I have been designing a 66/64 PCI interface and now is trying to fit my design into a Altera 20K chip.
After some efforts, we found PCI interface AD pins Tco(clock to output) is at 13ns level. It is due to AD_OE delay. AD_OE has to drive 64 pins at the same clock, no ways to shorten its Tco by placing near all pins. An engieer from Altera suggested me to assert 1 clock before AD data is sent to AD.
The following is an exerpt from his email:
"If you are unable to place both the OE reg and the OUTPUT reg near the pin to meet timing, what you might have to do is assert the OE signal one clock earlier to when the data is asserted.
Since your design will own the bus, you should be able to assert the OE signal without any problems, and then on the next clock edge your data can be successfully transferred onto the bus meeting all Tco requirements."
I would like to carefully listen to everyone's opinion and experiences on whether his method is workable or any other tricks, for example, using reverse clock to get another half clock time for AD_OE.
Micro Memory Inc.
9540 Vassar Avenue
Chartsworth, CA 91311