??? 10/18/06 15:38 Read: times |
#126672 - well, that's a big maybe ... Responding to: ???'s previous message |
Peter Dannegger said:
Richard Erlacher said:
Trading off board space and cost between the SSI/MSI vs. CPLD's, these days, the CPLD always wins out. A 36 MC CPLD costs But there are also many points, where the CMOS logic always wins out: 1. CPLD,FPGA are often only 3.3V or below, CMOS can work with 5V, 12V or 15V. E.g. for 15V you need only a level converter CD4504 for the serial bus and some CD4094 for the outputs and CD4021 for the inputs. That's true enough but probably seldom arises. What sort of application for which one would consider programmable logic would work with 4xxx-series CMOS? What kind of serial interface uses unipolar 15 volts? 2. placing a common serial bus and some CMOS circuits exact on this place, where the IOs are needed may cause an easier pcb as a single big FPGA in the middle of the pcb and tons of wires around it. Very true, but small CPLD's are different. A TQFP-44 is about the size of your little fingernail. You can buy them in 5-volt, as well as in 3-volt types. 3. FPGA need often more current than CMOS logic circuits. Peter FPGA is a different story than CPLD. These normally only "tolerate" 5-volts, below which it's hard to do "useful work." Further, they generally live in dense, high-pin-count packages (though TQFP-100 is pretty small, about the size of your thumbnail) and require an external "boot" rom. I mentioned the 36-macrocell 5-volt CPLD only because I just bought a few of them and they're cheaper than the HCMOS logic I'm replacing with them, though only by a little. One really good thing, though, is that I can program them in-situ, hence, don't have to socket 'em, should I ever go to production with such an arrangement. In my case, I'll program them and subsequently stick them in a socket. With that XC9536, I can build the equivalent of an 8255 in a fixed configuration, i.e. I leave out the soft configuration and set the features by programming the device. If I need to change something, it's programmable in-situ. Your points are well taken. There certainly are cases where it's more reasonable to use 4xxx-family logic. Those are performance-limited, though. RE |