Suggestions on circuit design for LTE-M

[diminatto]

Hello,

I´m design a board with Nordics NRF9160 to collect some data from a few sensors, i have already made a few boards for testing, but i´m having a problem with the LTE radio on it, it´s having a hard time obtaining and mantaining a connection to the tower, it takes a long time for making one and when it does the signal is so bad i can´t talk with the tower, or the connection drops after a few seconds. It also  has GPS wich can get a fix and it´s working. 

I´m using a flat antenna just like ANT-LTE-RPC-UFL from Linx, the antenna is connected via a IPX connector.

My board is quite small 24 mm (0.9') x 45 mm (1,77´), so i´m thinking maybe my ground plane is too small, i don´t have much experience with RF.

Here is my board:



Any suggestions on what could i do to try to solve my problem with the signal ?

[Smile2016]

try to directly connect the antenna, no tuning/matching elements from NRF9160 to IPX. make sure the PCB stackup is for 50 Ohms. stick the antenna to a plastic enclosure as per datasheet. compare the return loss to datasheet plot if you have a VNA. check power supply and decoupling to support peak currents during Tx

[diminatto]

It´s already connected without any tuning just a 0 ohm resistor, i think i may have fogotten to select controlled impedance during my order in the fab. I gonna buy a NanoVNA to check it up, currently i dont have one. Power supply is good.

Thanks!

[Smile2016]

It´s already connected without any tuning just a 0 ohm resistor, i think i may have fogotten to select controlled impedance during my order in the fab. I gonna buy a NanoVNA to check it up, currently i dont have one. Power supply is good.

Thanks!

How many layers? 4 layer or 2 layer PCB? I assume you have a GND reference plane in layer2. what's the trace width and Layer1 to Layer2 distance ?

[Kean]

Some vendors have a free service where application engineers review your PCB layout for their RF modules.
I would guess that it is better for them to review designs for free early on, than deal with it later in the design stage and have disappointed (ex-)customers.

I'm not sure if Nordic offer it, but u-blox have done it for me with one of their nRF52 based modules.

[diminatto]

It´s already connected without any tuning just a 0 ohm resistor, i think i may have fogotten to select controlled impedance during my order in the fab. I gonna buy a NanoVNA to check it up, currently i dont have one. Power supply is good.

Thanks!

How many layers? 4 layer or 2 layer PCB? I assume you have a GND reference plane in layer2. what's the trace width and Layer1 to Layer2 distance ?

It´s a 4 layer, trace width is 3mm and distance is 0.2104mm of prepeg.

[diminatto]

Some vendors have a free service where application engineers review your PCB layout for their RF modules.
I would guess that it is better for them to review designs for free early on, than deal with it later in the design stage and have disappointed (ex-)customers.

I'm not sure if Nordic offer it, but u-blox have done it for me with one of their nRF52 based modules.

I asked for a review before production, they said that it sould have worked, probably the mistake was while ordering.

[Smile2016]

assuming is 0.3mm you meant for trace w.... in that case with a prepeg Er~4.4 you should be not far from 50 Ohms... is most probably not the layout impedance . trace is short enough too... suspecting the power supply or antenna itself?
[/quote]


It´s a 4 layer, trace width is 3mm and distance is 0.2104mm of prepeg.
[/quote]

[diminatto]

assuming is 0.3mm you meant for trace w.... in that case with a prepeg Er~4.4 you should be not far from 50 Ohms... is most probably not the layout impedance . trace is short enough too... suspecting the power supply or antenna itself?

For the power supply i´m using a bench unit, i tested this antenna on a LTE module from Quectel and it connects just fine, maybe the antenna has problems with the band i´m using, i´m using B3 (1.800 Mhz nominal) .

The seller says the antenna ranges from 600 Mhz to 6Ghz, i´m kinda lost here lol

[Smile2016]

I'm talking width as in image ...3mm will give you less than 15 Ohms way off

[diminatto]

I'm talking width as in image ...3mm will give you less than 15 Ohms way off

Ohhh, sorry i typed it wrong and did a broken quote from your answer, it´s 0,3 mm of width and not 3mm.

When i was creating the board i used some online calculators to get this value, for confirming i gave a look at Nrf9160 Feather that has a open design and it also has 0,3mm traces for RF signals.

[Smile2016]

Some vendors have a free service where application engineers review your PCB layout for their RF modules.
I would guess that it is better for them to review designs for free early on, than deal with it later in the design stage and have disappointed (ex-)customers.

I'm not sure if Nordic offer it, but u-blox have done it for me with one of their nRF52 based modules.

seems PCB is ok, unless CM did not respect the stackup. what does AT commands related to NR signal quality return? are you using same network and SIM with Quectel setup? a Nordic NRF9160 EVK might help to narrow down...

[diminatto]

Some vendors have a free service where application engineers review your PCB layout for their RF modules.
I would guess that it is better for them to review designs for free early on, than deal with it later in the design stage and have disappointed (ex-)customers.

I'm not sure if Nordic offer it, but u-blox have done it for me with one of their nRF52 based modules.

seems PCB is ok, unless CM did not respect the stackup. what does AT commands related to NR signal quality return? are you using same network and SIM with Quectel setup? a Nordic NRF9160 EVK might help to narrow down...

It´s the same network i´m using with the Quectel modem but not the same SIM, the one i´m using with the Nrf is specific for NBIot and CAT-M

I did a %CONEVAL command to equests an evaluation of the LTE radio signal state and an estimation of the energy efficiency, this is what it returned:

Connection pre-evaluation successful
rcc_state = RRC connection in RRC CONNECTED state during measurements
energy_estimate = Bad conditions. Difficulties in setting up connections. Maximum number of repetitions might be needed for data.
rsrp = -104 db
rsrq = -16 db
snr = -4db

some cell ID data..

band = 3 (i locked because of local regulations)
ce_level = CE Level 1. Medium number of repetitions.
tx_power = +23 dbm
tx_repetitions = 16
rx_repetitions = 8
pathloss = 124 dB

[diminatto]

Seeing the SNR now after you suggested the power supply could be the cause i´m starting to think it could be bad, gonna try to connect directly to a battery and see if something changes.

[diminatto]

Same problem

[Smile2016]

Same problem

your hardware seems ok to me. is there other band/s available in your area  besides B3? might the basestation not be well reachable in B3? antenna might slightly help if tuned for B3. is queqtel MODEM limited to the same band just curious ?

[diminatto]

your hardware seems ok to me. is there other band/s available in your area  besides B3? might the basestation not be well reachable in B3? antenna might slightly help if tuned for B3. is queqtel MODEM limited to the same band just curious ?

Yeah, right now for troubleshooting i removed the band lock, it´s pre certified in B3 and B23. I tryed in multiple locations so i doubt it could be any problem related to not beign able to reach the basestation. It wasn´t, i locked band on the Quectel modem and it worked.

I´m trying to think about what to try next, if i put the sim card on my phone it detects a network with strong RSSI.

[Kean]

Have you tried building and testing a second sample?
There is the possibility of an assembly issue, and you might want to rule that out.
You could also try extending the ground plane by scraping away some of the solder mask around the edges and soldering on some thin copper sheet/tape.

[RadioNerd]

Based on your description I would first check all components in the RF signal path.
Is everything electrically connected? all solder joints OK? All components oriented the right way? I would especially check the SMD coax receptacles. With these IPEX (or whatever standard it is) receptacles it can somtimes happen to mount them rotated by 180 degrees by mistake which will short the RF line...

[diminatto]

Have you tried building and testing a second sample?
There is the possibility of an assembly issue, and you might want to rule that out.
You could also try extending the ground plane by scraping away some of the solder mask around the edges and soldering on some thin copper sheet/tape.

I made three samples, they all have the same problem, gonna try adding some more ground.

[diminatto]

Based on your description I would first check all components in the RF signal path.
Is everything electrically connected? all solder joints OK? All components oriented the right way? I would especially check the SMD coax receptacles. With these IPEX (or whatever standard it is) receptacles it can somtimes happen to mount them rotated by 180 degrees by mistake which will short the RF line...

There´s only one 0 ohm resistor in the path, already checked with a multimeter and everything is connected from the modem to the coax on the antenna. I soldered them incorrectly the first time lol, but they have a little indentation on them pointing the grounded side and now they are ok.

I have three samples with the same problem, i already reflowed them to discard bad soldering.

[diminatto]

Have you tried building and testing a second sample?
There is the possibility of an assembly issue, and you might want to rule that out.
You could also try extending the ground plane by scraping away some of the solder mask around the edges and soldering on some thin copper sheet/tape.

I made three samples, they all have the same problem, gonna try adding some more ground.

Tryed adding some more ground, still have the same problem. Didn´t have copper sheet or tape so i welded to a PCB.

[diminatto]

A little bit more information i collected, if i try to connect to the network it answers me with reject_cause 15 "No Suitable Cells in Tracking Area". If i do a AT+COPS=? to list all the operators, it takes some time to give a response, but the command brings two, one of wich is the operator from the SIM card.