CC1352P: RF noise floor is inconsistent from board to board

Hi,

There will be a natural variance of the RSSI level up to 10 dB within a square meter, when receiving via an antenna due to multi-path propagation effects.

When the antenna is removed would expect that the variation should be much lower when the RSSI tests are made conducted compared to radiated.

Are you connecting a signal generator to the LaunchPad to make the RSSI tests ?

Regards,

   Richard

Since a low datarate works as suspected and a high datarate not I suspect that you get some extra noise in with higher RX BW. 

- Try with a 50 ohm termination of the antenna input to minimize the external noise. Do you see the same delta then?

- Try to turn off the DCDC (it's a radio button in SmartRF Studio. Do you see a difference if you do this? I have seen that the DCDC, if the layout is not done according to best practice, could increase the noise some. 

But without antenna I was expecting noise floor to be similar in all boards. I have tested those boards with RF signal generator. I sent -70dBm signal and all boards showed RSSI level correctly, around -70dBm. Without antenna, when I monitor noise floor, RSSI is different from board to board in range from -98dBm to -82dBm.  A snapshot from my statistics:

PCB 45 RF1 (CC50) -98 RF2 (CC52P) -93

PCB 46 RF1 (CC50) -97 RF2 (CC52P) -93

PCB 47 RF1 (CC50) -98 RF2 (CC52P) -91

PCB 48 RF1 (CC50) -98 RF2 (CC52P) -87

PCB 49 RF1 (CC50) -98 RF2 (CC52P) -85

PCB 50 RF1 (CC50) -96 RF2 (CC52P) -87

PCB 51 RF1 (CC50) -98 RF2 (CC52P) -83

PCB 52 RF1 (CC50) -98 RF2 (CC52P) -87

PCB 53 RF1 (CC50) -96 RF2 (CC52P) -91

PCB 54 RF1 (CC50) -98 RF2 (CC52P) -85

PCB 55 RF1 (CC50) -97 RF2 (CC52P) -89

PCB 56 RF1 (CC50) -97 RF2 (CC52P) -97

PCB 57 RF1 (CC50) -98 RF2 (CC52P) -89

PCB 58 RF1 (CC50) -98 RF2 (CC52P) -82

PCB 59 RF1 (CC50) -98 RF2 (CC52P) -86

PCB 60 RF1 (CC50) -97 RF2 (CC52P) -89

As said, it could be noise generated by the board or something else. And without antenna, does that mean that you have just removed the antenna but not terminated the antenna connector in a known impedance? If that is the case it could be that the impedance change causes the differences. 

I understand the importance of proper termination. I terminate input with 50ohm terminator but noticed that termination does not change the RSSI reading even a fraction. I tried with termination, without termination. In all cases result is identical. CC1352P circuitry is shielded with the grounded metal can from the rest of the board.  DC/DC is routed with the smallest possible current loop. Next to the corresponding CC1352P pins. 

I want to point out the fact that the same boards have no noise floor scattering when CC1352P is controlled from TI Smart RF Studio via XDS110 and during this measurement the rest of our custom board is active and running. So it should not be DC/DC routing issue or termination issue. 

Maybe there are parameters we need to play around with outside those that I can export and get in smartrf_settings.c  and smartrf_settings.h files?   

Tried to manipulate DC/DC option in RF Studio, but that did not make any difference. But I was not expecting to do, because our board controlled from RF studio has no issues, good noise floor figure.

Additional information:

Toke one PCB with pour performance, noise floor -83…84dBm. Replaced CC1352P chip on the board to new one from the IC reel and now the same board measures -89…90dBm noise floor.

I do not have the proof, but it feels like there is some parameters that heavily corelate to CC1352P silica die tolerances. And RF smart studio uses those coefficients in a different manner.  But those that are available in smartrf_settings.c  and smartrf_settings.h I triple checked to be the same. Maybe something different in #include DeviceFamily_constructPath(rf_patches/rf_patch_cpe_prop.h) or other resources that are part of SDK?  

I use “Smart RF Studio 7  2.24.0”

SDK for FW code development “simplelink_cc13x2_26x2_sdk_5_10_00_48”

NOTE: The same FW code and PCB board but just with CC1350 RF frontend does not have this issue. All boards measures aprox -98dBm noise floor. By saying "the same" of course I mean that everything is the same except the RF chip initialization and complementary code. But all RSSI averaging algorithms and strategy are the same. PCB has the same DC/DC converters, the same functional blocks, the same microcontrollers and board layout is identical. 

Regards

Aivaras

The settings in SmartRF Studio and sysconfig are the same. Meaning that if you select, as an example, 50 kbps, in both tools the radio has the same setup. You can also see what is used by the radio if you set a breakpoint at the radio setup in the code.

- The CC1352P and CC1350 are not pin compatible meaning that you can't use the same PCB for both. Also the fact that changing the chip changes the performance is a bit strange. Would you be able to post a screenshot of the RF part of the schematic/ layout to get a better idea? 

- Could you share how you are measuring RSSI in your FW?

Good morning,
Yes, they are not pin to pin compatible. I meant that those two PCB’s are identical except the part where RF transceiver is fitted. RF shielded section, containing RF transceiver is different but elsewhere the board is identical.

RSSI code 

RF_cmdRxTest.endTrigger.triggerType = TRIG_REL_START;
RF_cmdRxTest.endTime = RF_convertMilisecondsToRatTicks(500);
RF_CmdHandle cmd_handle = RF_postCmd(
    rfHandle,
    (RF_Op*)&RF_cmdRxTest,
    RF_PriorityNormal,
    &RADIO_rxCallbackScan,
    RF_EventRxEntryDone);
int8_t rssiNow = 0;
while((rssiNow == RF_GET_RSSI_ERROR_VAL) || (rssiNow == 0)){
    rssiNow = RF_getRssi(rfHandle);
    }
    RF_cancelCmd(rfHandle, cmd_handle, 0);
    RF_pendCmd(rfHandle, cmd_handle, RF_EventRxEntryDone)

Additional Information:

Our programmer narrowed down origin of this issue. Now all boards measure the same RSSI which is about -102dBm.  And I can see RF CW to those levels, when fed from RF generator. The only difference in code is small sleep routine added before actual measurement.

    RF_cmdRxTest.endTrigger.triggerType = TRIG_REL_START;
    RF_cmdRxTest.endTime = RF_convertMilisecondsToRatTicks(500);
    RF_CmdHandle cmd_handle = RF_postCmd(
            rfHandle,
            (RF_Op*)&RF_cmdRxTest,
            RF_PriorityNormal,
            &RADIO_rxCallbackScan,
            RF_EventRxEntryDone);
 
    int8_t rssiNow = 0;
    while((rssiNow == RF_GET_RSSI_ERROR_VAL) || (rssiNow == 0)){
        Task_sleep(1);
        rssiNow = RF_getRssi(rfHandle);
    }
 
    RF_cancelCmd(rfHandle, cmd_handle, 0);
    RF_pendCmd(rfHandle, cmd_handle, RF_EventRxEntryDone);

So it seems like RF_getRssi(rfHandle) returns improper RSSI value which depends from CC1352P  silica die. If we give some idle time with  Task_sleep(1) we get proper RSSI value.

Now I need to measure actual sensitivity of the system when working with real data packets, real communication. I am no longer confident that we get full sensitivity during communication and it is not tampered my bad noise floor. Unless someone from community sees this situation differently and can explain.

In general this community has more experience has seen more different scenarios then I, maybe someone can share opinion on this matter? Do I need to worry that during full speed 1Mbps communication I have that inconsistent noise floor as a bottle neck for sensitivity I can achieve? 

At first I suspected a HW issue but now I think you have a SW issue. I haven't looked at the timing here but I suspect that you in some cases are reading out RSSI when not in RX. It's possible to set a DIO high when the radio is in RX, I would do that and at the same time set a DIO just before/ after using getRSSI.

For packets it's easier, the RSSI of the last received packet can be appended to the receive buffer. Have you looked through https://dev.ti.com/tirex/content/simplelink_cc13xx_cc26xx_sdk_6_40_00_13/docs/proprietary-rf/proprietary-rf-users-guide/proprietary-rf-guide/index-cc13xx_cc26xx.html ?