610 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 56, NO. 2, APRIL2007 Accuracy Improvement of Real-Time Load-Pull Measurements Valeria Teppati, Member,IEEE, Andrea Ferrero, Member,IEEE, Daniela Parena, and Umberto Pisani Abstract—This paper describes a new procedure aimed to im- prove the effectiveness of real-time load-pull calibration. Load- pull measurement accuracy is strongly affected by calibration residual uncertainty. The novel methodology reduces this uncer- tainty contribution by means of error terms optimization. The proposed method has been tested with simulations and applied to actual measurement data. Considerable improvements have been achieved. Index Terms—Directional couplers, microwave devices, micro- wave measurements, microwave phase shifters, microwave power amplifiers, tuners, uncertainty. I. I NTRODUCTION L OAD-PULL systems offer a powerful tool for nonlinear device characterization and design by measuring the large- signal performances of a device under test (DUT) with different load impedance values [1], [2]. The enhancement of load-pull measurement accuracy is a must, especially when dealing with devices having high input and output reflection coefficients. Load-pull systems can use passive tuners or active loads [3], [4]; in both cases, impedance values and reflection coefficients are measured by a vector network analyzer (VNA), directly (real-time systems), or by means of tuner precharacterization. Let us consider a VNA-based real-time load-pull system, such as the one sketched in Fig. 1. The two reflectometers take the incident and reflected waves at ports 1 and 2 and provide them to the VNA for measurement. The reflection coefficient seen at port 2 (Γ L ) can be varied with an active (e.g., active loop) or passive (e.g., tuner) load tuning system. This system allows for real-time measurements of Γ L , Γ in , input and output power, and gain of an active DUT. The system is generally calibrated in two steps: 1) A tradi- tional two-port calibration is carried out at the DUT reference planes, and 2) a power meter measurement is performed for absolute power-level calibration [4], [5]. As a consequence, the overall uncertainty strongly depends on the accuracy of power-level measurements. This issue has been extensively investigated in [6] and [7]. In [6], a method for the evaluation of active real-time load-pull uncertainty for power, gain, and power-added efficiency (PAE) was given, whereas in [7], real- time and nonreal-time load-pull uncertainties were compared. This paper is focused on the development of an optimiza- tion methodology to improve real-time load-pull measurement accuracy. In Section II, we define the problem and describe the opti- mization technique. Section III shows the preliminary simula- Manuscript received July 11, 2006; revised October 30, 2006. The authors are with the Dipartimento di Elettronica, Politecnico di Torino, 29-10129 Torino, Italy (e-mail: valeria.teppati@polito.it). Digital Object Identifier 10.1109/TIM.2007.891049 Fig. 1. Simplified scheme of a VNA-based real-time load-pull system. tions, which are aimed to test the proposed method, whereas Section IV presents the measurement procedure, the experi- mental setup, and the measurement results. Eventually, some conclusions are briefly discussed. II. PROBLEM DEFINITION The real-time load-pull test set sketched in Fig. 1 allows for real-time measurement of input power P in , output power P out , gain G, and PAE of a DUT. In particular, operating gain is defined as G = P out P in = |b 2 | 2 -|a 2 | 2 |a 1 | 2 -|b 1 | 2 = |b 2 | 2 (1 -|Γ L | 2 ) |a 1 | 2 (1 -|Γ in | 2 ) (1) whereas the available gain is defined as G av = P out P av = |b 2 | 2 ( 1 -|Γ L | 2 )( 1 -|Γ S | 2 ) |a 1 | 2 |1 - Γ in Γ S | 2 (2) where a i and b i (i =1, 2) are the incident and reflected waves, and Γ in = b 1 /a 1 and Γ L = a 2 /b 2 are the reflection coefficients at the input and output ports, respectively. In [6], uncertainty contributions due to VNA measurement repeatability, power-level uncertainty, and connection repeat- ability were taken into account. In this paper, we will consider only the residual calibration uncertainty contributions that are independent from the power- level measurement, regardless of their origin. The basic idea of this paper is to enhance the measurement accuracy by optimizing the calibration coefficients (calset here- after) and exploiting a load-pull map of a thru device versus Γ L at single frequency. The gain of a thru device is equal to 0 dB, by definition, since a 1 = b 2 and b 1 = a 2 and should not vary with Γ L . If affected by uncertainty, instead, |G| dramatically increases with |Γ L |. This effect has been demonstrated in [6]–[8], and in this paper, it is exploited to optimize the calset. 0018-9456/$25.00 © 2007 IEEE