Ultra-fast Photodiodes: The Missing Link Between Wireless And Optical Networks
Wireless-over-fibre systems are needed to increase data rates to the speed demanded by next-generation wireless networks. One key component in these systems is the near-ballistic uni-traveling carrier photodiode, which can operate at ultra-high speeds with the addition of a p-type charge layer inside the collector, says Jin-Wei Shi from National Central University, Taiwan.
Nowadays, we rarely make a call on our mobile. Instead, we use it to stream YouTube and Facebook videos, watch Internet TV shows, play video games, download and store songs and movies, and take and share pictures. Unfortunately, all these types of data transmission consume radio bandwidth – lots of it. The reality is that we are in the midst of a famine of the radio spectrum.
The most useful wireless communication occurs between 0.3 to 3.5 GHz, a spectral band that enables compact antenna, produces low propagation loss and allows good penetration through buildings. However, the maximum capacity in these wireless channels is typically just several Mbits/s, way short of the bandwidth requirements for next-generation Gbit/s wireless networks.
The simplest way to rescue us from this famine is to boost up wireless carrier frequencies. Several currently ‘unlicensed’ (sub-) millimetre wave (MMW) bands are attracting attention for this purpose, including the Vband (60 GHz), D-band (120 GHz), and bands beyond 300 GHz. However, (sub-) MMW signals suffer substantial propagation loss, both in free-space and in transmission lines. This weakness, and their inherent straight-line path of propagation, hamper connections and synchronization between different parts of the communication system.
A promising way to overcome this problem is the MMWover- fibre (MOF) technique (see Figure 1 for two examples of MMW over-fibre communication systems). This approach has three key strengths: It employs flexible optical fibre, rather than rigid MMW waveguides; propagation losses fall by several orders of magnitude (in MMW waveguides the loss is typically 0.05 dB/cm at 100 GHz, compared with typically 0.1 dB/km for MMW carrier-wave signal distribution); and network coverage is increased significantly. However, there is a tremendous difference in frequency between the 1.55 μm signals used in optical networks, which have a frequency of 190 THz, and MMW signals that have wavelengths of a few millimetres and frequencies of tens or hundreds of gigahertz.
The first example provided in Figure 1 shows setups for additional electrical-to-optical and optical-to-electrical conversion in the central office and base station of an MOF system. The electrical-to-optical conversion in the central office usually employs 1.55 μm lasers and optical modulators. To let the optical wave have the desired MMW envelope for remote distribution through a low-loss fibre to several base stations, the electrical MMW is used for the input signal. This approach eliminates the huge propagation loss of the MMW signal that occurs along an electrical transmission line or in free-space. This optical signal is returned to the electrical domain at the base station. Down-conversion of the incoming optical signal extracts the MMW envelope, which is radiated over the last-mile to the end-user through an antenna.
A high-speed photodiode that can operate at the (sub-) MMW regime is the key component for converting the optical MMW envelope into an electrical MMW signal. Ideally this device also has a high output saturation power, so that it can generate high MMW power under intense optical power injection.
When these high-power MMW PDs are used in partnership with a high-power, erbium-doped fibre amplifier (EDFA), it is possible to minimize the burden imposed on the limited gain, noise, and saturation power performance of the next-stage MMW amplifier for wireless data transmission. What’s more, the optical signal processing technique can deliver a photo-generated MMW signal of superior quality to that generated from a chain of MMW mixers and amplifiers.
It is also possible to build an MOF system that does not require a MMW photodiode (see Figure 1(b)). In this case, the optical fibre links are used to distribute the optical data signals from the central office to each base station. In the base station, incoming optical data is transformed into electrical MMW signals for radiation to the end-user with an oscillator, mixer, amplifier, and antenna operating in the MMW bands.
Figure 1. A MMW-over-fibre communication system promises to improve wireless data transfer rates. Such a system can feature a common optical local oscillator MMW source shared by: (a) different base stations and (b) different electrical local oscillator MMW sources installed in each base station
With this approach, the photodiode only converts lowfrequency optical data; there is no optical MMW signal in this system. However, this benefit has to be weighed against more expensive and complex base stations, which must be synchronized and share the same optical MMW signal. Greater expense stems from the high-cost MMW ICs, which include the elements mentioned above, such as a mixer, oscillator, and amplifier for the MMW bands.
In addition, it is challenging to synchronize the different MMW oscillators at different base stations. This may be an issue for mobile users roaming among different base stations. Given this complexity, it is clearly better to build an MOF system that incorporates a MMW photodiode.
This device can be integrated with a MMW antenna to form a photonic transmitter (PT) at the base station of a high-performance MMW photonic-wireless link that also features a high-quality optical MMW source in the central office.
To create such a system, groups around the world have been developing of high-power, ultra-fast photodiodes. This effort had yielded several photodiode architectures, which have been trialled in MOF wireless communication systems deliver data rates exceeding 10 Gbit/s.
Photodiodes serving this application ideally excel in both speed and output power. One way to enhance the photodiode’s speed is to reduce its parasitic capacitance, which is possible by trimming the device’s absorption volume. However, when the absorption volume falls below about 1 μm3, the density of photogenerated free carriers is very high, and this increases the space-charge field and photocurrent density, which screen the external applied bias field. The upshot: A severe degradation in electrical bandwidth due to a reduction in the drift velocity of the photo-generated carriers (see Figure 2 (a) for an illustration of this point).
Figure 2. (a) With a single photodiode, intense light injection produces a high-output, photocurrent-induced voltage on the load,which has the opposite polarity to the bias voltage. This reduces the net electric field in the active layer of the device, thereby slowing down the carrier drift-velocity and speed performance. (b) Splitting the optical power between an array of photodiodes that are connected in parallel increases the output saturation current, but severely degrades the RC-limited bandwidth. (c) This restriction is lifted with distributed (travelling-wave) connections,which ensure matching of the velocity and phase of the injected optical wave and the photo-generated carrier wave
Researchers have developed two approaches that can produce an increase in power and electrical bandwidth. These methods are based on replacing a single photodiode with several smaller ones, or turning to a superior epitaxial structure.
With the first option, high power of the optical input can be shared between several miniaturized ultra-highspeed photodiodes. The power generated by each of them can then be combined with a low-loss electrical transmission line. This approach, known as the distributed or traveling-wave photodetector (TWPD), was first used in the era of the vacuum tube to improve the bandwidth performance of the vacuum tube amplifier. Parallel connections are the simplest way to combine the photocurrent from these small photodiodes. The major downside of this approach is a serious degradation of the RC-limited bandwidth, which stems from the hike in junction capacitance. But this can be avoided with a distributed structure (Figure 2 c). In this case, the MMW signal generated from each photodiode is coherently combined, leading to minimal broadening and distortion of the resultant signal. In an ideal case, the maximum bandwidth is as high as the bandwidth of each single photodiode.
Modifying the traditional epi-layer structure of the p-i-n photodiode is the other popular approach to improving the device’s power and speed. This can combat saturation of the device’s speed by quickening the rate at which photo-generated carriers are drained from the photodiode’s active layers. With this approach, space-charge density falls, driving down the induced electric field.
One of the best ways to do this is to speed up the carrier drift-velocity inside the photodiode. This is possible by switching to a uni-traveling carrier photodiode (UTC-PD) design (see Figure 3). This swaps an intrinsic photo-absorption layer with a p-type doped epilayer, and modifies carrier transport: Photo-generated holes relax directly into the p-contact metal without drift, diffusion or accumulation in the photo-absorption layer, and only electrons remain in the active carrier. This promises to greatly enhance carrier velocity, by eliminating holes, which are far slower at moving through the device than electrons.
Figure 3. (a) The low mobility of holes limits the speed of conventional p-i-n photodetectors. (b) The UTC photodiode overcomes the limitation of low hole mobility by getting these carriers to relax in the p-contact metal without drift, diffusion or accumulation in the photo-absorption layer. However, if the electron is subjected to a high electric field, it can easily be swept to another electron potential valley with a heavier effective mass and a slower drift-velocity (c) With the NBUTC photodiode, a p-type charge layer is inserted inside the collector to control the distribution of the electrical field and minimize intervalley scattering
Thanks to their superior transport properties, verticalilluminated type, waveguide type, and distributed type UTC-PDs can all deliver a high power-bandwidth product. However, there are still some weaknesses when the operating frequency of the UTC-PD is in the MMW (sub-THz) regime. In this frequency regime, the internal electron transient time, rather than the RClimited bandwidth, tends to limit device bandwidth. In these devices the electron drift-velocity can saturate under a high external applied electric field due to intervalley scattering (see Figure 3). To prevent this, a very small reverse bias voltage, such as -0.75 V, can be applied to the thin collector layer (~200 nm thick). This is necessary to maintain the overshoot drift-velocity of the electron and maximise the speed performance of the UTC-PD.
The downside of using a very small reverse bias voltage to realise a high drift-velocity in a sub-MMW UTC-PD is that it tends to be screened by the output AC voltage produced by this device (see Figure 2a). A small load resistance, such as less than 25 Ω can minimize the amplitude of the output AC voltage, but this addition slashes the output power of the photodiode, which normally operates under a standard 50 Ω load.
To overcome such problems, our group at National Central University in Jhungli, Taiwan, has developed a near-ballistic uni-traveling carrier photodiode (NBUTCPD). This outperforms a UTC-PD in the sub-MMW regime, in terms of the product of saturation current and bandwidth. One key feature of our photodiode is its ptype charge layer inside the collector layer, which can properly control the distribution of the electrical field and minimize intervalley scattering. Near-ballistic transport of electrons in the moderately high optimum reverse bias regime (-2 to -3 V) minimises external saturation. We note that a similar concept has also been reported in the near-ballistic collection HBT by researchers at the Electrical Communication Labs of NTT, Kanagawa.
The backbone of future wireless communication networks operating at Gbit/s speeds will be the optical fibre based network, which connects different base stations, each distributing large volumes of data. Every base station in the MMW wireless network will need many pico-cells (or femto-cells) to radiate the photogenerated MMW power from the photodiode to the lastmile for the end-user.
Each photodiode is integrated with an antenna to form a photonic transmitter that radiates the photo-generated MMW power. The high speeds and powers of UTC-PDs and NBUTC-PDs enable the construction of phototransmitters that can produce a range of radiation patterns, operate at 0.1 to 1 THz, and deliver output powers that are typically 20 dB higher than those associated with photo-transmitters based on conventional photodiodes. By pairing the UTC-PD based photo-transmitter with an end-fire taper slot antenna or a broadside patch antenna, researchers have shown that it is possible to excite WR-10, WR-08, and WR-03 rectangular waveguide-based horn antenna (see Figure 4). This has enabled the demonstration of 10 Gbit/s and 16 Gbit/s line-of-sight data transmission using centre frequencies of 120 GHz and 300 GHz by use of the WR-08 and WR-03 waveguides based photonic transmitter, respectively.
Figure 4. (a) Conceptual diagrams of a broadside patch antenna and an end-fire tapered slot antenna for rectangular waveguide (WR) excitations; and (b) conceptual cross-sectional view of UTC-PD based photomixers with integrated patch antenna and tapered slot antenna for WR-10 and WR-08 waveguide excitations, respectively. (re-printed by permission from A. Ueda, et al., IEEE Trans. Microwave Theory Tech., vol. 51, pp. 1455-1459. May, 2003 © 2003 IEEE and A. Hirata, et al., IEEE Trans. Microwave Theory Tech., vol. 52, pp. 1843-1850,Aug. 2004 © 2004 IEEE)
Our NBUTC-PD based photo-transmitters have a unique advantage over UTC-PD based photo-transmitters: An ultra-fast switching speed. This means that the photogenerated MMW power from the NBUTC-PD can be shut ‘on’ and ‘off’ very fast with a high extinction ratio, simply by switching the bias point. To realise this, we incorporate an additional input port, an intermediatefrequency input, into our device. We use this for injecting high-speed electrical data and for also modulating both the bias point and photo-generated MMW power of our novel photo-diode.
Our device can do more than just convert the incoming optical MMW envelope into electrical data: It can also up-convert incoming electrical data to the MMW regime. The entire device functions as an MMW mixer called the photonic-transmitter-mixer (see Figure 5). The superior modulation speed of our device stems from a combination of bias modulation in only the reverse bias regime, and a high extinction ratio MMW envelope, which predominantly originates from variations in electron drift-velocity in the collector layer under different reverse bias voltages.
Figure 5. (a) An end-fire quasi-yagi antenna for rectangular waveguide (WR) excitation; (b) topview of an NBUTC-PD based photonic transmitter with an integrated quasi-yagi antenna for WR-10 waveguide excitation; and (c) the device during measurement
However, in order to quench the photo-generated MMW power from the UTC-PD, it is necessary to push the device into near forward bias operation. This induces a slow minority carrier injection process, which severely degrades the modulation speed and limits the maximum data transmission rate. By use of the ultra-fast switching characteristic of the NBUTC-PD, we have demonstrated the photonic-transmitter-mixer for an MMW-over-fibre communication system operating at 20 Gbit/s (see Figure 6).
Figure 6. (a) The photonic-transmitter-mixer developed by National Central University, Taiwan, that is used in the last-mile MMW-over-fibre system; and (b) set-up of a photonic wireless linking system for 20 Gbit/s data transmission using a line-by-line pulse shaper with a central frequency around 100 GHz as the photonic MMW source
In such proposed scheme, it is possible to use another optical wavelength to provide the data signal for bias modulation on our photonic-transmitter-mixer, by converting this to electrical 20 Gbit/s data with another low-speed photo-diode.
Getting commercial traction
Today, the biggest opportunities for MMW wireless linkage in consumer electronics exist at 60 GHz, for indoor linking, and at 100 GHz, for outdoor linking. Applications include wireless HDMI broadcasting and linking, and high-speed data transfer between cell phones and Sony’s cyber-shot camera via Transfer Jet, a proximity-based, wireless technology that can transfer images, video, and other files between compatible devices held about an inch apart.
Up until now, photonic technology has not made an impact on these systems. Competition from the allelectronic approach is strong, and the cost of the MMW photonic wireless linking approach must fall. This situation could change, because increasing maturity of CMOS IC technology is making it possible to install MMW CMOS local-oscillator (LO) chips at each base station without using a synchronized LO signal.
However, the MMW wave that is produced above 60 GHz behaves like an optical-wave: Propagation of the signal is in a straight line, and obstacles in the way drive down its intensity. Realising good network coverage for the end-user could require the deployment of numerous remote antenna units, particularly for an outdoor wireless linking system.
The biggest advantage of the photonic technique over the all-electronic approach is the use of a fibre backbone to interconnect and synchronize the units. This minimizes interference and multi-path effects between the units. Other merits of the photonic technology include superior immunity to bad weather conditions, plus an improvement in data processing efficiency. However, these benefits will not be enjoyed unless there is widespread availability of ultra-high speed photodiodes and photonic-transmitter (mixer) modules.
© 2012 Angel Business Communications. Permission required.
M. Lazarus, IEEE Spectrum 47 26 (2010)
J.-W. Shi et. al. NPG Asia Materials 3 41 (2011)
H.-J. Song et. al. IEEE Trans. Terahertz Science Tech. 1 256 (2011)
Nan-Wei Chen et. al. IEEE Trans. Microwave Theory Tech. 59 978 (2011)
H. Ito, et. al. IEEE J. of Sel. Topics in Quantum Electronics 10 709 (2004)
F.-M. Kuo et. al. IEEE/OSA Journal of Lightwave Technology 29 432 (2011)
F.-M. Kuo et. al. IEEE Photonics Journal 3 209 (2011)