Back to search

Aristotle University Of Thessaloniki (AUTH)

Variability of low frequency fluctuations in sub 45nm CMOS devices-experiment, modeling and applications

Abstract

dc:description

Low frequency (LF) noise and fluctuations in MOS devices has been the subject of intensive research during the past years. The LF noise is becoming a major concern for continuously scaled down devices, since the 1/f noise increases as the reciprocal of the device area. Excessive low frequency noise and fluctuations could lead to serious limitation of the functionality of the analog and digital circuits. The 1/f noise is also of paramount importance in RF circuit applications where it gives rise to phase noise in oscillators or multiplexors. The development of submicronic CMOS technologies has led to the onset of new type of noises, i.e. random telegraph signals (RTS), yielding large current fluctuations, which can jeopardize the circuit functionality.However, the statistical variability in the transistor characteristics is one of the major challenges for upcoming technological nodes. The detailed knowledge of variability sources is extremely important for the design and manufacturing of variability resistant devices. Whereas the impact of random dopants, line edge roughness and oxide thickness variations is relatively well understood, the role of the polysilicon or metal gate material has only lately been investigated in simulations and experimental confirmation and quantification of its contribution is still lacking. In addition, the study of LFN variability behavior and maybe its relation with the other factors of device variations has never been done. Therefore, the research challenges and objectives of this thesis are centered towards the studies of low frequency fluctuations and noise in 32 nm CMOS technologies and beyond. More specifically, the objectives of the LF noise investigation is summarized in the following points: i) Detailed LF noise characterization of new CMOS technologies featuring high-κ metal gate stacks, channel pockets etc, ii) change of LF noise parameters from different technologies and iii) impact of LF noise and RTS fluctuations as a variability sources for analog and digital circuits. The first objective addresses the origin of the LF fluctuations in CMOS devices in terms of trap density and defect localization in the gate dielectric and along the channel for various architectures (pocket, Ge channel, FD-SOI etc). The second objective considers the LF noise variability resulting from huge dispersion of noise sources from device to device; this is conducted owing to statistical measurements of LF noise characteristics as a function of device area and technological splits. The third issue is focused on the impact of LF noise or RTS fluctuations on the operation of elementary circuits (inverter, SRAM cell) regarded as temporal variability source. Concerning the first objective mentioned above, the achievements of this work are summarized as follows: we developed a generic MOSFET compact CNF/CMF low frequency model based on a single equation for all operation regions with two physical parameters, namely the square root of the flat-band voltage spectral density fluctuations √SVfb related to the oxide trap density and Ω=αsc.μeff.Cox related to the effective Coulomb scattering coefficient αsc, the effective mobility μeff and the gate oxide capacitance Cox. Both parameters can be extracted experimentally from a plot of √SVfb versus Id/gm in linear and/or non-linear operation regions. The knowledge of these two parameters for a given CMOS technology provides a full description of the input gate voltage noise versus gate and drain voltages, and by turn of the drain current 1/f noise SId for any bias conditions. It is worth noting that the obtained constant value of the product αsc.μeff as a function of gate voltage, it clearly means that: (i) the conventional assumption of a constant αsc in the CNF/CMF model is not adequate, since in that case αsc.μeff should decrease at strong inversion because μeff degrades due to surface roughness scattering, and (ii) the reduction of the Coulomb scattering coefficient αsc at strong inversion due to screening would also not be consistent with a constant αsc.μeff product.The second objective of the thesis is the analysis of LFN in CMOS bulk technology nodes manufactured in STMicroelectronics the last 12 years. The above results can be summarized by plotting the volumetric trap density Nt versus equivalent oxide thickness of the investigated n- and p-MOS devices to have an idea about the evolution of LFN through almost all bulk technology nodes. The results shows that in both n-MOS and p-MOS devices from the 28nm technology node the oxide trap density Nt remains almost constant with the channel length L for all the “flavors” measured, i.e. Nt remains almost unchanged with channel length for this technology node except of the device with different oxide thickness, GO2. The different types of devices measured in this technology node are RVT, standard threshold voltage device, LVT, low threshold voltage device, SLVT, super low threshold voltage device, HPA special device for analog applications with no pockets inside the channel and finally the device with higher oxide thickness, GO2. The above observation was the moving force to plot the Nt values from all the technology nodes versus the equivalent oxide thickness and see the impact of EOT on the volumetric trap density. Finally, we have investigated the evolution of the effective Coulomb scattering coefficient α’=αsc.μeff with the equivalent oxide thickness for all the measured devices. The parameter α’ has been extracted from the experimental noise data using the CNF/CMF model. For both types of devices α’ remains almost constant through the evolution of the equivalent oxide thickness. Furthermore, the parameter a’ is higher in the p-MOS devices as expected.The LFN analysis in CMOS bulk devices is much more complicated than some people believe. As we have seen from the results of all technology nodes, our compact model is applied in most of the cases but there are still few ones where the model does not work properly. At this point, several questions give rise: What are the criteria to distinguish the CNF from the extended CNF/CMF model? Why in some cases the one or the other model applies and not in every case? What is the physical phenomenon that distinguishes the two models? All these questions need to be answered. Even though we have achieved until now to understand many issues regarding LFN, much more need to be done. A rather new idea is to consider that the two noise terms in CNF/CMF model are uncorrelated. What does that mean? It means that the physical phenomenon is the same, trapping/detrapping of carriers into slow oxide traps but the result is divided into two parts, the one affecting the carrier number within the channel and the other affecting the carrier mobility. This idea we believe that needs more investigation in order to fully understand the LFN mechanisms and the physics behind these mechanisms. On the other hand, the historical figures describing the volumetric trap density Nt for n- and p-MOS devices give us some new insights about LFN in MOSFETs. Starting from n-MOS devices, it is clear that for this type of devices the process affects seriously the Nt values. We cannot conclude any empirical rule connecting Nt values with EOT. The technology process plays a more important role than the equivalent oxide thickness as was previously believed. The situation is simpler for p-MOS devices. The impact of the technology for this type of devices is diminished compared to n-MOS. We can conclude that Nt is inversely proportional to EOT2. The general rule of circuit designers that p-MOS devices are less noisy than the n-MOS was until now true. However, for the 28nm technology node this rule is not applied, i.e. p-MOS devices are more noisy than n-MOS devices. The second part of the thesis includes the LF noise variability resulting from huge dispersion of noise sources from device to device; this is conducted owing to statistical measurements of LF noise characteristics as a function of device area and technological splits.The low frequency noise variability is analyzed in detail. Initially, we try to clarify the meaning of LFN variability in terms of its connection with a widely known phenomenon, the RTS noise. Then we continue with a representative example on how this phenomenon can become a major drawback in the functionality of digital circuits. Our approach to explain and model the LFN variability in CMOS bulk and FD-SOI technology nodes is based on a model already known which describes the RTS noise. This model can explain and predict the LFN variability behavior of the 28nm bulk CMOS devices with accuracy and in detail. We extended this model in the 28nm FD-SOI technology node in order to take into account other parameters such as the correlated mobility factor and the impact of the second interface. We have incorporated these parameters in the standard bulk model, trying to understand how the second interface introduced in FD-SOI affects the LFN variability behavior of the transistors. We developed an LFN variability model which incorporates for the first time the impact of the correlated mobility fluctuations in LFN variability. The comparison between the simulation data and the experimental results showed that the role of the back interface in the overall LFN variability of the devices is not strong. This result was anticipated since the equivalent oxide thickness of the back interface is much smaller than the front one. Further analysis in this domain should be carried out in order to verify the validity of the simulation model. In addition, a thorough investigation of the LFN variability through CMOS bulk technology nodes has been performed for the first time. The results reveal that the LFN variability shows better control for 28nm technology for both n- and p- MOS devices. Interestingly, these results are well correlated with the diminution of the static parameter variability (mismatch of threshold voltage). In addition, the experimental results revealed that the LFN variability is a more complicated phenomenon and it is highly associated with the process characteristics of the transistors and not only with the quality of the interface. The final part of this thesis deals with the research challenge of the impact of LF noise and RTS fluctuations as a variability sources on analog and digital circuits. We present a detailed characterization and modeling of the low frequency noise characteristics of CMOS inverters. The LF noise model has been developed within the carrier number fluctuations scheme of MOS transistor excess noise, using the concept of flat band voltage or threshold voltage power spectral density. It allows us to describe accurately the load current and output voltage LF noise characteristics as a function of input voltage obtained on inverters from a 45nm bulk CMOS technology.The developed LF noise modelling approach could constitute a useful tool for analyzing the impact of time domain fluctuations on the static and dynamic operation of CMOS inverters in VLSI circuits. In particular, it could be used for predicting the influence of dynamic fluctuations due to carrier trapping-detrapping on static noise margin and dynamic stability in SRAM cells.The impact of the dynamic variability due to low frequency fluctuations on the operation of CMOS inverters, which constitute the basic component of SRAM cell, has been investigated. The experimental methodology to characterize the effect of dynamic variability in a CMOS inverter is first established based on fast I-V measurements of the load current following the application of a ramp input voltage Vin(t). It is shown that, for small ramp rise times, the load current characteristics IDD(Vin) exhibit a huge sweep-to-sweep dispersion due to the low frequency noise. The impact of such dynamic variability sources on the inverter’s output characteristics Vout(Vin) is finally demonstrated, revealing a 20% noise margin reduction for the smallest inverter cell.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2013

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Ioannidis, Eleftherios
  • Ιωαννίδης, Ελευθέριος

Subjects

dc:subject × 12

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/36932

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Ioannidis, Eleftherios; Ιωαννίδης, Ελευθέριος. Variability of low frequency fluctuations in sub 45nm CMOS devices-experiment, modeling and applications. Aristotle University Of Thessaloniki (AUTH), 2013. http://hdl.handle.net/10442/hedi/36932