One doesn't often see the words dimmers
and 4GHz together. Explanation, please?
NIF has 192 beam lines, each a big pipe full of flashtube pumped
amplifiers. The input to each line is a fiberoptic blip of light at
about 1100 nm; it makes 4 free-space passes through the slabs and
emerges at the and, goes through a tripler, and enters the target
chamber as UV. It all converges on a little gold capsule full of
tritium and stuff.
The beamlines are grouped as 48 quads. In the MOR, the master
oscillator room, there's a master laser that feeds 48 of our modulator
boxes. Each box has a 4 gs/s 16-bit arb and a programmable 1 ps
resolution square-pulse generator, and a bunch of bias stuff. We drive
a 2-stage Mach-Zehnder optical modulator, one stage being a square
pulse gate and the other being an arbitrary waveform. The output of
our boxes gets optically amplified and is then run to the 192 big amps
out on the floor. The idea is to precisely shape the waveforms of the
light that ultimately whacks the target with about 1.5 MJ of UV.
What's a little unusual about the arbs is that they are triggered and
generate their one-shot waveform with a couple picoseconds of jitter.
The whole system - arbs, modulator, downstream amps - is fairly
nonlinear, so they fire the MOR stuff 24/7, at 960 Hz, and sample
downstream optical signals, and close a loop on the desired waveforms.
So everything up the the main flashtubes is run and tweaked
continuously.
The arb is done by sheer brute force. There are 140 gaussian impulse
generators, each a delay circuit, a gaasfet, and a shaping network.
Each is dac programmed for amplitude and time of peak, and all 140 are
fired at 250 ps intervals and summed. If we had to do it again, we'd
probably go with a ring of staggered-triggered medium-fast dacs,
mux'ed somehow.
We also generate some 90 ps wide fiducual pulses, for aligning shot
diagnostics.
The math of summing gaussian impulses is interesting. It has analogies
to making arbitrary shapes from power or Fourier series.
John
