A GPS receiver fails in two ways. The first is obvious: the battery runs out. The second is slower and harder to diagnose: the receiver spends so much energy searching for a signal that it never has enough left to do anything useful with one.
That second failure mode is what we kept seeing in San Francisco, and it turned out to trace back to an architectural assumption the industry stopped questioning a long time ago.
Figure 1. oneNav Test route in San Francisco Financial District
The Reacquisition Loop
In the Financial District, we watched competing receivers enter what we started calling the reacquisition loop. The device picks up a signal, starts building a fix, loses the signal behind a building, and begins searching again from scratch at full power. In a dense urban canyon, this happens dozens of times per minute. Buildings exceed 200 meters. Street corridors leave receivers with a narrow slice of visible sky, and signals that do arrive have often bounced off glass and steel facades before reaching the antenna. The receiver never really settles. It burns energy.
This is not a fringe scenario. It is the operating environment for a growing class of devices: last-mile delivery robots, wearable navigation systems, dismounted military operators, autonomous inspection platforms. For all of them, the urban canyon is the mission, and the reacquisition loop is the primary obstacle to completing it.
Why the Architecture Is the Problem
Most dual-band GPS receivers can’t acquire L5 without first locking onto L1. The L1 lock is a hard architectural dependency baked into the design. When L5 was first introduced, this made sense: L1 C/A was the only practical way to obtain the timing and ephemeris data a receiver needs to compute a position. L5 was a precision layer bolted on top of an existing system.
That constraint no longer exists. L5 civil signals now carry their own CNAV navigation message, which includes full ephemeris and satellite health data. A receiver with a warm-start almanac has everything it needs to acquire L5 directly. The dependency on L1 is a legacy of how the system was built, not a reflection of how it needs to work.
Figure 2. Architectural difference between L1/L5 and L5-direct GNSS receivers.
The deeper problem is that L1 is also the wrong signal for cities. It is weaker than L5 and far more vulnerable to multipath interference, the reflections that bounce off glass towers and arrive at the antenna corrupted or delayed. In a dense urban environment, L1 is constantly being blocked and degraded. Each time lock drops, the receiver enters a full-power search state: correlator banks active, AGC hunting, oscillator unsettled. It burns through the power budget chasing the signal it depends on most and can reliably receive least.
The conventional dual-band receiver is structurally committed to the worst signal for the environments it is increasingly asked to operate in.
What L5-direct Changes
L5-direct removes the dependency entirely. By acquiring L5 without ever requiring an L1 lock, the receiver skips the most power-hungry phase of GNSS operation. The effects compound across the hardware.
Single RF chain.
A conventional dual-band receiver runs two parallel paths, each with its own LNA, filter, and downconversion stage. L5-direct uses one, which removes 5 to 15 mW from the power budget before the baseband processor does any work.
Fewer reacquisition events.
L5 is transmitted at roughly 3 dB higher power than L1 and uses a signal ten times wider in bandwidth. The wider bandwidth sharpens the correlation peak, making multipath reflections harder to corrupt the measurement and making it easier to hold lock in difficult conditions. The reacquisition loop does not disappear entirely, but it becomes rare rather than constant.
Lower duty cycle.
In the San Francisco conditions where hybrid receivers were running at 50 to 100% duty cycle, essentially never sleeping, our L5-direct engine ran at around 10%. The baseband processor spends most of its time in a low-power state, waking briefly to confirm the fix before returning to sleep.
At 4.7 mW in continuous tracking mode, a platform that previously needed a battery recharge every few hours can now run for days. For a dismounted operator, a delivery robot, or a wearable medical device, this capability is the difference between a device that completes the mission and one that goes dark before it does.
The Tradeoff That No Longer Exists
Traditional GNSS architectures impose a linear tradeoff between power and accuracy. Reduce processing intensity to extend battery life and positioning precision drops with it. Every efficiency gain comes at a cost to performance. This is the constraint that has shaped receiver design for decades.
L5-direct breaks it. Rather than running at constant power, the receiver stays in a low-power state and bursts to 27 mW only when signal conditions require it. Accuracy stays flat regardless of how often those bursts occur. The system pays the power cost when the environment demands it, and not before.
Figure 2. The relationship between power and accuracy in L1 single band, L1/L5 dual band, and oneNav L5-direct receiver. The dots represent full tracking power, the arrows represent adaptive low power tracking and FTP stands for Frequency Track Power. Note two solid arrows for oneNav – the longer arrow includes all constellations, the shorter represents results with GPS and Galileo only.
The accuracy results reflect this. In matched urban canyon testing, L5-direct achieved a 2x improvement in 99th percentile positioning error over leading L1/L5 hybrid competitors. The 99th percentile matters more than the median here: it measures the worst-case fixes, the ones that happen in the deepest canyons, at street level, near the most reflective surfaces. Those are the conditions where a bad position has the most consequences, and where traditional architectures, caught between power and precision, struggle most.
Figure 4. Scatterplot of 99th percentile horizontal positioning error comparing L5-direct over leading L1/L5 hybrid competitors.
Part of that improvement comes from L5’s signal characteristics. The rest comes from an ML-based multipath classifier that identifies reflected signals and incorporates them into the position calculation rather than treating them as noise. When a reflected signal arrives inside the correlation peak of the direct path, a condition that defeats narrow-correlator techniques, the classifier recognizes it for what it is and uses that information to improve the fix. The result is a more accurate position computed from the full signal environment, not a filtered one.
San Francisco in Numbers
In our Financial District trials, oneNav’s L5-direct engine achieved a median Time to First Fix of 33 seconds. Competing hybrid receivers frequently took several minutes, caught in reacquisition attempts that our architecture bypasses by design.
Table1. Performance gap between oneNav and competitive devices in a pedestrian test in San Francisco Financial District.
Thirty-three seconds is fast for an environment where many receivers cannot produce a stable fix at all. But the more important number is the gap. For autonomous systems making navigation decisions, for operators working under time pressure, for any platform where position feeds directly into action, the minutes spent hunting for a signal are minutes the mission cannot afford.
The Design Principle
These gains follow directly from removing a structural dependency that was always the weakest point of the architecture. Every milliwatt saved in the RF front-end extends mission duration for battery-constrained platforms. For devices operating on 100 to 500 mAh cells, which covers most wearables, small robotics platforms, and dismounted electronics, the GNSS subsystem is consistently one of the largest contributors to power consumption. The gap between 4.7 mW and 40+ mW determines how long the device remains useful.
The solution was removing a dependency that had outlasted its justification.
oneNav L5-direct has been proven in silicon and extensively tested in deep urban environments. It is available now as a licensable IP for integration into wearable, robotics, autonomous navigation, and dismounted operator platforms.
About oneNav
Based in California, oneNav is developing L5-direct™ GNSS receiver technology for unmanned systems (UxS) platforms, munitions, consumer devices including smartphones, wearables, and tracking devices, and more, and has built a large L5-band patent portfolio. oneNav’s advanced technology features cutting-edge anti-jamming and anti-spoofing capabilities, delivering robust, reliable positioning even in contested and challenging signal environments. oneNav’s team comprises top GNSS experts from Qualcomm, Apple, Intel, SnapTrack, SiRF, Trimble, and eRide. With extensive experience in GNSS system architecture, multipath mitigation, signal processing, ASIC design, and AI/machine learning, oneNav engineers have designed and built billions of GNSS receivers on the market today and have collectively filed over 300 career GNSS patents.