These highly integrated circuits can help fortify the design of a tunable oscillator by providing isolation from output-power variations as a function of supply and temperature.
Base-station designers are faced with the daunting task of driving down product costs while achieving superior levels of radio performance. One of the most obvious solutions is to employ greater degrees of circuit integration throughout the receive and transmit lineups. The MAX9987/90 family of local-oscillator (LO) buffers/splitters have been specifically designed with this singular goal in mind. In addition, these these components help to improve the overall performance of the LO drive lineup by offering exceptional output-power variance control, isolation, and noise performance—all critical parameters for optimizing passive mixer designs. An overview of typical LO drive circuits follows, along with a description of how the MAX9987/90 family of parts can be optimized for virtually any LO drive application.
A typical LO lineup requires a buffer amplifier to isolate and drive a passive mixer from a voltage-controlled oscillator (VCO) with relatively low output power. Most passive mixers require drive levels ranging from +14 to +20 dBm. However, simple amplification of the VCO signal is not sufficient for optimizing mixer performance. A key requirement for any LO lineup is to maintain a nominal drive level despite temperature, voltage, and VCO drive variations. Failure to contain LO drive variance can lead to degradations in receiver (Rx) sensitivity and third-order-intercept-point (IP3) performance. For the transmit chain, LO drive variance can also impact output power, IP3, and corresponding adjacent-channel power ratio (ACPR).
Most of the variance encountered within an LO drive circuit is directly related to the VCO's output characteristics. The output power of a VCO can typically vary by as much as ±3 dB, depending upon temperature, frequency, and part-to-part differences. Table 1 provides a detailed look at each of these variance contributors. As can be seen from Table 1, VCO part-to-part differences are the most significant contributors to power variance in the LO drive circuit. However, a good LO drive circuit attempts to address all of the variances with one common solution.
Discrete solutions are typically used in today's high power diversity and single branch LO drive circuits (Fig. 1.) The overwhelming majority of these circuits use at least one amplifier that is driven hard into saturation. By pushing the amplifier(s) into compression, a relatively stable level of output drive is provided regardless of variations in input power, temperature, and supply voltage.
However, the drawback of these discrete solutions is that they are relatively bulky—especially when a designer uses lumped or distributed Wilkinson splitters as the representation of the power divider. Also, the parts count can be significant as noted in Table 2.
As shown in Fig, 1, the MAX9987/88 replaces four discrete amplifiers, a passive splitter and coupler, plus dozens of biasing components. This high degree of integration enables a designer to reduce the overall size of the LO drive circuitry by a factor of 2.5 times, while simultaneously cutting the parts count by as much as 41 percent. Table 2 provides a more detailed look at how well these integrated devices stack up against their discrete-component equivalents.
These components are ideal for cellular/Global System for Mobile Communications (GSM)/digital-cellular-system (DCS)/personal-communications-services (PCS) and Universal Mobile Telecommunications System (UMTS) base-station applications where dual, high-level LO drives are required for diversity transmit and receive lineups. Single-output versions, namely the MAX9989/90, can be similarly used for single-branch systems. At the heart of each device is the on-chip buffer circuit, which provides output-to-input isolation of 40 dB to prevent LO pulling, and output-to-output isolation of 30 dB to reduce branch-to-branch interference. As an added benefit, the MAX9987/90 feature an on-board PLL amplifier which provides a convenient +3-dBm output for prescaler feedback. Each member of the MAX9987/90 family comes in a remarkably small, pin-compatible 5 × 5-mm QFN-20 package.
The MAX9987/90 series of LO buffers/splitters were specifically designed to provide LO drive control of better than ±1 dB over a wide range of temperatures (−40 to +85ºC), input-power levels (±3 dB), and supply voltages (5 ± 0.25 V), all without the use of external calibration or control. Figure 2 depicts the basic relationship between output power and input power for the MAX9987/90's typical application circuit. As shown, the device is capable of providing ±1-dB variance control over a relatively large input-power swing of ±3 dB. The designer is tasked with providing a nominal level of input power for the MAX9987/90. After this nominal level is determined, all variance control—including part-to-part variations—is handled directly by the integrated circuit (IC).
The MAX9987/90 offers a nominal output level of +17 dBm (Fig. 2). Note, however, that the MAX9987/90 also possess a feature whereby the designer can precision-set the output power levels through the implementation of four external biasing resistors. In effect, these resistors determine the degree of biasing on the chip's internal amplifiers. The specified output power levels are adjustable from +14 to +20 dBm, depending upon the chosen resistor settings (Fig. 3).
For the majority of LO drive applications, ±1 dB of variance control is more than sufficient for optimizing mixer performance. However, in certain cases, a designer may find it desirable to limit this variance to even lower limits.
The technique presented below caters to such an application by extending the capabilities of the MAX9987/90 to yield nominal output levels that are accurate to within 0.05 dB. Such adjustments allow the designer to calibrate out part-to-part differences which lead to variances in input drive level. In the case of a typical LO drive circuit, the VCO's part-to-part variations of ±2 dB can be eliminated altogether. All that remains is a very manageable delta of less than ±0.5 dB over temperature and voltage, centered around the calibrated value of output power.
The calibration process is facilitated by the MAX9987/90's programmable output-power feature. Instead of using fixed resistors, it is possible to control the output power directly with a voltage applied to the reference pins. This type of control provides the option of modifying the output power at any time, and lends itself to dynamic adjustments which can be implemented during a calibration test. The proposed method, shown in Fig. 4, allows for testing and setting of the output power level in a production environment. Other possible implementations are suggested toward the end of this article.
For demonstrative purposes, the goal of the design shown in Fig. 4 is to set (with high accuracy) an output power level of +17 dBm on the MAX9989. Other output power levels are possible, depending upon the level of bias applied to the reference pin. In addition, this technique can be applied to any member of the MAX9987/90 family.
For the bench test of this circuit implementation, a constant +7-dBm RF source at 900 MHz was used to drive the MAX9989. Figure 5 shows the measured transfer function of RF output versus digital-to-analog converter (DAC) voltage for this particular circuit. Laboratory measurements of this circuit reveal that the output power of the MAX9989 can be fine-tuned with 0.05-dB accuracy. It should be noted that, for this particular circuit, a nominal level of +17 dBm (delivered to the load) corresponds to a DAC voltage of 320 mV. The coupler used on the calibration port taps off −7.4 dBm of power from the MAX9989, and hence the designer needs to drive the bias on the device a bit higher to compensate for the 0.3-dB coupler loss.
The following lists some key findings from the implementation presented in Fig. 4. If a 10-b DAC is used to set a voltage between 0 and 1.25 V, the control resolution will be:
Resolution = (voltage range)/2number of bits = (1.25 V)/210 = 1.2 mV.
The control slope is approximately 0.02 dB/mV, so the resolution is effectively 0.02 dB (more than sufficient for the 0.05-dB control target). It is possible to use an 8-b DAC to provide sufficient resolution, depending on the goal of the application. For measurement simplicity, the plots shown in Fig. 5 were generated using a DAC integrated within the MAX1407 (a data-acquisition system on a chip). Other stand-alone DACs, such as the two-channel, three-wire interface, 8-b MAX519, are suitable for this type of control as well.
The MAX1407 (in Fig. 4) has an internal reference at 1.25 V which is used for Maxim's internal testing. If another DAC is used, it is possible to use the MAX9989's internal 1.5-V reference source (available on pin 5 of the device).
A 1200-MHz coaxial lowpass filter was used to reject any second- or higher-order harmonic components that might be generated from the saturated amplifier. When measuring the load RF power directly, a lowpass filter should be used as well.
Further enhancements to the circuit in Fig. 4 are also possible; four additional possibilities include:
Regardless of how they are used, the MAX9987/90 are ideal parts for providing high levels of LO drive with exceptional output-power variance control. By using these devices, base-station designers can dramatically improve the performance of their LO drive circuit while only using a fraction of their current component count and board footprint.
ACKNOWLEDGMENTS
The authors would like to thank Elliott Simons and Mike Mellor for their technical insight and general support.
