Someone wrote me, that I'm wrong when demanding on high noise margins, because the micro AT89C2051 couldn't be fast enough to generate relevant ground bounce (ground noise). I was told, that rise and fall times aren't less than 20-30nsec with this type of micro (12-clocker with 24MHz maximum clock frequency).

Well, I don't have an AT89C2051 running at low clock speed, but an AT89S52, which is of the same type, and this one is much much faster!
The following plot shows the signal at P2.0, loaded by 10pF probe capacitance (10:1), while the micro is running at only 3MHz:




This plot was produced by a 50MHz digital storage scope at a sampling rate of 1Gs/sec and shows a rise time of 8.4nsec.
Keeping in mind, that intrinsic rise time of the scope is about 1/50MHz/pi = 6.4nsec, and that rise times are added geometrically, AT89S52 shows a rise time of about SQRT(8.4nsec^2 - 6.4nsec^2) = 5.4nsec!
Although being a rough approximation, it can clearly be seen that a modern '89C52` derivative can indeed show fast rise and fall times, which indeed can generate lots of nasty ground noise, even if the micro is clocked at only 3MHz!

Kai