Proceedings of the 43rd European Microwave Conference
Low-Jitter Electrooptic Sampling of Active mm-Wave Devices up to 300 GHz M. Jamshidifar, G. Spickermann, H. Schäfer Eberwein, and P. Haring Bolívar Institute of High Frequency and Quantum Electronics University of Siegen, Germany [email protected]
enable on-wafer characterization of mm-wave and sub THz devices with an unprecedented bandwidth.
Abstract—In this work we present an electrooptic sampling system for on-wafer characterization of ultrafast active electronic devices. The system capabilities are demonstrated on a prototype 65 nm CMOS nonlinear transmission line circuit. Device operation and sampling up to 300 GHz is achieved by diminishing the relative time jitter between source and sampler.
A schematic diagram of EOS system is shown in Fig. 1. The setup is configured to sample the electrical signals of the device under test (DUT) which is driven by a microwave signal. In this setup, in order to minimize the relative jitter effect on the measurements we use the laser source itself to generate the driving microwave signal. Therefore the source of electric field and sampler are both driven from the same laser source. For this purpose the femtosecond pulse of the laser photo excites a fast photodiode, generating a full comb of harmonics up to very high frequencies. Filtering one specific harmonic at approximately 10 GHz is used to excite the DUT. The injected signal into the DUT is shifted by an intermediate frequency IF = 50 kHz from the corresponding laser harmonic by using an IQ modulator as frequency up converter. The technique, termed Laser-Master Laser-Slave (LM-LS) configuration is described in more detail in . In this setup,
Femto second Laser
The development of ultra-short (femtosecond) pulsed lasers has enabled to increase the bandwidth and temporal resolution of electrooptic sampling (EOS) systems up to the THz frequency range . EOS being a purely optical sampling technique does not suffer from conventional electronic measurement limitations like transit or RC times, and is therefore recognized as a state of the art method to characterize ultrafast electronic devices far beyond classical electronic operation frequencies. A measurement bandwidth beyond 1THz has been demonstrated for the analysis of “passive” circuits (like e.g. coplanar waveguides), where the device under test is activated by the laser and sampled by the same laser e.g. by photoconductive pulsed excitation . EOS has extremely valuable advantages in areas where electronic instrumentation faces significant challenges. Continuous super-wide measurement bandwidths over hundreds of GHz and near field scanning capabilities of devices with electrooptic probing are attractive features for high frequency designers in order to evaluate their designed circuits.
λ/4 plate λ/2 plate
BS Beam 1
Despite the widely accepted advantages of EOS for passive circuit characterization, there are still significant challenges remaining when the device under test (DUT) is an “active” free-running or electrically activated circuit. In this case, the ultra -short pulse of the sampling laser cannot be used to excite the device. As such devices are typically driven by conventional microwave oscillators limitations become apparent limiting the fundamental capabilities of such an EOS system. Especially, the unavoidable presence of a finite jitter between electrical excitation and laser sampling can severely degrade the ultra-wide measurement bandwidth [2-4].
A RF RF+IF IQ IF
The present paper concentrates on presenting mechanisms to resolve the problems described above. In this paper we demonstrate a low jitter, high sensitivity EOS system which
978-2-87487-031-6 © 2013 EuMA
Balanced photodiode detector
Keywords—Electrooptic sampling; Nonlinear transmission line; mm-wave; THz
Optical path Electrical path BS: Beam Splitter BB: Base Band WP: Wollaston Prism PD: Photo diode IQ: IQ modulator
Fig. 1. Schematic diagram of the experimental setup with LM-LS configuration to minimize the relative jitter between source and sampler. The DUT is being injected by a 10 GHz signal which is extracted from 132th harmonic of the femto- second laser repetition rate (76MHz)
7 -10 Oct 2013, Nuremberg, Germany
shown in Fig. 1, the output signal of the DUT can be sampled with an electrooptic (EO) sampling head (probe) which includes a 50 μm thick LiTaO3 EO crystal attached to a GRIN lens . The goal of using GRIN lens is to reduce the beam size and therefore enhance the high spatial resolution to less than 5 μm. The EO probe which is mounted on XYZ stages  (stages are not shown in Fig. 1 for simplification) and can therefore be freely positioned over the DUT, enabling imaging with measurements of amplitude and phase of the device electric near field. The EO crystal, which is made from LiTaO3, senses the transversal electric field of the DUT based on the Pockels effect. In this effect the evanescent electric field of the DUT reaching into the EO crystal, alters the linear polarization state of the sampling laser beam into elliptical. The ellipticity corresponds to a phase difference between X and Y polarization components which is linearly proportional to the sampled electric field and can be transferred to an intensity modulation of laser, by separating both polarizations of the beam in a polarizing beam splitter (Wollaston prism in Fig. 1) and later on detected with base band (BB) electronics e.g. with a balanced photodiode detector (ellipsometer). The resulting signal which is a down converted replica of the RF signal of the DUT, after amplification with low noise current to voltage amplifiers can directly be measured with lock-in amplifier (see Fig. 1). The specifications of the setup are summarized in Table I. More details can be found in [3, 4]. III.
Maximum field Fig. 2. The EWB which terminates the NLTL. The maximum electric field is observed at the lower corner of the antenna and the EOS measurement is performed at this point.
antenna corner in an optical image. This antenna is connected to the NLTL as a terminator and radiates the NLTL signal. Fig. 3 shows NLTL signal in frequency domain which is measured by EOS at harmonics of the injected fundamental. The lock-in measurement bandwidth is set up at 10 Hz (time constant=100ms) bandwidth which strongly limits the noise and the signal is averaged over 1000 averages. In this configuration, the noise floor of the measurement is measured at -80 dB below the signal at the first harmonic and the sensitivity is limited by shot noise . Previously we have shown in presence of jitter, averaging would never gain specially at higher harmonics where the S/N reduces by harmonic number [3, 4].
The DUT in these measurements is a 65 nm CMOS nonlinear transmission line (NLTL) which is terminated by an on-chip extremely wideband (EWB) antenna . The NLTL is a realized by periodically placing CMOS varactors on a microwave CPW transmission line [6, 7]. The device design motivation is to generate broadband THz (sub) pulses from a continuous wave microwave injected signal. The performance of the device and its capability to generate at least 200 GHz has been confirmed with earlier EOS experiments with a 10 GHz spectral resolution . Due to nonlinearity of the NLTL, the output signal contains comb of harmonics of its input signal (fundamental). The fundamental signal in these measurements is chosen at 10 GHz with power of 18 dBm in order to enhance the level of detected signal. Since in this setup, the source and sampler are both driven from the laser, therefore the “relative jitter” is maximally reduced. The maximum signal is measured at the edge of on-chip antenna which has been found by scanning the DUT. Fig. 2 shows the point of the measurement on the
Fig. 3. Measured signal of a 65 nm CMOS NLTL and noise limitation of the experimental setup.
TABLE I. SPECIFICATIONS OF THE EOS EXPERIMENTAL SETUP Parameter Measurement Technique
Imaging Spatial resolution
-80dBm (Shot noise)