Novel Topology Supports Wideband Passive Mixers

Marc Goldfarb, Russell Martin, and Ed Balboni

Choices for frequency mixers capable of wide spurious-free dynamic range (SFDR) and broadband performance are often limited. Active mixers can support the bandwidth, but with limited SFDR. A passive mixer followed by an intermediate-frequency (IF) amplifier can achieve wide SFDR, but typically over limited bandwidth. Fortunately, a new mixer format from Analog Devicesessentially, broadband passive mixers with programmable performance parametersallows users to trade off different characteristics, such as noise figure and SFDR, as needed by an application. The first two members of the new mixer family, single-mixer model ADL5811 and dual-mixer model ADL5812, offer RF and local oscillator (LO) frequency ranges from about 700 to 2700 MHz. The two radio-frequency integrated circuits (RFICs) are supplied in 32- and 40-pin chip-scale packages, respectively, for ease of integration in a wide range of communications applications.

With increasing use of digital modulation formats in modern communications systems, it is often desirable to implement digital-predistortion (DPD) techniques in a system's transmit signal chain to achieve high third-order-intercept (IP3) performance over a wide bandwidth. Whether an active or passive mixer, it should also minimize leakage of local oscillator (LO) signals from the mixer's RF input port to prevent unwanted radiation from a communication system's antenna port.

An RF mixer is often characterized by two key parameters: SFDR and input IP3 (IIP3) performance. SFDR is a measure of the performance of the mixer limited on the one extreme by thermal noise floor, which is the sum of the noise floor of a single load and the single-sideband (SSB) noise floor of the mixer (kTB + NF), and on the other extreme by the mixer's IIP3. It is represented as a single number in dB as:

SFDR = (2/3)(IIP3 noise floor) = (2/3)(IIP3 kTB NF_SSB)

where:

k = Boltzmann's constant (approx. 1.38 x 10²³ J/K),

T = the absolute temperature of the load (K), and

B = the measurement bandwidth (Hz).

An RF mixer is also characterized by its blocking dynamic range (BDR), which is defined by the mixer's input 1-dB compression point (IP_1dB) at the upper extreme and the thermal noise on the lower extreme. It is also expressed in dB as:

BDR = (IP_1dB noise floor) = (IP_1dB kTB NF_SSB)

Note that in this expression for BDR, IP_1dB may be replaced by the power at which a blocker reduces the system noise figure by some amount, such as 3 dB. This value is dependent upon the amount of gain in the system preceding the mixer, so it is not used here as a metric for the mixer itself.

Mixers for communications purposes are also evaluated in terms of SSB noise figure under blocking conditions. This is the increase in apparent mixer noise figure as a result of cross modulation, occurring when a large blocking signal is in close proximity to the desired RF input signal and mixes with the phase-noise sidebands of the LO signal. When the blocker level is high enough or the LO amplifier has high phase noise, this can dramatically increase the system noise figure. Because the blocker is an in-band signal, it cannot be filtered. As a result, the system relies on the capabilities of the mixer to set the performance limits in this case. Allocation of gain in a communications receiver signal chain determines not only the SSB noise figure and IP3 requirements of the mixers, but must also be carefully considered with respect to the expected blocker profile.

Previously, designers had a choice between the very high SFDR of a narrowband passive mixer and the moderate SFDR of a wideband active mixer. But three technical advances have aided in the development of a wideband passive mixer design that covers 700 to 2700 MHz while also achieving input IP3 of +25 dBm, SSB noise figure of 11 dB, and conversion power gain of 7 dB. Additionally, this mixer design is well matched on both single-ended RF and LO inputs, and exhibits very low LO leakage at the RF and IF ports.

The first advance involved the development of a limiting LO amplifier capable of generating a high-voltage, near-square-waveform signal over a wide range of frequencies (Fig. 1). In a conventional sine wave LO amplifier, an inductive load in the form of a tuned transformer is placed in between the output of an amplifier stage and the LO input of a passive FET mixer. The inductance value and turns ratio are determined by the desired operating frequency and voltage swing required for the mixer. The peak voltage swing is determined by the required compression point of the mixer and may be limited by the breakdown of the field-effect-transistor (FET) mixer core in an IC mixer. Additionally, the bandwidth is limited by the parasitic circuit elements of the mixer's gates and the amplifier's output capacitance.

Page Title

The input IP3 in a passive mixer tends to be limited most directly by the rise time of the LO waveform in the voltage range where the FETs in an IC mixer are most rapidly switching. Since this waveform is a sine wave, and the amplitude is known, the derivative of the voltage waveform is also in the form of a cosine wave whose maximum rise time has a fixed relationship of ωA, where ω is the angular frequency (2π/T) and A is the peak amplitude of the LO signal. This type of amplifier typically has only a few hundred megahertz of bandwidth for a 2-GHz center frequencyperhaps a 25% fractional bandwidth. In contrast, a wideband mixer may operate over two octaves, with more than 130% fractional bandwidth.

It is difficult to introduce wideband tuning into the resonant tank LO sine wave amplifier since the voltage swing in the tank is large. A tuning or switching element with very high breakdown voltage generally has a poor on-to-off capacitance ratio. This ratio limits the bandwidth over which this type of amplifier can be tuned to a fractional bandwidth of generally less than 50%.

To achieve the necessary bandwidth, a wideband approach was introduced for the LO amplifier in these broadband passive FET mixers. Since it is acceptable (and desirable) for the amplifier to limit its output voltage, a CMOS inverter-based topology was applied to leverage the limiting and self-biasing characteristics of a CMOS inverter when creating the LO amplifier for these new passive mixers. The LO amplifier boasts exceptionally good rise time (on the order of 15% of the period of the highest operating frequency) to provide sharp edges, thus enabling high mixer IIP3. The only difficulty with this type of amplifier is that the design is typically limited by the charging current (i) and the capacitance (C) of the mixer cores according to i = C (dV/dt), where C is the total capacitance introduced by both the mixer gates and the amplifier transistors (Fig. 2).

While it is not possible to entirely overcome this physical limitation, it is possible to reuse current intelligently to minimize the average current required to operate the amplifier. Additionally, this form of amplifier naturally reduces the DC drawn from the power supply at lower frequencies as tr/T decreases. Because these structures are inherently single ended, it is also necessary to carefully construct the amplifier chain to ensure that the output waveform applied to the mixer is well balanced, minimizing generation of second-order distortion products.

The second significant technical issue involves creating a well-balanced RF signal to apply to the FET mixer (Fig. 3). In narrowband designs, an RF balun consisting of a magnetic or transmission line transformer is employed to achieve a balanced RF signal with low loss. While these transformers typically have low loss when properly implemented (on the order of 1dB), they exhibit only moderate bandwidth, with perhaps one octave of usable bandwidth. In this case, it is possible to implement a tuned, resonant balun structure to provide excellent RF balance, with a small incremental loss over a fixed balun. At the band center of the tuning range, an additional incremental loss of only 0.5 dB is observed compared to a fixed balun.

The third technical issue associated with a passive mixer and IF amplifier structure involves the broadband nature of the mixer core and processing of unwanted sidebands. While it is desirable for the core to translate desired signals over a wide bandwidth, unwanted sidebands may also be included. For example, in a mixer with 800-MHz RF input and 700-MHz LO, the difference IF output is at 100 MHz (800 700 MHz), but with an equally strong sim IF tone at 1500 MHz (800 + 700 MHz). The composite waveform at the mixer's IF output contains both tones superimposed. With both signals, the peak input to the following IF amplifier may result in amplifier compression, unless the amplitude of the unwanted sideband can be reduced into a load.

In a narrowband passive mixer, this is often accomplished by a simple highpass/lowpass resisitive-capacitive (RC) diplexer or other lowpass structure. It is important to recognize that while simple filtering may result in improved compression point, the bilateral nature of the passive FET mixer means that mismatching one of the output tones of the mixer will result in that impedance being Fourier transformed back to the input, resulting in an imperfect match and possible mismatch loss in the mixer. This can be solved through the use of tuned filter networks that provide the proper sum termination as a function of the RF and LO frequencies.

These three enhancements are instrumental to the performance of the ADL5812 dual mixer (Fig. 4) and the ADL5811 single mixer RFICs (Fig. 5). The ADL5812 dual mixer and IF amplifier combine a wideband LO amplifier, a tunable RF balun, and tunable sum termination filter in an RFIC housed in a 40-pin, 6-mm lead-frame chip-scale package (LFCSP). It is intended for the receiver chain in wideband wireless infrastructure applications or in software-defined radios (SDRs). All of the aforementioned functions are controlled via a three-wire serial port interface (SPI) to minimize the control interconnections. In addition to the RF balun and LPF settings, performance can be further optimized by adjusting the DC bias voltage to the passive mixer gates, as well as power-down functions.

Page Title

SPI Port Control

For optimum performance in a narrowband passive mixer, it is necessary to tune the center frequency of the LO amplifier, as well as the RF balun and sum frequency filtering, depending on whether the LO frequency is above or below the intended RFi.e., for upper- or lower-sideband operation. The combination of the wideband LO, along with the SPI port controllable balun and sum frequency filter, permits upper and lower sideband operation from the same part by simply reprogramming the SPI port. Changing frequencies and entire bands can be similarly implemented via the SPI port, as no external impedance matching components are required. To minimize power dissipation, each channel of the ADL5812 can be enabled or disabled independent of the other. For DPD transmit observation receivers or nondiversity applications, the single-channel ADL5811 is suitable for a single receiver chain in a multichannel or multiband platform. It offers all of the functionality of the ADL5812 in a single-channel, 32-pin 5 x 5 mm LFCSP.

This level of SPI programmability in either mixer provides unprecedented flexibility to the system designer. For example, if all of the settings of the RF balun were overlaid for either mixer, the resulting gain vs. RF frequency curve looks like Fig. 6. The balun setting can be optimized for each range of frequencies, along with the LPF and other settings, the resulting gain vs. frequency curve is now less than 0.5dB over a 4:1 bandwidth (Fig. 7). Similarly SSB noise figure and input IP3 can be optimized over a broad range of frequencies (Fig. 8).

The capability to trade off SSB NF for IIP3 performance also enables a system designer to reprogram the mixer as dictated by the circumstances: for example, for optimized noise figure in a low-distortion environment, and for optimized IIP3 in a high-distortion environment. Compared to a narrowband passive mixer, the IIP3 response is very flat over a wide bandwidth.

The mixers' square-wave-limiting LO amplifiers have some additional advantages. The lack of a large resonant transformer structure eliminates a coupling path for LO energy to leak to the IF and RF ports. In a narrowband passive mixer, the large sinusoidal LO swing can easily result in about -13 dBm LO leakage at the IF port. Since the narrowband LO limiting amplifier has generous gain and an internal LO voltage swing of about +27 dBm if referenced to a 50-O system, this is about 40 dB actual isolation. But the wideband passive mixers exhibit a dramatic improvement of more than 25 dB in isolation, resulting in low LO leakage of only about -35 dBm at each mixer's IF output (Fig. 9). More significant is the difficult-to-filter LO leakage at the RF port, which is typically -40 dBm beyond 2300 MHz.

In any communications systems design, mixer performance in the presence of a large blocker at the RF port must be considered. While it is not possible to eliminate the effects of a large blocker near the desired RF input, the effects can be minimized by good LO amplifier design to minimize the phase noise of the amplifier without introducing excessive gain.

At the same time, it is necessary to provide sufficient gain to fully limit the output at the minimum amount of LO input power. Figure 10 and Figure 11 show the performance of the new mixers with blocking inputs, as well as demonstrate the IIP3 performance as a function of LO input power at a range of LO frequencies. It is clear that the amplifier is driven into amplitude limiting for even modest LO input levels to 3 GHz without excessive noise figure under blocking.

As the input level increases, more stages of the LO amplifier enter limiting so blocking noise figure can be substantially reduced by increasing the LO drive above 0 dBm. Figure 10 and Figure 11 demonstrate the improvement that is possible by increasing the LO drive to the mixer. For operation where blocker signals are less of a concern, LO power can be reduced to 0 dBm or below without sacrificing intermodulation performance (Fig. 11).

The ADL5811 single and ADL5812 dual passive mixers deliver a wide bandwidth for a variety of applications, without sacrificing SFDR performance. They are supplied in standard chip-scale packages for ease of integration, with wide RF and LO ranges to simplify system design. Both the single and dual passive mixers support a variety of applications from existing receivers and DPD architectures to software-defined radios (SDRs), as well as many emerging communications applications.

Marc Goldfarb, Design Engineer
Russell Martin, Product Engineer
Ed Balboni, Design Engineering Manager
Analog Devices, Inc.
804 Woburn St.
Wilmington, MA 01887
(781) 329-4700

Acknowledgments

Bringing a revolutionary product to market requires an extensive and diverse group of technical professionals. The authors take great pleasure in acknowledging the efforts of the following experts: Kurt Fletcher spent endless hours crafting a circuit layout that would ensure excellent balance and performance. Srivatsan Parthasarthy was extensively involved in the design and qualification of the new wideband ESD structures. Rui Liu made extensive measurements and suggestions to streamline the applications PCB for these novel mixers. Jeff Vorderer developed production test capability for these new passive mixers. Simar Handa and Borko Eleta were instrumental in making extensive measurements. Wendy Dutile, Ed Gorzynski, and Chris Norcross spent many hours crafting intricate RF test boards for these mixers. Marianne Geraci coordinated all the business aspects of bringing this project together successfully.