GPS timeout and interval (with SUPL A-GPS)

[anicare-tero]

Hi,

OS: Windows 10
Editor: SEGGER Embedded Studio for ARM (Nordic Edition) V5.50c (64-bit)
nrfjprog version: 10.13.0
JLinkARM.dll version: 7.50a
nRF Connect SDK version: 1.6.0
SUPL client library version: v0.7.0
nRF9160 modem firmware version: 1.2.3

I'm having some problems with GPS and SUPL client library on our custom nRF9160 device. I'm using the nRF9160 GPS driver with a periodic GPS search mode and I have a similar gps_controller than Asset Tracker has. I also have PSM enabled.

In purpose, I tested the device inside a building so it won't get a GPS fix and see what the device is doing then. I'm not happy with the results. I see that timeout and interval is not working properly there.

Below is parts of the project configuration, code and also logs from the device.

prj.conf

# GPS
CONFIG_NRF9160_GPS=y
CONFIG_GPS_CONTROL_FIX_CHECK_INTERVAL=1200
CONFIG_GPS_CONTROL_FIX_TRY_TIME=180

# AGPS
CONFIG_AGPS=y
CONFIG_AGPS_SRC_SUPL=y
CONFIG_FPU=y

# Heap and stacks
CONFIG_HEAP_MEM_POOL_SIZE=16384
CONFIG_MAIN_STACK_SIZE=8192
CONFIG_APPLICATION_WORKQUEUE_STACK_SIZE=4096
CONFIG_SYSTEM_WORKQUEUE_STACK_SIZE=8192
CONFIG_IDLE_STACK_SIZE=2048
CONFIG_NRF_MODEM_LIB_HEAP_SIZE=2048
CONFIG_NRF9160_GPS_THREAD_STACK_SIZE=4608

gps_controller.c

/* GPS search configuration */
struct gps_config gps_cfg = {
	.nav_mode = GPS_NAV_MODE_PERIODIC,
	.power_mode = GPS_POWER_MODE_DISABLED,
	.use_case = GPS_USE_CASE_SINGLE_COLD_START,
	.accuracy = GPS_ACCURACY_NORMAL,
	.timeout = CONFIG_GPS_CONTROL_FIX_TRY_TIME,
	.interval = CONFIG_GPS_CONTROL_FIX_TRY_TIME +
		CONFIG_GPS_CONTROL_FIX_CHECK_INTERVAL,
	.delete_agps_data = false,
	.priority = true,
};

main.c

#if defined(CONFIG_AGPS)
/* Request A-GPS data no more often than every hour (time in milliseconds). */
#define AGPS_UPDATE_PERIOD (60 * 60 * 1000)
static struct gps_agps_request agps_request;
static struct k_work_delayable send_agps_request_work;
#endif /* CONFIG_AGPS */


#if defined(CONFIG_AGPS)
/**@brief Request AGPS data. */
static void send_agps_request(struct k_work *work)
{
	ARG_UNUSED(work);

	int err;
	static int64_t last_request_timestamp;

	if ((last_request_timestamp != 0) &&
	   (k_uptime_get() - last_request_timestamp < AGPS_UPDATE_PERIOD)) {
		LOG_WRN("A-GPS request was sent less than 1 hour ago");
		return;
	}

	LOG_INF("Sending A-GPS request");

	err = gps_agps_request_send(agps_request, GPS_SOCKET_NOT_PROVIDED);
	if (err) {
		LOG_ERR("Failed to request A-GPS data, error: %d", err);
		return;
	}

	last_request_timestamp = k_uptime_get();
}
#endif /* CONFIG_AGPS */


/**@brief Callback for GPS events. */
static void gps_handler(const struct device *dev, struct gps_event *evt)
{
...
	case GPS_EVT_AGPS_DATA_NEEDED:
		LOG_INF("GPS_EVT_AGPS_DATA_NEEDED");
#if defined(CONFIG_AGPS)
		/* Send A-GPS request with short delay to avoid LTE network-
		 * dependent corner-case where the request would not be sent.
		 */
		memcpy(&agps_request, &evt->agps_request, sizeof(agps_request));
		k_work_reschedule_for_queue(&application_work_q,
						&send_agps_request_work,
						K_SECONDS(1));
#endif /* CONFIG_AGPS */
		break;
...
	}
}

/**@brief Main function. */
void main(void)
{
...
#if defined(CONFIG_AGPS)
	k_work_init_delayable(&send_agps_request_work, send_agps_request);
#endif /* CONFIG_AGPS */
...
}

log

[00:00:24.650,848] [1B][0m<inf> gps_control: GPS config set: Timeout = 180, Interval = 1380[1B][0m
[00:00:24.659,881] [1B][0m<inf> application: Enabling GPS[1B][0m
[00:00:24.665,954] [1B][0m<inf> gps_control: GPS starting in 0 seconds[1B][0m
[00:00:24.673,675] [1B][0m<dbg> nrf9160_gps.enable_gps: GPS mode is enabled[1B][0m
[00:00:24.686,340] [1B][0m<dbg> nrf9160_gps.start: GPS operational[1B][0m
[00:00:24.692,901] [1B][0m<inf> gps_control: GPS started successfully. Searching for satellites 
to get position fix. This may take several minutes.
The device will attempt to get a fix for 180 seconds, 
before the GPS is stopped. It's restarted every 1200 seconds[1B][0m
[00:00:24.718,627] [1B][0m<inf> application: GPS_EVT_SEARCH_STARTED[1B][0m
[00:00:24.725,708] [1B][0m<dbg> nrf9160_gps.gps_thread: A-GPS data update needed[1B][0m
[00:00:24.733,642] [1B][0m<inf> application: GPS_EVT_AGPS_DATA_NEEDED[1B][0m
[00:00:25.693,237] [1B][0m<dbg> nrf9160_gps.gps_thread: Waiting for time window to operate[1B][0m
[00:00:25.702,117] [1B][0m<inf> application: GPS_EVT_OPERATION_BLOCKED[1B][0m
[00:00:25.740,905] [1B][0m<inf> application: Sending A-GPS request[1B][0m
[00:00:25.747,772] [1B][0m<dbg> agps.init_supl: Using GPS driver to input assistance data[1B][0m
[00:00:25.756,622] [1B][0m<inf> agps: SUPL is initialized[1B][0m
[00:00:25.884,735] [1B][0m<dbg> agps.open_supl_socket: ip 142.250.150.192 (c096fa8e) port 7276[1B][0m
[00:00:25.971,160] [1B][0m<inf> agps: Starting SUPL session[1B][0m

[00:00:28.325,286] [1B][0m<dbg> agps.supl_logger: SUPL session finished[1B][0m
[00:00:28.332,366] [1B][0m<inf> agps: SUPL session finished successfully[1B][0m
[00:00:28.339,599] [1B][0m<dbg> net_sock.z_impl_zsock_close: (application_work_q): close: ctx=0x20021200, fd=1[1B][0m
[00:00:35.718,322] [1B][0m<dbg> nrf9160_gps.gps_priority_set: GPS priority enabled[1B][0m
%CESQ: 255,0,255,0
[00]+CSCON: 0
[00][00:00:40.088,928] [1B][0m<dbg> lte_lc.at_handler: +CSCON notification[1B][0m
[00:00:40.096,038] [1B][0m<inf> application: RRC mode: Idle[1B][0m
[00:00:40.152,313] [1B][0m<dbg> nrf9160_gps.gps_thread: GPS has time window to operate[1B][0m
[00:00:40.160,797] [1B][0m<inf> application: GPS_EVT_OPERATION_UNBLOCKED[1B][0m
[00:00:40.168,334] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Tracking: 0 Using: 0 Unhealthy: 0[1B][0m
[00:00:40.178,222] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Seconds since last fix 40[1B][0m

[00:03:24.098,236] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Seconds since last fix 204[1B][0m
[00:03:29.107,727] [1B][0m<inf> application: GPS_EVT_SEARCH_TIMEOUT[1B][0m
[00:03:29.114,685] [1B][0m<inf> application: GPS will try to get fix again after 1200 seconds,
and will try again 2 more times[1B][0m
[00:12:15.414,703] [1B][0m<inf> application: GPS_EVT_SEARCH_STARTED[1B][0m
[00:12:15.421,661] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Tracking: 0 Using: 0 Unhealthy: 0[1B][0m
[00:12:15.431,549] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Seconds since last fix 735[1B][0m

[00:26:24.443,939] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Seconds since last fix 1584[1B][0m
[00:26:29.453,521] [1B][0m<inf> application: GPS_EVT_SEARCH_TIMEOUT[1B][0m
[00:26:29.460,479] [1B][0m<inf> application: GPS will try to get fix again after 1200 seconds,
and will try again 1 more times[1B][0m
[00:37:15.449,981] [1B][0m<inf> application: GPS_EVT_SEARCH_STARTED[1B][0m
[00:37:15.456,939] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Tracking: 0 Using: 0 Unhealthy: 0[1B][0m
[00:37:15.466,857] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Seconds since last fix 2235[1B][0m

[00:49:24.476,348] [1B][0m<dbg> nrf9160_gps.print_satellite_stats: Seconds since last fix 2964[1B][0m
[00:49:29.485,961] [1B][0m<inf> application: GPS_EVT_SEARCH_TIMEOUT[1B][0m
[00:49:29.492,919] [1B][0m<inf> application: Disabling GPS[1B][0m
[00:49:29.499,084] [1B][0m<inf> gps_control: GPS stopping in 0 seconds[1B][0m
[00:49:29.506,134] [1B][0m<dbg> nrf9160_gps.stop_gps: Stopping GPS[1B][0m
[00:49:29.521,514] [1B][0m<inf> gps_control: GPS operation was stopped[1B][0m
[00:49:29.528,503] [1B][0m<inf> application: GPS_EVT_SEARCH_STOPPED[1B][0m

I read this ticket and I think that almanac downloads might have something to do with this. But shouldn't it already have ephemerides and almanacs downloaded from SUPL?

Is there any way to avoid those downloads or make timeout and interval working properly?

Or am I doing something wrong on A-GPS request?

It's killing the battery pretty fast when GPS is active for so long.

Regards,
Tero

[anicare-tero]

Just FYI: I updated the SUPL client library to version v0.7.1 and disabled FPU from project configuration.

[Øyvind]

Hi Tero, 

Did your last comment make it better, worse or no change? Have you tested the A-GPS sample, to verify behavior? You should look at this sample for the correct usage.

Kind regards,
Øyvind

[Øyvind]

Hi Tero,

Sorry for the delayed answer. I've talked to our development team regarding the GNSS. 

 

anicare-tero said:

What those "@deprecated since v1.2.0." lines means?

 There is a new GNSS API developed to improve usability of the GNSS, and to move away from the GPS driver. This can be used with modem FW 1.2.3, but will improve usability on mfw 1.3.0. Currently, the GPS sample has been updated with this, while there is still ongoing work to introduce more imrprovement.
https://github.com/nrfconnect/sdk-nrfxlib/blob/100872a0e1052931c4bad505e709a92912d4743b/nrf_modem/include/nrf_modem_gnss.h

Also, please note that the SUPL library is now available as v 0.7.1

SUPL client library v0.7.1
Changelog:

Added soft-float build of the library.
Fixed cell ID encoding in the sent request, so that the server can determine the location of the device and send a reference location back to the device. The MCC table is used as a fall back in case no reference location is received from the server.
Reference location is now only injected if it was requested.
Fixed usage of an IPv6 address as the device ID.
Removed the dependency to the modem library (modemlib). This version works with both modem and old bsdlib (sdk-nrfxlib/bsdlib) library.

[anicare-tero]

Hi Øyvind,

Øyvind said:

Also, please note that the SUPL library is now available as v 0.7.1

anicare-tero said:

Just FYI: I updated the SUPL client library to version v0.7.1 and disabled FPU from project configuration.

Already updated that earlier.

Øyvind said:

There is a new GNSS API developed to improve usability of the GNSS, and to move away from the GPS driver.

So you suggest to stop using the GPS driver and to move to GNSS API? That would need a lot of work.

Things I can't do now:

1. Take the master branch of the SDK since we need a confirmed stable SDK as we are already in production.
2. Use modem FW 1.3.0 since we have still B0 revisions and our devices have modem FW 1.2.3 which is not updatable to 1.3.0.

I really need this bug to be fixed and it kind of should be for the GPS driver.

Why is the A-GPS breaking those timeouts and intervals?

[Øyvind]

Hi anicare-tero, 

anicare-tero said:

Already updated that earlier.

Yes, that is true. I forgot that detail earlier today.

 

anicare-tero said:

So you suggest to stop using the GPS driver and to move to GNSS API? That would need a lot of work.

No, I'm asking if it helps to use this to test the stability of the GPS. 

 

anicare-tero said:

Take the master branch of the SDK since we need a confirmed stable SDK as we are already in production.

 The link I provided to GNSS API in my last comment is from NCS v1.6.1. 

 

anicare-tero said:

Use modem FW 1.3.0 since we have still B0 revisions and our devices have modem FW 1.2.3 which is not updatable to 1.3.0.

 That is OK. The new GNSS API should support mfw 1.2.3 as well. Some features are marked to work only with mfw 1.3.0, but the rest should work. 

Kind regards,
Øyvind

[anicare-tero]

Hi Øyvind,

Is there any fix for GPS driver for this issue?

Is the reason of this bug known?

As long as there is that "SHOULD support or SHOULD work" with the GNSS API with modem fw 1.2.3, it's not an option for us to move from GPS driver to GNSS API.

[Øyvind]

Hi anicare-tero,

anicare-tero said:

As long as there is that "SHOULD support or SHOULD work" with the GNSS API with modem fw 1.2.3, it's not an option for us to move from GPS driver to GNSS API.

 Sorry, the correct wording is:

/** @brief GNSS woke up in periodic mode.
 *
 * @details This event is sent when GNSS receiver is turned on in periodic mode. This happens when
 *          GNSS starts acquiring the next periodic fix but also when a scheduled download starts.
 *
 * This event doesn't have any payload.
 *
 * @note This event is only supported by modem firmware v1.3.0 or later.
 */
#define NRF_MODEM_GNSS_EVT_PERIODIC_WAKEUP     7

"...only supported by modem firmware v1.3.0 or later." Meaning that it works with 1.2.3, however, some features are only available for mfw 1.3.0. 

But I fully understand that you do not want to move to this yet.  

anicare-tero said:

Is the reason of this bug known?

Have been talking to our nRF91 team to figure out what to do. Unfortunately, the problem is not in the GPS driver but in the modem firmware itself. This is a bug that our team is aware of, and are currently working on a fix. The problem is the scheduled download logic which will perform the download regardless of the interval set. The current version of the modem fw does not allow to disable the scheduled downloads which would be the solution to use with A-GPS. This means that the periodic mode of GNSS is unfortunately not functioning as intended. Thus, we encourage customers to use GNSS in single fix or continuous mode until a fix is introduced.

[anicare-tero]

Øyvind said:

Unfortunately, the problem is not in the GPS driver but in the modem firmware itself.

That's good to know you have located the problem.

Øyvind said:

This is a bug that our team is aware of, and are currently working on a fix.

I heard there would be coming out some "sleep bug fix". Not sure if it repairs this issue but I hope so.

[pemar13]

Hi Öyvind,

I "hijack" this thread as I am becoming increasingly confused about how to update our application and custom board to achieve better GPS performance and stay up to date with Nordic's "API" development.

I recently had this exchange with Håkon. Your compatibility matrix has now been updated, but it is not clear to me how it should be read. I read it "both ways", i.e.,

"Modem FW 1.3.0 REQUIRES 1.60 - 1.61 while Modem FW 1.2.3 REQUIRES 1.40 -1.5.1"

but this is neither in agreement with your statement above or with the fact that we have successfully used modem FW 1.2.3 with ncs 1.3.2 (following your suggestion here).

This is rather crucial, since we are in production with custom boards with nRF9160 revision B0 and will start producing boards with nRF9160 revision B1, especially since your modem download page states that modem FW 1.3.0 must ONLY be used for production with revision B1.

1. I hope you can clarify this.

2. Also, when you mention forthcoming bug fixes in the modem FW, can we expect a modem FW version 1.2.4 that works as expected using revision B0?

3. Finally, using the "old" GPS driver with ncs 1.6.1 and modem 1.2.3, I don't get any GPS_EVT_SEARCH_TIMEOUT using single-fix or periodic mode. Is this as expected or just me (this worked fine with ncs 1.3.2)

Best regards,

Per

[Øyvind]

Hi Per, 

pemar13 said:

I "hijack" this thread as I am becoming increasingly confused about how to update our application and custom board to achieve better GPS performance and stay up to date with Nordic's "API" development.

 I'm really sorry to hear that you are getting increasingly confused, I hope that we can clear the confusion. You can also respond to the thread/answer with me in ticket 264436. 

 

pemar13 said:

Your compatibility matrix has now been updated, but it is not clear to me how it should be read. I read it "both ways", i.e.,

"Modem FW 1.3.0 REQUIRES 1.60 - 1.61 while Modem FW 1.2.3 REQUIRES 1.40 -1.5.1"

Where did you find these requirements? As it states "This section shows compatibility between nRF9160 modem firmware and nRF Connect SDK versions", and as stated in the nRF Connect SDK release notes, under supported modem firmware: 

"See Modem firmware compatibility matrix for an overview of which modem firmware versions have been tested with this version of nRF Connect SDK."

We have tested the latest version of the nRF Connect SDK v1.6.0 towards modem fw 1.3.0, that does not mean it is required, it means that we have not tested latest NCS with modem fw versions below 1.3.0. 

pemar13 said:

This is rather crucial, since we are in production with custom boards with nRF9160 revision B0 and will start producing boards with nRF9160 revision B1, especially since your modem download page states that modem FW 1.3.0 must ONLY be used for production with revision B1.

 Yes, that is correct, the modem fw 1.3.0 is targeted nRF9160 SiP Revision 2 as per the release notes. If you have any concerns regarding this switch from B0 to B1, please do not hesitate to contact your local RSM who should be able to provide more answers.

That said, I cannot guarantee that NCS v1.3.2 will work with modem fw 1.3.0 you will need to test this yourself. 

 

pemar13 said:

2. Also, when you mention forthcoming bug fixes in the modem FW, can we expect a modem FW version 1.2.4 that works as expected using revision B0?

 I'm afraid I do not have those details. Please ask your RSM. 

 

pemar13 said:

Finally, using the "old" GPS driver with ncs 1.6.1 and modem 1.2.3, I don't get any GPS_EVT_SEARCH_TIMEOUT using single-fix or periodic mode. Is this as expected or just me (this worked fine with ncs 1.3.2)

Are you testing with the GPS sample or your application?

Kind regards,
Øyvind

[pemar13]

Hi Öyvind!

Many thanks for the quick response!

Øyvind said:

"See Modem firmware compatibility matrix for an overview of which modem firmware versions have been tested with this version of nRF Connect SDK."

I guess I was mislead by the table's preable "This section shows compatibility between nRF9160 modem firmware and nRF Connect SDK versions", possibly overinterpreting the word "compatibility". Maybe this should be rephrased using your words like "This section shows which which modem firmware versions have been tested with different versions of nRF Connect SDK", which, to me, has a much less stringent meaning.

Øyvind said:

Are you testing with the GPS sample or your application?

I am using the nRF9160 agps sample. Am I interpreting your above statements correctly in that you recommend using the methodology from the gps sample (i.e., using the new GNSS library) rather than the old GPS driver? In that case, an upgrade of our code to ncs v1.6.1 using the GNSS library with modem 1.2.3 for B0 and 1.3.0 for B1 would seem like the best way forward for us.

Edit: This seems related to the modem FW version. With 1.3.0 the timeout works fine in single-fix mode, but not with 1.2.3.

Best regards,

Per

[Øyvind]

Hi Per, 

pemar13 said:

Maybe this should be rephrased using your words like "This section shows which which modem firmware versions have been tested with different versions of nRF Connect SDK",

Noted. I will forward this feedback to the responsible person. 

 

pemar13 said:

Am I interpreting your above statements correctly in that you recommend using the methodology from the gps sample (i.e., using the new GNSS library) rather than the old GPS driver?

Yes, that is correct, but as mentioned, the mfw 1.2.3 is not tested with NCS v1.6.1, so we cannot guarantee all functionality. As long as it works for your application, then it's OK.

Kind regards,
Øyvind