Both bulk and Epi PIN diodes can satisfy RF-switching applications with their fast response times, long lifetimes, high linearity, and the simplified cascading of switches.
Costly mechanical switches can be justified when used in precision test equipment, such as vector network analyzers. For mass-produced consumer products like cable- or satellite-television (CATV/SATV) delivery systems, however, less-expensive electronic switches are a better fit. These switches are based on either transistors or PIN diodes. The semiconductor switches have no moving parts. As a result, they provide faster response times and longer life spans than their mechanical counterparts.
PIN diodes are often employed as the switching elements in single-pole, single-throw (SPST) and single-pole, multiple-throw configurations. The PIN diode behaves like a current-controlled resistor to all signals higher in frequency than 10 times the cutoff frequency (fc) of the diode, given by:
fc = 1/(2πτ), where τ s the minority carrier lifetime.
The PIN diode's junction resistance, Rj, can be changed from high to low by the application of a forward bias current. In addition, PIN diodes can be used in either series or shunt switching mode. The series-connected switch has an insertion loss, A, corresponding to:
In the shunt connection, the insertion loss becomes1:
where Zo is the characteristic impedance (typically 50 or 75 Ω in RF transmission systems).
The selection of a switch topology requires a trade-off between bandwidth and isolation requirements. Although a series switch has the benefit of low-loss transmission over a very wide frequency range, it also has poorer isolation. Shunt switches are usually used in conjunction with quarter-wave transmission lines, which are inherently narrowband. Compared to the series connection, however, these transmission lines provide superior isolation.
Both test instruments and CATV/SATV equipment demand RF switches that are capable of multi-octave operation without significant signal loss. A multicarrier environment like CATV/SATV imposes a stringent linearity demand on the switch. It must not introduce excessive distortion that could lead to interference between channels. Such interference could result in the degradation of signal quality.
To improve the isolation compared to a single PIN diode, two or more PIN diodes can be connected in series. This series connection also allows the sharing of the same bias current in order to conserve power. The beauty of a two-terminal switching element, such as the PIN diode, lies in the ease with which additional diodes can be cascaded in series. In contrast, the three-terminal transistor requires duplication of the control lines for each additional series switch element.
BULK VS. EPITAXIAL PIN DIODES
Circuit designers need to distinguish between the bulk and epitaxial (Epi) types of PIN diodes. These two different methods of constructing PIN diodes result in significant differences in RF behavior. Consequently, they affect the PIN diodes' suitability for differing applications. For their part, bulk diodes have a low doping density in the substrate. To turn on, they therefore require a high bias current. As a result, the bulk PIN diode is generally unsuitable for portable and other battery-operated applications. Its very thick and pure intrinsic (I) layer produces a long carrier lifetime, of 300 ~ 3000 ns. This long carrier lifetime is an essential parameter for low-distortion performance in both switch and attenuator applications.
In contrast, the I-layer of the Epi diode is highly doped. The Epi diode is well suited for low-current RF switching in current-constrained products. The carrier lifetime is much shorter (τ 5 ~ 300 ns). Unfortunately, this difference makes the epitaxial PIN diode much poorer in linearity than the bulk diode. As the linearity of PIN diodes generally deteriorates at low bias currents, this aspect practically rules out the Epi diodes from consideration as attenuators.
As previously mentioned, also determines the PIN switch's lower-frequency limit of usability due to its relation to the cutoff frequency, fc. Below 10 times the cutoff frequency, the PIN diode no longer behaves like a current-controlled resistor. When
the diode's behavior is unpredictable. It alternates between a current-controlled inductor and capacitor. If the frequency is further lowered to
the PIN junction of the diode acts as a conventional PN junction. In general, the bulk diode's thicker I-layer permits operation at a lower frequency than the Epi diode.
PIN-DIODE MODEL
Parasitic circuit elements (such as unwanted inductances and capacitances), which are inherent in both the diode chip and package, define the limits of switch performance. Within the confines of series-switch configuration, both package and die capacitance (Cp and Cj, respectively) combine to create a gradual degradation of isolation with increasing frequency. The package parasitic inductance, Lp, causes the switch's insertion loss to increase proportionately with frequency (Fig. 1). To improve the PIN-diode performance in the microwave region, manufacturers are constantly inventing smaller packages. These packages work to minimize parasitics. The industry-standard SOT323, SOD-323, and SOD-523 are reflections of the never-ending impetus to produce lower-parasitic PIN-diode parts in low-cost plastic packages.
Unfortunately, the PIN diode cannot be modeled on the ubiquitous workhorse of the computer-aided-design (CAD) world, SPICE. SPICE has no provision for minority carrier lifetime, τ, which is an important PIN-diode parameter. As a workaround, the PIN-diode chip can be modeled as a simple linear circuit consisting of two resistorsone fixed and one variableand one capacitor, as shown in Fig. 1.
The diode chip's current-dependent junction resistance can be approximated by:
where If is the forward bias current in mA.
The parameters A and K are constants that are obtained from curve-fitting the previous equation against the graph of the measured RF resistance versus forward bias current, If. An RF inductance-capacitance-resistance (LCR) bridge (e.g., the Agilent 4286A with the optional external bias accessory) provides a convenient and repeatable way to make this measurement. Rmin and Rmax represent the chip's contact and zero bias resistances, respectively. They are estimated from the minimum and maximum RF resistance shown on the aforementioned graph. In a packaged part, the diode-chip capacitance, Cj, can only be obtained by indirect inference. First, the zero-bias capacitance is measured at a low frequency (typically 1 MHz). As a result, the reactance of the package's parasitic inductance, Lp, becomes negligible. Subsequently, subtracting the package capacitance, Cp, from the measured zero bias capacitance gives Cj.
Usually, the diode manufacturer can provide these statistical data based on large sample sizes. The circuit designer is then spared much effort in extracting the parameters.
For this investigation, a bulk PIN diode (Avago HSMP-386Z, w = 22.5 m, τ= 300 ns, and fc = 0.3 MHz) was compared with an Epi PIN diode (Avago HSMP-389Z, w = 6.5 m, τ = 180 ns, and fc = 0.8 MHz). Although there are bulk diodes with thicker I regions than the above example, their disproportionately higher turn-on current makes them more suited to attenuator rather than switch applications. The diode chips were packaged in similar SOD-323 packages. Electrical connections were made using the same wire-bonding profile. The two different PIN diodes were then tested for insertion loss (IL), isolation (ISO), and third-order intercept (IP3)the parameters that are crucial to operation in wideband RF switching.
To prevent degradation in the signal-to-noise ratio, a series switch should have low insertion loss. This factor is especially critical in a weak-signal-reception system. Below 1 GHz, the diode switch's reactive components do not have a significant impact on the IL. This parameter is primarily dictated by the equivalent series resistance, Rs. Within the boundary of the diode's safe operating limits, it is possible to reduce the IL and Rs by increasing the bias current. The upper operating frequency limit is determined largely by the package's parasitic inductance, which causes a rapid worsening of the IL above 2 GHz.
Generally, the Epi PIN diode has a lower IL than the bulk type at a given value of If. In this comparison, the thin bulk diode needed approximately four times higher bias current than the Epi diode (20 mA vs. 5 mA) to achieve the same IL (Fig. 2 and Fig. 3).
In a series switch, the maximum usable frequency is determined by the decreasing isolation with frequency. The package and junction capacitances (Cp and Cj) allow higher frequencies to bypass the unbiased PIN diode's high-junction resistance. At the low-frequency end, the bulk PIN diode exhibits better isolation than the Epi PIN diode because of the former's higher resistance at zero bias (Fig. 4).
Of course, semiconductor switches do have an Achilles Heel: applications that combine multi-octave bandwidths and multiple carriers (not to mention systems requiring high-power handling). Junction nonlinearity creates even- and odd-order in-band distortion products that are impossible to filter in CATV networks. By comparison, mechanical switches generate inconsequentially small amounts of distortion under similar conditions. The low-band VHF channels (70 to 100 MHz), for example, generate second harmonics that can interfere with the high-band VHF channels (107 to 170 MHz). Generally, PINdiode switches are more linear than transistor-based switches.2
In the forward-biased PIN diode, harmonic and intermodulation distortion are created by the modulation of the I-layer charge density by RF currents. The distortion is influenced by frequency, stored charged, and junction resistance.3 The intercept point (IPn), a widely used figure of merit for switch linearity, is a fictitious point where the linear transfer function intersects with the power of the intermodulation product. Third-order intermodulation products, 2f1f2 and 2f2f1, are considered the most troublesome because they occur close to the desired signal. The third-order intercept point (IP3) of the PIN switch can be analyzed using the method of Caverly and Hiller4:
where f is in MHz, If is in A, and τ is in ns.
In field-effect transistors (FETs), which form a competing switch technology, the distortion characteristic cannot be changed by varying bias. The PIN switch therefore holds an obvious advantage, as it can raise the IP3 substantially with a small increase in bias current. Measurement of the PIN diodes' IP3 showed a reasonably good agreement with the predicted values (Fig. 5).
In conclusion, both bulk and Epi PIN diodes offer different properties that suit them for different niches in RF switching. The RF switches used in CATV systems are fine examples of stringent requirements for wide bandwidth and low cost. Compared to FET- or CMOS-based switches, PIN-diode switches have the advantage of higher linearity, which is especially critical in a multi-carrier environment. They also promise ease in cascading additional switches in series without the need to duplicate control lines. Being two-terminal devices, PIN diodes also are less complicated to model in simulators.
ACKKNOWLEDGEMENTS
The author would like to acknowledge his mentor, Ray Waugh, who continued to guide him even after his retirement. Special thanks are also due to colleagues Chong Wern Ian, Soon Chee Huei, Ho King Pieng, Alan Rixon, Ian Piper, and Robert Brophy of Avago Technologies for their assistance, encouragement, and constructive feedback.
REFERENCES
