CC2652R: Custom hardware RF erratic performance

Part Number: CC2652R
Other Parts Discussed in Thread: CC2592,

Hello.

We develop a custom board with the cc2652R and the cc2592 PA, we want to use it as a coordinator in a Zigbee network. We make a prototype run and now we are testing it. For testing purposes we use the zc_zone example. In the TI development board we achieved the goal of comissioning some other devices (third party sensors) sucessfully and we can receive the sensor data. When we adapt and test the same software on the custom hardware the software works fine, no errors show when we debug it. The problem is that, some times it works fine (comissioning and receiving sensor data), and some times it doesen't work at all. We have a Zigbee sniffer and when it doesent work, seems like the device can send packages but when a sensor sends it data or is in the joining the network process our hardwere rf doesent receive nothing.

We discover that putting our finger over the cc2652 seem to make it work everytime, and it stays working some time after we retire the finger, but some time after it stops working.

We checked different boards and the problem is the same. Obviously is a hardware problem, but when it works, it works really good (even without the finger on top of the IC). We even tried some bluetooth software on it (simple_perihperal), and it works prefect all the time. At this point we are a little bit lost, do you have any sugestion of what can be causing it or what we can look for in the hardware to solve it?

Thanks.

  • Hi Juan,

    Since pressing on the device seems to resolve the issue, do you think it may be soldering related? Have you tried reflowing the devices?

    Otherwise, you can submit you design for review here: https://www.ti.com/tool/SIMPLELINK-2-4GHZ-DESIGN-REVIEWS 

    Thanks,

    Alexis

  • Please check the load capacitor of CC2652R,48Mhz crystal load capacitor approximately 7pF ,we encounter the same time many years ago.

  • Hello.

    Based on you answer about the 48Mhz crystal load capacitors, we tested, using the SmartRF studio, the internal cap array and playing witth the values, without result.

    But using the SmartRF studio and using one of our boards and the cc2652 lanuchpad for RX and TX testing we observed that using the continuous TX and continuos RX the RSSI was very low, but using the "magic finger" we get very good RSSI both ways (our board seding, or our board receiving, using different finger postitions). Further research show that when the finger was near the capactor of the LNA and PA improved the RSSI of the RX or the TX RSSI respectively.

    Using that info we tried one last thing. Manually setting the LNA pin of the cc2592 high we get perfect RX performance and setting the PA pin we get perfect TX performance.

    Seems that something is wrong with the output LNA and PA signals of the cc2652, but we tested it in a oscilloscope in previous tests and we didn't find anything suspicious.

    This is the code we used to set the LNA and PA outputs, as in the documentation:

        /*Initialization for board related stuff such as LEDs
        following TI-RTOS convention
       */
       Board_initGeneral();

       IOCPortConfigureSet(LNA_ENABLE, IOC_PORT_RFC_GPO0, IOC_IOMODE_NORMAL);
       IOCPortConfigureSet(PA_ENABLE, IOC_PORT_RFC_GPO1, IOC_IOMODE_NORMAL);

    Is anything wrong?, we used the same code with the bluetooth examples without any problem.

    Thanks.

  • Hello.

    I just realized that i was using the SmartRF studio, without setting the output pins for enabling PA and LNA, so it was logical that the cc2592 wasent working properly. And manually setting the LA and the LNA pins high worked.

    But that opens another question, why using SmartRF studio the transmision and reception of our board works perfectly fine, and using the zc_zone example it has a erratic behavoir.

    The tests we did with the SmartRF studio were:

    1- Using the development board (cc2652R lauchpad) in TX mode, and our board as RX, strong signal and 0% error rate at a certain distance  (Setting the PA pin and LNA pin in SmartRF studio )

    2- Using the development board (cc2652R lauchpad) in RX mode, and our board as TX, strong signal (better than the test 1) and 0% error rate at a certain distance.  (Setting the PA pin and LNA pin in SmartRF studio ). The RSSI was better in this case, showing that the cc2592 was working correctly and validating our custom hardware (right?).

    Other question is: There is anything that the SmartRF studio is not using form the real hardware (RTC, 48Mhz clock, ...) that is making it work, but when it uses the real hardware it doesen?

    Thanks.

  • The truely is emulator influence crystal and load capacitor,you shall disconnect emulator whenever test your hardware.

  • Hi Juan,

    Have you tried exporting smartrf_settings.c/h generated from SmartRF studio into your project?

    Thanks,

    Alexis

  • Hello. 

    After some more research we found that the problem must be some type of noise in the JTAG pins. we use a 2-pin JTAG programming, and we have 5 pins 0.1 inch header with RST-VCC-GND-TCK-TMS pins. When we put our finger over the JTAG pins the device works flawlesly. We removed the header and the device improved, but some times it fails receiving packages. 

    The PCB tracks for the TMS and TCK lines are short, like 2 cm max. Do you know why this happens?, and if i can put some kind of cap for noise inmunity in the JTAG pins?

    Thanks.

  • Hi Juan,

    Its difficult to provide recommendations without seeing your design. If you want feedback on your schematic/layout, you can submit a request here:  https://www.ti.com/tool/SIMPLELINK-2-4GHZ-DESIGN-REVIEWS  

    Thanks,

    Alexis