Technical Insight
Component selection is vital for APD receiver performance
There is far more to an optical receiver than just the APD or pin diode - consideration of the surrounding components is crucial to good receiver performance, writes Joe Boisvert.
Fiber-optic receivers for telecommunication applications often employ avalanche photodiodes (APDs) as the photodetector to improve receiver sensitivity and increase link reach. A longer link reach translates into fewer transponders and reduced system deployment costs. The cost savings can be particularly significant in metropolitan-area networks, where many thousands of transponders will be deployed. Even a 1 dB improvement in receiver sensitivity can have a large impact on the overall system cost, which makes sensitive APD receivers market-driven products that are currently the focus of intense interest. While APD-based receivers offer the potential for higher sensitivity than their pin-based cousins, there are several careful choices that an APD receiver designer must consider before setting out to bring a new product to market.
Figure 1 is a block diagram of a typical receiver portion of a fiber-optic transponder. The receiver is denoted by the enclosed box, and typically contains at least a photodetector and a transimpedance amplifier (TIA). The TIA converts the output current from the photodetector into a voltage signal for the subsequent amplification and signal processing required to convert the analog signal to a digital data stream. Additional functionality such as power-supply conditioning (filter capacitor and resistor) may also be included in the receiver, depending on subsystem requirements. How much additional functionality to include in the receiver plays a key role in determining the designer s cost versus value-added trade space, and can have a significant impact on package assembly complexity.
APD versus pin receiversWith a pin diode as the photodetector, the input optical power from the fiber is converted to a current with a response efficiency of approximately 0.8 A/W at 1550 nm. An APD provides first-stage amplification through an internal gain mechanism known as avalanche multiplication. An APD will have a typical response of about 8 A/W at its optimal operating point near a gain of 10 (the responsivity simply scales with the internal gain). The photodetector output current establishes the signal level. The noise level is set by the combination of noise from the photodetector, TIA and photodetector bias supply.
Typically, the TIA noise dominates at telecommunication frequencies greater than about 1 GHz, so increased optical response from an APD leads to a larger signal-to-noise (S/N) ratio and improved sensitivity of about 6-8 dB over a pin receiver. The improved signal response does not come without problems, however, because avalanche gain amplification introduces additional noise known as excess noise (F), which increases with the internal APD gain. Increasing the APD gain increases S/N until at some gain the S/N ratio reaches a maximum, and any further increase in APD gain has a deleterious effect. At that point receiver sensitivity is optimized. Because the noise floor at telecommunication bandwidths is strongly dependent on the TIA noise, the specific improvements in sensitivity will hinge not only on the designer s choice of an APD, but also on the selection of the TIA and on passive bias supply filtering components.
Choosing an APDMost telecommunication APDs use separate absorption and multiplication regions with InGaAs as the light absorbing layer, and another wider bandgap layer, such as InP or InAlAs, as the material in which avalanche multiplication takes place. The use of two different semiconductors for absorption and multiplication functions is essential in this technology, because the electric fields that would be required to achieve reasonable gains in InGaAs alone result in very large tunneling dark currents that degrade receiver sensitivity. The wider bandgap InP or InAlAs layer is able to sustain much higher electric fields without producing significant dark currents.
The choice of an InP versus InAlAs APD is influenced by several factors. InP APDs have been available for several years, and have been fielded in many networking transponders. However, because of the way that multiplication is achieved, the excess noise factor is not as low as can be achieved in InAlAs APDs (F ~ 5-6 in InP; F ~ 4 in InAlAs at an internal gain (M) of 10). InAlAs APDs have only recently started to be used in telecom APD receivers, but have already set world record sensitivities and appear to be poised to supplant InP APDs in very-high-sensitivity applications. Table 1 lists some of the important figures of merit for these two APD material technologies.
Another parameter that the receiver designer must consider is the bias required for the APD to operate. Because InGaAs-based APDs operate close to the avalanche breakdown electric field, they typically require 30-60 V of bias that must be supplied though an external, filtered bias supply. As discussed in the following section, the bias supply noise as well as passive component size, performance and cost are all functions of the APD operating bias - lower is better. A good rule of thumb is to keep the APD bias to less than 50 V, because this is a break point in the maximum DC voltage ratings and sizes of filter capacitors and resistors. Spectrolab InAlAs APDs are designed to operate at 30-40 V. At a gain of 10 these APDs have about 65% quantum efficiency, resulting in a responsivity of 8 A/W.
Because APDs work with high internal fields, they can be sensitive to changes in operating temperature. When holding the APD bias constant, an increase in temperature will decrease the avalanche gain. A temperature compensation circuit on the APD bias supply is typically used if the required operating temperature range is large enough to significantly impact on receiver performance. The change in breakdown voltage with temperature is about a factor of 10 less in InAlAs APDs (10 mV/ºC versus 100 mV/ºC for InP), which greatly relaxes the performance requirements of the temperature control circuitry. Figure 2 shows the gain versus temperature characteristics of Spectrolab InAlAs APDs.
Figure 1 is a block diagram of a typical receiver portion of a fiber-optic transponder. The receiver is denoted by the enclosed box, and typically contains at least a photodetector and a transimpedance amplifier (TIA). The TIA converts the output current from the photodetector into a voltage signal for the subsequent amplification and signal processing required to convert the analog signal to a digital data stream. Additional functionality such as power-supply conditioning (filter capacitor and resistor) may also be included in the receiver, depending on subsystem requirements. How much additional functionality to include in the receiver plays a key role in determining the designer s cost versus value-added trade space, and can have a significant impact on package assembly complexity.
APD versus pin receiversWith a pin diode as the photodetector, the input optical power from the fiber is converted to a current with a response efficiency of approximately 0.8 A/W at 1550 nm. An APD provides first-stage amplification through an internal gain mechanism known as avalanche multiplication. An APD will have a typical response of about 8 A/W at its optimal operating point near a gain of 10 (the responsivity simply scales with the internal gain). The photodetector output current establishes the signal level. The noise level is set by the combination of noise from the photodetector, TIA and photodetector bias supply.
Typically, the TIA noise dominates at telecommunication frequencies greater than about 1 GHz, so increased optical response from an APD leads to a larger signal-to-noise (S/N) ratio and improved sensitivity of about 6-8 dB over a pin receiver. The improved signal response does not come without problems, however, because avalanche gain amplification introduces additional noise known as excess noise (F), which increases with the internal APD gain. Increasing the APD gain increases S/N until at some gain the S/N ratio reaches a maximum, and any further increase in APD gain has a deleterious effect. At that point receiver sensitivity is optimized. Because the noise floor at telecommunication bandwidths is strongly dependent on the TIA noise, the specific improvements in sensitivity will hinge not only on the designer s choice of an APD, but also on the selection of the TIA and on passive bias supply filtering components.
Choosing an APDMost telecommunication APDs use separate absorption and multiplication regions with InGaAs as the light absorbing layer, and another wider bandgap layer, such as InP or InAlAs, as the material in which avalanche multiplication takes place. The use of two different semiconductors for absorption and multiplication functions is essential in this technology, because the electric fields that would be required to achieve reasonable gains in InGaAs alone result in very large tunneling dark currents that degrade receiver sensitivity. The wider bandgap InP or InAlAs layer is able to sustain much higher electric fields without producing significant dark currents.
The choice of an InP versus InAlAs APD is influenced by several factors. InP APDs have been available for several years, and have been fielded in many networking transponders. However, because of the way that multiplication is achieved, the excess noise factor is not as low as can be achieved in InAlAs APDs (F ~ 5-6 in InP; F ~ 4 in InAlAs at an internal gain (M) of 10). InAlAs APDs have only recently started to be used in telecom APD receivers, but have already set world record sensitivities and appear to be poised to supplant InP APDs in very-high-sensitivity applications. Table 1 lists some of the important figures of merit for these two APD material technologies.
Another parameter that the receiver designer must consider is the bias required for the APD to operate. Because InGaAs-based APDs operate close to the avalanche breakdown electric field, they typically require 30-60 V of bias that must be supplied though an external, filtered bias supply. As discussed in the following section, the bias supply noise as well as passive component size, performance and cost are all functions of the APD operating bias - lower is better. A good rule of thumb is to keep the APD bias to less than 50 V, because this is a break point in the maximum DC voltage ratings and sizes of filter capacitors and resistors. Spectrolab InAlAs APDs are designed to operate at 30-40 V. At a gain of 10 these APDs have about 65% quantum efficiency, resulting in a responsivity of 8 A/W.
Because APDs work with high internal fields, they can be sensitive to changes in operating temperature. When holding the APD bias constant, an increase in temperature will decrease the avalanche gain. A temperature compensation circuit on the APD bias supply is typically used if the required operating temperature range is large enough to significantly impact on receiver performance. The change in breakdown voltage with temperature is about a factor of 10 less in InAlAs APDs (10 mV/ºC versus 100 mV/ºC for InP), which greatly relaxes the performance requirements of the temperature control circuitry. Figure 2 shows the gain versus temperature characteristics of Spectrolab InAlAs APDs.
