.

--
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.

Looks like you have an equivalent serial rate of 320 MHz, plus some sort of framing overhead so that you can find bit/byte boundries. You can certainly do this with an electrical interface over 10m and save a lot of money over optical.

What time frame are you looking at to implement the solution? Reason I ask is that a small company called Core Foundry is finalizing a new FPGA development system called PROTEUS. Their intent is to create a modular FPGA development platform that can be customized to an application simply by changing removable plug-in modules. The first incarnation is a small aluminum box, about 7"w x 9"d x 1.5"h that has room for 4 I/O modules out the front. I've seen modules for dual T1/E1 (so up to 8 ports can fit in a box), a T3/E3/STS-1 single port electrical interface, and a multirate SFP module with CDR that can handle 30Mbps through 3.2 Gbps. This might be the module you care about as it will accept any SFP plug-in, including the copper GbE SFP modules from Molex and others.

The I/O modules plug into a main board that has a simple serial interface on it to the outside world. Behind the I/O modules are connectors to accept FPGA boards. A single-wide board connects one I/O module to one FPGA. A double-wide board connects one FPGA to two I/O modules, and a quad-wide board connects one FPGA to 4 I/O modules. Behind that are connectors for a more powerful microcontroller board that is optional. This will have the

10/100, USB and more serial ports on it. The first uC board available will be based on soft FPGA cores as opposed to hard uC chips.

The initial FPGA boards will be Cyclones by Altera, as they are readily available. All parts sizes will be available. Spartan 3 boards will most likely follow next year whenever they are production qualified` and readily available. Beyond that they've talked about plans for Stratix and Virtex FPGA boards, and perhaps some non-FPGA boards with DSP chips, or maybe combinations of DSP and FPGA.

For your situation, a single SFP I/O module and a single Cyclone 1C3 FPGA module should suffice. Probably wouldn't need the uC board at all. The FPGA board has ribbon cable connectors on top so that you could cable into it (with the box lid removed). Then serialize and format the data and shoot it out the SFP optical port (or maybe an SFP GbE port, but you'd have to do more work on the data up-front to make the transcievers happy). Mapping your data into an OC-12 payload would be trivial, on the other hand.

The PROTEUS system is in hardware testing now, and I don't know when it will become available, or at what price points. They don't have any info posted on their website yet either, although they say that info will be posted in October. I'm looking to be an early customer myself.

I'm not a wireless expert by any stretch, but Infineon makes a nice looking BT module, ROK104001, that looks like it would make BT simple. I think in small qty they are $20 or so. I heard about it from a local Insight sales rep.

would

boards?

boards in

an

other

expansion

transmission.

MHz