High Sampling Rate Research Articles

Signal Processing for Novel Noise Radar Based on de-chirp and Delay Matching

Modern radar technology requires high-quality signals and detection performance. However, traditional frequency-modulated continuous wave (FMCW) radar often has poor anti-jamming capabilities, and the high sampling rates associated with large time-bandwidth product signals can lead to increased system hardware costs and reduced data processing efficiency. This paper constructed a composite radar waveform based on noise frequency modulation (NFM) and linear frequency modulation (LFM) signals, enhancing the signal’s complexity and anti-jamming capability. Furthermore, a method for optimizing the processing of echo signals based on de-chirp and delay matching is proposed. The locally generated LFM signal is used to de-chirp the received echoes, resulting in a narrowband difference frequency noise signal. Subsequently, delay matching is performed in the fast time domain using the locally generated NFM signal according to the number of sampling points in the traversal processing period, allowing for the acquisition of target delay information. While reducing the analog-to-digital (A/D) sampling rate, the detection performance for wideband echo signals under high sampling rates is still maintained, with sidelobe levels and range resolution preserved. Accumulating this information in the slow time domain enables accurate target detection. The effectiveness of the proposed method is validated through simulation experiments.

A hybrid local-global neural network for visual classification using raw EEG signals

EEG-based brain-computer interfaces (BCIs) have the potential to decode visual information. Recently, artificial neural networks (ANNs) have been used to classify EEG signals evoked by visual stimuli. However, methods using ANNs to extract features from raw signals still perform lower than traditional frequency-domain features, and the methods are typically evaluated on small-scale datasets at a low sample rate, which can hinder the capabilities of deep-learning models. To overcome these limitations, we propose a hybrid local-global neural network, which can be trained end-to-end from raw signals without handcrafted features. Specifically, we first propose a reweight module to learn channel weights adaptively. Then, a local feature extraction module is designed to capture basic EEG features. Next, a spatial integration module fuses information from each electrode, and a global feature extraction module integrates overall time-domain characteristics. Additionally, a feature fusion module is proposed to extract efficient features in high sampling rate settings. The proposed model achieves state-of-the-art results on two commonly used small-scale datasets and outperforms baseline methods on three under-studied large-scale datasets. Ablation experimental results demonstrate that the proposed modules have a stable performance improvement ability on multiple datasets across different sample rates, providing a robust end-to-end learning framework.

Development of an ultra-low-cost wireless data acquisition system for monitoring strain and deflection

Strain gauges are gaining popularity for monitoring stress and strain within structural elements and for detecting and identifying structural damage. Although they are relatively inexpensive, they are not as widely used as accelerometers in structural health monitoring. This is due to several factors, such as the high cost of data acquisition systems required to collect strain data and the need for a large number of gauges to monitor a structure. Therefore, this research aims to develop a low-cost wireless strain gauge data acquisition system with acceptable accuracy and a high sampling rate. This system is composed of commonly available electronic components and open-source software, with a cost of $20. The system’s precision in measuring displacement was assessed in a laboratory experiment, and further experiments were conducted on an aluminum beam to investigate its response under both static and dynamic conditions. Additionally, the accuracy of the system for measuring deflection and strain in reinforced concrete elements was assessed through a bending test on a reinforced concrete beam. The findings indicated that the developed data acquisition system can accurately measure displacement, with a root mean square error of 0.1 mm, as well as strain values within the range of a few micro-strains.

Investigating the suitability of the matched fiber Bragg grating approach for guided wave based structural health monitoring

Fiber Bragg grating (FBG) sensors are thought to be ideal sensors for structural health monitoring (SHM). Amplitude based techniques such as matched filters and the edge-filtering have been proposed to fulfill the high sampling rates necessary for guided waves (GW) sensing. The current research for the first time shows the inherent robustness the matched filter technique provides to the sensing system. The matched system is realized through a FBG deployed on a low cost micrometer screw gauge which allows flexibility in matching the reflected wavelengths of the FBGs. Based on the results it is shown that, in the remote configuration the matched FBG approach is robust under changing temperature conditions. This concept has been shown analytically and experimentally. Furthermore, the approach is then extended for multiplexing of the sensors and the deployed system is used for detecting and localizing damage in the plate.

Impact of Receiver Bandwidth on the Performance of Nonlinear Volterra Equalizer in IM/DD Systems

Nonlinear electrical equalization is an effective means to compensate for nonlinear waveform distortions in short-haul intensity-modulation (IM)/direct-detection (DD) optical transmission systems. Nonlinear distortions contain broader spectral components than the original signal. Thus, a wide receiver bandwidth and high sampling rate are beneficial to effective compensation of nonlinear distortions. However, this leads to poor signal-to-noise ratio at the receiver. In this paper, we investigate the impact of receiver bandwidth on the performance of 2nd-order nonlinear Volterra equalizer in IM/DD transmission systems. Through the computer simulation and the experimental verification performed with 4-ary pulse amplitude modulation signals, we show that an optimum receiver bandwidth for bit-error ratio performance increases with the amount of nonlinear distortions in IM/DD systems. We also show that the sampling rate should also be increased to avoid the aliasing of the nonlinear distortion components. The cost of the receiver increased by the wider receiver bandwidth and higher sampling rate could be offset in part by the non-integer fractionally sampled nonlinear Volterra equalizer.

Performance of synthetic DAS as a function of array geometry

Distributed Acoustic Sensing (DAS) can record acoustic wavefields at high sampling rates and with dense spatial resolution difficult to achieve with seismometers. Using optical scattering induced by cable deformation, DAS can record strain fields with ones of meters spatial resolution. However, many experiments utilizing DAS have relied on unused, dark telecommunication fibers. As a result, the geophysical community has not fully explored DAS survey parameters to characterize the ideal array design. This limits our understanding of guiding principles in array design to deploy DAS effectively and efficiently in the field. A better quantitative understanding of DAS array behavior can help improve the quality of the data recorded by guiding the DAS array design. Here we use array response functions as well as beamforming and back-projection results from forward modelling calculations to assess the performance of varying DAS array geometries to record regional and local sources. A regular heptagon DAS array demonstrated improved capabilities for recording regional sources over segmented linear arrays, with potential improvements in recording and locating local sources. These results reveal DAS array performance as a function of geometry and can guide future DAS deployments.

Concurrent Validity and Interunit Reliability of 25-Hz GNSS Units for Profiling Sprinting Performance.

Dennison, L, Duthie, GM, Ehrmann, F, and Psarakis, MA. Concurrent validity and interunit reliability of 25-Hz GNSS units for profiling sprinting performance. J Strength Cond Res XX(X): 000-000, 2024-Profiling sprint performance by assessing within-race velocity or time is crucial for understanding an athlete's capabilities and identifying areas for improvement. While traditional gold standard systems provide valid and reliable measurements, they are often costly, laboratory-based, or impractical for field-based settings. New Global Navigation Satellite System (GNSS) units (25 Hz) with higher sampling rates may address some of these limitations. The purpose of the project was to evaluate the concurrent validity and interunit reliability of 25-Hz GNSS units by examining their agreement with laser devices and timing gates for velocity measurements and sprint/interval times. Concurrent validity was assessed during a real track and field training session. Thirty subjects each completed 3 to 6 maximal 40-meter sprints, for a total of 106 sprints, while being assessed simultaneously through all systems. Interunit reliability was assessed by placing 3 GNSS units on a motor vehicle and completing 60 accelerations from 0 to 60 m. Low mean bias (<1%) and typical error less than <2% for all measurements demonstrate excellent agreement between GNSS and criterion devices. The units demonstrated good reliability for the 0-10 m interval time (intraclass correlation coefficient [ICC] = 0.86), excellent reliability for all remaining 10 m to 60 m intervals (ICC = 0.91-0.99), sprint times 0-60 m (ICC = 0.97), and maximal sprint velocity (ICC = 1.0). These GNSS units provide an efficient and feasible alternative to traditional measurement devices. For coaches, this technology offers a scalable method to concurrently assess the sprint performances of multiple athletes during training and competitions, enabling evidence-based decisions to guide athletic development.

Typical Scenario Load Identification Based on Feature Fusion and Transfer Learning

Background: The electricity demand is continuously increasing. However, various institutions, enterprises, and individuals exhibit many irregularities in their electricity usage, leading to significant wastage of electricity. To achieve effective energy management, researchers are attempting to analyze and regulate users' electricity demands by monitoring their load usage through Non- Intrusive Load Monitoring (NILM) technology. The accuracy of load identification in this technology will greatly impact the results of load monitoring. Although there are currently many articles and patents related to NILM, they utilize a large amount of computational resources and require high sampling rates from devices, yet the results are still unsatisfactory. Therefore, it is necessary to improve the accuracy of load identification in data with relatively low sampling frequencies. Objective: To improve the accuracy of load identification with low sampling frequency data, this paper proposes a typical scenario load identification method based on feature fusion and transfer learning. Methods: This method adopts the fusion of current and power factor angles to provide abundant identification information for NILM, effectively reducing the situation of single-feature overlap of different loads. By inputting the fused feature data into GoogLeNet and utilizing transfer learning for training, not only is the accuracy improved, but also the training time and the requirement for the sampling rate of training data are greatly reduced. In addition, selecting typical scenario loads can monitor loads in a targeted manner, reduce the waste of computing resources caused by irrelevant loads, and more effectively guide electricity usage strategies. Results: The proposed load identification method was tested on the low sampling frequency dataset used in this paper. It achieved an overall load identification accuracy of 94.61% across three scenarios, improving accuracy by 3% to 7% compared to other models. Conclusion: The simulation results indicate that this method achieves high load identification accuracy at low sampling frequencies. It also exhibits good generalization ability. This method not only reduces the performance requirements for monitoring equipment but also enhances monitoring efficiency.

Investigating the effective temporal resolution in a task-based functional MRI experiment at 7 T MRI using a dynamic phantom

Abstract An increasing number of human fMRI studies aim to discern the time delays between evoked responses under different stimuli conditions in different brain regions. To achieve that, a primary goal is to acquire fMRI data with high sampling rates. This task is now possible with ultra-high field (≥7 T) MRI and the advancement of imaging acceleration methods. Consequently, it becomes imperative to understand what is the actual or effective temporal resolution (ETR) that is realized in given settings of an fMRI experiment. In this study, we utilized a dynamic phantom to reliably repeat a set of scans, generating a “ground truth” signal with controllable onset delays mimicking fMRI responses in a task-based block-designed fMRI. Here, we define the ETR and quantify a scan’s ETR using the dynamic phantom. The quantification was performed for various scanning parameters, including echo time (TE), repetition time (TR), voxel size, and contrast-to-noise ratio (CNR). We further show that combining data from multi-echo EPI can improve the ETR (i.e., reduce it). In addition, parameters of the fMRI paradigm were examined, including the blocks’ length and density. As tissue properties (e.g., level of iron deposition) affect the CNR and thus change the ETR, we examined the signal rise mimicking not only the cortex, but also the basal ganglia (known for its high iron deposition). Combining multi-echo data, the estimated ETR for the examined scans was 151 ms for a cortex-mimicking setup and 248 ms for a basal ganglia-mimicking setup, when scanning with a sampling time (i.e., TR) of 600 ms. Yet, a substantial penalty was paid when the CNR was low, in which case the ETR was even larger than the TR. A feasibility set of experiments was also designed to evaluate how the ETR is affected by physiological signal fluctuations and the variability of the hemodynamic response. This study shows the viability of studying time responses with fMRI, by demonstrating that a very short ETR can be achieved. However, it also emphasizes the need to examine the attainable ETR for a particular experiment.

Real-time monitoring and control of a PV-fed enhanced cubic voltage gain converter for DC microgrid

In this manuscript, a novel high-gain DC-DC converter with an ultra-step-up voltage gain value of 22.2 is introduced along with a custom-developed web application for remotely monitoring and controlling the power converter. The proposed converter is synthesized using two stages – stage 1 yields a cubic voltage gain and stage employs a diode-capacitor gain cell (DCGC) to reduce the voltage stress on the switch. The proposed gain extension hypothesis is experimentally validated through an 18 V to 400 V, 175 W prototype converter which delivers 175 W to the load at a 93.5% efficiency. The proposed converter exhibits excellent dynamic response when regulating the output voltage to 400 V over a wide range of input voltage and load current variations; the overshoots and undershoots are also negligible. Further, the maximum voltage stress on the switch is only 37.5% of the output voltage. For remotely controlling and monitoring the converter under real-time conditions with a high sampling rate, a web application is developed using React.js. The STM32 microcontroller is programmed to transmit data serially to the server, which then interacts with the web application using hypertext transfer protocol (HTTP) and WebSockets. The effectiveness of the developed interface is also practically verified by controlling the proposed converter in various modes viz., open-loop, soft-start, constant voltage, constant current, soft-stop, load regulation and overvoltage protection modes. Based on the comparison with several converters the proposed converter possesses unique advantages. Additionally, its web-based remote monitoring and control features are preferable for DC microgrid application.

From Labquakes to Megathrusts: Scaling Deep Learning Based Pickers Over 15 Orders of Magnitude

AbstractThe application of machine learning techniques in seismology has greatly advanced seismological analysis, especially for earthquake detection and seismic phase picking. However, machine learning approaches still face challenges in generalizing to data sets that differ from their original training setting. Previous studies focused on retraining or transfer‐learning models for these scenarios, but require high‐quality labeled data sets. This paper demonstrates a new approach for augmenting already trained models without the need for additional training data. We propose four strategies—rescaling, model aggregation, shifting, and filtering—to enhance the performance of pre‐trained models on out‐of‐distribution data sets. We further devise various methodologies to ensemble the individual predictions from these strategies to obtain a final unified prediction result featuring prediction robustness and detection sensitivity. We develop an open‐source Python module quakephase that implements these methods and can flexibly process input continuous seismic data of any sampling rate. With quakephase and pre‐trained ML models from SeisBench, we perform systematic benchmark tests on data recorded by different types of instruments, ranging from acoustic emission sensors to distributed acoustic sensing, and collected at different scales, spanning from laboratory acoustic emission events to major tectonic earthquakes. Our tests highlight that rescaling is essential for dealing with small‐magnitude seismic events recorded at high sampling rates as well as larger magnitude events having long coda and remote events with long wave trains. Our results demonstrate that the proposed methods are effective in augmenting pre‐trained models for out‐of‐distribution data sets, especially in scenarios with limited labeled data for transfer learning.

Instrumentation techniques for the measurement of gunfire (Part 3)

This paper is the third in a series concerning near-field measurement of sound pressure from small arms gunfire. It has been long established that exposure to gunfire events can lead to permanent threshold shift for an unprotected shooter. Close to a shooter's ear, the peak sound pressure level from a rifle or pistol can be 155-to-160 decibels with a pulse rise time on the order of ten microseconds. The U.S. Department of Defense has considered two approaches for assessing exposure limits from small arms gunfire - one involves the unweighted peak sound pressure and the other a series of A-weighted sound exposure events integrated to obtain an equivalent eight-hour dose. This dose is then compared to hearing damage risk criteria established for continuous noise signals. The first approach places great demands on the measurement system since the necessary bandwidth extends well into the ultrasonic range, thereby requiring fast slewing from the amplifiers combined with high sampling rates from the digitizer. The second approach requires the A-weighting filter to have specified time and frequency domain properties in the ultrasonic region. The paper discusses some laboratory experiments using actual gunfire signals to simulate the outcome from both approaches.

Vehicle and Pedestrian Traffic Signal Performance Measures Using LiDAR-Derived Trajectory Data.

Light Detection and Ranging (LiDAR) sensors at signalized intersections can accurately track the movement of virtually all objects passing through at high sampling rates. This study presents methodologies to estimate vehicle and pedestrian traffic signal performance measures using LiDAR trajectory data. Over 15,000,000 vehicle and 170,000 pedestrian waypoints detected during a 24 h period at an intersection in Utah are analyzed to describe the proposed techniques. Sampled trajectories are linear referenced to generate Purdue Probe Diagrams (PPDs). Vehicle-based PPDs are used to estimate movement level turning counts, 85th percentile queue lengths (85QL), arrivals on green (AOG), highway capacity manual (HCM) level of service (LOS), split failures (SF), and downstream blockage (DSB) by time of day (TOD). Pedestrian-based PPDs are used to estimate wait times and the proportion of people that traverse multiple crosswalks. Although vehicle signal performance can be estimated from several days of aggregated connected vehicle (CV) data, LiDAR data provides the ability to measure performance in real time. Furthermore, LiDAR can measure pedestrian speeds. At the studied location, the 15th percentile pedestrian walking speed was estimated to be 3.9 ft/s. The ability to directly measure these pedestrian speeds allows agencies to consider alternative crossing times than those suggested by the Manual on Uniform Traffic Control Devices (MUTCD).

Unlimited sampling beyond modulo

Analog-to-digital converters (ADCs) act as a bridge between the analog and digital domains. Two important attributes of any ADC are sampling rate and its dynamic range. For bandlimited signals, the sampling should be above the Nyquist rate. It is also desired that the signals' dynamic range should be within that of the ADC's; otherwise, the signal will be clipped. Nonlinear operators such as modulo or companding can be used prior to sampling to avoid clipping. To recover the true signal from the samples of the nonlinear operator, either high sampling rates are required, or strict constraints on the nonlinear operations are imposed, both of which are not desirable in practice. In this paper, we propose a generalized flexible nonlinear operator which is sampling efficient. Moreover, by carefully choosing its parameters, clipping, modulo, and companding can be seen as special cases of it. We show that bandlimited signals are uniquely identified from the nonlinear samples of the proposed operator when sampled above the Nyquist rate. Furthermore, we propose a robust algorithm to recover the true signal from the nonlinear samples. Compared to the existing methods, our approach has a lower mean-squared error for a given sampling rate, noise level, and dynamic range. Our results lead to less constrained hardware design to address the dynamic range issues while operating at the lowest rate possible.

A threshold activation-based simplified Lv's transform algorithm for transient multi-component linear frequency modulation signals analysis.

The high sampling rate in modern digital systems generates a large scale of data. To address the computational burden, this paper proposes a threshold activation-based simplified Lv's transform (SLVT) algorithm to analyze the transient multi-component linear frequency modulation signals. Only the signal arrival can trigger the signal analysis. This mechanism alleviates the computation pressure because of the sparsity of signals. The threshold activation mechanism enables transient signal detection and sampling rate adjustment, thereby enhancing the efficiency and effectiveness of the analytical process. The simplified Lv's transform (LVT) removes redundant computations on stretch weighting and the Discrete Fourier Transform (DFT) in Lv's transform. SLVT uses the efficient Bluestein chirp-z algorithm to implement the stretch keystone transform. The comparison results show that SLVT reduces the computational complexity of the original LVT by at least 30.8%. This algorithm exhibits superior performance compared to other advanced signal processing methods, such as discrete chirp Fourier transform, fractional Fourier transform, and Radon Wigner transform algorithms, in terms of parameter extraction accuracy, computational complexity, and execution time. Moreover, the implementation of a field programmable gate array accelerates SLVT computing by a factor of 116 in comparison to the CPU (Central Processing Unit) platform.

Comparative Evaluation of Neural Network Models for Optimizing ECG Signal in Non-Uniform Sampling Domain

Electrocardiographic signals (ECG) are ubiquitous, which justifies the research of their optimal storage and transmission. However, proposals for non-uniform signal sampling must take into account the priority of diagnostic data accuracy and record integrity, as well as robustness to noise and interference. In this study, two novel methods are introduced, each utilizing a distinct neural network architecture for optimizing non-uniform sampling of ECG signal. A transformer model refines each time point selection through an iterative process using gradient descent optimization, with the goal of minimizing the mean squared error between the original and resampled signals. It adaptively modifies time points, which improves the alignment between both signals. In contrast, the Temporal Convolutional Network model trains on the original signal, and gradient descent optimization is utilized to improve the selection of time points. Evaluation of both strategies’ efficacy is performed by calculating signal distances at lower and higher sampling rates. First, a collection of synthetic data points that resembled the P-QRS-T wave was used to train the model. Then, the ECG-ID database for real data analysis was used. Filtering to remove baseline wander followed by evaluation and testing were carried out in the real patient data. The results, in particular MSE = 0.0005, RMSE = 0.0216, and Pearson’s CC = 0.9904 for 120 sps in the case of the transformer patient data model, provide viable paths for maintaining the precision and dependability of ECG-based diagnostic systems at much lower sampling rate. Outcomes indicate that both techniques are effective at improving the fidelity between the original and modified ECG signals.

Accelerated model-based T1, T2* and proton density mapping using a Bayesian approach with automatic hyperparameter estimation.

To achieve automatic hyperparameter estimation for the model-based recovery of quantitative MR maps from undersampled data, we propose a Bayesian formulation that incorporates the signal model and sparse priors among multiple image contrasts. We introduce a novel approximate message passing framework "AMP-PE" that enables the automatic and simultaneous recovery of hyperparameters and quantitative maps. We employed the variable-flip-angle method to acquire multi-echo measurements using gradient echo sequence. We explored undersampling schemes to incorporate complementary sampling patterns across different flip angles and echo times. We further compared AMP-PE with conventional compressed sensing approaches such as the -norm minimization, PICS and other model-based approaches such as GraSP, MOBA. Compared to conventional compressed sensing approaches such as the -norm minimization and PICS, AMP-PE achieved superior reconstruction performance with lower errors in mapping and comparable performance in and proton density mappings. When compared to other model-based approaches including GraSP and MOBA, AMP-PE exhibited greater robustness and outperformed GraSP in reconstruction error. AMP-PE offers faster speed than MOBA. AMP-PE performed better than MOBA at higher sampling rates and worse than MOBA at a lower sampling rate. Notably, AMP-PE eliminates the need for hyperparameter tuning, which is a requisite for all the other approaches. AMP-PE offers the benefits of model-based recovery with the additional key advantage of automatic hyperparameter estimation. It works adeptly in situations where ground-truth is difficult to obtain and in clinical environments where it is desirable to automatically adapt hyperparameters to individual protocol, scanner and patient.

Dual-Wavelength LiDAR with a Single-Pixel Detector Based on the Time-Stretched Method.

In the fields of agriculture and forestry, the Normalized Difference Vegetation Index (NDVI) is a critical indicator for assessing the physiological state of plants. Traditional imaging sensors can only collect two-dimensional vegetation distribution data, while dual-wavelength LiDAR technology offers the capability to capture vertical distribution information, which is essential for forest structure recovery and precision agriculture management. However, existing LiDAR systems face challenges in detecting echoes at two wavelengths, typically relying on multiple detectors or array sensors, leading to high costs, bulky systems, and slow detection rates. This study introduces a time-stretched method to separate two laser wavelengths in the time dimension, enabling a more cost-effective and efficient dual-spectral (600 nm and 800 nm) LiDAR system. Utilizing a supercontinuum laser and a single-pixel detector, the system incorporates specifically designed time-stretched transmission optics, enhancing the efficiency of NDVI data collection. We validated the ranging performance of the system, achieving an accuracy of approximately 3 mm by collecting data with a high sampling rate oscilloscope. Furthermore, by detecting branches, soil, and leaves in various health conditions, we evaluated the system's performance. The dual-wavelength LiDAR can detect variations in NDVI due to differences in chlorophyll concentration and water content. Additionally, we used the radar equation to analyze the actual scene, clarifying the impact of the incidence angle on reflectance and NDVI. Scanning the Red Sumach, we obtained its NDVI distribution, demonstrating its physical characteristics. In conclusion, the proposed dual-wavelength LiDAR based on the time-stretched method has proven effective in agricultural and forestry applications, offering a new technological approach for future precision agriculture and forest management.

Assessment of hybrid machine learning models for non-linear system identification of fatigue test rigs

The prediction of system responses for a given fatigue test bench drive signal is a challenging problem, since highly dynamic loads from measurement campaigns must be reproduced accurately. Linear frequency response function models are commonly used for this system identification task, but energy intensive experimental iterations are required to account for system non-linearities. Two novel hybrid modeling strategies are suggested, which augment existing approaches using non-linear Long Short-Term Memory networks. These are trained and deployed on short subsequences of measurement data and recombined using a windowing technique, which enables their application to measurement data with high sampling rates. In addition to fatigue test bench commissioning, these hybrid models can also be employed in the field of virtual sensing. The approach is tested using non-linear experimental data from a servo-hydraulic test rig and this dataset is made publicly available. A variety of metrics in time and frequency domains, as well as fatigue strength under variable amplitudes, are employed in the evaluation. It is shown, that hybrid models can successfully use frequency response function models as a linear baseline estimate, which is further improved by Long Short-Term Memory networks to enable non-linear predictions.

THz-TDS transflection measurements as a process analyser for tablet mass

Tablet content and content uniformity are essential for the market release of the drug product. For tablets, content and uniformity are determined by the weight ratio of active pharmaceutical ingredient in the tablet and the tablets’ total mass. Novel process analytical technology tools for the control of the ratio of the active pharmaceutical ingredient have been proposed and implemented, but more robust, sensitive, and fast sensors for the control of tablet mass are desirable. In the presented study terahertz time-domain spectroscopy (THz-TDS) is proposed as a potential process analyser for tablet mass. THz-TDS is based on pulsed terahertz signals, which are mapped in the time-domain. Thus, the signal amplitude and arrival time are recorded. THz-TDS measurements of a tablet with a reflection setup result in two signals – a frontside reflection and a backside transflection. The presented study demonstrates that an increase in the tablet mass results in an increase in the time delay of the backside transflection. This is a result of the high refractive index of the solid fraction compared to air. It is suggested that the transflection can be used as an indirect measure of tablet mass for which root mean squared errors of around 1 mg were found. The potential to measure tablets at high acquisition rates (50 Hz) is explored and considered feasible. Additionally, it has been demonstrated in previous work that the time delay of the frontside reflection allows a simultaneous assessment of the tablet height. The presented methodology opens the possibility of in-line monitoring of tablet mass as part of a content and content uniformity strategy at high sampling rates in the production environment. Further, as tablet height and mass can be assessed simultaneously, monitoring and control of the compaction process based on a comprehensive assessment of physical tablet attributes can also be envisioned.