Optik 124 (2013) 1493–1498

Contents lists available at SciVerse ScienceDirect

Optik
journal homepage: www.elsevier.de/ijleo

Performance comparisons between PIN and APD photodetectors for use in optical
communication systems
Osayd Kharraz, David Forsyth ∗
Photonics Technology Centre, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310 Skudai, Johor Bahru, Johor Darul Ta’zim, Malaysia

a r t i c l e i n f o a b s t r a c t

Article history: In this report, a performance comparison of the conventional PIN photodiode with the Avalanche Photo-
Received 25 November 2011 diode (APD) in an optical communication system is presented. The effects of bandwidth, gain, extinction
Accepted 8 April 2012 ratio, shot noise and thermal noise are compared and studied in detail. It was shown that the Q factor
produced by each detector is heavily affected by the thermal noise in the PIN device, and by both the
thermal and shot noise in the APD. It was also found that the APD’s gain plays a significant role, and
Keywords:
the shot noise has to be carefully dealt with. Additionally, the relationship of receiver sensitivity with
Q factor
thermal and shot noise was investigated and compared.
PIN
APD © 2012 Elsevier GmbH. All rights reserved.
Thermal noise
Shot noise
Gain
Extinction ratio

1. Introduction is merely to transport data signals to and from the device under
scrutiny, reliably over long distances. We include in our system an
Much interest has recently been expressed in investigating the attenuator, where we can simulate real fiber span. Our desired Q
basic PIN and the more complex, expensive (about 4× the cost of the factor is approximately 7, corresponding to a very low BER of 10−12 .
PIN) and voltage-hungry APD receiver performances, mainly due Throughout this work we use the Q factor as metric to assess all
to the on-going and high-pressured commercial demands for cost- cases.
cutting in systems incorporating these ultra-fast receivers [1–7].
APD photodetectors have been shown as the better candidate for 2. Theory
long haul communications, due to their internal gain availabil-
ity [1,8–11]. However, with the recent rapid developments in the In the PIN photdiode, the net noise is given by:
technology of both PIN and APDs, the bandwidth of both is now 2 4kTB
(iPIN ) = 2e(Īph + IB + ID )B +. (1)
realized at 1 THz rates or higher, and this opens the door to new R
design considerations. In the PIN photodiode, thermal noise plays The thermal (Johnson) noise is assumed to be 4kTB, where k is
the dominant role in the performance of the receiver. In the APD, the Boltzmann constant and T is the absolute photodiode tempera-
2 ) = 2e(I )B, where e is the
ture (K). Shot noise is assumed to be (iPIN
both the thermal and shot noise is significant. D
Since optical communication systems are quite complex and charge of an electron (C) and ID is the dark current (nA).
difficult to analyze, it is useful to predict the effects of various In the APD, the net noise is given by:
parameters and characteristics of the photodetectors used in opti- 4kT
2 2
cal systems prior to their practical their construction [12]. This work (iAYAL ) = (ithermal ) + (ia2v ) = 2e(Iph + IB + ID )BM 2 F + B (2)
R
simulates the effects of varying parameters on the performances of 2 2
The shot noise is assumed to be (iAYAL ) = (ithermal ) + (ia2v ) =
both types of device. Such simulation, as a tool, can accurately pre-
2
2e(ID )BM F, where e is the charge of an electron and ID is the
dict the performance of any proposed system to be implemented by
analyzing the device under development in order to eliminate any dark current. Due to the statistical nature of the avalanche process,
performance degradation. The objective of the fiber optic link used avalanche photodiodes generate excess noise (F):
 1

F = M + 2 − (1 − k) (3)
M
∗ Corresponding author. Fax: +6 07 5566272. This excess noise factor (F) is a function of the carrier ionization
E-mail address: [email protected] (D. Forsyth). ratio, k, where k is usually defined as the ratio of hole-to-electron

0030-4026/$ – see front matter © 2012 Elsevier GmbH. All rights reserved.
http://dx.doi.org/10.1016/j.ijleo.2012.04.008
1494 O. Kharraz, D. Forsyth / Optik 124 (2013) 1493–1498

Fig. 1. The relationship between gain and excess noise factor at different ionization
ratios.

ionization probabilities (k < 1), and M is the multiplication gain. We

have regenerated a conventional representation of Eq. (3) by using
Matlab (Fig. 1). This illustrates the rise in statistical noise behavior
with increasing APD multiplication gain.

3. Design methodology

The main components of the optical link are shown in Fig. 2.

A pseudo-random-bit-sequence (PRBS) generates pulsed electri-
cal signals, which are then used as a modulation input. Signals
are split into two outputs signals, which are detected by both the
PIN photodiode (PIN-PD) and APD photodiode (APD-PD), and are
Fig. 3. (a) Q factor and (b) sensitivity degradation at both receivers output with ER;
converted into electrical output signals again. These electrical sig- bit rate = 10 Gb/s, APD gain = 3, input power = 0 dB, attenuation = 30 dB.
nals are filtered through a low-pass Bessel filter, which shapes
the voltage pulse. Its purpose is to reduce the noise and dis-
tortion without introducing much inter-symbol interference (ISI) rise and fall time for the laser signal were both equal to 0.05 bit. The
[12]. insertion losses of 1 dB and depth 100 dB are taken in consideration
The key components of the simulation set-up are: PRBS gener- for a 4th order low pass filter. Both PIN and APD photodiode have
ator, pulse generator, laser diode source, attenuator, PIN PD, APD a responsivity set to 1 A/W and a dark current of 10 nA. Binary data
PD, low-pass filter, and BER and eye diagram analyzers for assessing are produced by PRBS generator, and then converted into electrical
the performances of the link. The laser diode source’s nominal cho- pulses by pulse generators, and passed into a directly modulated
sen wavelength was 1550 nm, with output power set to 0 dBm. The laser. The electrical pulses generated in pulse generator have an

Fig. 2. Systematic layout of the simulation set-up.

 O. Kharraz, D. Forsyth / Optik 124 (2013) 1493–1498 1495

amplitude 1 a.u. A low-pass Bessel filter, with cut-off frequency 0.75 electrical signal into optical signal form, where we can specify
bit rate and depth 100 dB, is employed after the photodetectors to the extinction ratio. We sweep the figure of the extinction ratio,
remove distortion in the electrical signals. and then calculate the sensitivity and Q factor at the receiver.
In high-speed optical communication systems, large modula- Fig. 3(a) illustrates the improvement seen in both devices with
tion depths are required along with large modulation frequencies increasing extinction ratio. The optimum value is about 30 dB for
and minimal power losses, in order to avoid undesirable effects. both devices. Fig. 3(b) shows how the extinction ratio affects both
A high modulation depth is desirable because the optical signal receiver sensitivities. Both PIN and APD receiver sensitivities are
exhibits a greater margin to resist the noise. The extinction ratio reduced and degraded. However, the optimum gain considered
is a measure of the modulation depth of a source transmitter, here helps the APD to maintain its superiority. Otherwise, the APD
and expresses the proportional relationship between the power sensitivity is expected to be degraded more.
levels of the binary 1 and 0 signals, while transmitting a PRBS sig- Next, we determined the minimum received powers (dBm)
nal. The effect of the Mach-Zehnder modulator extinction ratio which were required to achieve our desired goal Q factors of 7, using
variation on the Q-factor and BER was also studied. Obviously, the set-up shown in Fig. 4(a). The transmitter power was selected
the signal is degraded much by the decrease in the extinction again to be around 0 dBm, and the bit rate 10 Gb/s. In the PIN, the
ratio. sensitivity was shown to be about −19.9 dBm when the optimized
attenuation was 17.25 dB. In the APD, the average sensitivity was
4. Results and discussion higher, at around -24 dBm, when the optimized attenuation was
21.5 dB. The eye diagrams in this case are shown in Fig. 4(b) and
Results of the simulations are observed by connecting the (c). They show the maximum Q factor with respect to the received
oscilloscope; BER and eye diagram analyzer to 3R-Regenerator power for PIN and APD, respectively.
(Fig. 2). Fig. 5(a) and (b) shows how the signal is degraded by both the
A major component of the transmitter is the modulator. Here, thermal and shot noise in the PIN photodetector. The former affects
we use a Mach–Zehnder arrangement having two inputs, one the performance of the PIN quite heavily, while the shot noise
for the laser diode and the other for the data. It converts the affects it only lightly.

Fig. 4. (a) Set-up for goal attainment attenuation optimizations of Q-factor of 7, (b) eye from PIN Q = 6.9 and (c) eye from APD Q = 6.9.
1496 O. Kharraz, D. Forsyth / Optik 124 (2013) 1493–1498

Fig. 6. (a) APD with only shot noise; Q factor = 9.13 dB and (b) APD with both thermal
Fig. 5. (a) PIN with only shot noise; Q factor = 9.7 dB and (b) PIN with both thermal and shot noise; Q factor = 14.4 dB.
and shot noise; Q factor = 2.56 dB.

are illustrated in Fig. 9. We can readily conclude that the APD is

much more tolerable to a wide range of bandwidth.
For the APD case, the situation is a little bit different, as we can Obviously, there are some situations where APD fails to be supe-
see from Fig. 6(a) and (b). The performance is affected greatly by rior over PIN, i.e. where shot noise is high. This may occur in many
the shot noise. practical applications where the gain factor of APD becomes a dis-
Practically, the PIN yields a higher signal-to-noise ratio (SNR) advantage. The PIN performs better as the thermal noise effects
than the APD, due to the APD shot noise. However, as far as the shot decrease. Whereas, in the APD, as shot dominates over thermal
noise is concerned, the APD generates a higher SNR than the PIN at noise its performance becomes worse (Fig. 10). When the dark cur-
equal optical power source. This improvement in the SNR is due to rent is high, the shot noise dominates the system and the APD is
the APD gain, which multiplies the photocurrent by a multiplication more affected. Specifically, when the dark current is at value of
factor M. 85 ␮A then this can be very practical in Germanium photodetec-
From our data in Fig. 7, photodetector systems incorporating tors, where the dark current is normally around that value. From
APDs have a Q factor higher than their PIN counterparts at a mul- Fig. 10, the cross-over point in our case is around 82 ␮A.
tiplication factor of 3. This is the threshold point at which the shot
noise affects the performance of our APD that much more, so that it
is essential to look for the optimum multiplication factor to get the
best APD gain figure. Hence, an APD multiplication factor of more
than about 8 starts to degrade the system’s performance to grants
the superiority to PIN.
Thermal noise in both the PIN and APD affects their sensitivities.
From Fig. 8(a), it can be seen that as the thermal noise increases,
both sensitivities increase linearly. Moreover, we can readily inves-
tigate the relationship between both parameters in Eqs. (1) and (2).
Fig. 8(b) shows that the PIN is superior when the thermal noise
begins to dominate.
Since a low-pass electrical filter determines a receiver’s band-
width, it is also highly crucial to investigate the dependence of high
speed performance on the bandwidth of APD and PIN. By extracting
the bandwidth parameter of Fig. 2 from the low-pass filter to get
the receiver sensitivity of −17 dBm and Q factor of 7, we found that
APD bears a wider range of bandwidth compared to the PIN. The
APD maintains a good performance at 40 GHz, while the PIN per-
forms well at low bandwidth of 10 GHz. The effects of bandwidth Fig. 7. APD gain with Q factor compared with PIN.
 O. Kharraz, D. Forsyth / Optik 124 (2013) 1493–1498 1497

Fig. 10. Q factors plotted with dark current.

5. Conclusions

In summary, we have compared and analyzed the performances

of PIN and APD optical receivers for use in optical systems. The gain
of the APD photodetector, which results in higher SNR, has made
it more suitable for long haul communications, as a high speed
receiver in high bandwidth applications and bit rates, where the
cost is inevitable. Moreover, it is a good candidate in high-sensitive
network access applications. We found that noise increases with
the bandwidth of both photodetectors, hence it is important to opti-
mize APD gain value. This specific value is required to be equal to
the secondary electronic pre-amplifier input noise. This is the great
advantage of the APD over the PIN – where its gain does not partic-
ipate in the system noise. Indeed, it helps to minimize and cancel
out other generated noise in the system.
The conventional PIN is much easier and cheaper to fabricate,
Fig. 8. Thermal noise vs. receiver sensitivity (a) and Q factor (b) for both APD and and highly reliable all the time. In fact, the APD fails to be supe-
PIN. rior candidate in certain cases: e.g. when the dark current is high
enough to cause high shot noise. The APD is more sensitive to dark
current than the PIN because it has more layers, so the depletion
region will be smaller. In summary, the PIN is more suited to low-
bandwidth and short distance applications.
However, the overall general superiority of the APD perfor-
mance comes at a price, not just in monetary terms, but in the many
trade-offs occurring within the device mechanisms. In conclusion,
useful information on situations where the PIN may be just good
enough for the task in hand can be made available to designers from
simulations like those done here.

Acknowledgments

The authors wish to acknowledge the administration at the Uni-

versiti Teknologi Malaysia (UTM), especially the Human Resources
Department for their financial support. We would also like to show
our appreciation towards the Photonics Technology Center (PTC)
and Faculty of Electrical Engineering at UTM for providing us with
the software and facilities to accomplish this work.

References

[1] J.C. Campbell, Recent advances in telecommunications avalanche photodiodes,

J. Lightwave Technol. 25 (1) (Jan 2007) 109–121.
[2] G. Wang, T. Tokumitsu, I. Hanawa, K. Sato, M. Kobayashi, Analysis of high
speed p–i–n photodiode S-parameters by a novel small-signal equivalent cir-
cuit model, IEEE Microw Wireless Compon. Lett. 12 (10) (Oct 2002) 378–380.
[3] Hui Nie, High performance, low-cost PIN, APD receivers in fiber optical net-
works and FTTx applications. Wireless and Optical Communications, 2005, in:
Fig. 9. Q factors plotted with photodiode bandwidths. International Conference on 14th Annual WOCC, 22–23 April 2005, p. 94;
S. Jung, M. Moon, H.J. Kim, H. Park, S. Ryu, A simulation study of silicon avalanche
1498 O. Kharraz, D. Forsyth / Optik 124 (2013) 1493–1498

photodiodes, in: IEEE on Nuclear Science Symposium Conference Record, vol. International Conference on Numerical Simulation of Optoelectronic Devices,
2, October 29 2006–November 1 2006, pp. 1064–1067. 1–4 September 2008, pp. 37–38.
[4] J. Rue, M. Itzler, N. Agrawal, S. Bay, W. Sherry, High performance 10 Gb/s PIN [9] A.H. You, S.L. Tan, T.L. Lim, P.L. Cheang, Multiplication gain and excess noise
and APD optical receivers, in: Proceedings of the 49th Electronic Components factor in double heterojunction avalanche photodiodes, in: IEEE Interna-
and Technology Conference, 1999, pp. 207–215. tional Conference on Semiconductor Electronics, 25–27 November 2008, pp.
[5] L. Gomez-Rojas, N.J. Gomes, X. Wang, P.A. Davies, D. Wake, High performance 259–262.
optical receiver using a PIN photodiode and amplifier for operation in the [10] D. Hu, B. Xu, X. Zhou, F. Guo, Optimization and analysis on several impact factors
millimeter-wave region, in: 30th European Microwave Conference, October of high-gain separate absorption, grading, charge and multiplication avalanche
2000, pp. 1–3. photodiodes, in: Symposium on Photonics and Optoelectronics, 14–16 August
[6] W. Chen, S. Liu, PIN avalanche photodiodes model for circuit simulation, IEEE 2009, pp. 1–4.
J. Quant. Electron. 32 (December (12)) (1996) 2105–2111. [11] K.F. Li, D.S. Ong, J.P.R. David, P.N. Robson, R.C. Tozer, G.J. Rees, R. Grey, Avalanche
[7] A. Ebberg, R. Bauknecht, M. Bittner, M. Grumm, M. Bitter, High performance photodiode (APD) noise dependence on avalanche region width, in: 5th Device
optical receiver module for 10 Gbit/s applications with low cost potential, Elec- Research Conference Digest, 23–25 June 1997, pp. 170–171.
tron. Lett. 36 (April (8)) (2000) 741–742. [12] G.P. Agrawal, Fiber-optic Communication Systems, John Wiley and Sons, Inc.,
[8] W. Lei, F.M. Guo, W. Lu, D.Y. Xiong, Z.Q. Zhu, J.H. Chu, Based simulation of 2002.
high gain and low breakdown voltage InGaAs/InP avalanche photodiode, in: