So for the foreseeable future, surveyors need to pair their tablets to external GPS or GNSS units to achieve better than 5-meter accuracy during remote field surveys. To show how the accuracy of a tablet or phone compares to an affordable 2.5-meter accuracy external GPS unit, we setup field testing that would accurately compare the internal GPS chips of the iPad Air and iPhone 6 to the Bad Elf GPS Pro (BE-2200) and newer Bad Elf GPS Pro+ (BE-2300).
GPS and GNSS Devices Compared
|Device||Receiver Type||Factory Accuracy Rating|
|iPad Air||“Assisted” GPS and GLONASS||Not Available|
|iPhone 6||GPS and GLONASS||Not Available|
|BE-2300||GPS and GLONASS||2.5-meter|
To compare these units, we used a local National Geodetic Survey (NGS) marker as a known location to compare points collected with each type of device. Specifically, we used NGS marker AI2002 as a reference point, as it has fairly good sky view with the nearest tall trees located 40 meters south of the site.
Each device was tested on the same day at the same marker. Using the iPad and iPhone 6 required us to turn off the cellular data and wi-fi to ensure the internal GPS was forced into use when collecting points with a GIS app. We made sure only the device being tested was turned on and nearby to minimize interference.
To collect test data, we:
- Allowed the test device 10 minutes to “warm up” by collecting the most current data almanac, if applicable
- Held the device at arm-length and head height, flat on the palm
- Faced south to take advantage of the WAAS satellites
- Recorded 1 point each 60 seconds for 10 minutes until testing was done (N=10 per device)
- Finally, the device was turned off and the next test device was activated and we started over at Step 1
The data was exported from our GIS app as shape files. A project was created with Quantum GIS (QGIS) ver. 2.4.0-Chugiak, free open source GIS software. All data was converted to NAD83 (HARN) (EPSG:4152). We used the MMQGIS plugin to calculate hub distance between the known location of AI2002 and our test data points.
For ease of comparison, we averaged the 10 records for each device and calculated the 95th Percentile. This is an important number because most GNSS devices give their accuracy rating as an accuracy number that is achieved 95% of the time in clear-sky scenarios. This means that in clear sky views, you can likely expect an accuracy slightly better than the factory rating most of the time. Remember, this assumes your location has a clear view of the sky and isn’t being interfered with by something. Placing your GPS receiver or antenna on a metal structure, next to another device sending and receiving signals, or partially blocking the device’s sky view with your own body will interfere with and lessen the accuracy and precision.
If you need a reminder of the differences between accuracy and precision, please see our previous blogpost.
Accuracy Test Findings
N=10 for each device; all measurements in meters
|Device||Factory Accuracy Rating||Highest Accuracy||Lowest Accuracy||Test Average||Test 95th Percentile|
|iPad Air||Not Available||2.10||6.61||4.84||6.59|
These results are a good example of the accuracy differences between the internal GPS chips of cellular devices and those of dedicated stand-alone GPS/GNSS devices. Keeping in mind that a sample size of 10 for each device is small, but these data points were collected under normal field conditions in the same method we have used to correctly collect hundreds of points over the years. Also, expect these accuracy ratings to degrade during work in woodlands with tree canopy causing interference with satellite signals.
With that said, we were shocked to see how similar the accuracy ratings were for the iPad Air and iPhone 6. It was also surprising to see the Apple devices achieve better than 7-meter accuracy for the 95th percentile. Many times, we have seen an iPhone’s internal GPS return accuracy ratings as poor as 10+ meters.
The Bad Elf GPS Pro (BE-2200) and newer Bad Elf GPS Pro+ (BE-2300) did great with both units being better than their rated 2.5-meter accuracy.
Next one would ask, “why should I pay more for the BE-2300 over the BE-2200 if their accuracy is so close?”. Good question. We would still recommend going with the BE-2300 because of that model’s added GLONASS capability. By adding these extra satellites, the BE-2300 should perceivably be able to acquire more accurate data and better precision in real-world field settings with various levels of interference such as topography and tree canopy because of the larger number of available satellites the GLONASS capability affords.
The data provided in the table above shows that the 95% confidence for the BE-2200 was within 0.24 meters of the BE-2300, and both units were better than the manufacturer rated 2.5-meters. This is a relatively small difference, however, the BE-2300 did show better precision than the BE-2200. The plotted data in the figure to the right shows BE-2300 data was more precise, all within 0.8 meters of each other. The greatest distance between BE-2200 data points was 3.0 meters.
See the figure; (click on the image and then expand to easily read the labels)
BE-2300 is represented with black diamonds, and
BE-2200 represented by red triangles