CC2642R: firmware behaves differently when device is powered reset

Hi Jameel,

Can you please share your SDK version and how you are initializing the GP Timers?  Do you have any logic analyzer or oscilloscope to validate your findings?  Are you using a TI LaunchPad or custom PCB?  What amount of precision are you expecting and how much error is introduced?  Have you tried using the Timer TI Driver instead?  Please note that the active JTAG connection could allow for a more accurate clocking than when it is disconnected, and the GPTimer clock is dependent on the MCU system clock. If very high-accuracy outputs are needed, the application should request using the external HF crystal.  Further resources are provided below:

https://dev.ti.com/tirex/explore/node?node=AD.iYbxHpXBYYxLP-HYhuQ__BSEc4rl__LATEST 
https://dev.ti.com/tirex/explore/node?node=AHzXNTeU-RUKjCfno2Lumg__BSEc4rl__LATEST 

Regards,
Ryan

Hi Ryan,

thank you for the quick reply, I am using simplelink_cc13x2_26x2_sdk_5_20_00_52, I am initializing the timers this way 

GPTimerCC26XX_Params_init(&posPhaseTimerParams);
posPhaseTimerParams.width = GPT_CONFIG_16BIT;
posPhaseTimerParams.mode = GPT_MODE_ONESHOT_UP;
posPhaseTimerParams.debugStallMode = GPTimerCC26XX_DEBUG_STALL_OFF;
posPhaseTimerHandle = GPTimerCC26XX_open(CC26X2R1F_GPTIMER0A, &posPhaseTimerParams);

I have 7 of these, running up to CC26X2R1F_GPTIMER3A

and I set the load value like this

BIOS_getCpuFreq(&freq);
GPTimerCC26XX_Value LoadVal_per1microsec = freq.lo / 1000000; // clock cycles per 1 micro second 
GP_Val = (LoadVal_per1microsec) * (80) - 1; 
GPTimerCC26XX_setLoadValue(posPhaseTimerHandle, GP_Val);

using custom PCB,  trying to get 80 ,microseconds minimum with +- 20% max, while connected with the Jtag, i get 86 microseconds, without it, it just jumps between 100 us and 140 (should note that if I remove the jtag after flashing the firmware it still works fine, the problem is introduced when the device is turned on after shutdown.

the application does request using the external HF crystal by using  Power_setDependency(PowerCC26XX_XOSC_HF) at the start of the task function.

I am using oscilloscope  

Thank you for the additional information.  The code appears to be correct and I am further interested in your custom PCB.  How is the device powered on after shutdown (battery, external power supply, etc.)?  Does the same behavior occur after a software reset (toggling the RST line)?  Have you monitored the HF crystal startup stability?  Also, can you re-create this behavior on a LAUNCHXL-CC26X2R1?

Regards,
Ryan

Hi Ryan,

The device is powered by a battery, 3.7 Volts 

I will have to check toggling the reset line and see what behavior I get, as well as the startup stability of the HF crystal 

I don't have a LAUNCHXL-CC26X2R1 at the moment to test it. 

Try testing with a dedicated power supply as well, or using the debugger voltage source if not the battery.  It is also worth reviewing SWRA495.

Regards,
Ryan

After reset still the same behavior

I want to understand why after removing the debugger first time the device works fine, does this mean the accuracy is not affected by the jtag lines ? 

power supply shows the same issue, tested it on multiple boards as well same behavior.

what is the main difference between loading the firmware using jtag and turning the device off and on ?

Thank you Ryan, disabling power saving solved the issue

Best regards,

Jameel