I've been doing some simple test sources, basically just one or other carefully selected LED driven by fast CMOS. Regular hex inverter packages (both 74AC04 and 74AC14) look great with resistive loads--clean
250 ps -- 300 ps edges, as measured with an 11801C sampling scope with SD-26 head.(There's a bit of residual inductive rise after the fast edges on the resistive load tests, but we can fix that.)
They're built dead-bug fashion with SO14 packages wired up like this:
10nF +5 C C | *---C C-------* | C C | GND | A *--|>o--* in | | 432R 0---|>o---|>o---*--|>o--*---RRRR---0) To sampler | | | *--|>o--* GND | | *--|>o--* | GNDWith 82 pF to ground from node A, they fall completely apart--the edges show an initial fast rise/fall of a volt or so, followed by a ringy mess for the next 10 ns.
In real life the LEDs won't be faster than 2 ns, and will be driven via a resistor, of course. However, I might well want to use a speedup cap of the same order, so it's very likely to be worthwhile spending an extra buck or so and using LVC1G04 gates in SC-70, with individual bypass caps.
Cheers
Phil Hobbs