EFR32xG22-Power-Consumption-Optimization

Introduction

One of the new features highlighted in EFR32xG22 is the Low System Energy Consumption which can reach 1.40 µA EM2 DeepSleep current with 32 kB RAM retention and RTC running from LFRCO. This document discusses how to measure the minimum current consumption in EFR32xG22 EM2, as well how to reduce current consumption.

Discussion

According to the EFR32MG22 data sheet, the typical test conditions should be: VREGVDD = 3.0 V. AVDD = DVDD = IOVDD = RFVDD = PAVDD = 1.8 V from DCDC Voltage scaling level VSCALE0 in EM2 with TA = 25 °C.

When creating an "SoC - Empty" project, the initial sleep current can be measured at around 2 µA in EM2 which is not 1.4 µA mentioned in the data sheet. This is mainly because certain peripherals, such as VCOM and Debug mode are enabled in SoC Empty project for development convenience. Users can disable these functions to reduce consumption.

The figure below shows the result tested from Energy Profiler in Simplicity Studio with two current measuring results. The left one represents the total average current which includes the significant current increase during reset. The right one is calculated from the selected range from the user read values.

In accordance with "AEM Accuracy and Performance" section from UG172, the AEM is capable of measuring currents in the range of 0.1 µA to 95 mA. For currents above 250 µA, the AEM is accurate within 0.1 mA. When measuring currents below 250 µA, the accuracy increases to 1 µA. Even though the absolute accuracy is 1 µA in the sub 250 µA range, Energy Profiler is still not accurate enough to measure low-power consumption especially in Deep Sleep mode.

As seen in the figure above, the radio board voltage is at around 3.3 V which does not correspond to 3.0 V mentioned in the data sheet because, when using AEM mode, a low noise 3.3 V LDO on the main board is used to power the radio board.

To get more accurate results, the following discussion and test will strictly obey the testing conditions in the data sheet as well as using a high-accuracy DC analyzer instead of Energy Profiler.




The DC Power Analyzer used in this article is N6705B from Agilent, whose ammeter accuracy is up to 0.025% + 8 nA. It also provides Data logger function with Measurement Interval from 20 µs to 60 s so that the average value of the current consumption can be easily calculated.


The following section discusses how different testing conditions and peripherals affect current consumption.

Supply Voltage

First, input voltage will be compared when supply voltage is 3.0 V and 3.3 V.

The figure below shows testing results of "SoC Empty Project" with different supply voltages. The upper line is the current consumption with 3.0 V supply voltage while the lower line is with 3.3 V. You can find from the table below that with 3.0 V supply voltage will consume higher current consumption than with 3.3 V supply voltage. This is because the device will maintain constant power in EM2. According to the formula P = U x I, voltage is inversely proportional to current under constant power.

Debugger

Debug connectivity can be enabled by setting the EM2DBGEN field on the EMU_CTRL register and will consume about 0.5 µA extra supply current. To reduce current consumption, comment out the line below.

//Force PD0B to stay on EM2 entry. This allow debugger to remain connected in EM2
//EMU->CTRL |= EMU_CTRL_EM2DBGEN;

DCDC

A DC-DC buck converter is a type of switching regulator that efficiently converts a high-input voltage to a lower-output voltage, it covers a wide range of load currents, and provides high efficiency in energy modes EM0, EM1, EM2 and EM3. For more information about DCDC, see AN0948.

  // Enable DC-DC converter
  EMU_DCDCInit_TypeDef dcdcInit = EMU_DCDCINIT_DEFAULT;
  EMU_DCDCInit(&dcdcInit);

DCDC is enabled by default in the SoC Empty project. The figure below shows the current curve comparison with and without DCDC usage after disabling debug mode in the SoC empty project. You can see from the average current that using DCDC can result in current savings.

External Flash

The external flash “MX25R8035F” equipped in the BRD4182A radio board is standby by default. Typical current draw in standby mode for the MX25R8035F device used on EFR32 radio boards is 5 µA, which makes observing the difference between VS2 and VS0 voltage scaling levels difficult.

Fortunately, JEDEC standard SPI flash memories have a lower current deep power-down mode, in which the typical current draw can be up to 0.35 µA is typically 0.007 µA. With the command below will put the MX25 into deep power down mode.

  /* Disable external flash memory*/
  MX25_init();
  MX25_DP();

MX25_init initializes the SPI Flash and calling MX25_DP sends the byte necessary to put the Flash into DP mode.

Voltage Scaling

Voltage scaling helps to optimize the energy efficiency of the system by operating at lower voltages when possible. Three supply voltage operating points are available, as shown below:

VSCALE SettingDECOUPLE VoltageOperating Conditions
VSCALE21.1 VEM0/EM1 Operation up to 80 MHz, EM2 and EM3
VSCALE11.0 VEM0/EM1 Operation up to 40 MHz, EM2 and EM3
VSCALE00.9 VEM2 and EM3 Only

The voltage sale level for EM2 and EM3 is set using the EMU_CTRL_EMU23VSCALE field. The lowest sleep current will be obtained by setting EMU23VSCALE to VSCALE0.

  EMU_EM23Init_TypeDef em23_init = EMU_EM23INIT_DEFAULT;
  em23_init.vScaleEM23Voltage = emuVScaleEM23_LowPower;
  EMU_EM23Init(&em23_init);

emuVScaleEM23_LowPower mode (vscale0) and emuVScaleEM23_FastWakeup (vscale2) are two voltage scaling modes in EM2 and EM3. Current reduction between different scaling modes will be shown in the subsequent sections.

Radio RAM Retention

The EFR32xG22 device contains several blocks of SRAM for various purposes including general data memory (RAM) and various RF subsystem RAMS (SEQRAM, FRCRAM). Frame Rate Controller SRAM(FRCRAM) and all part of Sequencer SRAM(SEQRAM) may be powered down in EM2/EM3 if not required. To control retention of these areas, set FRCRAMRETNCTRL or SEQRAMRETNCTRL in SYSCFG_RADIORAMRETNCTRL to the desired value.

  /* Disable Radio RAM memories (FRC and SEQ) */
  CMU_ClockEnable(cmuClock_SYSCFG, true);
  SYSCFG->RADIORAMRETNCTRL = 0x103UL;

Note : The command above can only be implemented in a MCU project. The wireless stacks won't work as expected if the FRCRAM and SEQRAM are disabled.

Disable different Radio RAM will result in different reductions. The following table lists the current draw measured in the MCU project with different RADIO RAM retention as well as 32 KB RAM.

FRC RETENTIONSEQ RETENTIONBOTH RADIO RAM RETENTIONNO RADIO RAM RETENTION
VSCALE22.33 µA2.66 µA2.75 µA2.22 µA
VSCALE11.68 µA1.89 µA1.97 µA1.60 µA
VSCALE01.24 µA1.38 µA1.43 µA1.19 µA

GPIO

All unconnected pins on the EFR32 should be configured to Disabled mode (high impedance, no pull resistor) where the reset state of the I/O pins is disabled as well, which is done by setting the GPIO gpioModeDisabled. See the GPIO setting in MX25_deinit(), which is used to disable SPI communication.

MX25_deinit();

Also, if you are reproducing the EM2 current consumption test using some of the example come with our SDK (either MCU or Wireless), also check the status of the VCOM. Enabling VCOM will increase the current consumption. De-assert the VCOM Enable pin and the TX and RX pins when not needed.

//initVcomEnable();

Power Domain

The EFR32xG22 implements several independent power domains which are powered down to minimize supply current when not in use. Power domains are managed automatically by the EMU. It includes lowest-energy power domain (PDHV), low power domain (PD0), low power domain A (PD0A) and auxiliary PD0 power domains (PD0B, PD0C, and so on). When entering EM2 or EM3, if any peripheral on an auxiliary low power domain (PD0B, PD0C, etc.) is enabled, that auxiliary low power domain will be powered causing higher current draw. Otherwise, the auxiliary power domain will be powered down.

The entire PD0B will be kept on in EM2/EM3 if any module in PD0B is enabled on EM2/EM3 entry. Therefore, ensure that the High Power peripherals are disabled when entering EM2.

Heating Impact

Note that the temperature has a major impact on the consumption. The recommended ambient temperature for this test is 25°C, as documented in the data sheet.

Note that you don't have to follow the condition to reserve full RAM and use LFRCO. You can either disable RAM retention or use a different oscillator for even lower power consumption.

SRAM Retention

RAM is divided into two 24 KB and 8 KB banks, beginning at address 0x20000000 and 0x20006000 respectively. By default, both banks are retained in EM2/EM3. Sleep mode current can be significantly reduced by powering down a bank that does not need to be retained. RAMRETENCTRL in the SYSCFG_DMEM_RETNCTRL register controls which banks are retained in EM2/EM3.

/* Disable MCU RAM retention */
// EMU_RamPowerDown(SRAM_BASE, SRAM_BASE + SRAM_SIZE);
/* Power down BLK0 0x20000000 - 0x20006000: 0x01; BLK1 0x20006000 - 0x20008000: */
CMU_ClockEnable(cmuClock_SYSCFG, true);
SYSCFG->DMEM0RETNCTRL = 0x01UL;

Disabling different RAM will result in different reductions. The following table lists the current draw measured in the MCU project with different RAM retention and no RADIO RAM retention.

32 KB RAM24 KB RAM8 KB RAM
VSCALE22.22 µA2.20 µA1.66 µA
VSCALE11.60 µA1.59 µA1.24 µA
VSCALE01.19 µA1.19 µA0.98 µA

Note: No RAM retention does not make sense (achievable but wake up fails).

Low Frequency Oscillator Setting

The LFRCO is an integrated low-frequency 32.768 kHz RC oscillator for low-power operation without an external crystal. It provides precision mode on certain part numbers which enable hardware that periodically recalibrates the LFRCO against the 38.4 MHz HFXO crystal when temperature changes to provide a fully internal 32.768 kHz clock source with +/- 500 ppm accuracy. With temperature variations, PLFRCO(LFRCO in precision mode) will autonomously run frequent calibrations which results in consumption increase.

The Low-Frequency Crystal Oscillator (LFXO) uses an external 32.768 kHz crystal to provide an accurate low-frequency clock. Using LFXO instead of PLFRCO will reduce the current consumption.

CMU_LFXOInit_TypeDef lfxoInit = CMU_LFXOINIT_DEFAULT;
CMU_LFXOInit(&lfxoInit);
CMU_OscillatorEnable(cmuOsc_LFRCO, false, false);
CMU_OscillatorEnable(cmuOsc_LFXO, true, true);
CMU_ClockSelectSet(cmuClock_LFXO, cmuSelect_LFXO);

According to the EFR32xG22 data sheet, MCU current consumption using DC-DC at 3.0 V input in EM2 mode, VSCALE0 is shown below:

Note : Entering EM2 mode immediately after reset may brick the device and the debugger may no longer be attached. To fix this, set the WSTK switch next to the battery holder to USB (powers down the EFR). Execute Simplicity Commander with command line parameters "./commander.exe device recover" and then immediately move the switch to the AEM position.

Reference

Setting up

The example project adopted most of the strategies mentioned above to reduce energy consumption. Because the low-power methods implemented in the MCU project and wireless project are quite different, the experiment will be run separately in these two domains.

Hardware Environment 1 WSTK Main Development Board 1 EFR32xG22 2.4GHz 6 dBm Radio Board (BRD4182A Rev B04)

Software Environment Simplicity Studio SV4.x Gecko SDK v2.7.x

Note : This document focuses on the strategies to reduce current consumption. The low power strategies can also be implemented in Simplicity Studio SV5.x with Gecko SDK 3.x.

BLE Project Example

Example Experiment

1.Create a new SoC - Empty application project with Bluetooth SDK using version 2.13.6 or newer.

2.Open app.c and comment out the code in system_boot to ban advertising to measure the sleep current in EM2.

      case gecko_evt_system_boot_id:

//        bootMessage(&(evt->data.evt_system_boot));
//        printLog("boot event - starting advertising\r\n");
//
//        /* Set advertising parameters. 100ms advertisement interval.
//         * The first parameter is advertising set handle
//         * The next two parameters are minimum and maximum advertising interval, both in
//         * units of (milliseconds * 1.6).
//         * The last two parameters are duration and maxevents left as default. */
//        gecko_cmd_le_gap_set_advertise_timing(0, 160, 160, 0, 0);
//
//        /* Start general advertising and enable connections. */
//        gecko_cmd_le_gap_start_advertising(0, le_gap_general_discoverable, le_gap_connectable_scannable);
        break;

3.Comment out EMU_CTRL_EM2DBGEN in init_mcu.c to disable debug in EM2.

// EMU->CTRL |= EMU_CTRL_EM2DBGEN;

4.Comment out VCOM in main.c to disable VCOM.

// initVcomEnable();

5.Build the project and download to your radio board xG22.

Experiment Results

The experiment results show the sleep current consumption in two minutes. You can see from the overall statistics in the table at the bottom and the average current consumption is about 1.65 µA.

Because the testing is done in a wireless BLE project (SoC empty project), Radio RAM (both FRC and SEQ) should be retained even in EM2, which consumes about 0.25 µA extra supply current. Therefore, the testing result will be higher than 1.4 µA. If the wireless radio functions are not required, xG22 can reach consumption lower than 1.4 µA in the MCU project.

MCU Project Example

Example Experiment

  1. Import a MCU project from the GitHub example. Choose "File -> import" and browse to import the project below. C:\SiliconLabs\SimplicityStudio\v4\developer\sdks\gecko_sdk_suite\v2.7\peripheral_examples\series2\emu\em23_voltage_scaling\SimplicityStudio

  2. Replace main.c with the file attached in this article.

  3. Build the project and download it to your radio board xG22.

Experiment results

You can see from the testing results in the MCU project that the current consumption can reach lower than 1.4 µA.

Usage

Enabling or disabling different peripherals has different impacts on the current consumption. To reduce current draw, use the adjusted voltage to optimize the energy efficiency of the system. Additionally, adopt different strategies depending on your requirements to reach the minimum consumption.

To reproduce and check the test results, see the above example section.

Note: Although Energy Profiler is not accurate enough for low-power measurements, it is able to detect changes in the current consumption as small as 100 nA. It is always recommended to use higher accuracy equipment if applicable.

Source

main.c