AN-1203 Ultra-Low Power Wi-Fi Connected IoT Node

Transcription

1 AN-1203 Ultra-Low Power Wi-Fi Connected IoT Node Many applications in the Internet of Things (IoT) market require an ultra-low power design because of limited power availability. Some IoT devices are battery operated, so batteries must be periodically replaced. Other devices are powered by energy harvesting, meaning the amount of available power is limited and dependent on external conditions (light, wind, heat, etc.). In both cases, ultra-low power is essential. To achieve ultra-low power consumption, the complete circuit should be asleep (consuming low power) most of the time. It should wake up periodically to acquire data or take action, communicate over a wireless connection, and then go back to sleep. For long periods between waking (e.g. hours), sleep mode energy consumption may exceed active mode consumption. This application note describes a Silego GreenPAK CMIC acting as a wake-up circuit for a Wi-Fi connected IoT node. GreenPAK CMICs can help keep sleep mode consumption ultra-low, as a microcontroller unit (MCU) can offload some functions to the GreenPAK while the MCU goes to sleep. Serial inputs to the GreenPAK come from MCU in the form of a deep sleep signal and sleep time duration, and the GreenPAK output is the wake-up signal for the MCU (or power down signal, depending on the MCU type). The ESP8266 low-cost Wi-Fi chip was selected as the MCU for this application note. The basic diagram is shown in Figure 1. ESP8266 Power Management Figure 1. Ultra-low power IoT node basic diagram The Espressif Systems ESP8266 is a low-cost Wi-Fi chip with full TCP/IP stack and MCU capability. This small size module allows microcontrollers to connect to a Wi-Fi network and make simple TCP/IP connections using Hayes-style commands. Espressif has released a software development kit (SDK) that allows the chip to be programmed, removing the need for a separate microcontroller. ESP8266-based modules have demonstrated themselves as a capable, low-cost, networkable foundation for facilitating end-point IoT development. ESP8266 s flexibility and price point have driven its explosive adoption in recent designs. Engineered for mobile devices, wearable electronics, and IoT applications, ESP8266 achieves low power consumption with a combination of several proprietary technologies. The power saving architecture features three modes of operation: active mode, sleep mode, and deep sleep mode, thus allowing battery-powered designs to run longer. Table 1 shows the differences between the three sleep modes and power down mode.

2 Item Modem Sleep Light Sleep Deep Sleep Power Down Wi-Fi OFF OFF OFF OFF Clock ON OFF OFF OFF RTC ON ON ON OFF CPU ON Pending OFF OFF Current ma ma μa 1 μa Table 1. ESP8266 low power modes In Deep Sleep mode, the chip will turn off Wi-Fi connectivity and data connection, leaving only the RTC module on, to enable periodic wake-ups (consuming μa, depending on the configuration options). In Power Down mode, even the RTC module is turned off. This means chip wake-up must be external, but the current consumption is only 1 μa. Wake-up Circuit Implementation If we turn off the RTC by switching to Power Down mode, and apply an external wake-up circuit consuming only microamps, we can significantly reduce sleep mode current consumption. When a GreenPAK CMIC is applied as an external wake-up circuit, the ESP8266 s internal RTC can be powered off during sleep periods, reducing the sleep current from ~50 μa to ~5 μa (between 2 μa and 10 μa, depending on the GPAK selected). Thus, the total sleep mode current is ~10 times less than using Deep Sleep mode. There are several different development boards for the ESP8266: NodeMCU, Adafruit Feather, and SparkFun Thing are among the most popular. For this implementation, we used the Adafruit Feather Huzzah, although any development board offering access to the CHPD pin and at least three GPIO pins (one for SLEEP and two for I2C) would be suitable. The wake-up circuit is directly connected to the host MCU as shown in Figure 2. This wiring diagram is applicable to all GreenPAKs and all MCUs with a Power Down pin. SDA and SCL wires (used for I2C communication) are optional, used to change the sleep period with I2C enabled GreenPAKs and I2C capable MCUs. POFF is the Power Off output. Figure 2. Connection to Host MCU The wiring diagram for a demo circuit using GreenPAK SLG46537V and Feather Huzzah ESP8266 is shown in Figure 3. This wiring diagram includes a pushbutton for forced wake-up (EWU, External Wake-up), and an additional pin for reading the wake-up cause (LWC, Last Wake-up Cause). Note also that ESP8266 GPIO16 is wired to RST to enable the internal Deep Sleep mode. The GreenPAK wake-up circuit solution is compatible with other ESP8266 sleep options, so the GreenPAK solution becomes yet another sleep mode: let s name it Ultra Deep Sleep. All existing sleep options Modem Sleep, Light Sleep, and Deep Sleep will work with unchanged parameters and performance, on the same device, along with the Ultra Deep Sleep option.

3 Figure 3. GreenPAK Connection to Adafruit Feather Huzzah ESP8266 The timing diagram for the circuit is shown in Figure 4. Signals are shown in positive logic for clarity. Programmable Sleep Period Figure 4. Timing diagram When using the internal ESP8266 Deep Sleep mode, the user can set the wake-up timer in MCU code. A similar feature can be made available using the GreenPAK I2C interface if required. By writing a value to the GPAK CNT/DLY register, the duration of the sleeping period can be (re)programmed from ESP8266 code. The ESP8266 Arduino Library contains Deep Sleep code, enabling the programmer to easily initiate a Deep Sleep period: ESP.deepSleep(sleepTime); where sleeptime is in microseconds. For Ultra Deep Sleep, code is provided to enable similar functionality: ultradeepsleep(sleeptime); where sleeptime is also in microseconds. The ESP8266 will send the sleep signal on the GPIO pin defined in line: #define ULTRADEEPSLEEPPIN 12 Software for I2C communication is based on the Silego library for GreenPAK5 I2C commands by David Riedell (more information on that Arduino library can be found in application note AN-1107: How to Use Silego's Arduino Library with GreenPAK).

4 Additional functions are provided to initiate Ultra Deep Sleep over I2C and read the current timer setup: void ultradeepsleep_i2c(sleeptime); int readsleeptime(); If you are initiating Ultra Deep Sleep over I2C, don t forget to reset the signal back HIGH after wake-up, because sleep is edge-triggered: silego.writei2c(virtual_inputs, 0x01); MCU Offload The MCU can offload some functions to GreenPAK to perform while the MCU goes to sleep. The sleep timer is only the first such function. The pushbutton for forced manual wake-up is a good example of an MCU offload function. As it is asynchronous (the user can push it anytime), keeping the MCU on to await this possible input would be a waste of power. However, the amount of power that GreenPAK uses to scan the pushbutton is negligible, since the circuit is fully static. In the demo circuit, a debounce filter is included to illustrate possible performance improvements at no extra cost. The debounce filter consumes power only upon pushbutton input, so no power is wasted while waiting for input. The debounce filter removes glitches and noise so the MCU gets processed and filtered input. A full list of potential MCU offload functions is out of the scope of this application note, so we ll just mention two: multiple external wake-up sources, and battery voltage monitoring. The pushbutton is one possible source of external wake-up. Some applications feature multiple sources, like radio wake-up, wake-up upon light or sound, etc. There may be a need to prioritize external wake-up sources, or they may be dependent on each other. It is easy to include all these functions within GreenPAK. To know when to initiate controlled shutdown procedure, battery powered IoT devices often need to know when the battery is almost empty. If the MCU wakes up periodically to monitor the battery voltage, that will shorten the battery life, and if the designer applies long sleep periods to save power, the moment of battery depletion might be missed. This function may be fully offloaded to GreenPAK, thanks to the available analog comparators (ACMPs). GreenPAK will simply wake up the MCU when the battery depletion threshold is reached. The ACMP consumes significant current if always on; however, current can be reduced to the microamp range by wake/sleep as presented in Silego app note AN-1025: 1.0uA Battery Voltage Monitor. Imagine your own MCU offload functions, like monitoring multiple power supply sources, or controlling sensors during sleep so that sensor data is ready to read right after the MCU is awoken. With GreenPAK in the picture, simple yet critical functions of your IoT node may be alive and armed, even when there is not enough power for the MCU to run. Last Wake-up Cause In some IoT applications, users may be interested in what caused the last wake-up ( what woke me up? ), for statistics or other purposes. Since the MCU is asleep at the moment of wake-up, it can t acquire the wake-up cause. GreenPAK stores the cause in a flipflop block so the MCU can read it during startup or even later. In the demo circuit, the MCU can read the last wake-up cause on the GPIO pin and/or over I2C communication: int readlastwakeupcause(); int readlastwakeupcause_i2c(); If the GPIO pin is used to read the last wake-up cause, its location is defined in the line: #define LASTWAKEUPCAUSEPIN 13 GreenPAK Design The basic circuit design includes a wake-up timer with programmable sleep duration and a pushbutton for manual wake-up. The internal GreenPAK design schematic is presented in Figure 5.

5 Figure 5. GreenPAK design The awake/asleep state is stored in a resettable DFF block so the sleep period can be interrupted by an external signal (the pushbutton). The DFF state equals the npoff (Power Off) signal: 1 = awake/power on, 0 = asleep/power off. The wake-up timer itself is built with one edge-triggered CNT/DLY block in Delay Mode. The falling edge of the npoff signal is delayed for a sleep period, so the timer is started at the moment the device goes to sleep. When the timer expires, it pulls the nset input of the DFF low to set the DFF/nPOFF and wake up the device. The awake period is not timed, so the sleep period can be started by the falling edge of the SLEEP input. The falling edge is inverted to reset the DFF and switch to the sleep state. The polarity of the signals is adjusted for ESP8266 and can be easily changed to suit other MCUs or modules. A couple of additional logic gates enable alternative sleep triggers (GPIO pin or I2C message) and wake-up triggers (external or timer). One extra DFF captures the wake-up cause: 0 = external wake-up, 1 = timer wake-up. Main features: The design enables initiating the sleep period using both the SLEEP pin or an I2C message Duration of the sleep period is programmable by I2C message Manual wake-up is enabled by pressing the pushbutton Pushbutton input is debounced Debouncing time constant is also programmable through I2C MCU can read current sleep period value through I2C Internal oscillator runs only during sleep periods, thus drawing no current during awake periods Zero external components Design stores the cause of last wake-up event and signals it on a pin MCU can read last wake-up cause on GPIO pin or over I2C The basic design was created with the SLG46537V. Though the design is very simple, it fits into any I2C-enabled GreenPAK. If we remove I2C features, the design can fit into any non-i2c GreenPAK and still offer a wake-up timer and forced wake-up functions. Note that with the presented design, the circuit will ignore sleep commands while external wake-up is active, so keeping the pushbutton pressed will stop the MCU from going to sleep. However, this will not stop it from going to other sleep modes like Deep Sleep or Light Sleep. Once the pushbutton is released, sleep mode is enabled again. Optional Design Modifications and Optimizations

6 For a fixed sleep period and no manual forced wake-up, the circuit design can be simplified. A simple wake-up timer packed into the tiny SLG46108V is presented in Figure 6. Note that even though SLG46108V is the smallest GreenPAK, there are still enough free blocks to implement external wake-up with debouncing filter. Figure 6. Simple wake-up timer design If GPIO pins are at a premium and your circuit design already uses I2C for communication with other peripherals, the SLEEP pin can be avoided by using I2C to initiate the sleep period. Just remove the SLEEP input and wire the timer input to the I2C block. Note that I2C-enabled GreenPAKs have a One Shot option for CNT/DLY blocks, so the DFF is not necessary and the whole wake-up timer can be implemented in just two blocks, including I2C programmable sleep duration. The presented demo circuit powers off the MCU instantly upon receiving the sleep input. If the application requires some delay to be introduced (for example, to let the debug messages pass through serial debug channel) that can be easily done by applying a DELAY block on SLEEP input. The proposed wake-up circuit may be applied to other wireless modules/mcus besides ESP8266, for example Particle Photon or Raspberry Pi, with little or no modification. For MCUs or modules with no power off input option, the power supply can be cut using GreenPAKs with integrated PFET switch available in SLG46116V/117V or SLG46125V. Circuit Performance The main performance concerns are related to sleep period duration and power consumption, shown in Table 2 for different GreenPAK parts. The minimum sleep period is not stated because it is in the microseconds range for all GreenPAK parts, and such short sleep periods are not expected in real world applications. If required, the sleep period can be made even shorter by using the 2 MHz oscillator (instead of 25 khz or LF OSC) for any GreenPAK selected. Only the first row in Table 2 gives the performance of the demo circuit in this app note, based on SLG46537V. Other rows present estimated performance of similar circuits based on other GreenPAK parts. Data in the table is valid for wake-up circuit design similar to the demo circuit presented in this app note, containing only the wake-up timer and a couple of logic gates. If surplus blocks are utilized to add other functions, current consumption may increase, depending on the design. Green fields emphasize performance marks outscoring other parts in that particular domain.

7 Sleep period duration Tolerance I2C Current draw PFET Max 1) resol 2) deb 3) 25 4) abs 5) ton 6) toff 7) years ms years % % μa μa A SLG4653xV > >100 ±2.5% ±12% yes no SLG46108V SLG46110V SLG46116V SLG46117V ~ days ±2.2% ±8% no no ~ days no SLG46140V > >100 ±20% ±20% no no SLG4662xV > >100 ±20% ±20% no no SLG4672xV > >100 no no SLG4658xV > >100 yes Table 2. Circuit Performance It is obvious from Table 2 that the maximum sleep period could be considered unlimited, since it is in the years range. Thus, maximum sleep period should be considered only if the application uses several CNT/DLY blocks for other purposes. Note that in the demo circuit only one CNT/DLY block is used for the timer, and the maximum sleep duration in that case is around twenty minutes. To increase the maximum sleep duration, cascade another CNT/DLY block and update the ultradeepsleep(sleeptime); function accordingly. Performance trade-offs are possible. For example, the SLG46140V exhibits poor ±20% tolerance when using the LF oscillator, but the current draw is exceptionally low at 0.7 μa. Switching to the 25 khz oscillator will improve tolerance, but current draw will rise to ~5 μa, leveling with other GreenPAK parts. Note the versatility of SLG4658xV parts. These parts offer both I2C communication and an integrated PFET power switch, while other parameters match other GreenPAK parts. Design Testing The GreenPAK Emulation Tool included in GreenPAK Designer Development Suite was used to test the CMIC internal design. The logic signal generator was used to simulate the sleep input. Pushbutton input was configured to Button mode to simulate manual pushbutton input. Test configuration for the first test was named Unit test.

8 Figure 7. Unit test configuration I2C tools were used to check internal signal states while debugging, including the states of internal counters. Output signals generated by the GreenPAK circuit are accessible on test pins of the GreenPAK Universal Development Board. For testing purposes, the ESP8266 IoT node was assembled on a breadboard and connected to the GreenPAK Universal Development Board using patch wires. The popular DHT22 temperature/humidity sensor was wired to the Adafruit Feather Huzzah to sense environmental conditions. The schematic diagram is shown in Figure 8. Figure 8. Final test schematic diagram GreenPAK current consumption was measured using a Fluke 45 multimeter. For this configuration, the GreenPAK Universal Development Board must be set to External VDD. The test configuration for final testing was named Final test. A photo is presented in Figure 9.

9 Figure 9. Test Setup Photo The IoT part of testing was based on the Adafruit IO platform, suitable for the Adafruit Feather Huzzah ESP8266 board. A tutorial on how to connect to Adafruit IO can be found on the Adafruit website. Figure 10. Adafruit IO IoT Service Platform Several tests were performed with the DHT22 temp/humidity sensor (please see attached Arduino IDE code). Since the ESP8266 will go to sleep after it publishes the acquired temp/humidity data, only the first test in code will be executed, so comment out all the tests before the one you want to run. Test 1: Standard Deep Sleep at ten second intervals: ESP8266 wakes up, collects temperature and humidity, publishes to Adafruit IO, and then goes into Deep Sleep for ten seconds. Test 2: Ultra Deep Sleep at ten second intervals: same as Test 1, but with Ultra Deep Sleep for ten seconds instead of regular Deep Sleep. Test 3: Ultra Deep Sleep at one, two, three,, ten second intervals to verify (re)programming of the sleep time: ESP8266 wakes up, collects temperature and humidity, publishes it, reads the last wake-up cause and current sleep duration, increases it by one second (up to ten seconds maximum) and goes into Ultra Deep Sleep. If the last wake-up cause is external wake-up, the sleep period is set back to one second. Note that in Test 3, the ESP8266 must read the previous timer value from GreenPAK in order to increase it by one second for the next sleep period. This is because powering down the ESP8266 clears all registers, including the internal RTC RAM.

10 Replace the test code with your own tests or with your application code. // Test 1 // Standard deep sleep at 10 sec interval Serial.println(F("ENTERING DEEP SLEEP MODE!")); ESP.deepSleep(sleepTimeS * ); //Sleep mode, for power save // Test 2 // Ultra deep sleep at 10 sec interval Serial.println(F("ENTERING ULTRA DEEP SLEEP MODE!")); ultradeepsleep(sleeptimes * ); //Ultra Sleep mode, for ultra save // Test 3 // Ultra deep sleep at 1,2,3..10 sec interval, external wakeup resets to 1 if (readlastwakeupcause()==low){ Serial.println(F("External wake-up! Setting to 1.")); sleeptimes = 1; } else { sleeptimes = readsleeptime(); if (sleeptimes < 10) sleeptimes++; } Serial.print(F("ENTERING ULTRA DEEP SLEEP MODE!")); Serial.println(sleepTimeS); ultradeepsleep(sleeptimes * ); //Ultra Sleep mode, for ultra save Power consumption (current consumption) during the sleep periods was measured during ESP8266 Deep Sleep mode (GreenPAK wake-up circuit idle) and with the GreenPAK wake-up circuit. The measured current consumption of SLG46537V at 3.2 V power supply is around 5 μa, almost 40% lower than the 8.4 μa estimate in the datasheet. Measurements show that power consumption during sleep periods is reduced ~10 times with the GreenPAK wake-up circuit applied. GreenPAK Part Selection Once you decide to use GreenPAK as a wake-up timer circuit, there are many options available depending on your application requirements. A quick selection guide is presented in the table below. If you just need the basic wake-up timer features, cells in the first column contain the best parts per criteria shown on the left. If you need more features (like I2C, dual supply, or integrated power switch) they are not available with the parts in the first column, so check out the parts in other columns. Selection criteria Basic features only With I2C options With Dual supply Integrated power switch Lowest power SLG46140V 0.7μA@3.3V SLG46533V 6.6μA@3.3V SLG46621V 1.3μA@3.3V SLG46116V/7V 5.3μA@3.3V Lowest cost SLG46108V $0.135 SLG46534V $0.224 SLG46121V $0.187 SLG46116V/7V $0.241 Smallest size SLG46108V 1.0x1.2mm SLG46533V/4V/7V 2.0x2.2mm SLG46121V 1.6x1.6 mm SLG46116V/7V 1.6x2.5mm Table 3. GreenPAK part selection For parts available in different packages, the smallest package is shown.

11 Key Advantages and Commercial Viability In many IoT applications, the MCU does not run continuously and peripherals may be idle most of the time. For these applications, sleep mode may represent the lion's share of power consumption and is the vital parameter to consider. The overall power consumption can be lowered by keeping the MCU in the various low power modes it supports. If the solution needs lower sleep current than the lowest power mode an MCU supports (ESP8266 consumes ~50 μa in Deep Sleep mode), GreenPAK is a commercially viable solution. Ultra-low sleep current may be required to prolong battery life, reduce power capacity of power harvesting sources, etc. Key advantages over MCU integrated sleep modes include lower sleep power and longer sleep periods. Even for MCUs with ultra-low sleep modes in the microamps range, if periodic wake-ups are needed for some functions, the average power consumption will be higher than if those functions are offloaded to GreenPAK. Key advantages over specialized ICs include lower cost, no external components, serial communications programmable sleep period, and smaller size. For example, the LTC2956 wake-up timer with pushbutton control from Linear Technology offers high performance, yet the GreenPAK solution offers similar or better performance and some additional features at a fraction of the LTC2956 price. Key advantages over a discrete solution include smaller size, no external components, serial communications programmable sleep period, and longer possible sleep periods. Overall advantages include surplus logic for additional MCU offload functions. In the case of the ESP8266 module, if the application sleeps less than two seconds, Light Sleep mode is preferred to Deep Sleep due to the low wake-up time (<3 ms) from Light Sleep mode. If the application sleeps for more than two seconds, then Deep Sleep mode is recommended. A similar note applies for Deep Sleep/Ultra Deep Sleep preference. ESP8266 needs two to three seconds to wake-up from Deep Sleep or Power Down. If the application is in Deep Sleep for less than a hundred seconds, then power consumption during sleep periods amounts to less than 1% of total power consumption, and power savings introduced by Ultra Deep Sleep are minor. When sleep periods are in hours, the total average power consumption could be cut by 50% and battery life nearly doubled. Check the power profile of your circuit design and estimate the power savings to determine if introducing a GreenPAK Ultra Deep Sleep circuit is commercially viable for your application. Conclusion ESP8266 s built-in Deep sleep mode offers ~50 μa current consumption during sleep periods. GreenPAK can put ESP8266 into power off mode where ESP8266 current consumption goes below 1 μa. GreenPAK s current consumption can be kept at less than 5 μa using the 25 khz oscillator for timing, so the total sleep mode current is ~10 times less than using ESP8266 s Deep Sleep mode. The GreenPAK solution offers many benefits: sleep mode current is reduced, longer sleep periods are possible, and surplus logic blocks are available for additional functions. 1) Absolute maximum sleep period duration with all CNT/DLY blocks cascaded. 2) Min step with no clock predivider/divider applied except for SLG46116V/7V and SLG46108V/10V where /8 clock predivider and /64 clock divider are applied to achieve longer max sleep period. 3) Maximum sleep period duration when one 8-bit CNT/DLY block is used to debounce pushbutton input (as in the demo circuit), and all other CNT/DLYs cascaded to form wake-up timer. 4) Sleep period tolerance at outside temperature 25 C and 3.3 V±10% power supply range based on GreenPAK internal clock frequency limits, roughly presented for convenience. For exact data please refer to relevant Silego datasheets. 5) Absolute sleep period tolerance over entire -40 C to +85 C temperature range and 1.8 V to 5.5 V power supply range based on GreenPAK internal clock frequency limits, roughly presented for convenience. For exact data please refer to relevant Silego datasheets. 6) Estimated total GreenPAK current consumption with sleep timer on (oscillator running) at 3.3 V power supply. For estimated current at other power supply levels please refer to relevant Silego datasheets.

12 7) Estimated total GreenPAK current consumption with sleep timer off (oscillator stopped) at 3.3 V power supply. For estimated current at other power supply levels please refer to relevant Silego datasheets. About the Author Name: Vladimir Veljkovic Background: Vladimir Veljkovic has over 25 years experience in professional grade electronics, primarily embedded, real-time, and distributed systems. His professional career evolved from analog & mixed signal (high power uninterruptible power supply systems) to complex HW/SW electronic systems (large scale telecom switch). He is versed in mixed HW/SW development processes with a focus on system architecture design, automated testing, and quality assurance. In recent years, he s been involved in the IoT industry. Contact: appnotes@silego.com Files AN-1203 Ultra-Low Power Wi-Fi Connected IoT Node.gp3- (10 KB) AN-1203 Ultra-Low Power Wi-Fi Connected IoT Node.gp5- (40 KB) AN-1203 Ultra-Low Power Wi-Fi Connected IoT Node.pdf- (1.0 MB) AN-1203.zip- (1.0 MB) See full list of Application Notes