RF Design Magazine
RSS    Save to Del.icio.us  Digg This


Accurately measuring phase and delay error in direct I/Q modulators
Apr 1, 2006 12:00 PM  By Peter Stroet

Direct-conversion I/Q modulation is an attractive solution for advanced RF transmitters, enabling complex modulation at lower cost. A direct I/Q modulator can generate an RF waveform of arbitrary phase, frequency and amplitude. In a typical application, the I/Q modulator produces a single-sideband output signal. For an ideal modulator with perfect 90° phase shift between the I-mixer and Q-mixer local oscillators, and with no other undesired phase or gain impairments, the modulator output will contain only the desired sideband. In practice, non-idealities arise from other sources of phase error such as baseband I/Q delay mismatch and DAC skew.

The error terms of interest in this analysis are the I/Q skew error due to the baseband signal source (φDGEN), the I/Q baseband skew error due to the modulator (φMOD), and the LO quadrature error (φLO) (Figure 1). Note that time delay errors, such as φDGEN and φMOD, result in a frequency-dependent phase error which can not be compensated by a frequency-independent I/Q phase adjustment. Consequently, they may limit the effectiveness of calibration procedures and the extent of image suppression.

In order to achieve the best image rejection for a broadband communications channel (e.g., W-CDMA), it is important to understand these error sources. This article provides a measurement method to determine the sources of RF and baseband phase error. The method requires a baseband I/Q signal generator with a user-adjustable phase (φGEN) between its I and Q channel outputs. Using this signal source, a sequence of three measurements are made. The individual error terms, φDGEN, φMOD, φLO, are derived from these measurements.

First measurement

As shown and defined in Figure 1, there can be phase differences between the I and the Q path for the baseband generator (φDGEN) and within the I/Q modulator (φMOD). There may also be a phase error φLO between the quadrature LO signals, LOI and LOQ, in the modulator. Consequently, an unwanted upper sideband signal at (ωLO + ωBB) is present at the RF output along with the desired lower sideband signal at (ωLO - ωBB). Using trigonometric identities and small angle approximations, it can be shown that the upper sideband suppression is given by:

RSB=20·log(φLOGEN1DGENMOD)-6.02 [dB] (1)

Note that the phases φ are in radians. The image term can be minimized by adjusting the signal generator phase setting to

φGEN1= -φLODGENMOD. (2)

Second measurement

This configuration differs from that of Figure 1 in that the differential baseband connections to the modulator's I inputs are reversed (Figure 2). The controllable signal generator phase, φGEN2 is adjusted to null the image. The desired signal in this case is the upper sideband signal (ωLO + ωBB), and the image signal is at (ωLO - φBB).

The lower sideband suppression is given by:

RSB=20·log(φLOGEN2DGENMOD)-6.02 [dB] (3)

The image is minimum for:

φGEN2 = φLODGENMOD. (4)

Third measurement

In this configuration (Figure 3), the I and Q differential inputs are exchanged. The controllable signal generator phase, φGEN3 is adjusted to null the image. The desired signal is now the upper sideband frequency component (ωLO + ωBB), and the image is a lower sideband signal at (ωLO - ωBB). The lower sideband suppression is given by:

RSB=20·log(φGEN3DGENLOMOD)-6.02 [dB] (5)

The image is minimum for:

φGEN3 = -φLODGENMOD (6)

Calculation of phase impairments

We can solve equations, (2), (4) and (6) to give:

φLO = (φGEN2 - φGEN1)/2 (7)

φDGEN = -(φGEN2 + φGEN3)/2. (8)

φMOD = (φGEN3 - φGEN1)/2 (9)

Applying the method

As an example, the techniques described here were used to determine the error terms for an LT5528 direct I/Q modulator. A Rohde and Schwarz AMIQ I/Q modulation generator was used as the baseband signal source. For five different LT5528 examples, φLO, φDGEN and φMOD were measured using 5 MHz and 10 MHz baseband frequencies. From the measured phase errors, equivalent delay errors can be calculated. The results in Table 1 show that the baseband signal generator phase error φDGEN is dominant in this setup and is equivalent to a delay error of about 300 ps, compared to the LT5528's baseband phase error φMOD, which is equivalent to a delay of only about 25 ps - 30 ps. In this example, the baseband signal source phase error φDGEN is the major error source. In an actual application, it should be carefully characterized and compensated.

ABOUT THE AUTHOR

Peter Stroet is a design engineer with Linear Technology Corp., Milpitas, Calif.

Table 1. Phase error measurement results of the LT5528.
BOARD BASEBAND FREQUENCY = 5 MHz BASEBAND FREQUENCY = 10 MHz
φLO φMOD τMOD φDGEN τDGEN φLO φMOD τMOD φDGEN τDGEN
UNIT DEGREE DEGREE ps DEGREE ps DEGREE DEGREE ps DEGREE ps
1 1.415 0.035 19.4 -0.55 306 1.445 0.035 9.7 -1.0 278
2 -0.60 0.055 30.6 -0.585 325 -0.65 0.07 19.4 -1.02 283
3 0.245 0.025 13.9 -0.59 328 0.285 0.055 15.3 -0.785 218
4 0.19 0.04 22.2 -0.59 328 0.20 0.06 16.7 -0.86 239
5 0.045 0.055 30.6 -0.61 339 0.08 0.1 27.8 -1.08 300


RSS    Save to Del.icio.us  Digg This




 
Back to Top