??? 01/19/05 08:18 Read: times |
#85274 - Head banging Responding to: ???'s previous message |
Russell Bull said:
1. Large chip count - we have suggested other means that address this. Doing it serially isn't necessarily slow. We still don't know how 'fast' you want this. You want it simple - you have a few choices - hc logic - large chip count, GALs - less chip count, CPLDs low chip count, FPGA low chip count and doing it serially at high speed. The actual microcontroller and clock frequency being used have yet to be discussed, and are of critical importance here. An address decoder built from multi-level HC logic is likely to deliver unacceptably slow performance for all but the most modest systems (i.e., clock/12 devices running at low clock frequencies), outside of being truly unwieldy on the board. Even PAL devices might not deliver high enough performance for the fastest 8051-based systems. I recently found myself adding MOVX stretch cycles to a DS89C420 (clock/1 device) design running at 29.491 MHz because the ATF22V10C-10PC PAL being used as an address decoder was too slow. A great deal of caution is required here, but before caution may be exercised, the facts must be on the table. 4. Fan out problem - I think capacitance might be an issue here. GALs,CPLDs or FPGAs address this problem. It might be, but I am no expert in this area. Bi-directional bus drivers (i.e., 74xx245/74xx645) could be used to alleviate this problem though should it arise, provided their propagation delay can be tolerated. 5. What would I do? Probably some totally different to what you're proposing - at this point I don't know what it is I'm supposed to be designing! 10 years ago I may have used discrete logic, nowadays programmable logic rules - if I make a mistake, I just reprogram it. But then again, it might be sensible to use a number of small and cheap micros serially connected to solve the problem or I do the whole thing in a fpga. How many of these I have to make also would influence my design decisions. Even better than a string of cheap micros serially connected, however, is a string of serial I/O expanders. For example, consider Maxim/Dallas' offering in this area: http://www.maxim-ic.com/products/...anders.cfm This is quite a rich set of offerings, including both I2C/SMBus and SPI interfaces. Barring any very high performance requirements, something like these are likely to be quite convenient. --Sasha Jevtic |