Min-Hyeong Kim
High-Speed Circuits and Systems Laboratory
E.E. Engineering at YONSEI UNIVERITY
2011. 5. 11.
[ Contents ]
1. Abstract
2. Background
- SACM APD structure
- Ionization/multiplication coefficient
3. Device structure
4. Measurement results
Dark current
Excess noise factor & Gain-Bandwidth product
III. Receiver sensitivity & BER
5. Conclusion
1. Abstract
 Monolithic Ge-Si SACM APD
operating at 1300nm
(separate absorption, charge and
multiplication avalanche photodiodes)
 Gain-BW product : 340GHz
 K_eff : 0.09
 A receiver sensitivity
: -28dBm at 10Gb/s
 Si material properties allow for high gain with less excess noise than InPbased APD and a sensitivity improvement of 3dB or more.
 With Si, an even higher gain–bandwidth product could be achieved based
on a simple layer structure with relatively large process tolerances.
2. Background
Ⅰ. SACM APD (separate absorption, charge and multiplication APD)
InAlAs-based APDs
InAlAs-based APDs
Si-based APDs
(This work)
2. Background
Ⅱ. Ionization/Multiplication coefficient
 K : Ratio of the ionization
coefficients of electrons and holes.
 A low k value is desirable for
high-performance APDs.
Impact Ionization probability = W
Multiplication coefficient M
 1  W  (W ) 2  (W )3 
1  W
( for W  1)
→ Excess Noise factor F(M)  kM  (2 
)(1  k )
3. Device structure
 Punch through voltage -22V
 Breakdown voltage -25V
with Responsivity 5.88A/W
Designs for a floating guard ring (GR) with various distances (1–3 mm)
between the guard ring and the mesa edge were introduced to reduce
the surface electric field strength at the silicon/insulator interface to
prevent premature breakdown along the device perimeter.
4. Measurement results
Ⅰ. Dark current
 When the reverse bias increase, not
only the gain becomes large but also
the dark current increases.
 It is because of (1) junction leakage
current (generation and recombination)
and (2) tunneling current.
T  e L
 The breakdown voltage is -25V, and
here at the dark current of 10uA.
 All these measurements are supported
at 1300nm wavelength.
4. Measurement results
Ⅱ. Excess noise factor
& Gain-Bandwidth product
F(M)  kM  (2  )(1  k )
 After measurement of excess noise
factor, the k value is calculated about
to 0.09 by using above equation.
 All measured devices had a gain–
bandwidth product over 300 GHz. The
highest gain–bandwidth product
obtained was 340 GHz.
 The 3dB BW was measured using
Agilent 8703A Network Analyzer. The
bandwidth is limited by RC and transit
time effect.
 As the gain is increased beyond 20,
the bandwidth dropped owing to the
avalanche build-up time effect.
4. Measurement results
A gain of 10
& -20dBm input optical power
Ⅲ. Receiver sensitivity & BER
 A data rate of 10Gb/s
 Using a pseudo-random binary
sequence(PRBS) and extinction
ratio(ER) of 12dB.
 In this set, the input optical
power(Receiver sensitivity) was
** Sensitivity in a receiver is normally defined as the minimum input
signal Si required to produce a specified signal-to-noise S/N ratio.
(So, it is a function of the SNR or BER.)
5. Conclusion
• To improve more,
(1) Reducing the dark current of the APDs.
: Better control of the germanium profile with respect to the electric field
distribution in the device can reduce the tunneling current.
(2) Reducing the value of k_eff.
: Studies have shown that k_eff can be reduced by optimizing the
multiplication region thickness. By this, we believe that a sensitivity of
approximately -32 dB m could be achieved.
• Demonstrate a monolithically grown, CMOS-compatible Ge-Si
SACM APD device with a gain–bandwidth product of 340 GHz
and a k_eff of 0.09 at 1300nm wavelength.
• The optical receivers built with this Ge-Si APDs demonstrated a
sensitivity of -28 dBm at 10Gb/s.

similar documents