Evaluating the Suitability of Commercial Off-the-Shelf Software Radio Peripheral Front-Ends for High-Fidelity Satnav Monitoring Applications

  • NAVIGATION: Journal of the Institute of Navigation
  • July 2026,
  • 73
  • navi.780;
  • DOI: https://doi.org/10.33012/navi.780

Abstract

Signal quality monitoring is an essential portion of the Global Positioning System (GPS) enterprise. Through monitoring, accuracy and normality are achieved for each satellite and signal. GPS users rely on this information from monitoring stations to ensure safe and successful operation. Traditional monitoring has been performed with hardware receivers or custom radio frequency front-ends. These receivers have known and measured impacts on the distortions of the received signal. This research aims to elucidate and quantify the amount of distortion introduced by the front-ends of general-purpose commercial off-the-shelf software-defined radio (SDR) peripherals. Monitoring stations outfitted with SDRs are more flexible and cheaper than those without SDRs and can quickly adapt to newly launched or reconfigured navigation satellites. This research quantifies the amount of signal distortions introduced by a handful of SDR front-ends when subjected to consistent navigation signals from a high-quality simulator.

Keywords

1 INTRODUCTION

The global reliance on positioning, navigation, and timing (PNT) has never been higher. Ever since the inception of the Global Positioning System (GPS) in the 1970s, an increasing number of applications and industries have relied on navigation and timing. The degradation or denial of this technology can be disruptive, for example, by preventing informed strategy decisions during a Formula 1 race (Formula 1, 2025) or leading to more severe outcomes such as a fatal aircraft crash. In either case, there is a need to increase the robustness and resiliency of PNT to prevent such events from occurring.

Numerous countries and companies have invested in techniques for increasing the robustness and resiliency of GPS for more than 20 years. These efforts began with the GPS modernization program, which aims to deploy advanced signals with increased accuracy and enhanced encryption (GPS.gov, n.d.). Increased resiliency can be achieved by increasing the number of PNT sources used by the receiver. GPS modernization aims to increase resiliency by doing just that: increasing the number of PNT sources by launching new satellites with new signals. However, this increase is slow and expensive. There are hundreds of alternative PNT sources in space, one such category being global navigation satellite systems (GNSSs).

Many countries and companies are not utilizing GNSSs for several reasons, one of which is trust. GPS has historically been the sole source of PNT for many applications, including military applications, while GNSSs have been largely underutilized because non-GPS GNSSs are not operated and controlled by the United States. The usage of all GNSSs, particularly for military applications, provokes an inherent level of healthy skepticism surrounding their usage (Air & Space Forces Association, 2025). If the goal is to leverage foreign GNSSs for military applications, some level of trust must be established. The establishment of trust must rely on determining what normal operation looks like for each constellation and satellite so that abnormal behavior can be detected and communicated to users.

For decades, the civil aviation industry has performed continuous monitoring of navigation satellites. Civil aviation continuously monitors unencrypted GPS signals and informs users about the accuracy and operating behavior of GPS (Walter & Enge, 2004). Extending this concept to determine trust among the military requires continuous monitoring of all GNSSs to establish normality for each system and satellite so that anomalous events can be detected and users alerted.

Continuous monitoring of all GNSSs will require hardware and software to be developed and deployed around the world. There exist commercial receivers that can process all visible GNSS signals but are not suitable for this task owing to a lack of agility and flexibility. New satellites are being launched at a cadence never seen before, and the satellite architectures themselves are more dynamic (Air Force Research Laboratory, n.d.). Consequently, monitoring stations must be designed with this same flexibility in mind, which can be accomplished by using commercial-off-the-shelf (COTS) software-defined radio (SDR) products such as the Universal Software Radio Peripheral (USRP).

The flexibility of SDRs, such as the tunability of the center frequency or bandwidth, makes them uniquely suitable for a flexible monitoring station. However, challenges exist when using SDRs to perform monitoring. SDRs introduce their own errors into the received signal in the form of signal distortions (Gunawardena & van Graas, 2013), which causes errors in the measurements of the receiver. Measurement errors interfere with the task of monitoring and hinder the establishment of normality for GNSS trust. These signal distortion errors are typically measured and calibrated for commercial receivers when they are manufactured. This same process is not performed for SDRs, and for many SDR platforms, the severity of this error in GNSS signals remains unknown.

The goal of this research is to quantify the amount of signal distortion caused by a handful of COTS SDRs prevalent throughout the PNT community.

2 BACKGROUND

There exists a large body of knowledge focused on understanding, measuring, and removing error sources that can cause signal distortions. As the name suggests, signal distortions are deformations to the signal caused by effects such as multipath or hardware biases (Thoelert et al., 2014; see also Gunawardena & van Graas, 2014; Phelts & Akos, 2004).

A basic GPS signal being transmitted from a satellite is shown below:

STX(t)=hSV(t)∗(2AC(t)D(t)cos⁡(2π(fL1+fD)t+φ))1

where:

hSV (t) is the transfer function of the satellite due to variations in satellite hardware

A is the amplitude of the signal

C(t) is the pseudorandom noise (PRN) spreading code

D(t) is the navigation data (50 Hz for coarse/acquisition [C/A] signals)

fL1 is the L1 carrier frequency (1575.42 MHz)

fD is the L1 carrier frequency Doppler shift

φ is the carrier phase

* denotes a convolution operation

As this signal propagates through the atmosphere, channel effects arise that cause deformations in the signal. The signal received at an antenna at the surface of the Earth is as follows:

sRX(t)=hENV(t)∗hSV(t)∗(AC(t)D(t)cos⁡(2π(fL1+fD)t+φ))2

where hENV (t) is the transfer function encompassing all channel effects, such as ionosphere (Comberiate, 2012) and multipath.

This signal sRX (t) is received by an antenna and sent through a radio frequency (RF) front-end (RFFE), represented by the following:

rIF(t)=hRFFE(t)∗[hENV(t)∗hSV(t)∗(AC(t)D(t)cos⁡(2π(fL1+fD)t+φ))]3

where hRFFE (t) is the transfer function of the RFFE.

Signal distortions are caused by the transfer functions hSV, hENV, and hRFFE, as shown in Equation (3), and will deform the correlation triangle (also known as the correlation function). The correlation function is a fundamental component of GNSS signal processing and is used in the formation of pseudorange measurements. Distortions in the correlation function cause errors in the pseudorange measurement, which translate to an error in the user position. Other applications utilize the correlation function for monitoring and situational awareness applications (Brenner et al., 2009; see also Phelts, 2001), and a distorted correlation function can negatively impact those monitoring applications. Therefore, it is crucial to determine how signal distortions impact the correlation function.

An example correlation function is shown in Figure 1. The black triangle in Figure 1 represents a theoretical “perfect” correlation function. This case is not possible in real life owing to band limiting and error sources but is shown here for illustrative purposes. The red distorted triangle is a more realistic view of a correlation function. As shown in Figure 1, the red function is nonuniformly distorted as δ changes. These nonuniform deformations in the red function are caused by the three transfer functions shown in Equation (3). Receivers aim to take pseudorange measurements where the code delay is zero on the correlation function, which is typically accomplished by generating three replicas of the incoming code that are aligned to the incoming signal with different delays. These three replicas are called early, prompt, and late replicas, which refers to their delay with respect to the prompt replica. Ideally, the prompt replica is aligned to the incoming code, which would maximize the correlation and place the prompt tap at the peak of the ideal correlation function in Figure 1. The prompt replica is aligned by equalizing the early and late replicas on either side of the correlation function such that E-L=0 and taking the midpoint as the prompt replica alignment. Under ideal but impossible conditions for the black triangle in Figure 1, the prompt replica is centered in the correlation function where the code delay is zero. However, when the triangle is distorted via the three previously discussed transfer functions, the midpoint of the equalized early and late replicas on the distorted red function is no longer at a code delay of zero, resulting in a pseudorange error.

Correlation function highlighting an illustrative ideal triangle and a distorted triangle
FIGURE 1

Correlation function highlighting an illustrative ideal triangle and a distorted triangle

Research into understanding and removing or mitigating the severity of error sources on signal distortions is applicable to all receivers, particularly monitoring station receivers. Station receivers are tasked with monitoring all applicable GNSS signals for the purpose of detecting and alerting when anomalous behavior occurs (Gps.gov, 2019). A lack of understanding regarding hSV, hENV, or hRFFE can lead to false alerts at monitoring stations when the signals and satellites are operating nominally.

Monitoring all GNSSs to establish normality for trust will require the deployment of new monitoring stations. SDRs are uniquely suited for this task because of their flexibility. However, the signal distortions hRFFE caused by SDRs have not been well studied for navigation signals. Without a measurement and calibration of the SDRs, each monitoring station will have a unique correlation function, owing to their own hardware variations. Consequently, a measured pseudorange error at a specific correlator spacing from one front-end will be different from that of other front-ends, even those of the same make and model, and will prevent the usage of differential corrections from one front-end to another. Additionally, signal distortions caused by hardware variations have a nonuniform effect on the correlation function, which results in pseudorange errors as a function of correlator spacing.

Figure 2 illustrates the purpose of equalizing the correlation function across receivers in a differential configuration using three GNSS receivers: one located at a monitoring station and two others aboard aircraft. The goal of the monitoring station is to send pseudorange corrections to the incoming aircraft, which is only possible if all front-ends have the same effect on the correlation function. In this example, the monitoring station and the top aircraft have measured and calibrated out the impact of their front-ends on the correlation function at d = 0.1; therefore, an error correction ϵ sent from the monitoring station can successfully be used by the top aircraft. Conversely, the aircraft on the left in Figure 2 uses a different correlator spacing value (d = 0.2) without a calibrated correlation function, which prevents the usage of ϵ from the monitoring station.

Scenario highlighting the importance of equalizing correlation functions between receivers for utilization of correction factors from monitoring stations to user equipment
FIGURE 2

Scenario highlighting the importance of equalizing correlation functions between receivers for utilization of correction factors from monitoring stations to user equipment

The equalization of correlation functions among all GNSS receivers is critical for a scenario such as that shown in Figure 2. This approach ensures that all receivers, whether they are located at a monitoring station or elsewhere, have the same sensitivity to anomalous events and reduces the likelihood of false positives or missed detections. Besides slight hardware variations in front-ends of the same make and model, there exist two main front-end architectures used by SDRs, each with their own advantages and disadvantages for signal distortions: superheterodyne and direct conversion.

Superheterodyne front-end architectures employ various mixing stages to down-convert an incoming signal to an intermediate frequency (IF). An example of a single-conversion superheterodyne architecture is shown in Figure 3.

Single-conversion superheterodyne IF front-end architecture (Morton et al., 2021)
FIGURE 3

Single-conversion superheterodyne IF front-end architecture (Morton et al., 2021)

A signal-conversion superheterodyne RFFE architecture is advantageous for GNSS signal monitoring applications because there are no in-phase (I)/quadrature (Q) gain or phase imbalance issues. This architecture also has a reduced number of analog components compared with a dual-conversion architecture (Morton et al., 2021). However, this architecture is not without its drawbacks: it has a higher power consumption and an IF bandpass filter, which can cause signal distortions as a function of temperature (Gunawardena & van Graas, 2015; see also Guerrero & Gunawardena, 2017).

In contrast, direct-conversion architectures employ a single down-conversion stage to down-convert a signal from RF to baseband (0 Hz IF). An example of a direct-conversion architecture is shown in Figure 4.

Direct-conversion zero IF (baseband) front-end architecture (Morton et al., 2021) (LNA: low-noise amplifier)
FIGURE 4

Direct-conversion zero IF (baseband) front-end architecture (Morton et al., 2021) (LNA: low-noise amplifier)

The main advantage of a direct-conversion front-end architecture is that it can be implemented on an integrated circuit, making chip-level GNSS receivers a reality. The main disadvantage of direct-conversion front-ends, in relation to PNT, lies in the signal distortions caused by this type of architecture. The mixing stage for the real and imaginary paths employs a 90° phase offset (shown in blue in Figure 4) to generate the sine and cosine. The phase offset is not exactly 90°, which causes a phase imbalance between the real and imaginary streams. Additionally, the real and imaginary paths have separate gain stages (red in Figure 4), which are also not equivalent to each other. This scenario causes unique amplitude imbalances on the real and imaginary streams.

3 DIFFERENTIAL CODE BIAS

In this study, receiver hardware biases were measured and quantified via the differential code bias (DCB). DCBs have long been used for determining the contribution of various error sources to the pseudorange, such as the ionosphere, satellite biases, and receiver biases (Montenbruck et al., 2014). A pseudorange observation can be modeled as follows:

P=r+trcvr−tsat+T+I+B4

where r is the geometric range between the satellite and receiver, trcvr and tsat are the receiver and satellite clock offsets, respectively, T and I are the troposphere and ionosphere delays, respectively, and B is a hardware-induced additive bias due to the satellite and receiver hardware. The DCB is computed by differencing pseudorange measurements, which is typically achieved by differencing pseudorange measurements from two different signals, as reported by Montenbruck (2014). However, in this research, the DCB is computed by differencing pseudorange measurements using the same signal but with different early-minus-late (E-L) correlator spacings, δ. Applying this approach for two different correlator spacings results in the following equation:

dcbδ1,δ2=Pδ1−Pδ2=rδ1+tδ1rcvr−tδ1sat+Tδ1+Iδ1+Bδ1−…      rδ2−tδ2rcvr+tδ2sat−Tδ2−Iδ2−Bδ25

All remaining terms cancel other than the bias term, because changing δ yields a different bias owing to the nonuniform distortion of the correlation function caused by the transfer functions in Equation (3). Thus, we obtain the following:

dcbδ1,δ2=Pδ1−Pδ2=Bδ1−Bδ26

Each bias term is a combination of distortions due to the satellite (BSV), environment (BENV), and receiver RFFE (BRFFE):

Bδ1=Bδ1SV+Bδ1ENV+Bδ1RFFE7

In practice, the bias term is more complicated than a simple sum of biases (Montenbruck et al., 2014) because of the compounding effects of the transfer functions described in Equations (1)–(3).

Figure 5 visually demonstrates what the DCB measures. As previously mentioned, pseudorange measurements are ideally taken where the code delay of the correlation function is zero (green line), which is accomplished by taking measurements at the mid-point of the early and late correlators (blue dots). Because of signal distortions, the pseudorange measurement will not occur at a code delay of zero and will be taken elsewhere (orange).

Schematic of how the DCB is computed and how distortions to the correlation function cause pseudorange errors
FIGURE 5

Schematic of how the DCB is computed and how distortions to the correlation function cause pseudorange errors

The bias terms in Equation (6) can be rewritten using the early replica and delays in Figure 5:

Bδ1=E1+d18

Bδ2=E2+d29

dcbδ1,δ2=Pδ1−Pδ2=E1+d1−E2−d210

where di=δi2.

4 EXPERIMENTAL SETUP

The goal of this research was to quantify the impact of various USRP RFFEs on signal distortions. Thus far, few studies have been published on how COTS front-ends, such as USRPs (Peng & Morton, 2013), impact signal distortions. Many applications utilize custom RFFEs (Gunawardena & van Graas, 2011). These custom front-ends can be expensive, take a long time to develop, and rely on skillsets that are becoming increasingly harder to find.

The experimental setup utilizes a Spirent GSS9000 as the high-fidelity signal source. This source enables full control over signal generation and allows modeling of additional errors, such as errors caused by henv or hSV, to be eliminated or minimized so that the effects of hRFFE can be isolated and measured. The simulator was configured to generate a single signal from a single stationary satellite at a time without the modeling of channel effects. To quantify the impact of binary phase shift keying and binary offset carrier (BOC) signals on signal distortions, the following signals were generated: GPS C/A PRN 6, GPS C/A PRN 7, GPS C/A PRN 8, and Galileo E1b PRN 3. The three GPS C/A PRNs were chosen because they provide coverage of the three different families of C/A autocorrelation functions (Liu et al., 2006).

The simulator RF output and a 10-MHz reference clock were fed into a USRP X310 SDR front-end (Ettus Research, 2024). The RF output was fed through a 1-to-4 splitter before feeding the four individual RF streams to the USRP. Digitized RF data from the USRP were collected and recorded by a collection computer using a high-speed data link and solid-state drive. Samples were complex 16-bit I and 16-bit Q collected at a sample rate of 50 MHz for a duration of 210 s per collection.

USRP X310s require the selection and installation of RF daughterboards, which are transceivers that include all of the analog components for the front-end down-conversion process. Various daughterboards are available, and the selection of daughterboard defines the characteristics of the front-end, such as the center frequency tuning range, bandwidth, and number of receive and transmit channels. This research investigates the impact of the TwinRX and UBX daughtercards on signal distortions.

TwinRX daughtercards are superheterodyne architecture front-ends. In this architecture, the TwinRX utilizes multiple down-conversion stages to down-convert the signal from an RF to an IF. There is also an analog-to-digital (ADC) component on the TwinRX that digitizes the signal. The samples output from the ADC are real samples only. These digital samples are ingested by the X310 field-programmable gate array and are down-converted further to complex samples at baseband (0 Hz IF). Figure 7 shows the TwinRX architecture from Ettus. Highlighted in blue is the surface acoustic wave (SAW) filter. As discussed earlier, this component is temperature-sensitive and will cause signal distortions. TwinRX daughtercards, as the name suggests, have two independently tunable receive channels and zero transmit channels.

Experimental setup showing the USRP X310 outfitted with two TwinRX daughtercards
FIGURE 6

Experimental setup showing the USRP X310 outfitted with two TwinRX daughtercards

TwinRX RF schematic (Ettus Research, 2023) (BW: bandwidth; HB: high band; LB: low band; LO: local oscillator; SPDT: single pole, double throw)
FIGURE 7

TwinRX RF schematic (Ettus Research, 2023) (BW: bandwidth; HB: high band; LB: low band; LO: local oscillator; SPDT: single pole, double throw)

UBX daughtercards are direct-conversion architecture front-ends. These front-ends convert a signal from an RF directly to baseband with a single down-conversion stage. Before the down-conversion stage, the signal is split into I and Q channels, which results in complex samples. Figure 8 shows the UBX diagram from Ettus. Highlighted in blue is the sine and cosine generation for the I and Q channels, which causes the phase imbalance. Independent gain stages for the I and Q channels also cause amplitude imbalances but are not shown in Figure 8.

UBX RF schematic (Ettus Research, 2021) (LPF: low-pass filter; RX: receiver; TX: transmitter)
FIGURE 8

UBX RF schematic (Ettus Research, 2021) (LPF: low-pass filter; RX: receiver; TX: transmitter)

Six TwinRX daughtercards and six UBX daughtercards were analyzed in this research. Data from multiple cards of the same make and model were analyzed to quantify intra-daughtercard variations and to determine if results from a TwinRX or UBX were applicable to other TwinRX or UBX daughtercards. Two daughtercards were installed in the USRP X310, and the unit was powered on for 30 min so that the daughtercards could reach a steady-state temperature. A temperature chamber was unavailable during this experiment for precise control; thus, this approach was applied to reduce any temperature effects. Simulation and data collection were accomplished in the following order: GPS C/A PRN 8, 6, and 7 and Galileo E1b PRN 3. The process was repeated for the remaining 10 daughtercards using the same power-on procedure and data collection order so that each scenario had similar temperature effects as the same scenario for other daughtercards. For example, for daughtercard 1, data were collected for GPS PRN 8 first. For daughtercard 2, data were also collected for GPS PRN 8 first. This order was followed for all subsequent daughtercards to ensure that temperature effects for all GPS PRN 8 data collections were similar.

5 RESULTS

I/Q data were processed by an SDR known as Pychips (Gunawardena, 2021). Pychips was configured to track the simulated signals using a coherent integration period of 20 ms, a noncoherent integration period of 50 ms, a second-order phase-locked loop (PLL) with a bandwidth of 18 Hz, and a first-order delay-locked loop (DLL) with aiding from the PLL. The PLL utilized a costas discriminator, whereas the DLL utilized a normalized E-L envelope (Kaplan & Hegarty, 2017). Pychips was configured with 128 correlators, each with a unique E-L correlator spacing, δ, in the range of [0, 1] chips. Each correlator was processed by the DLL discriminator and generated pseudorange measurements, which were used to compute the DCB. Pychips was configured to compute and save pseudorange measurements at a rate of 1 Hz per correlator. From the 128 correlators each with a unique δ, 128 pseudorange measurements per second were obtained. Thus, for a single collection, a total of ~25600 pseudorange measurements were generated.

The DCB, as described in Equation (10), computes the difference in pseudorange measurements from a pair of correlators. This study used δ1 = 0.1 chips and δ2 ∈ [0, 1]. The δ1 value was chosen because current GNSS receivers typically use a correlator spacing value at or around 0.1 chips. By definition, DCB = 0 when δ1 = δ2, which leads to all DCB results having a zero crossing at δ = 0.1. For a given δ and collection, the DCB was computed for each of the 200 pseudorange measurements. These 200 DCB measurements were averaged, resulting in a single DCB measurement for the given δ and collection.

Figure 9 shows the DCB results from all daughtercards for only the GPS C/A PRN 6 simulated scenario. Recall that TwinRX daughtercards have two receive channels per daughtercard; for this reason, the figure has a channel 0 and channel 1 for the same TwinRX daughtercard. Channels from the same daughtercard are shown in the same color (e.g., red for TwinRX 1), but the second channel is a pastel shade. At first glance, it is difficult to distinguish TwinRX from UBX, despite the daughtercards using different front-end architectures. It was originally hypothesized the UBX daughtercards would induce worse signal distortions owing to the phase and gain imbalances typically found in a direct-conversion architecture. If the UBX daughtercards caused more signal distortions, they would exhibit larger DCB values compared with the TwinRX. However, this does not seem to be the case. In fact, the three daughtercards with the largest average DCB value are all TwinRX (2, 4, and 6), and the two daughtercards with the lowest average DCB value are all UBX (5 and 6). There is also no noticeable similarity between the two channels from a single TwinRX. Consequently, any knowledge regarding signal distortion for channel 0 cannot be applied to channel 1.

DCB for all daughtercards for the GPS C/A PRN 6 scenario
FIGURE 9

DCB for all daughtercards for the GPS C/A PRN 6 scenario

Figures 10 and 11 show the DCB for all daughtercards for the GPS C/A PRN 7-only and GPS C/A PRN 8-only scenarios. These plots present trends similar to those in Figure 9: TwinRX 2 channel 0 has the largest average DCB, whereas UBX 6 has the lowest average DCB. There is also no clear difference between the different GPS C/A PRNs, despite each one having a slightly different autocorrelation function. Figure 11 does show that the DCB of UBX 3 decreases as δ → 1.

DCB for all daughtercards for the GPS C/A PRN 7 scenario
FIGURE 10

DCB for all daughtercards for the GPS C/A PRN 7 scenario

DCB for all daughtercards for the GPS C/A PRN 8 scenario
FIGURE 11

DCB for all daughtercards for the GPS C/A PRN 8 scenario

The results are more interesting for the Galileo E1b case, as shown in Figure 12. The Galileo E1b DCB values are generally smaller in magnitude compared with any of the three GPS signals, particularly at δ = 0.2. This trend is due to the sharper correlation peak caused by the BOC component present on E1b. It is worth noting that the best- and worst-case daughtercards in Figure 12 are the same as the best-and worst-case daughtercards from the GPS scenarios.

DCB for all daughtercards for the Galileo E1b PRN 3 scenario
FIGURE 12

DCB for all daughtercards for the Galileo E1b PRN 3 scenario

The results shown in Figures 9–12 can be used to calibrate out the signal distortions caused by the RFFE. For example, let us consider the blue curve with the × markers in Figure 12 (TRX 2 CH 0). For the desired δ, the DCB values should be saved and used to correct the pseudorange measurement. For example, if δ = 0.3, an error of ~0.38 m should be removed from the pseudorange measurements for Galileo E1b signals. The same process can be performed for the GPS C/A signals.

The DCB values shown in Figures 9–12 are the average DCB for each of the 200 pseudorange measurements for a given δ and collection. The figures do not give any indication of the stability, or spread, of the DCB over those 200 pseudorange measurements. Figure 13 shows all 200 DCB measurements for a single collection, prior to the computation of the average.

All 200 DCB measurements for the GPS C/A PRN 6 scenario using TwinRX (TRX) 4, channel 1
FIGURE 13

All 200 DCB measurements for the GPS C/A PRN 6 scenario using TwinRX (TRX) 4, channel 1

Figure 13 contains 200 unique DCB realizations for each of the 200 s of the data collection. The first DCB measurement is colored dark blue, and as time progresses, the DCB measurements fade to dark red. Using Figure 13, the stability of the DCB across time was computed. For improved visualization, δ = 0.2 was chosen as the E-L correlator spacing value to present the variability. Using the example plot in Figure 13, a boxplot was generated using the DCB values in Figure 13 for δ = 0.2. This approach resulted in a single boxplot that describes the spread, skewness, and outliers. Figure 14 presents the result of performing this analysis for all data across the Galileo E1b PRN 3 and GPS C/A PRN 7 scenarios using δ = 0.2.

Boxplots showing the spread with the mean (green triangle) and median (red line) of the DCB at δ = 0.2 for the Galileo E1b PRN 3 and GPS C/A PRN 7 scenarios
FIGURE 14

Boxplots showing the spread with the mean (green triangle) and median (red line) of the DCB at δ = 0.2 for the Galileo E1b PRN 3 and GPS C/A PRN 7 scenarios

Figure 14 displays boxplots that indicate the spread of DCB results for each daughtercard. The figure clearly shows that the Galileo E1b results have a lower average than the GPS C/A PRN 7 results. This difference is due to the sharper correlation peak of the Galileo E1b signal. As discussed earlier, the blue boxplot (TRX2, Ch0) corresponds to the worst-performing daughtercard, owing to this daughtercard having the largest DCB value. Figure 14 shows that TRX2, Ch0 has the largest average DCB value but not the largest spread of DCB values. TRX6, Ch1 has the largest spread for the Galileo E1b scenario, whereas UBX 3 has the largest spread for the GPS C/A PRN 7 scenario. A smaller spread is more important because a stable daughtercard with a large average DCB value can be calibrated. Large fluctuations in the DCB make it difficult to calibrate the daughtercards.

It is worth noting that according to Figure 14, the best-performing daughtercards are all UBX daughtercards. UBXs 1, 5, and 6 all consistently have a lower average DCB and smaller spread, despite being a direct-conversion daughtercard with gain and phase imbalances. UBX 1 has a small spread compared with all other daughtercards. Results for the GPS C/A PRN 6 and 8 scenarios are not shown here, but their boxplots present the same trends.

6 CONCLUSIONS

This research analyzed the impact of hRFFE on signal distortions. RF signals of GPS C/A PRNs 6, 7, and 8 and Galileo E1b PRN 3 were generated using a high-quality commercial simulator. Data were recorded by two different front-end architectures of USRP X310 daughtercards: TwinRX, which has a superheterodyne architecture, and UBX, which has a direct-conversion architecture. To quantify the impact of the front-ends on signal distortions, the DCB was computed for each scenario and daughtercard. For δ1 = 0.1, the DCB results show that the average worst-case DCB values are in the range of 0.8−0.9 m. It is worth noting that the USRP X310 mainboard is likely to contribute some amount to the overall DCB measurements. However, because the variance between daughtercards of the same model is on the order of 0.5 m, most of the DCB variation can be attributed to the analog components contained within the daughtercards themselves. The common-mode contribution from the X310 mainboard could be isolated in future studies by swapping a daughtercard between the two available slots on the X310 and collecting data after the system has reached a steady-state temperature.

Interestingly, the results show that the best-performing daughtercards are UBX, despite their direct-conversion architecture, which is known for introducing gain and phase imbalances from the real and imaginary channels. However, a manufacturer-recommended calibration procedure was followed for the UBX daughtercards; this procedure aims to minimize gain and phase imbalances. No such procedure exists for the TwinRX daughtercards because of their superheterodyne architecture. An analysis was not performed to determine the impact, if any, of the internal calibration procedure on signal distortions, as this was considered to be part of the setup and configuration process for the UBX-160 daughtercards. A manual calibration process for the daughtercards could be performed in addition to or instead of the built-in USRP utility. Future work could measure the DCB before and after the self-calibration script is performed to determine its effectiveness in reducing gain and phase imbalances.

Results from this research can be used to make more informed decisions about which USRP model and daughtercard should be used for various applications. Applications such as monitoring require front-ends with stable signal distortion effects that can be measured. Front-ends with a large variability, such as some of the TwinRX front-ends, are not suitable for detailed monitoring applications because of their propensity to introduce signal distortions with a large variance. Other applications might be able to tolerate the signal distortions, and therefore the pseudorange errors, introduced by these USRP X310 daughtercards. Regardless of the application, a metric such as the DCB can allow for the calibration and removal of pseudorange errors caused by hRFFE. While this research focused on USRPs owing to their proliferation within the PNT community, this analysis is applicable to any front-end. These results should not be interpreted as an endorsement of USRPs over potential alternatives. The conclusions informed by data collected from the USRP X310 may or may not be extensible to other SDR platforms. An evaluation of each unique SDR should be performed to determine their impact on signal distortions.

A question remains regarding the severity of the USRP-induced signal distortions. Prior studies (Brenner et al., 2009) have reported a 1-sigma pseudorange error bound budget of ~0.1 m solely from satellite-induced signal distortions using an overall error budget of ~1 m. Those researchers measured the amount of pseudorange error caused by signal distortions from GPS satellites in the 2000-2004 timeframe, with the worst case being ~0.15 m, by using multiple correlators spread across the correlation function. Similarly, this work aimed to quantify the amount of pseudorange error caused by signal distortions using multiple correlators, but with a focus on distortions from the receiver front-end instead of the satellite. Based on the overall error budget of ~1 m, the amount of signal distortion induced by the USRP front-ends is significant for monitoring applications and should be measured and calibrated.

A potential downside to this research is its lack of temperature control or analysis. As shown in Figure 13, the DCB measurements tend to decrease, or fade to red, as time increases. This trend indicates a dependency on temperature. For the TwinRX daughtercards, such as the one shown in Figure 13, it is known that SAW filters are sensitive to temperature and are the largest contributor in a superheterodyne architecture to signal distortions, as shown by prior research (Guerrero & Gunawardena, 2017). Future work should measure, vary, and record the temperature of the front-end during data collection. This approach would allow DCB measurements to be obtained as a function of δ and temperature.

HOW TO CITE THIS ARTICLE:

Carroll, M.J., Gunawardena, S., Machin, T.I., Schubert Kabban, C.M., & Rushanan, J.J. (2026). Evaluating the suitability of commercial off-the-shelf software radio peripheral front-ends for high-fidelity satnav monitoring applications. NAVIGATION, 73. https://doi.org/10.33012/navi.780

ACKNOWLEDGMENTS

The author’s affiliation with The MITRE Corporation is provided for identification purposes only and is not intended to convey or imply MITRE's concurrence with, or support for, the positions, opinions, or viewpoints expressed by the author.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

REFERENCES

  1. ↵
    Air & Space Forces Association. (2025, March 14). PNT at a crossroads. Retrieved August 15, 2025, from https://www.afa.org/agenda/pnt-at-a-crossroads
    Google Scholar
  2. ↵
    Air Force Research Laboratory. (n.d.). Navigation Technology Satellite–3 (NTS-3). Retrieved September 1, 2025, from https://afresearchlab.com/navigation-technology-satellite-3-nts-3-program/
    Google Scholar
  3. ↵
    Brenner, M., Liu, F., Class, K., Reuter, R., & Enge, P. (2009, September). Natural signal deformations observed in new satellites and their impact on GBAS. In Proceedings of the 22nd International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS) (pp. 1100–1111). https://www.ion.org/publications/abstract.cfm?articleID=8519
    Google Scholar
  4. ↵
    Comberiate, J., Kelly, M., Dyrud, L., & Weaver, G. (2012, September). Space weather effects on GPS systems. In Conference Proceeding at Applied Physics Laboratory, USA.
    Google Scholar
  5. ↵
    Ettus Research. (2021). UBX. Retrieved August 10, 2025, from https://kb.ettus.com/UBX
    Google Scholar
  6. ↵
    Ettus Research. (2023). TwinRX. Retrieved August 10, 2025, from https://kb.ettus.com/TwinRX
    Google Scholar
  7. ↵
    Ettus Research. (2024). USRP X3x0 series. In USRP Hardware Driver and USRP Manual. Retrieved August 15, 2025, from https://files.ettus.com/manual/page_usrp_x3x0.html
    Google Scholar
  8. ↵
    Formula 1. (2025). Vasseur assesses “difficult weekend“ for Ferrari. Retrieved August 30, 2025, from https://www.formula1.com/en/latest/article/vasseur-assesses-difficult-weekend-for-ferrari-at-silverstone-as-he-explains.1CeX5k1huNCoEAKkH1lBJS
    Google Scholar
  9. ↵
    GPS.gov. (n.d.) GPS modernization. Retrieved August 28, 2025, from https://www.gps.gov/gps-modernization
    Google Scholar
  10. ↵
    GPS.gov. (2019). Control segment. Retrieved August 28, 2025, from https://www.gps.gov/control-segment
    Google Scholar
  11. ↵
    Guerrero, J. M., & Gunawardena, S. (2017, February). Characterization of timing and pseudorange biases due to GNSS front-end filters by type, temperature, and Doppler frequency. In Proceedings of the 2017 International Technical Meeting of the Institute of Navigation (pp. 418–444). https://doi.org/10.33012/2017.14911
    Google Scholar
  12. ↵
    Gunawardena, S. (2021, January). A high performance easily configurable satnav SDR for advanced algorithm development and rapid capability deployment. In Proceedings of the 2021 International Technical Meeting of the Institute of Navigation (pp. 539–554). https://doi.org/10.33012/2021.17808
    Google Scholar
  13. ↵
    Gunawardena, S., & van Graas, F. (2011, September). Multi-channel wideband GPS anomalous event monitor. In Proceedings of the 24th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS) (pp. 1957–1968). https://www.ion.org/publications/abstract.cfm?articleID=9744
    Google Scholar
  14. ↵
    Gunawardena, S., & van Graas, F. (2013, September). An empirical model for computing GPS SPS pseudorange natural biases based on high fidelity measurements from a software receiver. In Proceedings of the 26th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+) (pp. 1341–1358). https://www.ion.org/publications/abstract.cfm?articleID=11229
    Google Scholar
  15. ↵
    Gunawardena, S., & van Graas, F. (2014, September). Analysis of GPS-SPS inter-PRN pseudorange biases due to receiver front-end components. In Proceedings of the 27th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+) (pp. 2611–2624). https://www.ion.org/publications/pdf.cfm?articleID=12324
    Google Scholar
  16. ↵
    Gunawardena, S., & van Graas, F. (2015, January). GPS-SPS inter-PRN pseudorange biases compared for transversal SAW and LC filters using live sky data and chipshape software receiver processing. In Proceedings of the 2015 International Technical Meeting of the Institute of Navigation (pp. 393–403). https://www.ion.org/publications/abstract.cfm?articleID=12638
    Google Scholar
  17. ↵
    Kaplan, E. D., & Hegarty, C. (Eds.). (2017). Understanding GPS/GNSS: Principles and applications. Artech House.
    Google Scholar
  18. ↵
    Liu, F., Brenner, M., & Tang, C. Y. (2006, September). Signal deformation monitoring scheme implemented in a prototype local area augmentation system ground installation. In Proceedings of the 19th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS) (pp. 367–380). https://www.ion.org/publications/abstract.cfm?articleID=6855
    Google Scholar
  19. ↵
    Montenbruck, O., Hauschild, A., & Steigenberger, P. (2014). Differential code bias estimation using multi-GNSS observations and global ionosphere maps. NAVIGATION, 61(3), 191–201. https://doi.org/10.1002/navi.64
    Google ScholarOpenURL
  20. ↵
    Morton, Y. J., van Diggelen, F., Spilker Jr, J. J., Parkinson, B. W., Lo, S., & Gao, G. (Eds.). (2021). Position, navigation, and timing technologies in the 21st century: Integrated satellite navigation, sensor systems, and civil applications (Vol. 1). John Wiley & Sons.
    Google Scholar
  21. ↵
    Peng, S., & Morton, Y. (2013). A USRP2-based reconfigurable multi-constellation multi-frequency GNSS software receiver front end. GPS Solutions, 17(1), 89–102.
    Google ScholarOpenURL
  22. ↵
    Phelts, R. E. (2001). Multicorrelator techniques for robust mitigation of threats to GPS signal quality [Doctoral dissertation, Stanford University].
    Google Scholar
  23. ↵
    Phelts, R. E., & Akos, D. M. (2004, December). Nominal signal deformations: Limits on GPS range accuracy. In 2004 International Symposium on GNSS/GPS (Vol. 6, p. 8).
    Google ScholarOpenURL
  24. ↵
    Thoelert, S., Vergara, M., Sgammini, M., Enneking, C., Antreich, F., Meurer, M., Brocard, D., & Rodriguez, C. (2014, September). Characterization of nominal signal distortions and impact on receiver performance for GPS (IIF) L1/L5 and Galileo (IOV) E1/E5a signals. In Proceedings of the 27th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS+) (pp. 3113–3128). https://www.ion.org/publications/abstract.cfm?articleID=12394
    Google Scholar
  25. ↵
    Walter, T., & Enge, P. (2004, September). Modernizing WAAS. In Proceedings of the 17th International Technical Meeting of the Satellite Division of the Institute of Navigation (ION GNSS) (pp. 1683–1690). https://www.ion.org/publications/abstract.cfm?articleID=5849
    Google Scholar
Loading
Loading
Loading
Loading
PDF
Back to top

In this Issue

NAVIGATION: Journal of the Institute of Navigation

NAVIGATION: Journal of the Institute of Navigation

Vol. 73

January 2, 2026

Table of Contents Index by author

Jump to section

Keywords

GPS, signal deformation, signal quality monitoring (SQM), software defined radios (SDRs)