要約

This application note describes the features of the MAX2683 3.5GHz upconverter. A typical schematic is given with the matching components for 3.55GHz output, 1.6GHz LO and 350MHz input. The noise figure is ~12.5dB, and the conversion gain is 8.6dB. The integrated circuit (IC) can also be used for a downconverter. Links to tables of S-parameters are provided to aid the design engineer.

Additional Information:

• Wireless Product Line Page
• Quick View Data Sheet
• Applications Technical Support

Introduction

The MAX2683 is a double-balanced active mixer based on the Gilbert cell that is capable of accepting RF inputs up to 3.5GHz and producing IF outputs up to 3.6GHz. It features an adjustable bias control, conversion gain, insensitivity to mismatch, and superior isolation in a very compact format.

This application note presents a brief description of the mixer, design tips, and typical performance features for the MAX2683.

Upconverter Review

A fundamental property of mixers is frequency conversion. This property is put to use in virtually all transmitters. For typical operation, a modulating signal operating at a frequency of f_MOD is injected into one port of the mixer, and a local oscillator (LO) signal at a frequency of f_LO is injected into a second port. The resulting output radio frequency (RF) signal is upconverted to a frequency of f_MOD + f_LO. Frequency conversion results from a multiplication of the modulated f_MOD waveform, cos(f_MOD * t), and LO waveform. From trigonometry, we have the following:

Cos(f_MOD * t) * cos(f_LO * t) = 1/2 cos(f_LO - f_MOD) ± 1/2 cos(f_LO ± f_MOD)

In this ideal multiplication, the output of the mixer contains only signals at frequencies f_LO- f_MOD and f_LO+ f_MOD; that is, the original modulation signal f_MOD and the local signal f_LO are completely suppressed at the upconverter output RF port.

The Gilbert cell active mixer is based on an emitter-coupled-pair amplifier. The operation of this amplifier is best understood by dividing the modulated signal into its common-mode and differential-mode components. The modulated signal enters one side of the pair, while the opposite side is AC-grounded through a capacitor. From symmetry, the common-mode component shifts the current between the two branches and, for small signals, acts as a standard common-emitter amplifier. The MAX2683 employs four cross-coupled devices to the basic amplifier to multiply the modulated signal by ± at the LO rate and to achieve the desired double-balanced mixer characteristics. The combination of these devices with the emitter-coupled pair completes the basic Gilbert cell. As with the modulated signal input, the LO is injected in a single-ended fashion with the opposite side AC-grounded through a capacitor. The positive LO voltages cause the outer set of the device to be on, resulting in a multiplication of the modulated signal by ± at the LO rate, while negative voltages cause the inner pair to be on, also multiplying the modulated signal by ± at the LO rate.

Product Design and Performance Features

The MAX2683 operates from a single +2.7V to +5.5V supply. The device is available in an ultra-small 16-pin TSSOP-EP package with an exposed paddle for the special application up to 3.6GHz. It uses a double-balanced Gilbert cell architecture with single-ended RF and LO inputs and differential open-collector output ports. Differential output ports provide a wideband, flexible interface for either single-ended or differential applications. The MAX2683 features an adjustable bias control, set with an external resistor, that lets the user trade supply current for linearity to optimize system performance. A logic-level control enables an internal frequency doubler on this device, allowing the external local oscillator source to run at full or half frequency. An internal LO filter reduces LO harmonics and spurious mixing. Figure 1 is a simplified block diagram of a MAX2683 application.Figure 2 is a pin description of the MAX2683. The details of consequent performance features are described below.

Figure 1. Simplified block diagram of a MAX2683 application.

Figure 2. Pin description of the MAX2683.

DC Bias

The MAX2683 requires a DC bias. Whereas conventional passive mixers use AC signals to create device conduction, active Gilbert cell mixers require a DC power supply. DC bias is applied to the device in the form of a voltage V_CC. Enough voltage must be applied to cause the transistors in the Gilbert cell to conduct, otherwise the desired switching action will not occur. The minimal voltage required for the mixer to operate is 2.7V. As V_CC is increased, the simple bias scheme allows the transistors to turn on harder. The gain of the mixer increases, as does the compression point. Because change in bias affects linearity, such changes alter the levels of harmonic and spurious signals produced by the mixer. Bias changes also affect f_t of the transistors in the chip and hence the frequency range over which the mixer operates. The linearity and supply current of the MAX2683 is externally programmable with a single resistor, BIAS, from BIAS to GND. A nominal resistor value of 1.2kΩ will set the supply current of 55mA. Decreasing the resistor value improves linearity at the cost of increasing the supply current. Increasing the resistor value decreases the supply current while degrading linearity. Use a resistor value in the range of 820Ω to 2.0kΩ.

Gain

The MAX2683 has conversion gain, so in conventional use the output signal will be at a higher power level than the input signal. Most of the gain in the MAX2683 comes from the emitter-coupled amplifier in the Gilbert cell. The amount of gain achieved will vary with the frequency, the temperature of the operation, the oscillator signal, and the bias level. In order to optimize the gain and the linearity, a properly designed PC board is an essential part of any RF/microwave circuit. Keep RF signal lines as short as possible to reduce losses, radiation, and inductance. Use separate, low inductance vias to the ground plane for each ground pin. For best performance, solder the exposed pad on the button of the device package to the board ground plane. The differential open-collector RFOUT- and RFOUT+ ports require external pullup inductors to V_CC, as well as an output matching network for optimum gain performance. The S-parameters of modulated signal input, LO input, and RF output are shown in Table 2. Designers can refer to that table to develop optimized matching circuits to meet their system specifications.

Oscillator Signals

The MAX2683 requires low oscillator drive levels. In the mixer based on the Gilbert cell, the primary function of the LO signal is to switch the conduction path between the outer and inner transistors of the cross-coupled quad. This requires relatively little power. In general, the spurious response of a Gilbert cell mixer will improve at a lower oscillator drive level. Increasing the LO power to the MAX2683 upconverter will saturate (actually, "quasi-saturate") the transistors of the quad and emitter-coupled pair, and decrease linearity. As the LO drive level is decreased from a nominal characterization value, there is a 5dB to 10dB range over which conversion gain is not affected significantly. When the LO drive level is reduced still further, conversion gain will "roll off." No sinusoidal LO signals can have frequency components at many (harmonically related) frequencies. A typical LO input power is -5dBm at 50Ω matching for the MAX2683.

Operating Frequency Range

The MAX2683 operates over a very wide frequency range. It can operate as a downconverter or an upconverter. The frequency of the modulated signal through the Gilbert cell quad can go up to 3.8GHz. The output frequency range can reach 3.6GHz if a proper output matching network is provided. The MAX2683 features an internal LO frequency doubler that allows the external LO to run at full or half frequency. Running the LO at half frequency has the benefit of reducing unwanted LO leakage through the amplifier to the antenna. An internal LO band pass filter is integrated after the frequency doubler to help reduce LO harmonic content and spurious mixing. To enable the LO frequency doubler, drive ENX2 to a logic low level and connect the half-frequency external LO to the LOX2 port. To disable and bypass the LO frequency doubler and LO filter, drive ENX2 to a logic high level and connect the full-frequency external LO to the LOX1 port. Disabling the LO doubler has the benefit of reducing the supply current by 15mA. The maximum frequency range of LOX1 is up to 3.9GHz, and the LOX2 frequency range is up to 1.95GHz.

Noise Figure

The Gilbert cell structure is not a low-noise configuration. The mixer noise figure comes primarily from the shot noise of the four collector-cross-coupled transistors, the noise of both transistors in the emitter-coupled pair, and the thermal noise of both feedback resistors used with the emitter-coupled pair. The switching action of the LO can affect the mixer noise figure when there is very low input LO power. The typical noise figure of the MAX2683 is close to 12.5dB.

Matching Circuit

Three ports need to be matched properly in order to achieve optimum performance. Table 1 provides a complete S-parameter for three ports, the frequency range covered from 50MHz to 6GHz. The designer can refer to this table to choose the best matching circuit to meet system specifications. This application note includes an application schematic that shows a typical matching circuit for three ports. The input port is matched to 350MHz frequency, the LO port is matched to 1.6GHz, and the output port is matched to 3.55GHz.

Table 1. S-Parameters of the MAX2683

Table 1.1. F Input S₁₁-Parameter

Linearity and Dynamic Range

The Gilbert cell structure does not yield a mixer with a high dynamic range. The following two equations describe linear dynamic range and spurious free dynamic range:

Linear dynamic range = P_1dB - [NF + G + 3dB -114 dBm + 10 log₁₀(BW)]
Spurious free dynamic range = 2/3 [IP₃ - G - NF -10 log₁₀(BW) + 114 dBm]

where P_1dB is the output power of the mixer at 1dB gain compression (in dBm), NF is the noise figure of the mixer (in dB), G is the conversion gain of the mixer (in dB), BW is the bandwidth of the mixer (in dB), and IP₃ is the output third-order interception point (in dB). These equations show that dynamic range is a function of noise figure, output compression point, interception point, and gain. Because the MAX2683 has moderate dB conversion gain, its dynamic range is not very low. The linearity of the MAX2683 is externally programmable with a single resistor. Increasing or decreasing that bias-resistor value will change the linearity performance of the MAX2683. There is a trade-off between linearity and supply current when the value of the bias resistor is changed.

Typical Application

Figure 3 shows a typical upconverter application circuit. As depicted in the figure, the mixer is a multiplier based on the Gilbert cell with an RF input amplifier. Double-balanced mixers such as this offer good port-to-port isolation and low LO free through at the output. RF output port input is configured for differential operation. However, the RF input and LO input can be driven in a single-ended operation. The LO and RF input are 50Ω. The mixer output requires an external matching network to convert high output impedance into lower impedance to meet system requirements. A balun or impedance matching transformer is required for such impedance transform and differential to single-ended transform. The test data of that application circuit is shown in Table 2.

Figure 3. Application Schematic of the MAX2683.

Table 2. Test Data of a MAX2683 Application

(Test conditions: V_CC = +5.0V, R_BIAS = 1.2kΩ, /ENX2\ = GND, f_RFIN = 350MHz, P_RFIN = -20dBm, f_LO = 1600MHz, P_LO = -5dBm; all input/output ports terminated in 50Ω; RFOUT+ and RFOUT- matched to single-ended 50Ω load; T_A = +25°C, unless otherwise noted.)

Notes

Note 1. The device has been fully characterized at this specified frequency range. Operation outside this range is possible but not guaranteed.
Note 2. IIP3 is measured with two tones at 350MHz and 351MHz, -20dBm per tone, f_RFLO = 1.6GHz.
Note 3. IIP2 is measured with f_RFIIN = 350MHz, P_RFIN = -20dBm, f_LO = 1.6GHz.
Note 4. The input match is optimized for best return loss at f_RFIIN = 350MHz.