Cryptography Reference
In-Depth Information
The data pulse passes through the unbalanced Mach-Zehnder interfer-
ometer where one arm applies phase shift modulation to the pulse. A D-to-
A converter drives an electro-optic modulator with an analog voltage that
produces the Basis and Value phase shifts clocked from the transmitter elec-
tronics. In the other arm an adjustable air gap delay line allows fine-tuning
of the interferometer differential delay. After exiting the interferometer the
data pulse is attenuated to achieve the requisite mean photon number. A po-
larizer then removes mistimed replicas of the data pulse that may have been
generated by misaligned polarization-maintaining components in the inter-
ferometer. At the transmitter output, the data pulse is combined with the sync
pulse in a DWDM (dense wavelength division multiplexing) optical filter.
At the receiver the sync and data pulses are separated with a DWDM
filter and the sync pulse is detected with a PIN-FET receiver. This signal is
shaped in a pulse thresholding circuit that produces two outputs: a 100 ns
TTL-level clock signal sent to the receiver electronics and a 4 ns NIM-level
APD gate-timing pulse that triggers the APD gate-pulse generators and the
pulse generator driving the APD output line gates. The output line gates are
timed to pass only the demodulated data signal from the APDs and block
noise due to spurious pulse reflections. An adjustable delay line in the NIM
pulse interconnection allows fine-tuning of APD gate-pulse timing.
The data pulse passes through a fiber delay loop to adjust its timing
with respect to the sync pulse and then through a circulator that is the input
to the interferometer demodulation circuit. This interferometer is a folded
version of the conventional Mach-Zehnder design and is independent of the
input polarization to accommodate the uncontrolled incident polarization at
the receiver. Faraday mirrors at the ends of the unequal-length arms reflect
light so that the polarization of the light returning to the beam splitter is
the same for each arm, producing interference with high visibility [5]. The
Basis is clocked out of the receiver electronics and applied to the electro-optic
modulator through a D-to-A converter to produce a phase shift of either 0 or
π
/2. A pair of cooled APDs, biased above avalanche breakdown only during
the time a data photon is expected to arrive, detect the interferometer outputs,
one from the beam splitter and the other from the circulator. After gating to
select only the data pulse, the APD signals are shaped by threshold detectors
and passed as 0 or 1 to the receiver electronics.
A phase-correcting feedback signal, derived by the receiver from training
frames sent by the transmitter, is used to maintain phase stability between the
transmitter and receiver interferometers as path lengths change with temper-
ature and stress. This phase-correcting signal is applied to the receiver inter-
ferometer electrooptic modulator through the transmitter electronics. Phase
correction is also necessary when a transmitter and receiver first connect dur-
ing a startup or switching operation to obtain the phase-matched condition
needed for low quantum bit error rate. See [6] for a discussion of BBN's algo-
rithms for automatic path-length control.
It is by now well-known that certain conventional InGaAs APDs can be
operated in the single-photon regime if properly cooled and gated. Like many
