Nrf9160 GPS long time to fix with healthy RF electronics

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Mon, 19 Apr 2021 13:33:04 GMT

Will add a comment;

After more internal testing from our engineers the results seems to point out hat you might ave a sub-optimal antenna design.

Customer's Ext-LNA + Antenna seem to give about 15...17dB gain (is it expected result with given antenna gain+ LNA gain?), which is close internal HW Ext-LNA (SKY65943-11) gain derived from RSSI rise w/ and w/o Ext LNA
From conducted results SNR 22.5dB at -90dBm compared to customer's radiated data SNR ~15dB at -90dBm ant power: Customer DUT's SNR missing several dB's, which could be explained by noise coupling to GPS antenna or antenna gain (again antenna gain/efficiency and LNA gain would be good to know)
Test with internal HW show that Ant pwr level vs SNR with and without external LNA behave as expected i.e. no SNR dips with external LNA and total RX SNR is maintained nevertheless SIP's NF rise due to AGC steps (that is important if very high gain Ext-LNA is placed in GPS frontend)
These CW tests are just to check RX path validity and check possible unwanted noise, true GPS signal at DUT antenna lies well below thermal noise <-113dBm/1.023MHz which is amplified by ext-LNA before RFIC

So our recommended prosess for designed new HW is as follows.

1) Conducted RF performance verification in lab: verify HW path from antenna feed to nRF9160 SIP GPS input is performing as expected in terms of ext-LNA control, in/out matchings and impedance control of routing.

2) Passive GPS antenna measurements: measure antenna radiation pattern and efficiency (=aka passive antenna measurements) to understand a bit how the device might perform. Do we have e.g. these results available?

3) Combined performance: to verify conducted and antenna solution performance adds-up as expected when combined. Here e.g. EMI from other parts of HW design (LEDs, other active ICs,..) may cause new issues not detected in 1) nor 2). (in proper anechoic chamber = consistent/known environment)

Input for such testing is given in https://infocenter.nordicsemi.com/topic/struct_nrf91/struct/nrf91_guidelines.html?cp=2_0_7
=> https://infocenter.nordicsemi.com/topic/nwp_034/WP/nwp_034/nwp_034_rf_perf_test.html
=> https://infocenter.nordicsemi.com/topic/nwp_033/WP/nwp_033/nwp_033_intro.html

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Tue, 23 Mar 2021 19:20:54 GMT

Hi Bjørn,

One thing that sticks out is the dips/drops that you experience. R&D replicated a the test and could not reproduce the results.

The result is perfectly aligned with thingy results – the SNR gets better as the signal level increases.

There is 15.5dB headroom in the Gain control design, so it is expected that SNR starts to get worse after -60 (if -75 is used in AT command and SIP input sees -75 dBm too), but all the way from -75 to -60 it seems to get linearly better if tested like this with cable (which seems to have about 1.5 dB loss). Since I don't have LNA on board the signal generator has 15 dB power more than the customer.

Signal Gen dBm	AT cmd	SNR	RSSI
-77	-75	409	-20037
-76	-75	425	-19796
-75	-75	439	-19546
-74	-75	453	-19286
-73	-75	463	-19055
-72	-75	481	-18775
-71	-75	486	-18535
-70	-75	505	-18284
-69	-75	515	-18004
-68	-75	534	-17764
-67	-75	539	-17513
-65	-75	561	-17013
-63	-75	571	-16512
-61	-75	579	-16001
-59	-75	593	-15471
-58	-75	583	-15230

So other then the dips the results you have is not bad, so on thing that might be a cause of this difference is the tuning of the antenna. Have you tested, tuned and matched the antenna with propper RF equipment ?

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Thu, 18 Mar 2021 20:11:16 GMT

Hi, (18.3.2021)

We added some more test, this time of the original GSP product with standard LNA (3dB lower compared to the previous test).

The product is located in the GTEM test chamber and with RF power to the antenna from -90dBm and increased to -70dBm. The purpose is to check if the LNA or NRF9160 hits saturation at a certain RF input power level.

These tests where run with AT command «at at%xrftest=2,1,-75». Some tests where repeated with the highest input power and with «at at%xrftest=2,1,-55».

Pin (dBm)	SNR	RSSI	SNR(dB)	RSSI(dBm)
-90	232	-19155	15	-75
-88	239	-18254	15	-71
-86	261	-17804	16	-70
-84	293	-17333	18	-68
-82	232	-16863	15	-66
-80	349	-16392	22	-64
-78	366	-15911	23	-62
-76	388	-15391	24	-60
-74	232	-15020	15	-59
-72	299	-14870	19	-58
-70	390	-14760	24	-58

Testet with XRFTEST=2,1,-75

SNR at the highest input power gave some strange data. Hence the test was repeated with the highest input power levels but with XRFTEST=2,1,-55.

Pin (dBm)	SNR	RSSI	SNR(dB)	RSSI(dBm)
-74	401	-14877	25	-58
-72	400	-14376	25	-56
-70	418	-13875	26	-54

THINGY TESTS:

The last test I did was on the Thingy with external antenna, , SAW and LNA. Antenna type Taoglas 574R. The J2 is connected to the extern

Pin (dBm)	SNR	RSSI	SNR(dB)	RSSI(dBm)
-90	261	-19706	16	-77
-85	337	-18555	21	-72
-80	412	-17313	26	-68
-75	472	-16041	30	-63
-70	371	-14970	23	-58

It has been demonstrated in earlier posts that the Thingy GPS has somewhat better GPS performance. But from these test it is not super clear why. We have seen that the external antenna used on the Thingy has better Xpol performance compared to the Product. A good Xpol antenna will give 3dB for free compared to a linear receiver antenna.

Thank you for your support on this case Jonathan.

Regards

Bjorn

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Wed, 17 Mar 2021 21:30:16 GMT

Hi Jonathan,

I made a few other measurements and the primary objective was to investigate if the receiver goes into saturation.

I used AT commands through the JLINK interface to communicate with the product. The prodcut was positioned inside my GTEM cell. The signal source could be ste to have different levels of RF power.

Test one: Using RF power levet at the GPS antenna port of -90dBm and command "at at%xrftest=2,1,-75". This gave the result 271,-18825 (SNR=17dB and RSSI=-73.5dBm).

Then the RF signal power was increased from -90dBm up to -72dBm (using command "at at%xrftest=2,1,-75"):

Antennapower(dBm) SNR RSSI SNR(dB) RSSI(dBm)

1) RF Power = -90dBm -90 274 -18805 17,1 -73,5

2) RF Power = -87dBm -87 318 -18104 19,9 -70,7

3) RF Power = -82dBm -82 393 -16873 24,6 -65,9

4) RF Power = -80dBm -80 412 -16392 25,8 -64,0

5) RF Power = -78dBm -78(edit from -7 434 -15881 27,1 -62,0

6) RF Power = -76dBm -76 455 -15351 28,4 -60,0

7) RF Power = -74dBm -74 325 -15020 20,3 -58,7

8) RF Power = -72dBm -72 384 -14850 24,0 -58,0

(Table 1)

I also tested with "at at%xrftest=2,1,-57" with antenna power of -72dBm followed by "at at%xrftest=2,1,-53" with antenna power of -67dBm. This gave SNR 495(31dB)/RSSI -14352(-56dBm) and SNR 538(34dB)/RSSI -13093(-51dBm).

I have added two graphs from "table 1" above.

It appears that SNR goes into saturation but only at significant higher power than -90dBm. I cannot see evidence that the Product LNA goes into saturation at -90dBm to the antenna (in the GTEM cell).

(Edit 18.3.2021: The LNA in these tests have been upgraded and has 3dB more gain. )

Thanks,

Regards

Bjørn

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Wed, 17 Mar 2021 16:40:58 GMT

Using this config : AT%XRFTEST=2,1,-75,0
the result should be close to 75dBm for the antenna power.

since you had a offset the SNR values you got from the measurement from the chamber test there is likely a offset in the SNR value not giving the true result due to the ~15dB offset from not accounting for the LNA.

If that maks sense ?

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Mon, 15 Mar 2021 16:58:16 GMT

Hi Jonathan,

Thanks for your answer.

I am not quite sure I understand the explanation on saturated LNA. We do have a LNA, SAW filter and an antenna upstream the gps port of the nrf9160. If we apply -90dBm to the antenna port (using the TEM wave produced by the GTEM cell) and the xRFTEST gives RSSI of -76dBm. then the gain is 14dB and roughly as expected.

My understanding of the AT%XRFTEST=2,1,-75,0 is that it should give 0dB gain?

Could you explain?

Thanks,

Regards

Bjørn

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Mon, 15 Mar 2021 09:54:08 GMT

Hi Bjørn,

Sorry for the slow reply, but what is the gain of the LNA used?

The camber test summarized:
And the values range (100 runs):
Antenna power:-19411 to -19371 so that would be -75.8dBm to -75.6dBm
SNR: range from 283 to 297, that would be 17.6dB to 18.5dB

Some feedback on the chamber test.

[quote user=""]Our preliminary conclusion is that the product (and the Thingy:91) has good RF performance (around 14dB gain including LNA, SAW and antenna gain).[/quote]

So you have to account for the gain when doing the test.

If you give -90 dBm power level in the command, have -90 dBm signal generator, and get -75dBm back as response to the AT command, it means that there is 15dB of gain between the signal generator and the nRF9160. The difference of 15dB in command's power and response indicates that the receiver is possibly saturated and the SNR result is possibly not accurate.

Thus, in order to take the LNA gain into account they should use command

AT%XRFTEST=2,1,-75,0

(set <param1> = 0)

Regarding the signal level setting there is also a note in the AT Guide:

Note: It is recommended to set the signal level defined in the GPS SNR ON parameters to the
expected signal level at the GPS antenna port of the nRF9160 SiP.

Then the SNR should also get higher. Expected level is closer to 30dB than 20dB. The RSSI in response should be close to the value given in input.

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Mon, 08 Mar 2021 16:52:44 GMT

Hi Jonathan,

Did you get a chance to comment on tha data logger file:

Chamber_test_Product_boresight_90dbm_vertical.log : log file from chamber test

thanks,

Regards Bjørn

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Thu, 04 Mar 2021 12:55:44 GMT

So the results indicate\shows that the better the CN0 value the quicker TTFF(Time to firt fix).Better antenna leads to better CN0 values witch leads to better TTFF.

Device 2 seems to be much better, both in terms of TTFFs and CN0s. Is there a difference in the two devices ?

The values you got for the thingy is as good as it gets and everything that is in the 40s range is very good.

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Thu, 04 Mar 2021 10:10:00 GMT

Hi Jonathan,

Thank you for your answer.

Could you comment the data fro the GTEM cell and comment on the SNR and RSSI ?

Thanks,

Regards

Bjørn

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Wed, 03 Mar 2021 13:20:33 GMT

So the values that you gett from your test in Drammen looks to be completely normal\expected results, and the values fromt he parking lot test with the two custom HW then Device 2 seems to be within the limit of normal expected results. except for the first iteration. Device 1 has a bit higher time in general but i asume that the two devices are identical ? The ground plane of the Device might change when connecting the device to laptop with a usb cable, then the behaviour of the GPS antenna might also change, that is somthing we have seen on the Thingy91. Are you able to test on Device 1 and 2 without a usb cable?

~20 seconds for GPS from cold start is as quick as it gets.

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Tue, 02 Mar 2021 15:13:57 GMT

Hi,

We also added some tests from a location in Drammen.

The test is on a Thingy:91 with external antenna and LNA. The Thingy is placed on the rooftop of the car and 15 cold start measurements were recorded (a cold start is a power on cycle when the unit is completely disconnected from any power supplies).

The next test is with the Thingly located on the dashboard of the car. This time we only took 5 measurements all from a cold start.

The last test was with the product located on the dashboard. 15 tests were recorded.

thingy (roof) : Product (dashboard)
1    33        52
2    37        46
3    21        32
4    21        52
5    33        43
6    37        41
7    38        44
8    39        48
9    37        42
10 36        42
11 29        48
12 43        46
13 38        52
14 36        42
15 38        35

      thingy (dashboard)
1    33
2    37
3    30
4    31
5    36

The log files from these measurement where created by connecting the JLink to the Thingy/Product. No USB ocnnection was used.

Thanks,

Regards

Bjørn

devzone.nordicsemi.com/.../Thingy_5F00_27_5F00_2_5F00_Sundland.zip

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Tue, 02 Mar 2021 14:56:55 GMT

Hi,

The client ran a test last week.

They made 15 consecutive tests with the GPS on a parking lot in front of the office. The products are powered and produce logs via a USB connector connected to a portable PC (this is slightly different compared to the tests made in Drammen).

The tests were executed in the following manner with two devices:

- Start both devices (giving them power)

- start logging

- Wait until both devices have fix.

- Power off

- Reiterate test

This was iterated 15 times with two in theory identical devices with the same SW setup (nrf SDK v1.4 base, using GPS driver in continues mode, no AGPS, modem fw 1.2.3)

These are the time to fix (in seconds, considering uptime of device) values we obtained:

Iteration    Device 1:        Device 2:
1                207                117
2                181              52
3              110                49
4                94                39
5                92                  36
6                44                  53
7                145            51
8                43                  37
9                93                  47
10              51                53
11              124                44
12              58              51
13              70                  37
14              89                32

Observations:

    - Large difference in time to fix between devices even though they are running at the same time (Device 1 took longer to get fix in all iterations except iteration 8 and iteration 10).

devzone.nordicsemi.com/.../25022021.zip

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Tue, 02 Mar 2021 14:51:11 GMT

Hi,

Thingy:91 with 1.4.0 2020.24.

Data is extracted using SWD by JLink from Segger.

Regards,

Bjørn

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Tue, 02 Mar 2021 08:09:46 GMT

Can you provide what version of the Thingy91 you are using ?
And are you using a USB cable on the thingy when running the test?

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Thu, 25 Feb 2021 16:58:13 GMT

We have made some additional tests using both the Thingy:91 with external antenna and the product we have designed based on your NRF9160.

The Thingy has been tested using an external antenna. The Coax connector for external antenna has no longer an internal switch so we selected to remove the dc block capacitance downstream of the connector towards the internal GPS antenna. This to try to keep the 50ohm transmission line from the NRF to the external antenna. The external antenna has a LNA and the test software is configured to switch this power on (confirmed by measuring that the bias voltage is present on the external coax line).

We have three configurations (four if you add the GPS cold start hit the pedal drive test .. : ) ). The first test was done putting the Thingy:91 w/ext ant. on the roof of my car. The car was parked at Sunland in Drammen with no cars nearby and farily open space around the site. The second test was run at the exact same spot but now with the product as the GPS receiver. The third test was done on the dashboard for both the Thingy and the product. The last test was executed by giving Thingy (and also the product) a cold start and then immidiatly driving away in order to test or get a reference measurement for a moving receiver.

I still have a few more tests to run before I will share the results. I will upload these test when Ithe extended test has concluded.

In addition to these tests I have added a more dedicated test on SNR and RSSI. I put the receiver into a suitable GTEM cell and configured the lab. Signal source to give -90dBm field power at the antenna of the test device. The GTEM is EM sealed and very little exernal radiowaves can kreep into the test chamber. Could you please evaluate the data in this log file and comment on the RSSI and SNR?

All data in the attached log files have been extracted using the JLINK and the JLinkRTTViewer.exe from Segger.

There are two files attached:

gps_driver_share.zip : Code for test in the field
gps_rf_test_share.zip : Code to test the gps in the RF chamber
Chamber_test_Product_boresight_90dbm_vertical.log : log file from chamber test

These datafiles have been modified to remove client source code.

devzone.nordicsemi.com/.../gps_5F00_driver_5F00_share.zip

devzone.nordicsemi.com/.../gps_5F00_rf_5F00_test_5F00_share.zip

devzone.nordicsemi.com/.../Chamber_5F00_test_5F00_Product_5F00_boresight_5F00_90dbm_5F00_vertical.log

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Mon, 22 Feb 2021 09:27:38 GMT

After going trough the GPS log, R&D had this to say:

I do not find any malfunction of GPS in the trace. It picks up PRN#30, #5 and #13 within a minute. It then picks up PRN#18 but fails to get its ephemeris due to extremely low CN0. It then picks up PRN#7 but before it can make a fix it drops #13 (PRN#13 is high up, so this is peculiar; is this due to a head or body of the user obscuring the satellite ?). After yet some time it finds PRN#14 and is able to make a fix using PRN#30, #5, #7 and #14.

Yes it takes a long time to find the satellites and get the fix, but what does stand out in the trace are the extremely low CN0 value*s. The best value seen is 34 dB, but mostly the CN0s are in the 20-30 dB range. Given that some of the satellites have a quite high elevation, *I would say the device is missing 10 dB SNR.

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Fri, 19 Feb 2021 14:37:41 GMT

Hi,

can you share the code when you have access to it?

And where is the teste done, does the device have access to the GPS signals from the start of the test? If the devise does goes in to a deep-search prolonged periods will happen. This can happen if the device for example is in doors at first then taken outside where GPS signals are readily available.

[quote user=""]I tend to conclude that the long time to fix can be explained by software. [/quote]

This is most likely software issue.

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Thu, 18 Feb 2021 15:14:48 GMT

I dont se anything obvious errors that would lead to this type of behavior, i will discuss this with R&D and get back to you.
Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Wed, 17 Feb 2021 11:12:15 GMT

An upload file needs to have an extension name. In this case the Thingy_3 file had to be renamed to Thingy_3.txt .

devzone.nordicsemi.com/.../thingy_5F00_3.txt

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Wed, 17 Feb 2021 10:28:16 GMT

[quote user="bjornsh_nrf"]I have attached one of the log files here.[/quote]

There does not seem to be a file attached. Could you try to attach again ?

Regards,
Jonathan

RE: Nrf9160 GPS long time to fix with healthy RF electronics

bjornshnrf — Tue, 16 Feb 2021 19:37:22 GMT

Hi,

The nrf9160 was updated with the modem software version 1.2.3. The GPS program was built with nrf-sdk 1.4 with some patches but none for the gps.

Your "thingy91_nrf52_connectivity_bridge_2020-10-30_5522ca2b.hex" hex file was downloaded to the nrf52 to get a bridge between the nrf9160 and the nrf52 in order to extract log data directly to a pc with usb2com driver. I have attached one of the log files here.

The nrf9160 software is based on a nordic printout gps and measures the gps continuously. I can upload the code when I receive it if of interest.

We experience fix time much higher than 36seconds.

Thanks,

Bjørn

RE: Nrf9160 GPS long time to fix with healthy RF electronics

JONATHAN LL — Tue, 16 Feb 2021 14:54:45 GMT

Hi Bjørn,

What modem firmware version was used ?

[quote user=""]We did a similar test with the Thingy:91 with the external antenna and got similar results.[/quote]

What SW did you use? have you tested with our gps sample?

We have listed time to fix here, with cold start and no AGPS it should be around 36 s in good conditions.

Regards,
Jonathan