Design Note HFDN-18.0 (Rev.1; 04/08)
Maxim Integrated Products
Page 3 of 9
Case 2. The laser slope efficiency is unchanged but
the threshold current is greater than nominal, as in
Figure 1, characteristic (c). This is the disaster
situation in which the laser operation falls below
threshold. Other scenarios can be investigated, with
various combinations of laser slope efficiency and
threshold, but it is obvious that, for specified average
optical power output, the disaster situation becomes
more likely as the nominal extinction ratio is made
larger.
2 Optical Feedback
via a Monitor Diode
Many laser assemblies include a monitor photo
diode. Photo diodes are essentially linear in their
relation between optical power input and reverse-
biased current. Incident photons generate hole-
electron pairs in the diode, and increase its reverse
leakage current above the “dark” value. Thus,
monitor current is a measure of laser optical power
output and, by incorporating the laser and monitor
into a suitable feedback system, it should be possible
to control the optical output. Figure 2 shows the
general idea.
The desired laser optical output waveform is first
scaled by a factor that is the inverse of monitor gain,
A
monitor
, where
.
)
(
)
(
monitor
output
power
optical
laser
d
nt
photocurre
monitor
d
A
=
(4)
This scaled waveform is used as input to a
“classical” feedback control system, in which the
forward path consists of a high-gain current
amplifier plus the laser acting in cascade, and the
monitor diode constitutes the feedback network.
Then, provided only that the loop gain is large at all
frequencies of concern, the actual optical output
waveform from the laser mimics the desired
waveform:
.
gain
loop
monitor
laser
amplifier
A
A
A
×
×
=
(5)
P
max
and P
min
, hence P
av
and r
e
, are all controlled and
held constant despite variations in the laser
characteristic.
The trouble is that photo diodes have a limited
bandwidth; details vary, but typical diodes behave
much like a low-pass filter with a cutoff around
100MHz. Diodes can of course be manufactured
with larger bandwidths, but at increased cost and
with other problems. When the data rate is low
(perhaps up to 100Mbps), the system can be made to
work. However a number of things go wrong at high
data rates:
•
Basically, the problems all stem from the fact
that the loop bandwidth must be of the same
order as the bit rate, for satisfactory reproduction
of the input waveform. Thus a 2.5Gbps data
waveform requires a bandwidth around 2.5GHz.
(2GHz or even 1.5GHz might be enough,
depending on the fidelity requirement, but the
order of magnitude is 2.5GHz.)
•
The monitor diode contributes a dominant pole
at about 100MHz to the feedback loop.
Therefore the requirements on the high-gain
current amplifier become extreme: it is difficult
to stabilize the feedback loop.
Figure 2. Feedback control of a laser