How accurate is USBL positioning on a small ROV?
About 5 to 12 metres absolute, depending on range - and for most of a working envelope the limit is the GPS on the boat, not the sonar.
Expect about 5 to 12 metres of absolute position accuracy from an acoustic positioning system on a small ROV, depending mostly on how far the vehicle is from the surface unit. Close to the boat you will be near 5 m; at 300 m slant range, closer to 12 m. Relative tracking - watching the vehicle move, holding station, returning to a marked spot - is considerably tighter than that.
The surprising part is why. For most of a typical working envelope, the limit is not the acoustics at all. It is the GPS receiver sitting on the surface.
Why a topside GPS sets the floor
An acoustic positioning system does not measure where your ROV is in the world. It measures where your ROV is relative to the surface unit, and then adds the surface unit’s own position to get a world coordinate.
That means two errors stack:
- Where the surface unit thinks it is. A standard GNSS receiver manages 3 to 5 metres circular error probable. Every subsea position inherits this.
- Where the acoustics think the vehicle is relative to the surface unit. Range measurement is excellent - tenths of a metre. Bearing is the weak axis, typically around 2 degrees.
Bearing error converts to distance error as the vehicle moves away. Two degrees is about 3.5 m of sideways error at 100 m slant range and about 10.5 m at 300 m.
Cerulean Sonar publish the resulting formula for their Omnitrack system, which is unusually candid - most manufacturers publish a single flattering number instead:
CEP ≈ √( 25 + ( sin(2.1°) × slant range in metres )² )
The 2.1° combines about 2° of angular measurement error with about 0.5° of heading error. The 25 is the GNSS error squared - five metres, squared.
Which means the two terms are equal when sin(2.1°) × range = 5 m, or a slant range of roughly 136 metres. Inside that radius, your position error is dominated by the GPS on the boat. Beyond it, the acoustics take over.
| Slant range | Error from acoustics | Total position CEP |
|---|---|---|
| 25 m | 0.9 m | about 5.1 m |
| 50 m | 1.8 m | about 5.3 m |
| 100 m | 3.7 m | about 6.2 m |
| 150 m | 5.5 m | about 7.4 m |
| 200 m | 7.3 m | about 8.9 m |
| 300 m | 11.0 m | about 12.1 m |
Calculated from Cerulean’s published formula using their 3–5 m GNSS figure. Arithmetic, not a measurement.
The practical consequence is worth stating plainly. If you are working inside 100 m and you need better than 5 m absolute, a more accurate USBL will not help you. RTK correction on the topside will. That is a different purchase, and often a cheaper one.
Resolution is not accuracy
The commonest way to be misled by an acoustic positioning datasheet is to read a resolution figure as an accuracy figure.
Resolution is the smallest increment the instrument reports. Accuracy is how close that report is to the truth. On the ROV Locator hardware inside Omnitrack, bearing resolution is 0.1° and slant range resolution is 0.1 m - but the IMU’s actual angular accuracy is 2°, twenty times coarser. Cerulean flag this directly in their own specification table, with the words “Resolution ≠ Accuracy” and a link to the accuracy page.
Reseller listings routinely strip that warning out and present the 0.1° figures under headings about precision. If a specification table shows 0.1° and does not mention 2° anywhere, you are reading resolution.
What measured results look like
Cerulean ran a controlled field test across an ice-covered freshwater lake, which is about as favourable as geometry gets: everything static, no wind, no current, holes cut in the ice at known distances. A Mk III system gave:
- At 110 m slant range: a scatter band about 10 m long, bearing error about ±2°, slant range error about 0.5 m.
- At 295 m slant range: a scatter band about 35 m long, bearing error about ±2.5°.
About 90% of the error in both cases was relative bearing measurement plus IMU heading error. Cerulean themselves describe the setup as not ideal for sonar - the bottom undulated and there was vegetation to the side.
Two things to take from this. The measured scatter is broadly consistent with the published formula, which is a good sign about the formula. And the error grows with range much faster than intuition suggests, because it is angular.
The options, and what separates them
Three technologies get sold to small-ROV operators, and they fail differently.
USBL (ultra-short baseline). One transducer array on the surface measures range and bearing to one transponder on the vehicle. Single compact surface unit, easy to deploy, and the industry default. Its weakness is that bearing accuracy depends on knowing which way the surface array is pointing, so heading measurement becomes part of the accuracy budget.
SBL (short baseline). Four separate receivers, or a four-receiver antenna, lowered into the water. Position comes from time-of-flight differences between them. Water Linked’s Underwater GPS G2 is the system most small-ROV operators will weigh against a USBL. SBL handles shallow water and acoustically reflective sites - near ship hulls, inside fish cages, alongside harbour walls, in tanks - better than USBL, which is Water Linked’s own stated position.
DVL (Doppler velocity log). Measures velocity over the seabed. Excellent short-term precision at high update rate, no absolute reference, and error accumulates over time. It is a complement to acoustic positioning rather than an alternative - which is the next question most people ask, and it has its own section below.
USBL against SBL, on published numbers
| Cerulean Omnitrack USBL | Water Linked UGPS G2 | |
|---|---|---|
| Principle | USBL | SBL |
| Maximum range | 300 m | 100 m or 300 m by version |
| Update rate | 1 Hz (up to 5 Hz on one configuration) | 2–4 Hz |
| Published angular accuracy | 2.1° combined, heading included | <1° horizontal angle |
| Heading source | Two GNSS antennas, built in | External compass needed in dynamic conditions |
| Surface deployment | One ring buoy, 8.6 kg | Topside case plus antenna or four receivers in the water |
| Subsea unit depth rating | 300 m | 300 m (A1 locator) |
| Long-term drift | None | None with A1; about 1 m per 6 h with U1 |
| EUR, European resellers | from €7,000 | listed by a Dutch reseller at up to about €8,100 excl. VAT across variants |
Read that angular row carefully, because it is the one that matters and it is not a straight comparison. The G2’s sub-1° figure is better than Omnitrack’s 2.1°. But Water Linked’s own documentation states that in dynamic or challenging environments an external compass must be supplied to achieve accurate position data - so the heading error is not inside their figure, and it is inside Cerulean’s. Omnitrack’s two GNSS antennas are that compass, built in and immune to magnetic interference.
Neither number is wrong. They are measuring different things, and a buyer comparing them side by side without noticing will reach the wrong conclusion.
What dual-GNSS heading actually buys you
A magnetic compass near a small ROV operation is in a hostile environment. Thrusters, batteries, a steel hull, a quay wall, sheet piling, the vessel itself - all of them pull on it. Magnetic calibration is fiddly, has to be redone when the environment changes, and degrades without announcing itself.
Two GNSS antennas on a fixed baseline derive heading from satellite geometry instead. Accuracy runs about 0.5°, no calibration is required, and it does not care what the boat is made of.
The catch is that it needs sky. Under a bridge, inside a dock, hard alongside a large hull, the GNSS compass loses its fix and a fallback magnetometer takes over - with the degraded heading you were trying to avoid. Cerulean are explicit that both the position and heading figures assume the compass is not shaded by the boat or by topside gear.
So: dual-GNSS heading is a clear win for open-water and structure-inspection work, and it is exactly where a magnetic compass is worst. It is not a solution for positioning inside enclosed structures, and nothing acoustic really is.
The adjacent question: do I also need a DVL?
This comes up immediately after accuracy, and the honest answer is that it depends on whether a human or an autopilot is reacting to the position.
A USBL fix arriving once per second is fine for logging where things are and geo-referencing sonar and video. It is too slow to fly on. Between fixes the vehicle keeps moving and the display keeps showing where it was, which is uncomfortable when you are trying to hold station against current and tether pull.
A DVL measures velocity at a much higher rate. Fuse the two and each covers the other’s weakness: the USBL supplies an absolute reference that stops long-term drift, and the DVL supplies smooth continuous motion between fixes. On ArduSub vehicles the autopilot performs the fusion itself.
Rough guide. Logging position and geo-referencing data: acoustic positioning alone. Holding station, flying survey lines, or anything autonomous: budget for both. A Cerulean Tracker 650 adds from €2,625 to a system starting at €7,000, so it is a meaningful increment rather than an afterthought.
What else costs you accuracy in the field
- GNSS shadow. Mount the compass where it can see sky. A wheelhouse or an A-frame between the antennas and the satellites degrades both position and heading, and the degradation is quiet.
- Multipath. Sound bouncing off the seabed, the surface or structure arrives late and corrupts the bearing. Worst in shallow water and near hard vertical faces. This is physics, not a product flaw, and it affects every acoustic positioning system ever built.
- Speed of sound. These systems convert time to distance using an assumed sound speed. Get it wrong - a thermocline, a river plume, brackish water - and every range is proportionally wrong. It is a configurable setting and worth setting.
- Range limits that are firmware, not physics. Omnitrack stops at 300 m because Cerulean cap it there to stop adjacent systems interfering with each other. It does not fade out. It stops.
- Geometry. A vehicle almost directly below the surface unit gives a strong solution. A vehicle at long range and shallow depth gives a poor one, because the bearing error is spread across a long lever arm. Move the boat.
In three lines
Acoustic positioning on a small ROV gives roughly 5 to 12 metres of absolute accuracy, and inside about 136 m of slant range the limiting factor is your topside GPS rather than the sonar.
Choose USBL with GNSS heading for open water and structure work where a magnetic compass cannot be trusted; choose SBL for shallow, enclosed or acoustically reflective sites.
Add a DVL if anything - pilot or autopilot - has to react to the position in real time.
Frequently asked questions
How accurate is USBL positioning?
Around 5 metres close to the surface unit and around 12 metres at 300 m slant range, for systems in the small-ROV price bracket. The error is dominated by topside GNSS at short range and by angular measurement at long range. Cerulean publish a formula for their Omnitrack system: CEP ≈ √(25 + (sin(2.1°) × slant range in metres)²).
Why is my USBL only accurate to 5 metres when the ROV is right underneath the boat?
Because the surface unit’s own GPS is only accurate to 3–5 metres, and every subsea position inherits that error. The acoustics are performing well; the limit is the GNSS. Adding RTK correction to the topside is what improves absolute accuracy at short range, not a better acoustic system.
What is the difference between USBL and SBL for a small ROV?
USBL uses a single surface transducer array measuring range and bearing, which makes deployment simple but puts heading accuracy into the error budget. SBL uses four receivers in the water and calculates position from time-of-flight differences, which handles shallow water and reflective environments better but requires more hardware over the side.
Does a USBL need a DVL as well?
Not for logging position or geo-referencing sonar and video. Yes in practice for holding station or autonomous survey lines, because a 1 Hz USBL fix is too slow to fly on. The two fuse well: USBL supplies the absolute reference, the DVL supplies smooth motion between fixes.
Is GNSS heading better than a magnetic compass for USBL?
For open-water and structure work, yes, and clearly so. Two GNSS antennas give around 0.5° heading with no calibration and complete immunity to steel, thrusters and quay walls. The trade-off is that they need sky - under a bridge or inside a dock they lose their fix and a fallback magnetometer takes over.
What is the maximum range of a USBL system for small ROVs?
Typically 300 metres in this class. Some systems reach 500 m by using a different timing-synchronisation method, which brings its own operational cost - periodic surfacing for GPS re-sync and accumulated clock drift. Accuracy degrades with range throughout, so maximum range and usable range are not the same figure.
Does resolution mean the same thing as accuracy on a sonar datasheet?
No, and conflating them is the commonest datasheet error. Resolution is the smallest increment reported; accuracy is how close it is to the truth. A system reporting bearing to 0.1° may have an actual angular accuracy of 2° - a twenty-fold difference.
Working out which system suits the job?
SepcoTech supplies the Cerulean Sonar positioning range across Europe with EUR pricing and no hidden import costs. Tell us the vehicle, the water depth and whether you need to hold station or just log position - we will name the configuration and say so if a simpler system would do.
SepcoTech A/S is an authorised Cerulean Sonar distributor. The hardware, warranty and documentation are Cerulean’s; the stock, the EUR quote and the support in your time zone are ours.
See the Omnitrack USBL Compare the positioning range Request a quoteTell us the job. We will name the configuration.
Tell us the vehicle, the water depth and whether you need to hold station or just log position. We will name the exact configuration - and say so if a plain ROV Locator would do the job for half the money.
Tell us what you are flying, where, and what the position data is for. Anything you have added to your quote list is carried into this form automatically.
Or call us on (+45) 6916 2400.