My gut feeling is simply: "don't do that." I don't have anything here with which I can test, unless I remove an oscillator from a board and feed in some sort of data stream.



No, it's a little board I built up, with an XC3S100E-100 in VQFP, a Camera Link transmitter and a Camera Link connector. It's a Camera Link pattern generator, intended to get me a head start in developing host software (LabView, actually) while the camera itself is in layout.

(Gaff-taping the Camera Link transmitter chip (TSSOP) to an eval board was a non-starter.)

The FPGA has a UART which talks to the CL serial port and a PicoBlaze to handle the serial port communication. The FPGA generates various patterns which it outputs to the Camera Link transmitter chip. One of the FPGA's I/O banks runs at 2.5V so I can directly connect the four CL camera control lines and the UART lines to the FPGA. The rest all runs at 3.3V, which the Nat Semi CL chip likes. Of course the FPGA needs a 1.2V core voltage.

The camera has an XC3S250E-FT256 and an SiLabs 8051 (so now we're on topic). The FPGA clocks the sensor and muxes the data and formats it for Camera Link. The micro handles the serial comms plus controlling the DACs that set various camera voltages, and its ADC provides some housekeeping. It was kind of a toss-up between using the 8051 vs using a PicoBlaze and external stuff.

-a