Ron Gatzke and Stephen P. Jurgiel also contributed to this article.

Passive FET modulators are commonly used in wireless base-transceiver-station (BTS) equipment. These zero-intermediate-frequency (zero-IF) modulators can provide excellent linearity with low carrier leakage, although the effects of reactive terminations can influence the carrier-leakage performance. Unwanted leakage can result from the loss of local-oscillator (LO) rejection nulling because of the terminating impedances. Fortunately, some simple diplexer circuits can provide sufficient carrier suppression for most wireless applications.

Direct-conversion transmitters rely on a single upconversion step to translate baseband signals to a desired transmit frequency. To meet the needs of modern wireless applications, these modulators should exhibit low transmit noise floor, good linearity, good single-sideband (SSB) rejection, and low LO leakage. Gilbert-cell-based active modulators feature low carrier leakage and operate with low drive levels.1 Unfortunately, such modulators fall short in noise and linearity performance. Passive-FET-based direct modulators have proven to meet the needs of multi-carrier UMTS and GSM base-station transmitters. Baseband ports can be directly driven by digital-to-analog converters (DACs) supporting both the in-phase (I) and quadrature (Q) channels.2,3

In a direct transmitter or receiver, the carrier (LO) is at the same frequency as the radio-frequency (RF) signal. The LO signal leakage at the output of a direct transmitter causes an undesired origin offset in the baseband signal constellation, resulting in a degradation of errorvector-magnitude (EVM) performance. To achieve the high linearity needed for cellular base-station applications, modulators tend to operate with higher levels of LO drive power. Ideally, a balanced (differential) circuit configuration can cancel LO signals and even-order harmonics even at these elevated drive levels. But ideal circuits are difficult to implement, and circuit imbalances (and LO leakage) tend to creep into modulators as a result of semiconductor process limitations and layout constraints.

LO leakage and even-order harmonics cannot be completely suppressed by design. However, introducing small DC offsets at the baseband inputs can cancel or null the leakage at a given LO frequency.4 Still, the effectiveness of such nulling can be hindered by changing operating conditions such as temperature, baseband drive, terminations, and frequency.

The output impedance of the DACs that drive these modulators at the LO frequency and its harmonics is not equal to the characteristic impedance of 50 ohms. Although DC offsets are introduced so that carrier leakage can be cancelled to less than –80 dBm, this is a frequency-sensitive cancellation; it depends on the external interconnections and reactive terminations offered to the LO frequency and its second harmonics.

The signal levels at the fLO and 2fLO frequencies coming from the I and Q ports of a FET-based modulator are finite (Fig. 1). Reflective terminations at the baseband ports return these signals back into the modulator. The reflected 2fLO signals blend with the mixer's internal LO signal and regenerate more leakage at the RF port as indicated in Fig. 2. The reflected LO signal at the IF port remixes with the carrier and generates DC products that perturb the offset voltage. The 2fLO signal also mixes with a sideband, fLO ± fIF, producing an image signal that requires suppression. Terminating these frequencies by properly designed dissipative networks reduces regeneration of carrier and image at the RF port. The RC network with a series inductor serves to absorb the fLO and 2fLO.

Figure 3 shows a plot of LO leakage in dBm versus frequency under different terminations at the baseband port for the MAX2022, a 1.7-to-2.2-GHz direct modulator integrated circuit (IC). The carrier leakage characteristic has a resonance when the baseband port has a 50-ohm termination. When the baseband port is connected to a cable terminated by a low impedance (short circuit) or a high impedance, many peaks and valleys of the leakage characteristics can be observed. This emulates the worst-case scenario when the driving circuit interconnect is routed via long traces; under such conditions, the LO leakage is being regenerated at RF as in Fig. 2. The leakage also becomes very sensitive to LO frequency.

Figure 4 shows the LO leakage of the MAX2021 nulled by DC offset introduced at I and Q ports at the midpoint of the GSM 900 band. In the absence of a diplexer, the LO leakage appears to be quite frequency sensitive. This is due to multiple reflections of fLO and 2fLO regenerating the leakage. With a diplexer in place, the LO leakage can be nulled for one frequency, one temperature, and one input drive level. For the example GSM modulator, nulling was at 940 MHz at room temperature when the modulator has output power of –1 dBm. Under those conditions, the LO leakage was controlled to better than –50 dBm. With the diplexer in place, the LO leakage into RF port is frequency insensitive even under open circuit conditions in the baseband port.

In summary, LO leakage from passive FET modulators tends to be frequency sensitive and termination dependent due to reflected carrier and harmonic signals when operating with large-signal drive levels. Fortunately, a properly designed diplexer can solve these reflection issues and help achieve control of LO leakage for high-frequency modulators.

REFERENCES

  1. U. Karthaus, N. Alomari, G. Bergmann, and H. Schumacher, "High dynamic range, high output power I/Q modulator in 50 GHz ft SiGe technology," IEEE RFIC Symposium Digest, 2004, pp. 539-542.
  2. R. Gatzke, "RF Modulator Enables Multicarrier Transmitters," Microwaves & RF, April 2005, p. 100.
  3. MAX2022, MAX2021, and MAX2023 data sheets, MAXIM Integrated Products, Sunnyvale, CA, Internet: www.maxim-ic.com.
  4. A. DeSimone and E. Nash, "Simplifying Direct-Conversion Tx Paths in Wireless Designs," CommsDesign, October 2002, Internet: www.commsdesign.com.