Not selling that service as an isolated item. It just happens in the context of another project. For example, we did a very advanced high power (1,500W) LED array. The project comprised mechanical, electronic design, microprocessor firmware, FPGA development and workstation software tool development.

Thermal design required running piles of FEA thermal simulations over a period of about three months, probably 12 hours per day. I feel it is better to charge per CPU-hour in these cases because it gives you the flexibility, when the tools support it, to give the client the option to rent additional CPU time to accelerate development.

On the FPGA front, depending on design complexity you can have simulations that run for hours before you can analyze results. Again, I don't feel it is proper to charge per man-hour when you can walk away and go do something else. You do have to charge for the processing time and use of hardware. We have 16 rack-mounted 8-core machines with tons of memory for simulation work.