A Comprehensive Analysis of LoRa Network Wireless Signal Quality in Indoor Propagation Environments

Abstract

This paper investigates how key Long-Range (LoRa) sensor network transmission parameters and the number and material composition of physical obstacles on the signal propagation path impact wireless signal transmission quality in indoor propagation environments. A dedicated test platform was developed to assess how different combinations of the LoRa transmission parameters, which include spreading factor, transmit power, transmit duty cycle, message payload size, and the quantity and material composition of physical obstacles, with the signal propagation path length influence critical signal quality indicators, specifically the signal-to-noise ratio (SNR) and the received signal strength indicator (RSSI). The developed experimental test platform was implemented for a real-world indoor LoRa network composed of LoRa end devices (DVs) and gateways (GWs), utilizing technologies such as Node-RED for service orchestration, InfluxDB for data storage, The Things Network (TTN) for LoRa wide-area network connectivity, and Grafana for data visualization. The results of the performed analyses reveal how different combinations of LoRa transmission parameters, specifically the number and material composition of physical obstacles encountered during signal transmission among the LoRa end DVs and GWs, affect wireless signal quality indicators, namely RSSI and SNR, in indoor propagation environments of LoRa sensor networks. The obtained findings contribute to the optimization of LoRa transmission parameter selection for reliable and efficient signal transmission in LoRa indoor sensor network deployment, such as in urban environments with obstacles of varying structural composition and density encountered on the communication paths of different lengths between the LoRa end DVs and GWs.

1. Introduction

In recent years, applications of the Internet of Things (IoT) have experienced substantial growth and are exerting an increasingly profound influence on various aspects of daily life. As the number of IoT devices continues to rise, the focus has shifted towards the development of wireless communication technologies that require minimal infrastructure, offer long-range connectivity, consume very low energy, and maintain cost efficiency. This demand has led to the emergence of Long-Range (LoRa) technology, which belongs to low-power wide-area networks (LPWANs) and represents a viable solution for IoT system implementation. Initially, LoRa technology was developed by the company Cycleo with the aim of being standardized as a type of LPWAN. However, in 2012, Cycleo and the corresponding LoRa technology were acquired by the U.S.-based company Semtech, which later became a founding member of the non-profit organization LoRa Alliance™ [1,2]. Established in 2015, the LoRa Alliance brings together major industry players to standardize and promote the LoRa wide-area network (LoRaWAN) protocol for LoRa networks.

National and regional practical implementation rollouts of LoRa technology based on LoRa Alliance recommendations started in Europe (EU), the United States (US), and Asia in 2016 and 2017. This has resulted in LoRa technology gaining prominence in practical implementations and becoming widely implemented in smart cities, utility metering, logistics, and agricultural IoT applications. Current deployment of LoRa technology demonstrates a strong presence in the global market, with over 350 million LoRaWAN end devices (DVs) and 6.9 million gateways (GWs) deployed worldwide by mid-2024 [3]. This represents notable growth compared to 2023, when approximately 300 million LoRa end DVs were in operation [4]. Globally, the number of LPWAN connections reached 1.3 billion in 2024 [5], and when compared to the estimated total IoT connections equal to 17.7 billion in the same year [6], such connections constitute nearly 7.5% of the entire IoT ecosystem. In addition, LoRa technology currently accounts for approximately 30% of all LPWAN connections in 2025 so far [7]. Assuming that this market share remains stable, the number of LoRa end nodes could reach 900 million by 2030 if predictions of IoT connections reaching 40 billion by 2030 hold true [8]. Alternatively, if the expansion of LoRa technology is equal to the current expansion of LPWANs, the number of LoRa connections could rise to 1 billion by 2030, with either an increase in LoRa’s market share or continued expansion of both the LPWAN segment within IoT and the IoT market as a whole.

LoRa technology enables effective communication over long distances. Data exchange occurs through LoRa radio-frequency wireless signals transmitted from the (user) end DV radio transmitter to the GW radio receiver, and vice versa. A key advantage of LoRa-based end DVs lies in their ultra-low energy consumption, which makes them particularly suitable for battery-powered sensor DVs (capable of operating for up to ten years without battery replacement) [9]. The energy demand for transmitting a data packet is minimal due to the small size of the packets and their low transmission frequency, which is typically only a few times per day. When end DVs are in sleep mode, their energy consumption is very low and is measured in milliwatts (mW) [10], which further contributes to DV battery longevity.

The long-range signal transmission of LoRa technology is manifested in the capability of a single LoRa GW to receive and transmit signals over distances exceeding 15 km in rural areas [11]. In the case of implementation of LoRa networks in dense urban environments, communication ranges typically extend from three to five kilometers [12,13], depending on the depth and location of the end DVs within indoor spaces. However, different causes can impact successful data transmission in wireless networks, and they may include periodic cell outages [14] or inter-cell interference [15,16]. Additionally, LoRa networks’ indoor wireless signal propagation is one of the most critical implementation issues for LoRa wireless communications, since LoRa technology was originally optimized for long-range outdoor IoT coverage. In the case of indoor signal propagation, several challenges appear. These challenges include signal penetration loss through floors and walls, since building materials (such as brick, reinforced steel, concrete, and metal-coated glass) introduce significant signal attenuation. Although LoRa’s chirp spread spectrum (CSS) modulation is designed to offer robustness against multipath fading, indoor signal propagation environments having strong fading caused by signal reflections from walls, ceilings, furniture, and other objects have a negative impact on indoor signal quality. In addition, path loss in non-line-of-sight (NLOS) signal transmission conditions, such as in basements, deep indoor rooms, or dense urban areas, is significantly higher than in outdoor LOS environments and requires transmission at higher spreading factors (SFs), which increase transmission latency and reduce wireless channel capacity. Transmission at higher SFs and a higher number of retransmissions caused by poor indoor signal quality drains the LoRa end DV battery faster. This conflicts with the main feature of LoRa technology, which is the existence of communication capabilities with low-power consumption of LoRa end DVs. Also, the presence of unpredictable noise-level variations, known as noise uncertainty in wireless signal communication paths [17,18], gives an additional contribution to degrading indoor wireless signal quality. Indoor noise uncertainty comes from the interference of environmental factors (e.g., other wireless devices (Wi-Fi, Bluetooth, etc.), varying bandwidth occupancy, people moving, different appliances (elevators, neon lighting, etc.), and it is non-Gaussian and time-varying. This makes LoRa indoor communication paths less predictable and less robust, especially when operating near the signal sensitivity limits.

To ensure the successful deployment of a LoRa network in indoor and dense urban environments, it is essential to optimally place LoRa GW and configure LoRa DV transmission parameters, such as spreading factor, transmit (Tx) power, message payload length, transmission interval (also known as duty cycle (DC) interval), etc., in accordance with a specific LoRa end DV use-case scenario. Moreover, the quality of the communication between the LoRa end DV and the LoRa GW on short-range indoor communication distances is significantly affected by the material composition, number, and density of physical obstacles through which the wireless signal must propagate. Therefore, this paper analyzes the impact of the material composition, number, and density of physical obstacles on the quality of the wireless communication link for different combinations of the LoRa end DV transmission parameters. The quality of the communication link between LoRa end DVs and GWs is expressed in standardized metrics such as signal-to-noise ratio (SNR) and received signal strength indicator (RSSI). These metrics are collected during one-day periods, processed, and analyzed on a newly developed real LoRa network test system. Therefore, the primary contribution of this study is the development of a real LoRa network test system for collecting standardized metrics such as RSSI and SNR, to express the quality of the wireless signal propagation in different indoor signal propagation environments. Another contribution of the paper is a presentation of the comprehensive assessment of the impact of different combinations of key LoRa end DV transmission parameters and the material composition, number, and density of physical obstacles, on wireless signal quality and reliability of LoRa network operating in indoor environments. The results of the analyses can be useful for optimal placement planning of LoRa GWs in terms of ensuring the best quality of wireless links for LoRa end DV allocation in indoor and dense urban environments.

The rest of the paper is organized as follows: Section 2 presents related scientific work on the use of LoRa technology in indoor environments. In Section 3, the main LoRa end DV transmission parameters are described. The architecture of the test network developed for collecting wireless signal quality parameters is presented in Section 4. Section 5 outlines the measurement methodology used for RSSI and SNR measurements, data collection, and data processing. The presentation and analysis of the results obtained from the performed measurements on the developed test network are provided in Section 6. Section 7 concludes the paper with the key findings and description of the future work.

2. Related Work

Several studies have examined how LoRa end DV positions, transmission parameters, signal propagation barriers, and DV to GW distances affect signal quality in LoRa networks. In work [19], through a series of experiments, the authors validated Semtech’s claims regarding LoRa’s capabilities. Their results confirmed communication ranges exceeding 10 km under line-of-sight (LoS) conditions. However, in NLOS environments, obstacles such as buildings and vegetation significantly degraded performance, and achieving long battery life required continuous transmission parameter tuning. Findings presented in [19] also indicated that a single LoRa gateway could support up to 6000 end DVs, while maintaining a packet reception rate (PRR) of 70% or higher. The study also proposed an algorithm for optimal transmission parameter selection in conditions of NLOS transmission, balancing coverage and energy consumption.

In work [20], the authors investigated the impact of end DV positioning on LoRa network performance, focusing on communication link range, signal quality, and packet loss. The experiments showed rapid wireless signal degradation when obstacles were present within a distance of 300 m between DV and GW. In forested and urban environments, communication failed beyond distances of 350 m due to high noise levels and dense materials. In open environments, signal quality remained consistently high. The work concludes that the large-scale deployments in open areas with LOS signal transmission can achieve communication over 15 km. However, obstacles such as houses, terrain, or vegetation can significantly reduce communication range in small-scale deployments that are often limited to single buildings or neighborhoods. Additionally, the authors of [20] found that falling below the received signal strength indicator (RSSI) threshold of −120 dBm leads to corrupted or lost data packets. In contrast, the successful packet transmission was even achieved for SNR at −14 dB. Therefore, packet loss and end DV placement are critical considerations during LoRa indoor network planning, as end DVs risk permanent data loss.

In work [21], the authors conducted an experimental study on the reliability of LoRa networks. Their analyses focused on the impact of transmission parameters on effective end DV data rate and energy consumption. In the paper, the authors also examined how both transmission parameters and environmental factors influence network reliability across indoor, outdoor, and underground transmission settings. Results showed that adjusting transmission parameters to increase reception probability, particularly near the edge of coverage, often leads to inefficient trade-offs by significantly reducing data rate. Additionally, environmental conditions, especially increased air temperatures, were found to considerably lower RSSI values and adversely affect data packet reception in LoRa networks.

In [22], the authors conducted a performance analysis of LoRa networks in indoor environments. They developed a test network comprising four end DVs and one LoRa GW, for monitoring RSSI values to evaluate wireless signal quality. The study found that at short distances, RSSI values were high for lower spreading factors. Increasing the spreading factor reduces packet loss at the cost of lower data rates, making it unsuitable for high-throughput IoT applications. For transmission at greater distances, interference levels were significantly higher, indicating the need for LoRa transmission at higher spreading factors.

The authors in work [23] developed a test network to analyze packet reception ratio (PRR) using two LoRa end DVs, with one mounted on a car rooftop and the other on a boat mast. The LoRa GW was positioned on a university rooftop antenna tower, 24 m above sea level. Results showed that above solid land, over 80% of packets were successfully received within 5 km, and over 60% between 5 and 10 km. Beyond 10 km transmission distances, most packets in the tested LoRa network were lost. Communication realized with signal transmission above sea, reached nearly 30 km in range, with approximately 70% packet delivery success for communication distances below 15 km.

In study [24], which was conducted within an indoor office environment using a network of six strategically placed LoRaWAN end DVs and a single indoor GW, the authors systematically measured RSSI and SNR under varying environmental signal transmission conditions. The environmental signal transmission conditions include variations in temperature, humidity, carbon dioxide (CO₂), air pressure, and particulate matter levels. Results showed that transient phenomena such as reflections, scattering, interference, occupancy of transmission patterns, and furniture rearrangement can cause signal attenuation variations up to 10.58 dB. These findings confirm that incorporating environmental conditions and transmission patterns into occupancy dynamics can significantly enhance signal attenuation prediction.

In study [25], the authors conducted an experimental investigation within an eight-story residential building to examine the impact of NLOS conditions on LoRaWAN network performance. With the receiver placed on the rooftop and the transmitter systematically moved across floors, RSSI and SNR were recorded at a fixed spreading factor (SF7). Results showed improved RSSI and SNR on higher floors due to shorter distances and fewer obstacles, with minimal packet loss. Compared to related studies, similar trends in signal degradation and increased packet loss on lower floors were observed. Despite these losses, LoRaWAN maintained reliable coverage across multiple floors and demonstrated resilience to indoor reflections and scattering.

Study [26] presents experimental results on LoRaWAN network performance in a multi-story building, based on over 86 million transmissions collected over nearly two years using 390 battery-powered sensor DVs and four gateways. The maximum recorded transmission distance was 64 m, typically limiting coverage to several dozen rooms. Based on these findings, the authors recommend dense LoRa GW placement: one every 30 m and one every five floors, aligning more closely with wireless-fidelity (Wi-Fi) router distribution, while offering significantly higher energy efficiency. They also note that previous indoor range estimates of 200–600 m coverage appear overly optimistic.

Although previous research explores different approaches to analyze the wireless signal quality and maximal distance ranges for indoor signal propagation of LoRaWAN technology, the impact of the combination of different transmission parameters and material composition, number, and density of obstacles in indoor LoRaWAN signal propagation environments has not been comprehensively studied. Therefore, this study differs from earlier research by concurrently analyzing the impact of a combination of set of essential LoRaWAN DV transmission parameters, such as Tx power, spreading factor, packet size (PS), transmission interval, duty cycle and non-transmission parameters such as obstacles number, material composition and density at the signal propagation path between LoRa GW and DV positions, on signal quality within indoor LoRaWAN signal propagation environments. The experimental setup was based on a newly developed LoRaWAN test network implemented using the state-of-the-art software tools and technologies, focusing on short-distance LOS and NLOS data transmission between multiple LoRaWAN end DVs and multiple GWs located in a 10-floor building.

3. LoRa Technology and Wireless Signal Propagation Characteristics

LoRa communication technology operates at the first physical layer (PHY) of the open system interconnection (OSI) network model (Figure 1). The LoRa technology at the PHY defines radio frequency (RF) modulation, signal encoding, transmit power, and carrier frequencies to establish wireless links, and serves as one of the foundational platforms for the practical implementation of IoT applications.

An enhancement to LoRa technology is the LoRaWAN protocol, which operates at the Media Access Control (MAC) sublayer of the 2nd data link layer of the OSI model (Figure 1).

LoRaWAN represents a communication protocol that governs how devices utilize the lower-layer LoRa hardware, specifically in terms of transmission methods (grouped in three different A, B, and C classes), message formatting, LoRa DV addressing, adaptive data rate (ADR) selection, DC control, and DV network joining and activation procedure [27]. As a network protocol, LoRaWAN supports bidirectional communication, device mobility, security, and standardized geolocation services. Unlike the proprietary LoRa specification, the LoRaWAN protocol is openly defined by the LoRa Alliance™ and freely available for public use (Figure 1). It is engineered to enable low-power devices to communicate with Internet-connected applications over long-range wireless links. Additionally, it facilitates interoperability among different vendor LoRaWAN DVs using either LoRa CSS or Frequency-Shift Keying (FSK) modulation schemes [27].

LoRaWAN does not operate at the network layer 3 or higher OSI layers. Those OSI higher-layer functions (such as routing, session management, application protocols) are typically handled outside the LoRaWAN stack, either by network servers, application servers, or Internet protocol (IP)-based systems operating on top of or in parallel to LoRaWAN.

3.1. LoRa Transmission Parameters

LoRa technology can be characterized through several key transmission parameters, which have a significant impact on wireless signal propagation, signal quality at the position of signal reception, and overall network performance. Some of such LoRa transmission parameters that are used in analyses in this work include bandwidth (BW), coding rate (CR), payload length (PL), carrier frequency (CF), spreading factor (SF), transmit (Tx) power, and transmission duty cycle (DC).

The BW parameter defines the width of the frequency band utilized for transmission within a LoRa communication network. Increasing the bandwidth accelerates transmission data rates. Also, transmission at higher BW reduces the time-on-air (ToA) of the signal, which denotes the time necessary for a signal to travel from an end DV to a network GW, and vice versa. However, transmission at increased BW also increases noise accumulation, which subsequently reduces LoRa GW sensitivity. While LoRa technology supports BWs ranging from 7.8 kHz to 500 kHz, LoRaWAN restricts BW usage to 125 kHz, 250 kHz, or 500 kHz [28].

The Tx power defines the output effective isotropic radiated power (EIRP) used by the LoRa node for signal transmission. LoRa chipsets generally support Tx power levels ranging from −4 dB to +20 dB [29], with maximal levels reaching up to +30 dB. For example, in the US, the frequency band US915 defined by (USA, FCC Part 15.247) allows maximum Tx power of +30 dBm (1 W), but typical LoRa devices use +20–+27 dBm. In the Asia-Pacific area, the frequency band AS923 (Asia, Japan, Australia, etc.) allows a maximum Tx power range of +16 dBm (40 mW) to +22 dBm, depending on the sub-region. In the European EU868 frequency band defined by (ETSI EN 300 220) [30], a maximum Tx power of +14 dBm (25 mW) is allowed for normal end DVs, and up to +20 dBm (100 mW) is allowed in some sub-bands with duty-cycle restrictions.

The CR refers to the forward error correction (FEC) ratio used in LoRa network protocols to detect and correct transmission errors without requiring retransmission, thus avoiding additional channel congestion. The coding rate can be set to 4/5, 4/6, 4/7, or 4/8 [31]. A higher CR enables correction of a greater number of transmission errors, at the cost of reduced effective throughput.

The LoRa network SF determines how many bits are encoded into a single symbol, with each symbol consisting of 2^SF chirps. LoRa communications support SF values equal to SF = 7–12 (denoted further in the paper as SF7, SF8, SF9, SF10, SF11, and SF12) [32]. Thus, a higher SF means that each symbol is spread over a longer time (more chirps), which results in longer transmission, but also results in improved signal resistance to noise and interference. A higher SF lowers the minimum required SNR threshold for successful signal demodulation, thereby enhancing LoRa receiver sensitivity and communication range. However, higher SF also increases transmission ToA, which has an impact on an increase in transmission latency, end DV energy consumption, and a decrease in transmission data rate.

The PS defines the actual size of the transmitted LoRa end DV message in bytes. Payload sizes in LoRa networks are adaptable and typically range from 1 to 256 bytes [33,34], depending on the application requirements. The maximum allowed payload size in LoRa networks also depends on the value of spreading factor (SF) used in transmission, since higher SFs (e.g., SF12) result in much longer ToA for each transmission, which is limited by DC of regional regulatory constraints (e.g., in the EU868, US915, AS923, etc., bands). To comply with these limits and avoid exceeding the maximum ToA, the LoRaWAN protocol reduces the maximum allowed PS as the SF increases for different regional frequency bands (for example, in the EU868 band, it equals 222 bytes for SF7 or 51 bytes for SF12).

The carrier frequency (CF) specifies the frequency of the wireless signal carrier used for LoRa transmission. Some LoRa chipsets support carrier frequencies ranging from 137 MHz to 1020 MHz, with a tunable resolution of 61.035 Hz [35,36]. LoRa operates in sub-GHz unlicensed Industrial, Scientific and Medical (ISM) frequency bands, whose ranges vary by geographic region. For example, some regional allocations include the EU868 band (863–870 MHz) and 433 MHz band in Europe (Figure 1), the US915 band (902–928 MHz) in the US, the AS923 band (923–925 MHz) in the Asia-Pacific area, etc. [27].

Duty cycle (DC) is a percentage of time a LoRa DV is allowed to transmit on a given frequency channel. Different regions in the world have different prescribed DC levels. For example, in the EU868 band, sub-bands impose 0.1%, 1%, or 10% DC limits, which means a LoRa DV can only transmit for 3.6 s, 36 s, or 360 s per hour on a given frequency, respectively. For example, due to hardware limitations, in the EU868 subband, having a frequency range of 868.0–868.6 MHz, Tx power above 17 dB is subject to a 1% DC restriction [29].

3.2. Characteristics of LoRa Wireless Signal Transmission

The link quality in LoRa wireless communication networks is expressed with the two fundamental indicators known as the received signal strength indicator (RSSI) and signal-to-noise ratio (SNR). The SNR is the ratio between the power of the received signal and the background noise (expressed in dB). A higher SNR in LoRa networks means the signal is clearer compared to the noise at the position of signal reception. Due to CSS modulation, LoRa can decode signals with a negative SNR of about –20 dB at high spreading factors (e.g., SF12) [37]. The RSSI expressed in dBm measures the absolute power level of the received signal at the receiver input. It reflects how strong the signal is when it arrives at the LoRa GW or end DV. Typical RSSI values in LoRa network range from about −40 dBm (indicating a very strong signal for DVs close to the gateway) to −140 dBm (indicating a very weak signal, but sometimes still decodable) [38,39,40]. The RSSI and SNR are used in adaptive data rate (ADR) algorithms to dynamically select the optimal balance between data rate and coverage, SF, Tx power, etc.

Figure 2 presents an illustrative simplified visualization of the impact of signal attenuation caused only by the free space signal losses for different signal transmission distances, PSs, SFs, BWs, ToAs, data rates, and RSSI sensitivity levels in a typical LoRa network. It is noticeable that transmission at lower SFs and higher BWs enables transmission at higher data rates, and vice versa. Also, higher data rates in LoRa networks can be achieved for lower signal attenuations (lower RSSI) when lower LoRa SFs can be used for signal transmission. According to Figure 2, LoRa devices utilize higher SFs to cope with increased signal losses in situations when the attenuation of the signal is higher (indicated by lower RSSI), and vice versa. Also, utilizing a higher SF for lower RSSIs results in a longer data transmission (ToA), and vice versa [41]. Thus, the ToA increases in case of transmission of messages with larger payload sizes or when RSSI decreases due to the need for transmission at a higher SF, in order to cope with the impact of wireless signal attenuation (Figure 2).

3.3. Overview of Path-Loss Modeling for Indoor Wireless Signal Propagation

Indoor or mostly indoor wireless signal propagation is highly influenced by complex environmental factors such as walls, floors, furniture, and human movement, leading to significant variability in signal attenuation. As a result, numerous indoor path-loss models have been developed to more accurately characterize signal degradation in such environments [44,45,46,47,48,49].

Table 1. Overview of the most relevant indoor path-loss models proposed in the literature.

Path-Loss Model	Mathematical Expression (Parameter Description)
Floor Attenuation Factor (FAF) Model [44]	$PL\left(d\right)=PL\left(d_0\right)+10·n_{samefloor}·{\mathit{log}}_{10}\left({\displaystyle \frac{d}{d_0}}\right)+FAF$ (1) nsamefloor: path loss exponent for same floor measurements; d: distance [m]; FAF: additional loss per floor [dB] (e.g., 12.9 dB for 1 floor, 18.7 dB for 2, 24.4 dB for 3 floors);
Multi-Wall Model (MWM) [45]	$L=L_{FS}+L_c+\sum _{i=1}^l{k}_{wi}{L}_{wi}+k_f^{\left[{\displaystyle \frac{k_f+2}{k_f+1}} -b\right]}{L}_f(2)$ L: total path loss [dB]; LFS: free space loss between transmitter and receiver [dB]; Lc: constant loss [dB]; Lwi: loss of wall type i [dB]; kwi: number of penetrated walls of type i; Lf: loss between adjacent floors [dB]; kf: number of penetrated floors; l: number of wall types; b: empirical parameter.
Indoor Dominant Path (IDP) Model [46]	$PL={PL}_0+10n\mathit{log}\left({\displaystyle \frac{d}{d_0}}\right)+{\sum }_i{L}_{W_i}+{\sum }_j{L}_{B_j}$ (3) PL: total path loss [dB]; PL0: path loss at reference distance d0; n: path loss exponent (typically 2 for free-space); d: distance along the path between access point and receiver [m]; d0: reference distance [m]; LWi: wall loss for wall i [dB], dependent on wall material (e.g., drywall 2 dB, concrete 10 dB, thick concrete 15 dB, glass 2 dB); LBj: bending loss for direction change j [dB].
Optimized Indoor Propagation Model [47]	$L=k_1+G_a+k_2+\mathit{log}f+k_3\mathit{log}R+n_w\left(k_{4}P1+k_{5}P2\right)+k_6{m}_f$ (4) L: predicted total path loss [dB]; Ga: antenna gain; f: operating frequency [GHz]; R: transmitter–receiver distance [m]; nw: number of walls between Tx and Rx; mf: number of floors; P1, P2: are associate with the angle of incidence to a wall; k1–k6: coefficients optimized using measurement data via minimum least square error fitting.
Empirical Indoor Path Loss Model with Padé Approximant [48]	$PL={PL}_0+10\gamma {\mathit{log}}_{10}\left({\displaystyle \frac{d}{d_0}}\right)+X+f\left(n_p,a,b\right)$ (5) PL(d): total path loss [dB]; PL0: free-space path loss at reference distance d0 = 1 m; γ: path-loss exponent which depends on floor index (np—number of floors), obtained from γ = 1.51 + 0.853·np − 0.109·np2; d: distance between transmitter and receiver [m]; $X:$random variable representing short-term fading, expressed as $X = \sigma ·x$where x is Rayleigh distribution and σ is the standard deviation of measured data; $f(np, a, b):$Padé approximant term modeling inter-floor attenuation, $f\left(n_p,a,b\right){\displaystyle \frac{a+\frac{ab{n}_p}{2}+\frac{a{b}^2{n}_p^2}{12}}{1-\frac{b{n}_p}{2}+\frac{b^2{n}_p^2}{12}}}$where parameters a and b were empirically determined.
Meaningful Indoor Path-Loss Formula [49]	$PL=PL\left(d_0\right)+10\alpha \mathit{log}\left({\displaystyle \frac{d}{d_0}}\right)+\beta d$ (6) PL(d): total path loss [dB]; PL0: path loss at reference distance d0 = 1 m; α: path-loss exponent representing wave divergence and guiding effects; β: specific attenuation coefficient [dB/m] due to obstructions such as walls, ceilings, furniture, and people. β can be estimated using β ≅ ${\displaystyle \frac{L_w\left(f\right)}{\alpha }}$, where Lw is the average wall-penetration attenuation; d: distance between transmitter and receiver [m].

presents an overview of the most relevant indoor path-loss models proposed in the literature, which are applicable for LoRa network wireless signal propagation modeling. The presented overview in Table 1 includes the Floor Attenuation Factor (FAF) model, the Multi-Wall Model (MWM), the Indoor Dominant Path (IDP) model, the Empirical Indoor Path Loss Model with Padé Approximant, the Optimized Indoor Propagation (OIP) model, and the Meaningful Indoor Path-Loss Formula, with their mathematical formulations, parameter descriptions, and environmental assumptions.

The Floor Attenuation Factor (FAF) model extends the path-loss models’ log-distance formulation by adding a constant attenuation term per intervening floor [44]. It quantifies vertical propagation losses in multifloor buildings, where FAF increases with the number of floors between transmitter and receiver. This model provides a practical correction for vertical signal penetration through reinforced concrete slabs and ceilings [44].

According to Table 1, the Multi-Wall Model (MWM) incorporates individual wall and floor losses into the path-loss equation, allowing the total attenuation to be expressed as a sum of empirical constants per obstacle type [45]. This model is widely adopted for indoor planning, as it explicitly accounts for heterogeneous building materials.

In addition, the Indoor Dominant Path (IDP) model estimates path loss by considering only the dominant propagation route between transmitter and receiver, including direct and single-reflection components [46]. It introduces specific attenuation terms for walls and building junctions (expressed as $\mathsf{\Sigma }{L}_W+\mathsf{\Sigma }{L}_B)$, which depend on the wall type, incidence angle, and frequency (Table 1). This approach reduces path-loss modeling complexity while maintaining high accuracy in complex office or residential layouts.

Another Optimized Indoor Propagation model refines empirical predictions through optimization of attenuation factors for walls, doors, and floors [47]. It integrates geometrical parameters such as the angle of incidence and material-dependent reflection coefficients, obtained by minimizing the mean prediction error against measurement data (Table 1). The model enables accurate prediction for WLAN and 2.4 GHz systems within complex multi-room environments [47].

Also, Barbosa et al. proposed an Empirical Indoor Model based on the Padé approximant, combining polynomial and rational components to better fit measured attenuation across multiple floors [48]. The path-loss exponent is parameterized as a function of the number of floors ${(n}_p)$ and empirically determined coefficients a and b, by allowing a smooth transition between the same-floor and multi-floor cases (Table 1). The model enhances predictive accuracy for complex building structures.

In addition, the Meaningful Indoor Path-Loss Formula introduces a physically interpretable attenuation coefficient $\beta \hspace{0.17em}[\mathrm{d}\mathrm{B}/\mathrm{m}]$, representing cumulative losses from walls, ceilings, furniture, and people (Table 1). It expresses path loss by linking the signal attenuation coefficient β to wall penetration loss and average room dimensions. This formulation unifies geometric spreading and obstruction effects within a single analytical framework [49].

4. Architecture of the Test Network

In order to conduct an experimental evaluation of the impact of the number of obstacles, their material composition, and density on wireless signal quality in indoor LoRa network implementations, a novel specialized LoRa network test system architecture was designed and implemented. Figure 3 presents the developed real LoRa test network architecture, which combines various hardware technologies and software tools that jointly operate to enable the collection, processing, storage, and visualization of measured LoRa wireless link quality data.

Table 2. List of test network components and protocols with corresponding versions/types.

Test Network Component	Versions/Type
HTTP protocol	HTTP protocol version 1.1
MQTT protocol	MQTT version 3.1.1
LoRaWAN protool	LoRaWAN version 1.0.x (1.0.1–1.0.4)
Node-RED (Node.js)	Node-RED version 1.3
InfluxDB (Flux)	InfluxDB version 1.8.5
Grafana	Grafana version 7.5
TTN stack/region	The Things Stack v2, EU handler (ttn-handler-eu)
Docker/Compose	Docker version 3.8
End DV transceiver module/ antenna models	HOPERF Microelectronics RFM95W/96W/98W [50]/omnidirectional anntena
GW1 transceiver module/antenna type	IMST GmbH iC880A [51]/omnidirectional antenna
GW2 transceiver module/antenna type	Larid Sentrius RG1xx [52]/omnidirectional antenna
GW3 transceiver module/antenna type	RAKwireless RAK831 [53]/omnidirectional antenna
GW4 transceiver module/antenna type	IMST GmbH iC880A [51]/omnidirectional antenna

lists test network components and protocols with corresponding versions/types used in analysis. According to Figure 3, LoRa end DVs communicate with each LoRa GW. End DVs are realized as handheld LoRa end DVs with an HOPERF RFM95W transceiver module [50] (Table 2). Also, to have a more realistic test network, different LoRa transceiver modules are used for the realization of the four LoRa GWs (Table 2).

Additionally, Figure 3 indicates that the developed experimental LoRa test network architecture utilizes the following protocols and network components: the Message Queuing Telemetry Transport (MQTT) protocol, the public LoRaWAN infrastructure provided by The Things Network (TTN), the Node-RED application for flow-based development, the InfluxDB time-series database for data storage, and the Grafana platform for statistical analyses and visualization.

MQTT is a lightweight messaging protocol designed for regulating message exchange among machine-to-machine (M2M) and IoT DVs. The MQTT protocol version 3.1.1. used in analyses employs message exchange based on a publish/subscribe model (Table 2), and it is designed to be scalable, reliable, and efficient in low-bandwidth, high-latency, unreliable, and resource-constrained networks. The main element of the message exchange based on the MQTT protocol is an MQTT broker (Figure 3), which is the central server that receives all messages, filters them, and distributes them to the appropriate connected clients based on topic subscriptions.

The TTN is an open-source networking platform that enables LoRa DVs integration and application registration within the global LoRaWAN network (Figure 3). The TTN used in analyses is based on Things Stack v2 with the EU TTN handler (Table 2), which also enables the management of the routing of IoT data between IoT DVs and applications. Within the TTN system, it is necessary to define ToA and DC parameters. For example, a 1% DC requires a 99% × ToA delay before the next transmission.

Node-RED, open-sourced by the JavaScript (JS) Foundation in 2016, is a flow-based programming tool. In the developed test network, Node-RED version 1.3, built on Node.js which can run on all state-of-the-art operating systems, is exploited as an open-source JavaScript runtime environment (Table 2). Node-RED simplifies integration between networked nodes through support of various application programming interfaces (APIs) and cloud services. Thus, Nod-Red offers a browser-based interface that allows users to construct information flows among applications visually by connecting functional nodes (Figure 3). These flows can be deployed with a single click, minimizing the need for extensive coding skills. It is also well-suited for connecting edge computing and sensing devices, as well as for cloud environments.

The InfluxDB developed by the company InfluxData is an open-source time-series database designed for high-performance storage and management of timestamped data (Figure 3). The InfluxDB version 1.8.5 used in analyses supports the storage of time series data with an associated timestamp and value (Table 2), and includes the InfluxDB User Interface (UI) and Flux scripting language for data manipulation and visualization. The UI facilitates dashboard creation and offers a script editor for performing more advanced queries on the database.

Grafana is an open-source web application for data analytics and interactive visualization (Figure 3). The Grafana version 7.5 used in analysis allows users to construct customizable visualization dashboards through advanced database queries and transformation functions (Table 2). Grafana does not require storing data in its own backend, and instead it aggregates data from multiple sources, offering a unified and dynamic visualization environment.

4.1. Design and Implementation of a LoRaWAN Data Processing System

Figure 4 presents the LoRa DV test messages processing flow in the real test system developed for analyzing the impact of obstacle material composition, density, and LoRa network transmission parameters on indoor LoRa network signal quality. The LoRa end DV message processing flow in the developed test network system comprises transmission, collecting, storing, and analyzing stream messages sent from LoRa end DVs. The test infrastructure presented in Figure 3 is composed of the two LoRa end DVs and four GWs (Figure 3). Communication between end DVs and GWs is realized through the exchange of LoRa messages, which are coordinated via the LoRaWAN protocol (Figure 4). According to Figure 3, the GWs are connected to the Internet through high-bandwidth wired connections.

Figure 4 shows that the GWs exchange data with the TTN platform via Hypertext Transfer Protocol -HTTPv1.1 (Table 2). All end DVs and GWs are pre-registered within the TTN infrastructure. The TTN handles LoRa message decoding and forwards the decoded payloads using the MQTT protocol to a locally set application server via a publicly exposed MQTT broker (Figure 4). At the application server, the Node-RED flow-based programming tool was used for enabling subscription to the relevant MQTT topic. Each received message undergoes further processing within Node-RED, where metadata is reformatted for storage in the InfluxDB database (Figure 4). Finally, the collected signal quality parameters (RSSI and SNR) are visualized in the Grafana dashboard tool. For the statistical analysis and result interpretation, the stored data are extracted from the InfluxDB database in a specific (CSV) format, which is suitable for further metadata analysis on the local host (Figure 4). Connection of the local host for configuration of NodeRed application, InfluxDB database, and Grafana tool is realized through direct communication of the local host over dedicated communication ports opened for each application on the application server that hosts each application (Figure 4).

4.1.1. Containerized Architecture of the Application Server

To ensure usability, operational efficiency, and service-oriented scalability, the application server was deployed as a Docker version 3.8 container-based virtualization platform (Table 2). The main architecture components of the application server are presented in Figure 5. They include a file system with Docker area containing Grafana and InfluxDB volumes, and virtualized services that are executed in separate InfluxDB, Grafana, and Node-RED containers. Thus, the application server was architected as a collection of containers that operate collaboratively using shared resources and communication channels (Figure 5). Each computational component is defined as a service, representing an abstract unit of the application, instantiated one or more times from a Docker image with specified configurations.

The Docker Compose tool was utilized to enable simplified installation, portability, networking, and isolation of the containerized system components. The application server was hosted on a Linux-based server in order to facilitate seamless interoperability with the Docker environment. All required container and orchestration configurations were defined within the DockerFile and docker-compose.yml files. The docker-compose.yml file, structured in YAML format, enables a platform-independent container-based application definition through specifying the application version, services, network settings, volumes, configurations, and secrets.

The application server developed for this test experiment contains three core services that include Node-RED flow-based programming tool, InfluxDB time-series database, and Grafana data monitoring and visualization tool (Figure 5). All containers of a specific service are identically instantiated. Under the services section of the container configuration, individual service parameters were defined, including container name, DockerFile path, port mapping, persistent storage settings, and network bridge connections (Figure 5). The services were deployed with the following standard port number assignments for Node-RED on 1880, for InfluxDB on 8086, and for Grafana on 3000 (Figure 4 and Figure 5).

The Volumes element in the file system specifies persistent data storage for InfluxDB and Grafana on the host system (Figure 5). For the InfluxDB service, this storage is used to save all collected data and backup copies in case of errors. In the case of the Grafana service, the corresponding volume stores the configuration settings of the Grafana dashboards.

4.1.2. Inter-Service Communication

Inter-service communication was established using Docker’s bridge networking technology (Figure 5). Operating at the data link layer of the OSI model, a bridge forwards traffic between network segments. In this context, Docker employs a software-based bridge that facilitates communication between containers connected to the same bridge network, while isolating them from other containers that are not connected to that network. The bridge driver autonomously configures routing rules within the host virtual machine, thereby preventing direct communication between containers on different bridge networks [54]. Each service communicates through a dedicated Ethernet interface (represented as veth (virtual Ethernet)) with a unique IP address (Figure 5). The veth interface acts as a virtual Ethernet tunnel, which is functionally equivalent to a real cable Ethernet connecting each service to the network bridge, thus enabling communication among all services in the same network.

4.1.3. Node-RED Application

At the application server side, the use of the Node-RED application version 1.3 enabled the development of a simple and user-friendly interface (Table 2), that defines an entire LoRa message metadata collection and process flow. The interconnection of all six Node-RED nodes participating in the developed Node-RED processing logic flow is presented in Figure 6.

The preconfigured TTN MQTT broker was integrated into the Node-RED environment through the MQTT client node (Figure 4). In Figure 6, the node labeled “TTN IN” functions as an MQTT client subscribed to the global topic (+/devices/+/up). This topic was chosen for its simplicity, as it aggregates all uplink traffic, avoiding the need for separate subscriptions to individual LoRa DV topics. During the configuration of the “TTN IN” node, the MQTT server Uniform Resource Locator (URL) and relevant API keys were also specified. When a message is transmitted by an end LoRa DV via TTN (Figure 4), it is received at the “TTN IN” node and passed along the processing flow (Figure 6). The “TTN-log” is a debug node used to locally log each received message, facilitating easier LoRa message error detection and functional verification.

The subsequent “Unpack” node extracts relevant metadata and parameters from the message payload for further processing (Figure 6). The “Transformation” node converts the unpacked metadata from a raw JSON format into a structure suitable for storage in the InfluxDB time-series database. An additional debug node, “influx-debug,” is included to verify the correct functioning of the previous nodes (Figure 6). The final node, “TTN-to-influx-metadata”, serves as the output node responsible for transmitting the processed JSON object to InfluxDB. Through Docker’s bridge network, this node connects to the InfluxDB service (Figure 5). Successful data transmission requires configuration of the target database name, measurement (table) name, service URL, and an authentication token, which must be pre-generated within the InfluxDB platform.

4.1.4. InfluxDB and Grafana Applications

Each LoRa end DV message received through a GW is recorded in the InfluxDB as a timestamped entry (Figure 4). Since one end DV transmission is received by four GWs (Figure 3), it results in four separate data points. The database table schema is dynamically generated based on the structure of the input JSON object, which represents one of the core advantages of InfluxDB. Following best practices recommended for InfluxDB development, metadata fields that are frequently queried and filtered are indexed for efficient querying by storing them as tags in InfluxDB. The following elements were stored as tags: application identifier (ID), LoRA end DV identifier (ID), carrier frequency, payload size, spreading factor, number of receiving GWs, acknowledgment status, gateway ID, and GW name. The application ID uniquely identifies the TTN application, the device ID identifies the end DV, and the GW ID identifies the receiving GW. These identifiers allowed precise mapping between the LoRa end DV transmitting a message and the LoRa GW that receives this message. In contrast, fields (which are not indexed) are used to store parameter values that do not significantly impact database query speed. The following elements were stored as fields: message reception time, SNR, RSSI, and message counter.

This InfluxDB structure allows for straightforward correlation of the SNR and RSSI values with transmission parameters such as the Tx power, SF, and PS. Data queries and filtering were executed using the Flux scripting language via the InfluxDB Data Explorer interface (Figure 4). Flux scripting language enables advanced data selection, transformation, and aggregation operations. The graphical Data Explorer environment supports the creation of custom dashboards for intuitive data visualization. Filtering operations in InfluxDB were performed using the filter function, specifying both the parameter to be filtered and the target value.

Following the metadata processing and storing in the InfluxDB, visualization of the obtained metadata in the Grafana tool is performed for identifying patterns in measured data and for supporting analytical conclusions (Figure 4). Grafana offers native support for the Flux scripting language, which enables the full use of its functions for advanced LoRa signal quality (RSSI and SNR) measured data visualization and analysis. Within the test network, Grafana dashboards were configured using the Grafana configuration channel to display parameter relationships between the two end DVs (Figure 4), filtered by GW and tracking parameter. The integration between the Grafana web-based visualization platform and the InfluxDB time-series database was streamlined through the use of a Docker bridge network (Figure 5). To configure Grafana’s data source, the InfluxDB URL was specified, with Flux as the query language, and basic authentication with the provided authentication token. The authentication token must be pre-generated within the InfluxDB service and added to the HTTP authorization header during setup. This configuration leverages the InfluxDB inherent isolation from other services and ensures security provided by the Docker environment (Figure 5).

4.1.5. Collected Metadata Structure

Every message received by any LoRa GW is logged with corresponding metadata. Figure 7 presents an example of the collected anonymized metadata for one LoRa end DV. According to Figure 7, a single transmission from one end DV produces metadata that contains information related to each GW (Figure 7), which is then used in subsequent analyses. Collected metadata includes the reception timestamp, carrier frequency, spreading factor, bandwidth, coding rate, and a list of GWs that received the corresponding message. Metadata related to each GW includes GW time, measured SNR/RSSI values, LoRa message counter state, and contains an additional set of GW-specific information (Figure 7).

Further processing involves extracting separate metadata entries for each GW. Specifically, for each of the four receiving GWs, an individual metadata record is generated, combining gateway-specific measurements (e.g., RSSI, SNR, timestamp, message counter) with shared message parameters. As a result, a single original record for one LoRa DV yields four distinct datasets (Figure 4), which are stored in the InfluxDB to facilitate detailed performance analyses for each GW individually.

5. Measurement Setup

Following the implementation of the developed test network presented in Figure 3, the real measurements of RSSI and SNR signal quality parameters for different combinations of LoRa DV transmission parameters were performed. The primary objective was to analyze how a specific combination of different LoRa transmission parameters and obstacles material composition and density located at the wireless signal transmission path, affects the RSSI, SNR, and overall LoRa network performance within a mostly indoor signal propagation environment. To ensure measurement accuracy, each measurement set (MS) of LoRa end DV transmission parameters was tested during a minimum period of 24 h.

5.1. Test Network Topology

The end DVs and GWs in the test network were strategically located in different parts of the faculty building, which hosts the allocation of LoRa network nodes. The illustrated visualization of the building with locations of DVs and GWs having allocations approximately proportionally scaled to their real space distances is presented in Figure 8. According to Figure 8, each LoRa end DV can establish a wireless connection with each GW. The faculty building is divided into three sections (A, B, and C) and has approximately 30,000 m² of indoor area spread over 10 floors.

Table 3. Names and locations of end DVs and GWs in the test LoRa network.

LoRa DV Name	LoRa DV Location	LoRa GW Name	LoRa GW Room Location
DV1	Lab. A507	GW1	Amphitheater A100
DV2	Lab. A509	GW2	Lab. A501
		GW3	Lab. A507
		GW4	Building roof

contains names, physical locations, and room number notations of the LoRa end DVs and GWs located in the faculty building. The first letter in the notation of LoRa DV location in Table 3 is the notation of the building section (A, B or C), while the first number in the notation of LoRa DV location indicates the floor number in which the room containing LoRa end DV is located (e.g., location A501 represents room 501 located in A buiding section at the 5th floor).

The end DV1 and DV2 were placed in section A of the building, located on the fifth floor, in laboratories A507 and A509, respectively (Table 3). Similarly, gateways GW1, GW2, and GW3 were also positioned within section A of the building. The GW1 was installed on the ground floor in amphitheater A100, while GW2 and GW3 were located on the fifth floor, in laboratories A501 and A507, respectively. The fourth gateway, GW4, was installed on the rooftop of the building in section C as an outdoor-mounted GW.

The communication between end DVs and GWs forms eight distinct LoRa wireless communication links (Figure 8). An overview of all communication links and distances between LoRa end DVs and GWs is presented in

Table 4. Overview of distances and obstacle material composition and densities for all combinations of LoRa end DV and GW connections.

LoRa Communication Link (Location of LoRa Nodes)	Signal Propagation Environment	Number/Material Composition of Walls on Communication Links Among Communicating Nodes	Approximate Direct Distance Length (m)	Density of Obstacles Degrading the Signal Propagation
DV1 (Lab. A507)–GW1 (Amphitheater A100)	Indoor and outdoor	2/Reinforced concrete/	62.00	Large
DV2 (Lab. A509)–GW1 (Amphitheater A100)	Indoor and outdoor	2/Plasterboard and3/Reinforced concrete	80.00	Large
DV1 (Lab. A507)–GW2 (Lab. A501)	Indoor	3/Plasterboard	24.20	Medium
DV2 (Lab. A509)–GW2 (Lab. A501)	Indoor	2/Plasterboard	6.70	Medium
DV1(Lab. A507)–GW3 (Lab. A507)	Indoor	0/LOS	1.00	Small
DV2 (Lab. A509)–GW3 (Lab. A507)	Indoor	2/Plasterboard	20.80	Medium
DV1 (Lab. A507)–GW4 (Building roof)	Indoor and outdoor	5/Reinforced concrete	61.40	Very large
DV2 (Lab. A509)–GW4 (Building roof)	Indoor and outdoor	5/Reinforced concrete	80.00	Very large

. Table 4 also indicates the obstacles’ material composition, the number of obstacles, the obstacle densities, and the distance between DVs and GWs, which impact LoRa wireless signal propagation on communication links. Figure 8 shows the placement of LoRa network nodes relative to their distances and the number of obstacles listed in Table 4.

As indicated in Table 4 and Figure 8, LoRa end DVs and GWs were strategically placed throughout the faculty building to test how the distance among LoRa nodes, material composition, and density of indoor obstacles affect LoRa signal quality.

5.2. Measurement Procedure

The measuring experiments involved variations in several key LoRa transmission parameters.

Table 5. Overview of the LoRa GWs, end DVs, and corresponding transmission parameters used in the analysis.

Parameter	Name/Velue
LoRa end devices	DV1, DV2
LoRa gateways	GW1, GW2, GW3, GW4
Transmit (Tx) power (dB)	2, 10, 20
Spreading factor	7, 8, 9, 12
Number of payload length (Bytes)	1, 25, 50, 100, 200
Transmit period/duty cycle (min)	2, 3, 4, 5, 6, 8, 10, 13, 17, 18, 30
Bandwidth (kHz)	125
Carrier frequency (MHz)	867–869
Coding rate (CR)	4/5

presents an overview of the corresponding transmission parameters used in the analyses performed with four LoRa GWs and two end LoRa DVs (Figure 8). The impact of different values of SFs on indoor wireless signal propagation quality was tested for SF values equal to SF7, SF8, SF9, and SF12. The LoRa DV Tx power was varied across Tx power levels equal to 2 dB, 10 dB, and 20 dB (Table 5). Additionally, the message PS of transmitted packets was equal to 1 B, 25 B, 50 B, 100 B, and 200 B.

The combinations of these transmission parameters enabled a comprehensive investigation of how the LoRa DV transmission parameters influence signal quality metrics, such as RSSI and SNR, in indoor propagation environments. The transceivers of the LoRa end DVs and the LoRa GWs were equipped with omnidirectional antennas (Table 2). Both LoRa node types (end DVs and GWs) operate at the EU 863 MHz–870 MHz frequency range (Table 5), since the use of this radio spectrum is allowed in the country in which tests were performed (Republic of Croatia). This frequency band is for the LoRa network defined by the country’s national regulation General License [55], which mandates the application of the ETSI standard EN 300 220 that defines technical requirements for Short-Range Devices (SRDs) operating in the 25 MHz to 1 GHz frequency range.

More specifically, the testing is performed for LoRa nodes transmission according to the EU frequency plan EU863-870 [56], at the frequency range of 867 MHz to 869 MHz, with a fixed bandwidth of 125 kHz and the coding rate equal to 4/5 (Table 5).

5.3. Measurement Sets of the LoRa End DV Transmission Parameters

Comparison of different measurement sets (MSs) of the LoRa end DVs transmission parameters is presented in

Table 6. Comparison of MS transmission parameters for DV1 and DV2 during the same period of the LoRa message transmissions.

Day	MS/DV1	DV1 Configuration (SF, Tx Power, PS, DC)	ToA (ms)	MS/DV2	DV2 Configuration (SF, Tx Power, PS, DC)	ToA (ms)
1 July	MS10/DV1	SF7, 2 dBm, 25 B, 4 min	61.7	MS12/DV2	SF12, 2 dBm, 25 B, 30 min	1482.75
30 June–1 July	MS10/DV1	SF7, 2 dBm, 25 B, 4 min	61.7	MS9/DV2	SF12, 20 dBm, 1 B, 30 min	827.39
1–2 July	MS11/DV1	SF9, 2 dBm, 25 B, 13 min	205.82	MS12/DV2	SF12, 2 dBm, 25 B, 30 min	1482.75
2–3 July	MS14/DV1	SF9, 10 dBm, 25 B, 13 min	205.82	MS13/DV2	SF7, 10 dBm, 25 B, 4 min	61.7
5–6 July	MS15/DV1	SF12, 10 dBm, 25 B, 30 min	1482.75	MS16/DV2	SF7, 20 dBm, 25 B, 5 min	61.7
6–7 July	MS17/DV1	SF9, 20 dBm, 25 B, 13 min	205.82	MS18/DV2	SF12, 20 dBm, 25 B, 30 min	1482.75
7–8 July	MS19/DV1	SF7, 2 dBm, 50 B, 6 min	97.54	MS20/DV2	SF9, 2 dB m, 50 B, 18 min	328.7
8–9 July	MS21/DV1	SF12, 2 dBm, 50 B, 30 min	2301.95	MS22/DV2	SF7, 10 dBm, 50 B, 6 min	97.54
9–12 July	MS23/DV1	SF9, 10 dBm, 50 B, 18 min	328.7	MS24/DV2	SF12, 10 dBm, 50 B, 30 min	2301.95
12–13 July	MS25/DV1	SF7, 20 dBm, 50 B, 6 min	97.54	MS26/DV2	SF9, 20 dBm, 50 B, 18 min	328.7
13–14 July	MS28/DV1	SF7, 2 dBm, 100 B, 10 min	174.34	MS27/DV2	SF12, 20 dBm, 50 B, 30 min	2301.95
27–28 June	MS3/DV1	SF12, 2 dBm, 1 B, 30 min	827.39	MS4/DV2	SF7, 10 dBm, 1 B, 2 min	25.86
14–15 July	MS30/DV1	SF7, 20 dBm, 100 B, 10 min	174.34	MS29/DV2	SF7, 10 dBm, 100 B, 10 min	174.34
15–16 July	MS32/DV1	SF7, 10 dBm, 200 B, 17 min	317.7	MS31/DV2	SF7, 2 dBm, 200 B, 17 min	317.7
16–19 July	MS33/DV1	SF7, 20 dBm, 200 B, 17 min	317.7	MS34/DV2	SF9, 2 dBm, 100 B, 30 min	553.98
19–20 July	MS35/DV1	SF9, 10 dBm, 100 B, 30 min	553.98	MS36/DV2	SF9, 20 dBm, 100 B, 30 min	553.98
20–21 July	MS38/DV1	SF9, 2 dBm, 1 B, 8 min	103.42	MS37/DV2	SF7, 2 dBm, 1 B, 4 min	25.86
21–22 July	MS39/DV1	SF12, 20 dBm, 1 B, 30 min	827.39	MS40/DV2	SF8, 20 dBm, 200 B, 30 min	563.71
22–23 July	MS41/DV1	SF8, 20 dBm, 200 B, 30 min	563.71	MS42/DV2	SF8, 2 dBm, 200 B, 30 min	563.71
23–26 July	MS44/DV1	SF8, 20 dBm, 200 B, 30 min	563.71	MS43/DV2	SF8, 20 dBm, 200 B, 30 min	563.71
28–29 June	MS5/DV1	SF9, 10 dBm, 1 B, 8 min	103.42	MS6/DV2	SF12, 10 dBm, 1 B, 30 min	827.39
29 June	MS5/DV1	SF9, 10 dBm, 1 B, 8 min	103.42	MS7/DV2	SF7, 20 dBm, 1 B, 3 min	25.86
29–30 June	MS8/DV1	SF9, 20 dBm, 1 B, 8 min	103.42	MS7/DV2	SF7, 20 dBm, 1 B, 3 min	25.86

. Data collection based on measurements of the LoRa wireless signal quality parameters lasts for a continuous period of one month. During this period, 42 distinct measurement sets (MSs) were generated, each defined by a unique combination of the LoRa DV transmission parameters configuration presented in Table 6.

Since each message sent by every LoRa DV was received over four different LoRa GWs (Figure 8), four sets of metadata were produced for each LoRa end DV transmission (Figure 7), resulting in a total of 24,397 raw metadata entries. Each measurement realized through continuous transmission of each LoRa DV message on the specific DV-GW communication link was conducted over a duration of approximately 24 h (Figure 8 and Table 4). However, the number of messages transmitted in the 24 h period differs (

Table 7. Comparison of packet delivery ratio and average GW packet delivery ratio for LoRa message transmission of all analyzed MSs.

Dataset/ Device	DV Configur. (SF/Tx/Power (dBm)/PS (B)/ DC (min))	No. of Transm. LoRa Message.	No. of Receiv. LoRa Message.	PDR (%)/ Average GWPDR (%)	Dataset/ Device	DV Configur. (SF/Tx Power (dBm)/PS (B)/ DC (min))	No. of Transm. LoRa Message.	No. of Receiv. LoRa Message.	PDR (%)/ Average GWPDR (%)
MS3/DV1	SF12/2/1/30	48	48	100.00/77.08	MS24/DV2	SF12/10/50/30	147	147	100.00/100.00
MS4/DV2	SF7/10/1/2	622	306	49.20/48.6	MS25/DV1	SF7/20/50/6	313	312	99.68/97.26
MS5/DV1	SF9/10/1/8	193	193	100.00/99.87	MS26/DV2	SF9/20/50/18	105	105	100.00/98.80
MS6/DV2	SF12/10/1/30	51	51	100.00/99.51	MS27/DV2	SF12/20/50/30	47	47	100.00/100.00
MS7/DV2	SF7/20/1/3	307	306	99.67/99.43	MS28/DV1	SF7/2/100/10	145	145	100.00/94.79
MS8/DV1	SF9/20/1/8	178	178	100.00/99.44	MS29/DV2	SF7/10/100/10	144	144	100.00/99.13
MS9/DV2	SF12/20/1/30	45	45	100.00/100.00	MS30/DV1	SF7/20/100/10	143	142	99.3/96.69
MS10/DV1	SF7/2/25/4	364	363	99.73/89.98	MS31/DV2	SF7/2/200/17	87	87	100.00/99.14
MS11/DV1	SF9/2/25/13	123	123	100.00/99.38	MS32/DV1	SF7/10/200/17	87	87	100.00/95.6
MS12/DV2	SF12/2/25/30	53	53	100.00/99.53	MS33/DV1	SF7/20/200/17	255	255	100.00/98.53
MS13/DV2	SF7/10/25/4	306	306	100.00/99.59	MS34/DV2	SF9/2/100/30	145	145	100.00/99.65
MS14/DV1	SF9/10/25/13	290	289	99.65/98,96	MS35/DV1	SF9/10/100/30	46	46	100.00/97.83
MS15/DV1	SF12/10/25/30	50	50	100.00/97.5	MS36/DV2	SF9/20/100/30	46	46	100.00/98.37
MS16/DV2	SF7/20/25/5	297	297	100.00/99.24	MS37/DV2	SF7/2/1/4	359	359	100.00/99.09
MS17/DV1	SF9/20/25/13	121	121	100.00/98.75	MS38/DV1	SF9/2/1/8	180	180	100.00/99.16
MS18/DV2	SF12/20/25/30	53	53	100.00/99.53	MS39/DV1	SF12/20/1/30	48	48	100.00/100.00
MS19/DV1	SF7/2/50/6	234	233	99.57/99.54	MS40/DV2	SF8/20/200/30	48	48	100.00/96.88
MS20/DV2	SF9/2/50/18	78	78	100.00/99.04	MS41/DV1	SF8/20/200/30	49	49	100.00/97.92
MS21/DV1	SF12/2/50/30	48	48	100.00/100.00	MS42/DV2	SF8/2/200/30	49	49	100.00/96.36
MS22/DV2	SF7/10/50/6	237	237	100.00/99.16	MS43/DV2	SF8/20/200/30	136	136	100.00/98.71
MS23/DV1	SF9/10/50/18	245	244	99.59/98.67	MS44/DV1	SF8/20/200/30	137	133	97.08/92.65

), due to EU LoRaWAN duty cycle restrictions and mandated TTN fair-use of ToA policy. More specifically, the TTN limits uplink airtime to 30 s per LoRaWAN node per day, and allows only 10 downlink messages (towards end DVs) per node per day. While these constraints do not apply to private LoRa networks, adherence to EU LoRaWAN regulations and TTN LoRaWAN limitations remains mandatory, and for that reason, they were respected during test measurements [57].

5.4. Measurement Setup of LoRa End DV Transmission Parameters

The test network configuration presented in Figure 3 and Figure 8 and the corresponding message process flow presented in Figure 4 enable measuring, data collection, and measured data monitoring in real time, with the ability to display metadata for each received message on a Grafana dashboard for the selected GW. An example screenshot of the Grafana dashboard interface presenting RSSI and SNR measurements in real time for all communication links between DV2 and each of the four GWs for a specific combination of LoRa end DV transmission parameters (example presented for MS9), is shown in Figure 9. The SNR plots on the left side of Figure 9 indicate low levels of ambient noise relative to signal strength. The RSSI plots on the right side of Figure 9 do not reveal significant variations in RSSI levels on the same GW, while significant differences among measured RSSI levels can be noticed among GWs. Following the completion of data collection for a given MS, further processing of the metadata was performed. For each DV-GW communication link, the raw RSSI and SNR data are processed to compute the following statistical metrics:

• Mean RSSI and SNR levels, which represent the arithmetic average of all recorded values.
• Minimum and maximum RSSI and SNR levels, which represent the lowest and highest measured values, respectively.

These statistics provide a basis for gaining conclusions regarding wireless signal quality and its variability across different LoRa DV transmission parameter configurations.

6. Results and Discussion

This section presents the analyses of the results acquired through the comprehensive measurement process performed on the developed test network presented in Figure 3 and Figure 8. The presented results are analyzed in terms of the impact on the RSSI and SNR levels of different combinations of the transmission parameters, such as the SF, Tx power, LoRa message PS, and DC (ToA). Also, this section presents packet delivery ratio (PDR) analysis for all MSs. Additionally, the impacts of the indoor wireless signal obstacles, their material composition, and signal propagation path length on wireless signal quality indicators (SNR and RSSI) are thoroughly discussed. For the purpose of analysis, only the results of specific MSs are presented and analyzed in this work.

6.1. Packet Delivery Ratio Analysis of the Test LoRa Measurements

Table 7 presents the packed delivery ratio (PDR) and average GW packet delivery ratio (GWPDR) comparison of LoRa message transmission for each MS. While the PDR represents the relationship between the overall successfully received by the network and the total sent LoRa messages for a specific MS, the GWPDR represents the average PDR for all four GWs for a specific MS (Table 7).

According to the PDR results presented in Table 7, almost all MSs have a 100% PDR of LoRa message transmissions, which means that each LoRa message sent by one end DV during the measurement period has been received by the network at the application server side through at least one of four GWs. The difference from these results is noticed only for MS4 and MS10, having PDR equal to 49.2% and 99.73%, respectively (Table 7). This PDR result is caused by the TTN restrictions related to the DC and ToA length of LoRa message transmission. Due to these TTN restrictions, the transmission of some LoRa messages was simply discarded by the TTN, since the set LoRa end DV transmission period (DC) and ToA transmission duration violates the TTN restrictions (equal to the 1% (36 min) per hour DC transmission limitation in the EU868 band).

The average GWPDR presented in Table 7 shows very high but mostly non 100% PDR results, which is a consequence of the packet losses in transmission between end DVs and GWs. The average GWPDR value lower than 100% means that neither of the four GWs in the test case performed for a specific MS, did not received during the measuring period, all LoRa messages sent from the LoRa DV. Main causes of loss of the LoRa messages during transmission are related to the wireless channel state fluctuations between LoRa end DVs and GWs, which can be impacted by inter-SF interference, material composition structure, density, and obstacle quantity including the movement of persons on the wireless signal transmission path.

The statistics of the overall successful LoRa data message transmissions per GW for all MSs and for all GWs are presented in Figure 10. According to Figure 10, the LoRa average PDR per GW (GWPDR) is somewhat lower for those GWs that are further from UDs (GW4 on the building roof and GW1 at the ground floor (Figure 8)), compared to those GWs that are located closer to the LoRa end DVs (GW3 and GW2). This is a direct consequence of the impact of the higher obstacle number, obstacle material composition, structure, and density, which are present on the communication path between LoRa end DV and GWs having higher spacing. More obstacles with dense material composition structure will have a stronger impact on the loss of some LoRa transmitted messages, which results in a lower GWPDR.

The obtained PDR and average GWPDR results presented in Table 7 also confirm that the performed measurements have not been significantly impacted by the inter-SF interference, which occurrence and impact were analyzed in [15,16], and that the impact of inter-SF interference can be neglected. The main reasons for this can be found in the following facts:

• The test network is composed of only two end DVs (Figure 8 and Table 4), which means that a very low number of end DVs and their density in the tested LoRa network reduce the chances for simultaneous transmission and inter-SF interference occurrence.
• The LoRA transmission parameters selected in simultaneous testing of two MSs (Table 6) are characterized by high diversity among transmission periods (DC) between LoRa messages (ranging from 2 min to 30 min). This diversity is based on combining either long DCs or DCs having very distinct transmission periods in terms of DC duration. This results in a low possibility of the simultaneous transmission of both the LoRa end DVs and consequently in no inter-SF interference.
• According to Table 6 and Table 7, the LoRA transmission parameters selected in simultaneous testing of two MSs (Table 6) are characterized by high diversity among the periods of transmission duration (ToA), which range between 25 ms and 2.3 s. This results in a low possibility of LoRa message transmission collisions among end DVs.

The measurement results visualized on the Grafana tool (presented as an example for MS9 in Figure 9) show very stable RSSI values for all measurements and give, during the measurement process, a visual confirmation about the lack of inter-SF interference in performing the measuring tests. Therefore, the lack of inter-SF interference confirms that the analyzed combinations of the LoRa end DV transmission parameters in each MS were appropriately selected for simultaneous measurements performed during a single one-month period (Table 6). It also confirms the validity of test results, since they were not impacted by significant inter-SF interference.

6.2. Impact of Tx Power on RSSI and SNR

To analyze the impact of the LoRa DV Tx power on wireless signal quality in indoor propagation environments, the measurement sets (MS12, MS15, MS18, MS21, MS24, and MS27) having LoRa transmission parameters with the fixed spreading factor SF12 and packet size equal to the 25 B and 50 B for three distinct Tx power levels equal to the 2 dB, 10 dB, and 20 dB were analyzed. Details of each MS with corresponding combinations of LoRa transmission parameters used in the analyses are presented in

Table 8. Overview of measurement sets with corresponding LoRa transmission parameters used in the analysis of the impact of LoRa DV Tx power on wireless signal quality.

User Device	Location	Measurement Set	Number of Collected Metadata	SF	Tx Power (dB)	Packet Size (B)	DC -Total Tx Period (min)
DV2	LAB A509	MS12	211	12	2	25	30
DV1	LAB A507	MS15	195	12	10	25	30
DV2	LAB A509	MS18	211	12	20	25	30
DV1	LAB A507	MS21	192	12	2	50	30
DV2	LAB A509	MS24	587	12	10	50	30
DV2	LAB A509	MS27	188	12	20	50	30

. The highest SF (SF12) was selected for analysis in all MSs, in order to capture the behavior of the LoRa network signal quality for transmission with the SF offering the highest robustness on indoor signal propagation degradation. The overall daily LoRa DV message transmission interval was set to 30 min, based on the ToA restrictions of the TTN.

The measuring results for measurement sets MS12, MS15, and MS18 are presented in Figure 11a, and the measuring results for measurement sets MS21, MS24, and MS27 are presented in Figure 11b. The results are presented for signal transmission between two end DVs and all GWs. Therefore, Figure 11a,b presents the mean, minimum, and maximum values of measured SNR and RSSI values for different combinations of transmission parameters having equal SF12, different Tx powers, and different LoRa message packet sizes equal to 25 bytes (Figure 11a) or 50 bytes (Figure 11b), respectively.

According to Figure 11, the LoRa end DV2 transmission at higher Tx power results in similar or slightly improved SNR and RSSI levels. The exception to this result is obtained for the communication links between the DV2 and GW1, showing a decrease in the SNR and RSSI values for an increase in the Tx power of the LoRa DV. These results, although unexpected, are a consequence of the impact of signal attenuation caused by human activity in the faculty building, since measurements for MS24 and MS27 have not been performed at the same time and day. These results reveal the important impact of human presence and movement on the attenuation of wireless signals in the LoRa indoor propagation environments.

In addition, the measuring results presented in Figure 11 for all communication paths of the DV1 are different than those of the measuring results obtained for the DV2, since the DV1 is positioned in a different location in the building (Figure 8), and has a different number and obstacle densities on the communication path with different communication path lengths (Table 4). However, the results presented in Figure 11 show that an increase in the Tx power for transmissions at the highest SF generally does not have a significant impact on the improvement in LoRa wireless signal quality, as most indoor communication paths have lengths of up to 80 m.

6.3. Impact of LoRa Message Payload Size on RSSI and SNR

The subsequent analysis examines the effect of different packet sizes on wireless indoor signal quality, indicated by RSSI and SNR metrics. Also, the impact of transmission interval (DC), which defines the period between the LoRa message stream transmission of subsequent LoRa messages from each end DV to GWs, is analyzed.

Table 9. Overview of measurement sets with corresponding LoRa transmission parameters used in the analysis of the impact of LoRa DV packet size on wireless signal quality.

User Device	Location	Measurement Set (MS)	Number of Collected Metadata	SF	Tx Power (dB)	Payload Size (B)	DC – Tx Period (min)
DV2	LAB A509	MS9	180	12	20	1	30
DV2	LAB A509	MS18	211	12	20	25	30
DV2	LAB A509	MS27	188	12	20	50	30
DV1	LAB A507	MS5	766	9	10	1	8
DV1	LAB A507	MS14	1140	9	10	25	13
DV1	LAB A507	MS23	959	9	10	50	18
DV1	LAB A507	MS35	180	9	10	100	30

presents the measurement sets (MSs) with the combinations of the configured LoRa transmission parameters used in the analyses. The overall daily LoRa DV message transmission interval for analyzed MSs was set in the range from 8 min to 30 min based on DC and ToA restrictions of the TTN.

Figure 12 presents the mean, minimum, and maximum measured values of RSSI and SNR for the communication path between two end DVs and every GW (Figure 8). Figure 12a shows the measured RSSI and SNR values for measurement sets MS9, MS18, and MS27 (Table 9), while Figure 12b presents the measuring results for measurement sets MS5, MS14, MS23, and MS35 (Table 9).

Hence, Figure 12a presents the RSSI and SNR values for different combinations of transmission parameters of the DV2 transmitting LoRa message packets with sizes ranging from 1 B to 50 B in the overall daily LoRa DV2 message transmission interval set on 30 min, with the SF12, and at the Tx power of 20 dB. According to the results presented in Figure 12a, an increase in the message packet size slightly contributes to the improvement of the SNR level and does not have a significant impact on the RSSI. Obtained results for SNR, although unexpected, are related to transmission at the highest SF12, which ensures the longest LoRa message transmit duration of the payload data (ToA). More specifically, for the transmission at the SF12, symbol duration is very long (e.g., 32.768 ms for transmission with the 125 kHz bandwidth), and since the SNR reported by the LoRa chipset is obtained through a quantization process (often in steps of the 0.25 dB), for shorter packets, this quantization and noise fluctuation in indoor propagation environmants can cause less stable LoRa receiver readings. However, for longer packets (having more payload symbols), the demodulator has more opportunities to refine correlation with the chirp sequence. As longer packets (25–50 B) spend more time for transmission than shorter packets, averaging over small-scale fading variations appears as a slight SNR improvement, and this phenomenon is presented in Figure 12a for all DV2-GW wireless communication links. This phenomenon of longer packets that appears to slightly “improve” SNR is also confirmed in [58] for transmissions in a Bluetooth network. It is confirmed as a phenomenon contributed by the effect of receiver-estimator averaging over more symbols, rather than the common effect related to obtaining longer physical power gain during transmission.

Figure 12b presents the measured SNR and RSSI values for combinations of transmission parameters of the DV1 transmitting LoRa message packets with sizes ranging from 1 B to 100 B (Table 9). The transmission is performed by the LoRa end DV1 for the overall daily message transmission interval set between 8 min and 30 min, with the SF9 and at the Tx power of 10 dB (Table 9). The phenomenon of decreased SNR for longer packet sizes was not detected in Figure 12b for any of the presented communication links, excluding the DV1-GW4 communication link with packet size transmission of 100 B. According to the results presented in Figure 12b, an increase in packet size does not have an impact on the SNR and RSSI levels. Transmitting longer packets with more bytes simply means that more symbols are transmitted, and while each symbol is transmitted with the same Tx power (10 dB) and the SF9, the instantaneous RSSI and SNR remain fairly constant for transmission of packets of different sizes (Figure 12b). The small variance in the SNR and RSSI levels presented for the DV1–GWs communication paths of most MSs in Figure 12b is only a consequence of the variations in wireless channel conditions, which, for practical indoor signal propagation, are not static. Therefore, the results presented in Figure 12 lead to the conclusion that the transmission interval and packet size do not affect signal quality, except in special occasions when longer transmission at high SFs can, due to averaging over small-scale fading variations, appear as a slight SNR improvement.

6.4. Impact of Spreading Factor on RSSI and SNR

In the subsequent analyses, the influence of the spreading factor on indoor wireless signal quality is analyzed.

Table 10. Overview of measurement sets with corresponding LoRa transmission parameters used in the analysis of the impact of LoRa DV SF on wireless signal quality.

User Dev.	Location	Measurement Set	Number of Collected Metadata	SF	TX Power (dB)	Packet Size (B)	TX Period (min)
DV2	LAB A509	MS7	1217	7	20	1	3
DV2	LAB A509	MS26	411	9	20	50	18
DV2	LAB A509	MS9	180	12	20	1	30
DV2	LAB A509	MS16	1175	7	20	25	5
DV2	LAB A509	MS36	181	9	20	100	30
DV2	LAB A509	MS18	211	12	20	25	30

presents an overview of the MSs with corresponding LoRa transmission parameters used in the analysis of the impact of LoRa DV SF on the wireless signal quality. According to Table 10, only DV2 is included in the analysis, and for all MSs, the Tx power was fixed at 20 dB. The MSs differ in the spreading factor, ranging from the SF7 to the SF12, packet size ranging from 1 B to 100 B, and maximal transmission interval, ranging from 3 to 30 min.

Results of the RSSI and SNR measurements for measurement sets MS7, MS26, and MS9 (Table 10) are presented in Figure 13a, while those for MS16, MS36, and MS18 (Table 10) are presented in Figure 13b. Measurement results for MSs having a fixed Tx power of 20 dB and three different SF values equal to SF7, SF9, and SF12, for the packet sizes equal to 1 B and 50 B are presented in Figure 13a, while those for the packet sizes equal to 25 B and 100 B are presented in Figure 13b. According to Figure 13, the increase in SF for LoRa DV transmission at the same Tx power and packet size has no impact on the RSSI and SNR values. However, measuring results presented in Figure 13 indicate that the combination of the transmission of longer packets (of 50 B or 100 B) at increased SFs (e.g., SF9) has an impact on the SNR and RSSI. Nevertheless, in theory, the level of SNR or RSSI does not depend on packet size or spreading factor, since transmission of LoRa DV with longer packets and higher SFs can not impact the RSSI or SNR level.

Therefore, the results presented in Figure 13 are a consequence of the presence of constant wireless channel fluctuations in indoor signal propagation environments. The LoRa radio receiver computes the RSSI and SNR based on averaging their measured values across all received symbols. As explained in the previous section, for longer packets (having more bytes in length), the averaging is performed over more symbols and for a longer time. This enables during the signal reception process, to smooth out the indoor channel fluctuations, often producing slightly higher and more stable reported SNR. This averaging can also have an impact on the RSSI, which is the phenomenon also noticed in [58]. The consequence of this is that the chip’s SNR estimator reflects this as improved robustness, and even though Tx power and bandwidth are the same, the improved RSSI/SNR for longer packets transmitted at higher SF is measured (Figure 13). Thus, longer packets can cause a slightly higher reported SNR/RSSI due to longer received signal averaging, while the higher spreading factors allow the receiver to decode at lower SNRs. Consequently, the chip’s estimator reflects this robustness in the form of the reported SNR that appears better. Therefore, the results presented in Figure 13 provide a clear illustration of the unpredictability of wireless signal quality in indoor propagation environments. Measurement results presented in Figure 13 lead to the conclusion that the transmission interval (DC) and SF do not have an impact on LoRa wireless signal quality. The exception to this is in special occasions when transmission of longer packets in combination with higher SFs appears as a slight SNR/RSSI improvement, due to averaging over small-scale fading variations.

6.5. Impact of Wireless Signal Propagation Environments on SNR and RSSI

The general conclusion related to the results presented in Figure 11, Figure 12 and Figure 13 is that specific LoRA transmission parameters or a combination of LoRa transmission parameters do not have a dominant impact on RSSI or SNR as common measures of the short-range LoRa indoor wireless signal quality. The dominant impact on LoRa indoor wireless signal quality for communication links of up to 80 m in the mostly indoor propagation environments (Table 4, Figure 8), is significantly related to the signal propagation environment itself. This means that the signal propagation environment in terms of the obstacle material structure and their densities and obstacle numbers on the communication paths, with the length of the communication links (paths), dominantly impacts indoor LoRa wireless signal quality.

According to Figure 11, Figure 12 and Figure 13, independently of the value of some specific LoRa end DV transmission parameter or a combination of the LoRa end DV transmission parameters, the communication links having propagation of wireless signal over longer communication distances with higher obstacle densities (Table 4, Figure 8), will have lower (more negative) RSSI and lower SNR velues, and vice versa. For example, the worst signal quality in terms of RSSI and SNR levels presented in Figure 11, Figure 12 and Figure 13 has been measured for communication links among LoRa end DVs and GW4.

In addition, according to Figure 10, the lowest GWPDR is noted for GW4. The GW4 has the largest GWPDR loss in comparison with other GWs’ GWPDR losses. These results were obtained for all analyzed measurement sets with corresponding LoRa transmission parameters. The communication links between LoRa end DVs and GW4 have the worst signal quality, since these communication links have the highest number of obstacles with very large obstacle density and the longest communication paths (Table 4, Figure 8). More specifically, a large number of reinforced concrete floors and reinforced concrete vertical bearing (facade) walls through which the LoRa wireless signal needs to propagate on communication links between LoRa end DVs and GW4 (Table 4, Figure 8), is a main cause of the weakest signal quality among all analyzed communication paths (Figure 11, Figure 12 and Figure 13). These floors and walls significantly attenuate the strength of the wireless signal, in a way that even an increase in LoRa DV Tx power from 2 dB to 20 dB (Figure 11) results in a minor contribution to RSSI and SNR improvements.

The results presented in Figure 11, Figure 12 and Figure 13 for the case of the communication links between the GW1 and LoRa end DVs show that the wireless signal quality in terms of RSSI and SNR values is somewhat better compared with that for communication links between the LoRa end DVs and GW4. These results were obtained for all analyzed MSs with corresponding LoRa transmission parameters. These results were also confirmed for the per GW PDR results presented in Figure 10 for GW1. The GW1 has the second-largest PDR loss in comparison with other GWs’ PDR losses. This is a consequence of the lower number and density of reinforced concrete floors and vertical bearing (facade) walls compared to the case of communication links between the DVs and GW4 (Table 4, Figure 8). However, the wireless signal quality for communication links between the GW1 and DVs is significantly lower when compared with the signal quality of communication links between the LoRa end DVs and GW2 or GW3 (Figure 11, Figure 12 and Figure 13). This is a consequence of the higher density and number of obstacles and the longer communication paths compared with communication links between the LoRa end DVs and GW2 or GW3 (Table 4, Figure 8).

The highest signal quality in terms of RSSI and SNR values has been measured on communication links between the LoRa end DVs and GW2 or GW3, for every combination of the LoRa DV transmission parameters (as presented in Figure 11, Figure 12 and Figure 13). Also, the PDR results presented for these GWs in Figure 10 show the lowest PDR losses. The GW2 and GW3 have the lowest average GWPDR loss in comparison with other GWs’ PDR losses. This is because these communication links have the lowest number and density of obstacles on the communication path, and they also have the shortest completely indoor communication paths (Table 4, Figure 8). However, there is no dominant advantage of any of the communication paths (between the DVs and GW2, and the DVs and GW3) in terms of signal quality (Figure 11, Figure 12 and Figure 13). This is a consequence of the short length of communication paths between the GW2 or GW3 and DVs, which are also located in relatively close proximity to each other. Such very short communication paths range from approximately 1 m LOS communication path for the GW3-DV1 communication link, to the 24 m long indoor GW2-DV1 communication link (Table 4, Figure 8), with a low number of obstacles having low density (realized with up to three plasterboard walls (Table 9)).

Therefore, short communication paths with a small number of obstacles and their low density result in good wireless signal quality of such communication links. The differences in combinations of the LoRa end DV transmission parameters generally do not have a significant impact on wireless signal quality on such short, dominantly indoor wireless communication paths. This means that for indoor LoRa communication paths of up to 25 m having no or a low number of low-density obstacles, any combination of LoRa transmission parameters can be used for achieving a satisfactory level of wireless communication link quality. For longer communication paths of up to 80 m with a higher number and density of obstacles, the tuning of the transmission parameters for signal propagation through specific structural conditions can bring some improvements. The study shows that substantial signal attenuation was detected for signal propagation through reinforced-concrete floors and especially reinforced-concrete vertical facade walls. Nevertheless, the study confirms that the implementation of LoRaWAN technology can be a viable solution for indoor or dominantly indoor urban deployments.

6.6. Correlation of Obtained Results with Indoor Signal Propagation Path-Loss Models

The experimental results obtained for RSSI and SNR values in indoor LoRa network test scenarios presented in Figure 11, Figure 12 and Figure 13 show strong consistency with the parameters and empirical coefficients used in the mathematical expressions presented in Table 1 for indoor path loss models. Specifically, the measured signal attenuation patterns correspond well with the path-loss exponents, floor attenuation factors, wall loss coefficients, and environment-specific constants employed in indoor path-loss models presented in Table 1, such as the Floor Attenuation Factor (FAF) model and the Multi-Wall Model (MWM). Similarly, the Indoor Dominant Path (IDP) model and the Optimized Indoor Propagation (OIP) model presented in Table 1, point to the need for accurate parameter value selection that can effectively capture the dominant propagation mechanisms and spatial configurations encountered during signal level measurements. Moreover, the empirical coefficients embedded in the Padé Approximant-based model and the Meaningful Indoor Path-Loss Formula were found to represent important parameters for precisely expressing the wireless signal behavior in complex indoor layouts with high fidelity (Table 1). These findings confirm the relevance and impact of appropriate modeling of obstacle material composition and obstacle number as the main path-loss model parameters for accurately describing the LoRa signal propagation in indoor environments.

7. Conclusions

This study presents a comprehensive analysis of LoRa network performance in indoor propagation environments. In the paper, a systematic examination of the impact of combinations of different LoRa DV transmission parameters and obstacle numbers, material composition, and obstacle densities at the communication paths on wireless signal quality is performed. The relevance of LoRa and LoRaWAN technologies within the IoT ecosystem is outlined in the paper with the description of the key LoRa end DV transmission parameters that include SF, Tx power, packet size, transmission interval, and their expected influence on the operating performance of the LoRa networks. Additionally, the paper describes the architecture of a developed dedicated experimental test network, which integrates LoRa end DVs, multiple LoRa GWs, and supporting test network components that include the MQTT protocol for the LoRa message exchange, the public LoRaWAN infrastructure provided by TTN, the Node-RED application for the LoRa message flow scheduling, the InfluxDB time-series database for data storage, and the Grafana platform for statistical analyses and visualization. This developed test platform enabled the collection, storage, and visualization of measured RSSI and SNR data across diverse LoRa test network MSs. The described test methodology enables performing controlled measurements over extended time periods, ensuring robust statistical analyses of the RSSI and SNR as principal indoor wireless signal quality indicators.

The measurement results demonstrated that the most dominant factor influencing the LoRa signal quality in indoor environments is the physical environment of the wireless signal propagation path, defined by the number, material composition, and density of obstacles between communicating LoRa devices. While variations in the message packet size and transmission interval showed negligible influence, increases in the SF and Tx power exhibited limited signal quality improvements under specific signal reception conditions, particularly when longer packets were combined with higher spreading factors. Nevertheless, these gains were minor compared to the substantial wireless signal quality degradation caused in the case of signal propagation through reinforced-concrete floors and vertical walls. The measurements confirmed that links traversing longer indoor distances with greater obstacle density consistently exhibited degraded RSSI and SNR levels, whereas shorter links with fewer obstacle numbers and densities maintained robust signal quality. These findings emphasize that careful planning of LoRa GW placement is crucial for ensuring reliable connectivity in complex indoor and urban environments.

Additionally, the results presented in this work confirm that the LoRa DV transmission parameter tuning alone cannot fully compensate for signal quality losses in challenging indoor signal propagation conditions, as environmental characteristics remain the predominant determinant of LoRa signal quality performance. In the design of the LoRa-based indoor and urban deployments, the study highlights the importance of considering both the LoRa transmission parameters configuration and the number and density of obstacles on the signal propagation path.

Future work will focus on expanding the scope of indoor LoRa signal quality measurements across additional buildings, obstacle types, and deployment scenarios. Moreover, subsequent research will integrate advanced adaptive algorithms and optimization frameworks for parameter selection, with the aim of further improving energy efficiency, reliability, and scalability of the implementation of the LoRaWAN systems in real-world indoor and urban environments.