The combination of a new design topology and proven half-micron enhancement-mode PHEMT semiconductor process help extend the performance of a Darlington amplifier.
Wideband Darlington amplifiers-offer versatile gain for a variety of wireless and wire-line applications, including in base stations, fiber-optic transceivers, cable-television (CATV) systems, and measurement systems. A Darlington feedback amplifier features multi-decade coverage, can be realizedin a small size, and is simple to use with few additional external components. What follows is a report on a Darlington amplifier that makes use an innovative bias topology and enhancement-mode pseudomorphic highelectronmobility-transistor (E-PHEMT) device process to achieve new levels of dynamic range for both 3.3- and 5.0-V designs.
Until now, silicon-bipolar junction transistors, silicon-germanium (SiGe) heterojunction bipolar transistors (HBTs), and indium-gallium-phosphide (InGaP) HBTs have been the predominant technologies used for RF Darlington amplifiers. Traditional bipolar and HBT Darlington amplifiers offer high linearity; their positive turn-on voltages allow ease of use through positive-single-supply operation. InGaP HBT-based Darlington amplifiers have gained in popularity for their high third-order-intercept (IP3) performance through cellular bands (about 2 GHz).1 Above 2 GHz, however, the IP3 performance rolls off fairly rapidly for these InGaP solutions, creating opportunities for other advanced semiconductor solutions, such as PHEMTs. Compared to InGaP HBTs, PHEMTS offer lower noise figures and more gain at higher frequencies.
High-linearity PHEMTs are usually depletion-mode devices that require negative gate bias voltage. However, the negative gate bias voltage has precluded the use of these devices in RF Darlington feedback amplifier topologies. Because of the need for a single positive supply in cellular handsets, enhancementmode PHEMTS are used in that application. But even with the availability of e-mode PHEMTs and e-mode/d-mode PHEMTs,2 stable monolithic biasing using PHEMT active topologies is still challenging due to threshold variations over process and temperature which is an order of magnitude greater than bipolar HBT technologies.
By combining enhancement-mode PHEMT devices with a new Darlington active bias topology, it has been possible to realize self-biased Darlington amplifiers. The approach provides amplifiers with robust performance over temperature, power supply and process variations, and achieves a higher IP3-bandwidth product than for InGaP Darlington solutions.
Figure 1a illustrates the conventional Darlington feedback amplifier topology with an enhancement-mode PHEMT implementation. The positive turn-on voltage of an enhancement-mode PHEMT, which can be as high as 0.5 to 0.6 V at moderate current densities, enables resistive biasing of the Darlington amplifier. An external resistor (Rdc) in combination with resistive feedback sets the quiescent bias current of the amplifier. For robust biasing that is insensitive to supply variations, a voltage drop of about 2 to 3 V is required across Rdc, forcing an inefficient use of a higher supply voltage. To compound matters, the threshold voltage of PHEMTs can vary much greater than their bipolar/HBT counterparts due to process and temperature variations. To achieve the best performance from a Darlington PHEMT implementation, a less process-, temperature-, and supply-sensitive bias scheme is needed.
Figure 1b presents the schematic diagram of the new bias scheme in which a mirror transistor, M3, is used to set the bias for the Darlington amplifier's output transistor. The current of the mirror is set through internal feedback resistor Rfb. This eliminates the need for the external set resistor of the conventional case (Fig.1a) and enables the Darlington amplifier to operate directly from the supply voltage. For example, a conventional 8- or 5-V Darlington may be operated at a reduced supply of 5 or 3.3 V, respectively. Without the 2- to 3-V overhead voltage across an external Rdc resistor, an amplifier's overall efficiency may be improved by as much as 40 percent. A filter is embedded in the amplifier to isolate and reduce the effects of bias circuit on RF performance. This new technique, for which a patent has been obtained, 3 helps maintain robust bias over temperature and voltage supply variations. The current-temperature sensitivity of this new bias scheme is summarized by:
A23 = the ratio of the area of amplifier transistor M2 to mirror transistor M3.
A similar relationship may be derived for a conventional resistively biased Darlington:
Plugging in typical values for the parameters of Eqs. 1 and 2, it is clear that the new bias approach can reduce the bias current sensitivity with respect to temperature by as much as an order of magnitude. By inspection, it can also be concluded that the sensitivity with respect to Vgs process threshold variation is also reduced by the same factor. These claims are supported by comparing the current-voltage (I-V) curves of a 5-V conventional resistive-bias Darlington amplifier to a 3.3-V activebias PHEMT Darlington amplifier (Fig. 2). The tightly spaced new 3.3-V design illustrates lower temperature sensitivity compared to the loosely spaced 5-V conventional resistive design, even while operating at a lower 3.3-V supply. The quiescent bias is maintained to within ±3 percent and ±11 percent for the new active and conventional resistive Darlington biases, respectively; this is nearly a factor of 4 improvement for the active case. Also, the similar slopes of the curves illustrate that voltage supply insensitivity can be maintained at lower supplies with the new bias scheme.
To demonstrate the third-orderintercept-bandwidth (IP3-BW) and low-voltage capability of the new PHEMT RF feedback Darlington amplifiers, both 5-V and 3.3-V enhancement-mode Darlington amplifiers were designed and fabricated using the 0.5-µm enhancementdepletion-mode (ED) PHEMT process2 from TriQuint Semiconductor (www.triquint.com) using the new Darlington self-bias technique.3 The process features an enhancement-mode PHEMT device transition frequency of 30 GHz and gate-drain breakdown voltage (BVgd) of greater than 14 V. Depletion-mode PHEMTs were also available in this process but were not used in these designs.
The table gives a summary of the measured results for both 5- and 3.3-V self-biased designs. The 5-V amplifier was designed for high gain and broadband high-IP3 performance while the 3.3-V amplifier was designed for wideband gain and to demonstrate robustness of the self-bias topology at lower supply voltages. The 5-V design achieved a gain of 20.2 dB and useable bandwidth from 100 MHz to 10 GHz. The 3-dB bandwidth is 4 GHz, although return loss was better than 12 dB across the full 10-GHz band. A high IP3 of +34.2 dBm was achieved at 2 GHz. The IP3 was maintained at better than +33 dBm across the 10-GHz band, using 90 mA bias current and a 5-V supply. The advantage of the new bias topology is that the supply voltage is directly applied to the drain of the output transistor making efficient use of the power supply available, unlike the conventional resistive self-biased Darlington discussed earlier. The output power at 1-dB compression (P1dB) is +21.8 dBm at 2 GHz. The use of PHEMT technology also allows reasonable low noise figure of ~ 3 dB.
To demonstrate the enhancementmode PHEMT gain-bandwidth capability as well as the capability of the new self-bias topology to robustly operate from a lower supply, a 3.3-V selfbiased Darlington was also designed. The table shows that the 3.3-V design achieves a much broader bandwidth of operation with an actual 3-dB bandwidth of 100 MHz to 14 GHz. The nominal gain is 12.3 dB and the return loss is better than 10 dB across the wide bandwidth. The IP3 performance is +33.1 dBm while the P1dB performance for this 3.3-V amplifier is +16 dBm, both at 2 GHz. This is achieved with a low supply voltage of 3.3 V and 75 mA current.
The impact of using a lower 3.3-V supply is minor for IP3 but was much greater for P1dB because the amplifier P1dB point (output voltage swing) as determined from the loadline is voltage limited. The noise figure is 3.5 dB, slightly higher than for the 5-V design, due to the use of a smaller feedback resistor required to achieve the broader gain bandwidth
Both 3.3- and 5-V self-biased amplifiers were fully characterized on-chip over temperature and supply variations. Figure 3 gives the gain-temperature characteristics of both the 3.3-V (from 0° to +85°C) and 5-V (from −20° to +85°C) designs. At +25°C, the nominal gain of the 5-V design is 20.2 dB with a 3-dB bandwidth of 4 GHz. The gain of the 5-V design varied by 0.81 dB at 2 GHz over the −20° to +85°C temperature range. This corresponds to a gain-temperature coefficient of −0.01 dB/C, which is typical of the performance, obtained from a more cumbersome externally regulated PHEMT MMIC amplifier solution requiring a separate regulator chip. The bias current changed from 90.0 mA at −20°C to 87.7 mA at +85°C for a total variation of 2.6 percent. The excellent Iddversus-temperature and gain-versus-temperature performance is attributed to the active bias scheme and should be indicative of similar current regulation over process threshold variation. The 3.3-V design also shows good temperature performance. At +25°C, the nominal gain is 12.3 dB with a 14-GHz bandwidth. Over the 0° to +85°C temperature range, the gain varied by 0.62 dB at 2 GHz. The corresponding-bias currents were 72.5 mA at 0°C, 72.0 mA at +25°C, and 72.0 mA at +85°C for a total variation of 0.7 percent. This variation in bias is even more impressive in view of the lower supply voltage. The corresponding gain-temperature coefficient is −0.007 dB/°C. This is exceptional for a fully monolithic PHEMT solution operating from a low 3.3-V supply. The reason for the different temperature range for both chips was the intermittent capability of the temperature chuck that was limited to condensation at the lower temperatures.
In addition, the supply voltage insensitivity of the lower 3.3-V design was also demonstrated-over a ±7.5-percent supply variation-from 3.05 to 3.55 V. The gain variation was 0.12 dB at 2 GHz and only 0.06 dB at 14 GHz over the supply voltage range. The gain insensitivity to supply is attributed to the high voltage overhead about the Rfb resistor that allows similar insensitivity as a higher supply conventional design using a large external resistor.
The IP3 and P1dB performance were also characterized over temperature for the high-IP3 5-V active-self-biased Darlington amplifier design. The IP3-versus-frequency characteristics over temperature are shown in Fig. 4. At +25°C, the IP3 is better than +33.6 dBm across a 10-GHz band with a peak IP3 of +35.9 dBm at 2.4 GHz. Over the −20° to +85°C temperature range, the IP3 varies by less than 1.9 dB across the band, and more typically exhibits variations of 1 to 1.5 dB. At +25°C, the P1dB is better than +22 dBm to 4 GHz and then falls off to +18.5 dBm at 10 GHz. Over the −20° to +85°C temperature range, the P1dB is typically better than +21.5 dBm to 6 GHz. The total P1dB variation over temperature is no more than 1 dB over this band. Although the active current bias was designed for stable current over temperature, the IP3 and P1dB temperature variations are reasonably good. The active bias may be used to further compensate IP3 and P1dB performance by incorporating a positive bias current slope with temperature. This would be straight forward by using the temperature coefficient of a Schottky diode with the current mirror topology. The IP3 and P1dB temperature behavior of the 3.3-V active bias Darlington amplifier is similar and omitted for brevity.
The motivation for using enhancementmode PHEMTs over existing high-linearity InGaP Darlington solutions is twofold. First, enhancement-mode PHEMTs allow the Darlington amplifier topology to achieve higher IP3-bandwidths for the same gain, supply voltage, current, and device transition frequency (about 30 GHz) as illustrated in Fig. 5. Figure 5 shows that the InGaP HBT active bias Darlington achieves higher IP3 at lower frequencies as expected from the high linearity of InGaP HBTs; however, the IP3 rapidly rolls off as frequency increases beyond about 2 GHz. The PHEMT Darlington, on the other hand, starts off with slightly less IP3 at low frequencies but can maintain IP3 flatness across a much wider band through 10 GHz. This is in spite of the fact that the gain bandwidth of the InGaP solution is 6 GHz1 which is actually higher than the 4-GHz 3-dB bandwidth of the PHEMT amplifier design. The resulting IP3-BW product of the InGaP HBT design is 14.3 W-GHz while the PHEMT obtains twice this with a value of 26.9 W-GHz. The PHEMT design achieves roughly a factor of 2 times better IP3-bandwidth product than the InGaP HBT implementation. The enhanced IP3-BW is believed to be partially due to the lower input and output capacitances of PHEMT for the same bias current and device transition frequency, which alters feedback phase as frequency increases. The assumption is that as the feedback phase departs from ideal 180-deg. negative feedback, the third-order intermodulation (IM3) products begin to vectorially add instead of cancel, therefore producing degradation of IP3 with increasing frequency. This explanation has only been partially verified by simulations because there are many other factors in the design that influence the IP3 behavior over frequency.
The second advantage of using enhancement-mode PHEMTs over InGaP HBTs is the lower threshold voltage of enhancement-mode PHEMTs (about 0.35 to 6 V) compared to the Vbe turn-on voltage of InGaP HBTs (about 1.35 V). This is because the lower turn-on voltage of enhancement-mode PHEMTs allow a lower Darlington knee voltage and better efficiency compared to InGaP HBT solutions. This is illustrated in Fig. 6, which compares the I-V characteristics of both enhancement-mode PHEMT and InGaP HBT Darlington pair devices. Figure 6 shows that the lower threshold voltage of enhancement-mode PHEMTs results in a low Darlington knee voltage of about 1 V compared to an InGaP Darlington knee voltage of about 2 V. This means that the PHEMTs can produce higher voltage swings from a lower supply voltage than the InGaP HBT Darlington amplifier, and this ultimately results in higher output power efficiency and higher linearity. Because the input bias voltage of the Darlington is about 2Vbe, the InGaP HBT Darlington is essentially inoperable from a supply voltage below 2.8 V (~2Vbe) while the enhancement-mode PHEMT Darlington-will be able to operate from a supply-as low as 1.5 V, yet another distinct advantage of the enhancement-mode PHEMT Darlington for low-voltage applications.
In conclusion, robust enhancementmode PHEMT Darlington RF feedback amplifier performance has been demonstrated using a new Darlington bias topology. Test results show low sensitivity to temperature and supply variations for Idd, gain, IP3, and P1dB. Enhanced IP3-BW performance over an equivalent InGaP design suggests that enhancement-mode PHEMT technology is more suitable for higherfrequency applications with the Darlington feedback topology for microwave frequencies above 2 GHz.
The author would like to recognize the support of Sirenza colleagues: T. Sellas for MMIC characterization; C. Kitani for layout; and J. Yee, K. Tan, S. Warren, and T. Hon for engineering support. Also thanks are due to the foundry team at TriQuint Semiconductor (www.triquint.com).