Fallouts in GNSS reception using AT#XGPS=1,1,1 and mobile networking disabled.

Question

Hi, 
 We are evaluating the NRF9151 for live tracking purposes. While testing with the nRF9151-DK we noticed the we had tracking fallouts using AT#XGPS=1,1,1 both during driving and stationary tests. So in order to only see how the GNSS itself acts we did testing, where after getting a fix the modem will go into GNSS only mode and disconnect from the mobile network. 
 With a stationary test we got a great result with 99.77% of GNSS positions arriving at 1s +/- 100ms intervals as expected, though still a few arriving with up to ~4s delay. The stationary test was done outside for more than 3 hours with only a plastic box on top to protect against potential rain. 
 With a driving test we got a few larger breaks in GNSS positions of more than 40s, though still ~98.65% of GNSS positions arrived at an 1s +/- 100ms interval. The large breaks though are noticeable in our application, where we send a position every 10s. The driving test was done with the modem on the passenger seat, while driving through an area with mostly fields, single story houses, mild tree cover at some roads and no tunnels or bridges. The test took about 45mins and went through the same area twice. 
 Is this expected performance and would it be expected to improve with an external antenna? 
 Thanks in advance.

Håkon Alseth · Accepted Answer

Hi, 
 
 Ideally, we would want the NMEA strings to look at each individual satellite's output, and thus check the SNR. 
 This can be done by setting the log output mode in SLM: 
 
 Still, we can look at the "accuracy" parameter here based on the notified value we get back: 
 https://docs.nordicsemi.com/bundle/ncs-latest/page/nrf/applications/serial_lte_modem/doc/GNSS_AT_commands.html#unsolicited_notification 
 which will give us an indication on the overall accuracy of the returned GPS data. 
 This "accuracy" parameter is effectively a sum of "dilution of precision": 
 https://en.wikipedia.org/wiki/Dilution_of_precision_(navigation) 
 
 You want this number to be as low as possible. If we look at the numbers while driving, especially the ones where you have gaps, the accuracy parameter is 20 to 50 area: 
 (11:22:16.340) #XGPS: 55.841264,12.206062,68.760078,24.440924,22.152363,258.476349,"2025-08-21 09:22:16"
(11:23:51.330) #XGPS: 55.834678,12.174893,49.940460,45.608971,20.690598,272.305969,"2025-08-21 09:23:51"
..
(11:28:03.314) #XGPS: 55.835753,12.189951,48.109840,28.087284,14.759391,66.664894,"2025-08-21 09:28:03"
(11:28:35.326) #XGPS: 55.836433,12.197599,83.675026,48.383675,12.404172,63.467396,"2025-08-21 09:28:35" 
 This indicates that satellites are going from tracked->untracked, and the total amount of tracked satellites is less than 4. 
 
 The driving test was done with the modem on the passenger seat, while driving through an area with mostly fields, single story houses, mild tree cover at some roads and no tunnels or bridges. 
 I suspect that the placement of the device within the car has a great impact. The view it has present when in the seat is fairly limited towards the windshield itself, as compared to being placed in the windshield. 
 Note that newer cars have windows and windshields that can attenuate signals better than older cars. 
 
 I would need to look at the NMEA strings in both scenarios, especially the GPGSV, as this holds the SNR for each satellite in view, to confirm this theory. 
 
 Kind regards, 
 Håkon

Fallouts in GNSS reception using AT#XGPS=1,1,1 and mobile networking disabled.

Top Replies