Every synchronous circuit has a speed limit. The maximum clock frequency fmax is the highest clock rate at which every register-to-register path still meets the setup constraint.
To find it:
- For each path, compute
tcq + tpd + tsu, using the path's criticaltpd. - Take the largest of these. That's the minimum clock period
Tmin. fmax = 1 / Tmin.
Unit shortcuts make step 3 quick: fmax in MHz = 1000 ÷ Tmin in ns, and fmax in GHz = 1 ÷ Tmin in ns.
What raises fmax:
- Faster logic on the critical path. Speeding up other paths does nothing.
- Pipelining: cutting the logic into shorter stages.
- Positive clock skew on the critical path, which subtracts from
Tmin.
What doesn't: anything about hold. The hold constraint has no clock period in it, so it neither limits nor is fixed by the clock speed.