Advanced Experimental Methods For Noise Research in Nanoscale Electronic Devices

1/f noise is a fluctuation in the bulk conductance of semiconductors and metals. This noise could be a fluctuation in the number of the free electrons or in their mobility. Many experimental studies on homogenous layers have proved that the 1/f noise is a fluctuation in the mobility. There is no theoretical model of mobility 1/f noise. The McWhorter model for 1/f noise in MOSTs simply adds generation-recombination spectra from surface states. According to this model, estimates of the noise magnitude give unrealistic values. More important, the McWhorter model is a model on number fluctuations, because GR noise always is a fluctuation in number. There is no experimental proof of number fluctuations in the 1/f noise of MOSTs.

We will examine possible sources of generation-recombination and 1/f noise in GaN/AlGaN 2D structures, quantum wells, and devices including contacts, bulk and quantum well itself and show that sources of g-r noise, and most probably of 1/f noise in HFETs are located in GaN or AlGaN layers within some distance from the 2D channel.

Some old, perennial questions in the field of low-frequency noise in solid are discussed in the light of some recent achievements concerning the investigation of the noise phenomena in nanomaterials and nanostructures. The main emphasis is placed on the very old yet still hot topic of 1/f noise. Possible new fashion is considered for three longstanding questions: Lorentzian superposition, surface vs. bulk and number vs. mobility fluctuation. Examination of the RTS noise behaviour in meso- and nanostructures does not support its fundamental character in the generation of the 1/f noise. Recent noise measurements on both Single- and Multiple-Walled Carbon Nanotubes (SWNT, MWNT) definitely relax the old dispute whether 1/f noise is a surface or a bulk effect: either surface or bulk or both of them can contribute to 1/f noise. As for the number fluctuation vs. mobility fluctuation controversy, noise measurements in nanomaterials favour mobility fluctuation hypothesis due to phonon scattering as the microscopic source of the 1/f noise.

It is a general conviction that any measured noise be stochastically continuous and weak stationary. Therefore, standard noise analysis uses the substitution of ensemble averages by time averages, and it considers likewise the autocorrelation function and the sample spectrum as an unbiased and complete characterization of the measured process. However, randomly distributed discontinuities make the constant variance turn into a random one. This contradicts the standard suppositions. We consider the random walk as a typical non-continuous process and derive the influence of the ‘variance of variance’ on the measured spectrum. In contrast to the standard analysis, sums of squares are no longer proportional to the chi-square-distribution, but to a distribution with a larger variance. When decomposing the data into fixed and random variance components, it can be shown that, despite independent increments, the random variance component produces a positive and time dependent expectation of the covariance. This is the source of the typically shaped non-zero autocorrelation function and the 1/f spectrum. The expectation of the autocorrelation at any given time difference is the product of the random variance component and a factor, which depends only on the total number of data and on the number of sampling intervals between the associated pairs of data. Consequently, the 1/f spectrum is no longer to be understood within the meaning of Parseval’s theorem. The larger the ratio of the random to the fixed variance component, the higher the 1/f increase onset frequency. ‘Almost smooth’ processes yield an ‘almost white’ spectrum, larger variance of increments generates a 1/f spectrum over a larger range of frequencies, and if the quotient between random and fixed variance components approaches 1, the 1/f spectrum will appear to extend over the full range of frequencies.

We develop a new approach of the quantum phase in an Hilbert space of finite dimension which is based on the relation between the physical concept of phase locking and mathematical concepts such as cyclotomy and the Ramanujan sums. As a result phase variability looks quite similar to its classical counterpart, having peaks at dimensions equal to a power of a prime number. Squeezing of that noise is allowed for specific quantum states. The concept of phase entanglement for pairs of phase-locked states is introduced.

We discuss specific noise phenomena in several ballistic mesoscopic structures and in networks of quantum dots or metallic dots interconnected by tunneling barriers, focusing on the techniques used for the numerical simulation and on the physical interpretation of the results.

We describe the transition from sub-Poissonian to super-Poissonian values of the zero-temperature shot noise power of a resonant double barrier of macroscopic cross-section. This transition occurs for driving voltages which are sufficiently strong to bring the system near an instability threshold. It is shown that interactions in combination with the energy dependence of the tunneling rates dramatically affect the noise level in such a system. Interaction-induced fluctuations of the band bottom of the well contribute to the noise and lead to a new energy in the Fano factor. They can enhance the noise to super-Poissonian values in a voltage range preceding the instability threshold of the system. This mechanism is different from the super-Poissonian enhancement due to the large effective charge.

The paper describes several models which may be of use for explanation of the origin and properties of experimentally observed types of noise. The stochastic models are analyzed first, including the autoregressive scheme and the moving averages on the n-th order. Galton board model and generalized baker map represent deterministic systems with rich set of properties. The results presented in the second half of the paper are intimately connected to the theory of deterministic chaos, and the crucial role is played there by the attractors in phase spaces of the systems.

In the first part of this paper different kinds of electrical noise in photodiodes are summarized. In the second part electrical and optical noise of laser diodes are presented taking into account the correlation between both noise sources. In the last part some prospective aspects for new research into noise of optical amplifiers are briefly reviewed.

Low-frequency noise characteristics of Fabry-Pérot (F-P) and distributed feedback, ridge-waveguide and buried-heterostructures InGaAsP/InP multiple-quantum-well lasers investigation has been carrier out. Mode hopping effect characteristic for F-P laser operation is caused by carrier gathering in barrier and cladding layers, and intensive optical and electrical noise during mode hopping is related with recombination in these layers. Defective laser diodes structures can be revealed by noise characteristic investigation, especially the correlation factor is more informative at threshold.

Microwave noise technique is applied to study ultrafast correlations in AlGaN/GaN and AlN/GaN two-dimensional electron gas (2DEG) channels subjected to a strong electric field applied in the plane of electron confinement. The experimental data are discussed in terms of hot-electron- energy dissipation on phonons, longitudinal-optical (LO) phonon conversion into other phonon modes, and hot-electron deconfinement. At high electric fields, the hot-electron energy relaxation is limited by the LO-phonon conversion. The LO-phonon conversion lifetime is estimated to be 350 fs in AlGaN/GaN channel. The lifetime is a useful parameter for extrapolation of hot-electron temperature beyond the field range where it is available from microwave noise experiments. The extrapolated hot-electron temperature is used to discuss hot-electron deconfinement noise.

The review describes the noise properties of the high-T _c superconducting (HTS) bolometers developed for applications in optical electronic devices of infrared and submillimeter wavelengths. The principle of bolometer operation and its noise theory are considered. The published results of bolometer noise modeling are discussed. Various sources of the excess 1/f-noise in HTS films are reviewed. Comparative analysis of noise characteristics of the most developed HTS bolometers for application is carried out.

Circuit-simulation-oriented equations (SPICE and BSIM3) for the 1/f noise are discussed and their fitting parameters are translated in the 1/f noise parameter α. The effect of scaling down on the 1/f noise is studied in the ohmic region as well as in saturation and sub-threshold. A prospective for scaling down is given for channel length L > 0.12 μm where velocity saturation becomes dominant. A relation is proposed between the 1/f noise corner frequency f _c , where the 1/f noise is equal to the thermal noise, and the unit current gain frequency f _I . Faster devices (with higher f _T ) are inherently noisier considering f _c . Approximately holds, 10-4f _T > f _c > 10-3f _T .

The down-scaling of MOSFETs brings about inversion-layer quantization. Another consequence is the enhancement of the direct tunneling current through the gate, which is a source of parasitic leakage and noise. While these effects are accounted for in the DC characteristics, so far, little attention has been paid to the impact on the low-frequency noise. In this work, the focus will be on the assessment of these quantum noise effects. It is shown that appropriate test structures are required, while more detailed information can be obtained by changing the vertical field, i.e., through a change in the bulk bias. It will also be pointed out that beside experimental data, there is a growing need for accurate noise models, which go beyond the generally accepted correlated mobility fluctuations approach.

It is shown that some peculiar features are typical for the drain current noise spectra of SOI MOSFETs with 2.5 nm gate oxide. In the frequency range 0.7 Hz≤f≤50 Hz a drain current spectral density is observed which follows a 1/f1.7 law for a broad range of operation conditions. It is demonstrated that this noise is only found in the front-channel current and is observed both in SOI and bulk MOSFETs. The model proposed considers this noise as being generated by carriers tunneling between the front channel and traps associated with the polysilicon gate/oxide interface and situated sufficiently close to the channel in the case of an ultra-thin gate oxide. When the absolute value of the gate voltage is equal or higher than 1 V, a Lorentzian component appears in the noise spectra measured in the linear regime. It is shown that the Lorentzian amplitude SI(0) can be described by the formula S_I(0)=Bτ(V_DS)2/L3 where B is a coefficient, τ is the Lorentzian time constant that decreases exponentially with increasing gate voltage, V_DS is the drain voltage and L is the channel length. The mechanism proposed for this noise is based on the idea that it originates from the filtered shot noise induced by majority carriers which are injected in the floating body of the transistor by electron valence-band tunneling across the ultra-thin gate oxide. Therefore, the appearance of both noise components can be regarded as thinoxide noise effects.

Recent developments in the novel generation of Non Volatile Memories (NVM) containing a plane of Si nano-crystals (Si-nc), embedded in the gate oxide MOS transistors, created a need for a better understanding of the basic physics of the charge/release phenomena on the Si-nc’s. Such studies are also important because the expected retention time is interesting for practical memory applications. Here, for the first time to our knowledge, low frequency noise (LFN) studies on MOS capacitors and transistors with the Si-nc’s are presented. The results obtained in the structures with and without Si-nc’s are compared. The implication of the Si-nc’s in the LFN generation and charge dynamics in the devices are discussed.

We investigate electron transport and shot noise in single and double barrier GaAs/GaAlAs semiconductor structures. Both structures evidence shot noise enhancement and suppression. The enhanced mechanism is due solely to the positive feedback between tunneling and space charge, and it is a precursor of current instability. Concerning the suppression mechanism, the standard sequential tunneling model does not explain a suppression with a Fano factor below 0.5, which is found in several experiments. By contrast the coherent tunneling model predicts shot noise suppression below 0.5 because of Pauli principle and/or Coulomb interaction in agreement with experiments. We conclude that shot noise suppression below one-half of the full Poissonian value is a signature of coherent tunnelling against sequential tunneling in double barrier resonant diodes.

A detailed study of low frequency noise drain-gate correlation in the presence of significant gate leakage current will be presented for ultrathin oxide MOSFETs. Measurements and a physical model for the correlation coefficient will be discussed. The correlation coefficient between the drain and the gate is derived on the basis of partition noise theory and the BSIM4 gate leakage current model with source-drain partition, and is in good agreement with correlation noise measurements as a function of the gate to the drain current ratio.

Attempts to explain the nature of the 1/f noise in AlGaN/GaN HFETs have involved three different mechanisms: occupancy fluctuations of the tail states near the band edges, fluctuations in the space charge regions surrounding dislocations, and electron tunneling from the 2D gas into adjacent GaN or AlGaN layers. Our experimental data favor the third mechanism. Three main arguments support this tunneling mechanism: (i) the observed temperature dependence of noise in the doped channel of AlGaN/GaN HFETs, which is explained by the tunneling model, (ii) a very weak temperature dependence of the 1/f noise in the temperature interval from 8 K to 300 K, (iii) the concentration dependence of the Hooge parameter typical for noise caused by the electron tunneling..

The continuing down-scaling of MOSFET’s make their noise properties to be more and more fundamentals, for some difficulties may happen when considering high frequency applications. Indeed, though the intrinsic noise performances of MOSFET’s are very promising, the technological process inherent to such devices when the device is scaling down makes de facto parasitic elements (such overlap capacitances, gate resistances) to limit greatly the f _max (maximum oscillation frequency) and then the noise performances of the device. Also, potential mismatch in Low Noise Amplifiers (LNA) along with the use of high losses Silicon Substrate may increase greatly the noise figure of the LNA, that will be much higher than the minimum noise figure of the stand alone device. As a general trend, the decrease of device’s dimensions along with the decrease of the DC supply voltage will require in the future to have the availability of more and more accurate thermal (and related diffusion) noise models. These noise models can be issued from device physics models but also can be extracted from experimental noise measurements; in such a case, one wants to perform accurate measurements of the noise parameters F _min , R _n and Γ _opt of the Device Under Test (DUT). A lot of methods are available in the literature, we will focus here upon the one used in routine in our laboratory, which is based on noise figure measurements in a 50 Ω environment and physical considerations. The presentation of the NF₅₀ method will be re-called prior describing the problems related to an accurate extraction of the noise sources. These problems will address, by the use of a physical noise modeling, a discussion related to the theoretical assumption (uncorrelated noise sources) made to extract the four noise parameters from the NF₅₀ data versus frequency. Due to the low resistivity silicon wafers used for such characterization, we will also pay attention towards the specific de-embedding noise procedure which needs to be used.

A general phenomenological approach — Flicker Noise Spectroscopy (FNS)-for revelation of information valuable parameters characterizing the arbitrary chaotic surfaces was developed to distinguish their patterns and describe quantitatively their functional properties. The method developed was applied to revelation of effects of a shungit filling agent in polypropylen matrix on the composite properties, revelation of hydrogen treatment effects on the cleavage surfaces of LiF monocrystals after their dissolution in water with quantitative evaluations of their anisotropy, analysis of activity of vacuum deposited porphyrins layers in a photosensibilized generation of singlet oxygen into gaseous phase. The approach elaborated can be used for developing the new control tools for processing scanning probe microscopy (SPM) data in nanotechnologies, microelectronics, production of polymeric materials with specific surface properties, and others.

The following topics will be considered: Noise characterisation of voltage amplifiers in series. Noise measuring set-up with dc excitation and single ended ac-amplifier or differential amplifier in a bridge configuration. Bias and sample criteria to observe 1/f noise in homogeneous samples submitted to homogeneous fields. Criteria for highest S _V value and highest corner frequency f _c between 1/f and thermal noise.Eddy current shielding (on wafer level measurements). Calculation of S _R from S _V or current spectra SI in different bias circuits. Contact noise reduction in a four probe configuration, by good sample design Noise correlation measurements

Recent theoretical and experimental findings have raised interest in the issue of shot noise suppression and enhancement in nanostructures. Several theoretical predictions have already been confirmed by means of sophisticated experiments, but further work is needed to improve the achievable sensitivity of noise measurements.We have been working on the integration of several different noise reduction techniques, with the objective of being able to measure the shot noise levels associated with currents of less than a picoampere. We combine the usage of correlation amplifiers, cryogenic cooling of the active elements and feedback resistors, correction techniques based on the substitution impedance method, and the precise evaluation of the transfer function of the amplifiers.

The paper describes the principles of operation and the expected performance of correlation spectrum analyzers in the detection and measurement of stationary noise produced by passive or active devices under test (DUT). It is shown that a sensitivity of few pV/vHz in voltage noise measurements and of few fA/vHz in current noise measurements may be reached with a properly long measurement time rarely exceeding few hours. The paper highlights the role of the DUT impedance and of the amplifiers noise sources in setting the sensitivity limit of the instrument as given by the generation of spurious correlated signals that feed both channels.

Large-area device noise models are based on ensemble averaging techniques. Typically, the physical mechanisms leading to fluctuations of microscopic entities, such as charge carrier number or mobility can be modeled by connecting these fluctuations to noise in a measurable device parameter, such as voltage or current. In the process, ensemble averaging of independent (or some times dependent) fluctuators is done to obtain an “average” power spectral density function that agrees with the experimentally measured one. This procedure breaks down in small area devices where the observed noise is basically a single electron phenomenon. Small devices can be manufactured with a single defect which in time domain show only two level switching signals known as burst noise and/or Random Telegraph Signal (RTS) noise. The RTS noise in sub-micron MOSFETs usually dominates over all the other noise sources and becomes a major noise generator for low frequency region of spectrum. With usual oxide and interface trap densities in the order of 1010 eV cm-2, active traps being located only within a few k _B T around the Fermi level, a sub-micron MOSFET of 1.0*0.15 μm dimensions will have only 78% chance of having an active electron trapping site. Therefore, it will be a hit or miss situation to observe RTS in the signal. It is essential to be able to perform accurate measurements of RTS both in time and frequency domains, not only to be able to understand, analyze and model the noise in these advanced devices, but also to use RTS measurements as a characterization tool for the interface and bulk traps responsible for these RTS events.

This paper investigates the emission and capture kinetics of random telegraph signals (RTS) in Quantum dots and submicron MOSFET structures. Emphasis is laid on the signals showing a capture process which deviates from the standard Shockley-Read-Hall kinetics. The proposed model distinguishes between primary processes consisting in quantum transitions of electrons between traps and the conduction or valence band and secondary processes, consisting in current modulation. If the RTS noise sources are quantum transitions of electrons between a shallow trap and the conduction band then the primary process is one-dimensional and it coincides with the secondary process — current modulation. For deep traps the primary process is a two dimensional g — r process. It is shown in this paper how to distinguish experimentally between the one or two-dimensional primary processes by measuring the probability density of the occupation time in both current states. These long-time measurements require a very stable power source and active shielding of low frequency magnetic fields.

Important points in designing a low noise amplifier in very low frequency region such as around 1mHz are to avoid the coupling capacitor at the front end, and to suppress the thermal drift. Practical examples are described.

Measurements of low-frequency noise in RuO₂+glass thick films at subkelvin temperatures are presented. Films were prepared by a standard “high temperature” process: 20 nm sized crystalline RuO₂ powder was mixed with 0.5 µm granular lead-borosilikate glass and organic solvent to give a paste, which was screen printed onto alumina substrates (with pre-fired AuPd contacts) and fired in a tunnel furnace. Noise measurements reveal that below liquid helium temperature and for low biasing voltages the low frequency excess noise is a pure resistance noise. At larger voltages its power spectral density depends sublinearly on voltage square. In the liner regime noise increases with decreasing temperature, approximately as T-2.3. Up to 4 T no dependence of noise intensity on magnetic field has been observed. These effects are discussed in terms of possible conduction mechanisms in the RuO₂+glass thick films: hopping, thermally activated tunneling and weak localization.

We investigated known methods of the LF (1/f) noise Gaussianity test. These are measurements of: (a) high order semi-invariants, (b) the histogram as the estimate of the probability density function, (c) the accuracy in the measurement of the noise intensity at the output of bandpass filter, (d) the correlation between intensities of the noise at outputs of non-overlapped filters, and (e) the complex bispectrum of the noise. These methods are sensitive to the non-stationarity of the noise as well. A special computeraided setup was designed for these measurements. Tests were performed for quantum well laser diodes manufactured in Nizhni Novgorod State University. We have found that the voltage noise in the diodes is non-Gaussian and seems to be non-stationary. Our results may be used for the check of radiation defects in semiconductor devices and for the investigation of the 1/f noise nature.

Preamplifier noise reduction in the view of varying source impedance, as in AE sensors, is discussed. First, the relation of the preamplifier noise and the source resistor is treated. Further, the input capacitance and the internal source capacitance influence are discussed. Eventually, the influence of preamplifier AC coupling on the noise is shown and analyzed.

High accuracy of an experimental investigation of noise features of semiconductor devices and materials in low frequency region is very important for reaching deep understanding of noise features It will enable precise extracting of essential noise parameters and modelling noise behaviour of real devices.In this paper measurement accuracy of 1/f and GR noise in semiconductor devices, as well as data processing for parameter extraction will be discussed. Theory of random process parameter estimation gives us a basis for evaluation of measurement errors. General analysis of statistical accuracy of noise measurement is made. From this analysis practical recommendations and formulas for accuracy evaluation are derived. Differences in 1/f and GR noise characteristics must be taken into account in measurement. Choice of the most appropriate measurement conditions for particular kind of noise is discussed. Another source of measurement error could be contribution of other interfering signal sources. As a rule 1/f and GR noise are present simultaneously and a task is to evaluate their parameters separately. In the measurement and data processing process discrimination of particular noise sources and subtraction of other background noise sources must be provided.

The radiofrequency interferometer, as compared to other methods for the measurement of amplitude noise and phase noise, shows higher sensitivity. In favorble onditions the background noise can be as low as—180 dB[rad2]/Hz at f= 1Hz off the carrier, up to microwaves (10 GHz) for real-time measurements. Exploiting correlation and averaging, a white noise floor of—204 dB[rad2]/Hz has been observed, not limited by the thermal energy k _B T₀ referred to the carroer power P₀. This article reports on the method, on the recent developments, and on the measurement of low-noise hard-to-measure devices.

It is shown that there is a wavelet technique that allows one to estimate the power spectral density of random noise in Laplace plane.

Two basic metrological noise models of a two-port: with noise sources at the input and with noise short-circuit currents, have been described. Noise parameters describing the models including a correlation between currents and/or voltages are measured using indirect and direct methods. Some noise reduction techniques are reported. Two-channel noise measurement system with computer-controlled biasing of a measured device using the LabVIEW software and noise signal processing procedures has been presented.

A probe system for wafer level noise measurements of submicron structures has been briefly presented. Significant influences on probe measurement have interference and inherent noise of the system, especially high level interference introduced by probes (contact resistance probe to wafer area fluctuations caused by vibrations and mechanical shocks in the measurement environment) and electromagnetic (EM) surroundings. Periodical interference caused by vibrations and EM fields can be separated from measured noise. The waveforms are recorded as a sequence of time series with duration being the multiple of the interference period and coherently summed up and finally subtracted from the registered data. It enables to analyse random and periodical components of the signal separately.

For the first time, to verify the origin of LF-noise (LFN) in the access resistance of metamorphic SiGe HMOS FETs, we used a thin film n-InSb magnetoresistor (MR) as a reference control. Hybrid MRs were fabricated on NiO×Fe2O3 ferrite substrates from Sn-doped MBE-grown n-InSb/i-GaAs heterostructures. The thickness of the InSb epilayers lie in the range 1.0-2.0 µm giving a room-temperature Hall mobilities of µ = 5.5×104 cm2/Vs at a carrier densities of 5×1016 cm-3. The device resistance could be changed by up to 500% in a magnetic field of B=325 mT. LFN spectra were measured at B=0 and 31mT. Results show that the current dependence of the PSD for this MR is described by the Hooge mobility fluctuation model, according to S _I /I2 =S _R /R2 =a _H ×µ(eR / fL2). A simple design for a calibrated noise reference is proposed.

A modular design of preamplifier for low frequency noise measurements with interchangeable first stage was chosen to improve reliability and to reduce the influence of connection cables on measurement results. In this communication we present the optimised preamplifier modules as the first stages for MOSFETs gate leakage and drain current noise measurements with input impedance 50 Ω – 108 Ω in the frequency range 1.0 Hz — 105 Hz. The best available commercial operational amplifiers (OPA’s) AD549, OPA637 and LT1028A were used for the first stage module at each of the three chosen impedance ranges. The noise characteristics of different OPA’s, which have been tested, are also presented.

In this work, two types of perovskite oxide materials with high temperature coefficients have been chosen as thermometers for use in bolometric applications: superconducting YBa₂Cu₃O_7-δ (YBCO) and colossal magnetoresistive La_2/3Sr_1/3MnO₃ (LSMO) thin films. Two different temperature ranges are concerned: around 90 K for YBCO, since they were operated at the superconducting-metal transition and around the room temperature for LSMO. The temperature coefficient ratio and the noise were carefully measured. Best NET values of 3×10-8 K.Hz-1/2 at 1 Hz, 90 K and 5 mA current bias for YBCO and 1.3×10-6 K.Hz-1/2 at 1 Hz, 300 K and 1.5 mA current bias for LSMO were obtained. Results are compared to literature and, in the case of room temperature applications, to other types of materials such as semiconductors (a-Si, a-Si:H, a-Ge, poly SiGe) and other oxide materials (semiconducting YBCO, VO_x, other manganite compounds). Finally, the possible use of these thermometers with such low NET characteristics for the fabrication of both membrane-type bolometers for mid-infrared detection and antenna coupled bolometers for THz applications is discussed.

Tracing the bi-stable mechanism of reverse-biased junction conductivity can be used as an efficient tool to study PN junction inhomegeneities. This conductivity mechanism is usually accounted for in terms of crystalline lattice imperfections, dislocations, or metallic precipitates in the PN junction region. Two methods have been suggested for practical application: first, a tentative method to distinguish between the bi-stable and the multi-stable conductivity mechanism, and, second, a noise current rms value versus DC current plot based method.

An instrument to improve power spectrum density (PSD) measurements, a highly sensitive preamplifier and a data acquisition unit are designed as well as a programmable power battery source. The preamplifier has 8 junction field effect transistor (JFET) inputs in parallel resulting in an input noise equivalent power of —186 dBV2/Hz. The data acquisition system consists of a low-pass filter, an A/D converter and a RAM, which transfer the data to the connected PC. The FFT or other signal processing is carried out within the PC. The battery power source is controlled by the PC, enabling us to obtain the device-under-test (DUT) bias DC required by the program. This measuring system works well at room temperature where the DUT and the power source could be put into a magnetic shielding box to reject spurious line noises. In the case where the DUT is placed in a cryostat to measure the temperature dependence of noise properties, however, the shielding of the DUT from the power line becomes a serious problem. Particularly in order to use a compressor cooling system, spurious noises come into the measuring system through the power lines of the compressor. In such a case the preamplifier must be placed as close as possible to the DUT, and the power line to the compressor is sometimes disconnected from the system.

A system for automatic, wafer-level, low frequency, LF, fluctuation measurements on semiconductor devices, involving a novel, programmable biasing amplifier, PBA, is presented. The PBA is computer-controlled and can remotely bias device terminals and measure currents flowing through them to the common ground. Its inputs are traxial with the appropriate guard potentials applied to their inner shields. The system is designed specifically for LF noise measurements on microelectronics structures. It can operate in manual and automatic modes. In the latter, the biasing voltage ranges and their increments are programmable. The program execution is carried out by step-wise biasing voltage variations, followed by measurements of appropriate currents and their fluctuations, with Fourier analysis, completed by the data storage. The program is written in LabView graphical language for a personal computer, equipped with a digital I/O and a data acquisition cards. No additional electronics is needed for the system operation. The system calibration by thermal noise of resistances is proposed. Some measurement results on the state-of-the-art microelectronics devices are discussed.