Low-energy ESP8266-based Board Sleeps Like A Log Until Triggered | Hackaday

Given

It seems that the tinker does need a simple device for "doing some interesting things on the Internet when the button is pressed". If you need this but don’t want to bend the Dash device as you wish, please check 

 instead.

trigBoard is a battery-based ESP8266 circuit board that includes some clever circuits that help it consume almost no power (less than 1 microampere!) while waiting to be triggered by a digital input. This input can be a reed switch, button or similar input, and you can configure the board as a normally open or normally closed switch.

Such a clever hardware with low power consumption is located

, We will also embed it after the break. To sum up: EPS8266 is completely dead for most of the time. Texas Instruments

 The power timer chip burns 35 nanoamps and wakes up the ESP8266 every hour to check the battery. The chip also has a manual wake-up pin, and it is this pin and more power saving circuits that are used to trigger actions based on external inputs.

Obviously, the microcontroller can somehow distinguish between waking up the battery check and pressing the button, so you don't have to worry about accidentally sending an alarm every hour. Set default firmware for use 

 Send notifications, but of course you can perform any operation that EPS8266 can perform. The code is in

.

The evaluation board also includes a standard micro-JST connector for LiPo batteries, and the battery can be charged via the micro-USB port. Complete TrigBoard 

On the Wiki, and the prefabricated equipment is 

After the break, the hardware walkthrough video of [Kevin] will be embedded.

[by

]

1uA? This is smaller than the trend of flared jeans.

This is the first article worth it!

I used it in a project and also measured total consumption. 65 nA when ESP is off. Cool product (

)

This is a great little board, and Kevin has put in a lot of work, and the results show.

that's nice

I personally use ATTINY for the wake-up circuit, but this is only because I have done it before, so I know how to do it, and I already have a lot of them, and they are easy to get.

and also:

TPL5111: US$0.77, ATTINY9: US$0.30 (unit price from Mouser)

However, the power consumption of running the watchdog timer is relatively high, about 5µA.

But using micro time instead of hardware time, you can also use more flexible wake-up options, such as multiple wake-up sources with filtering, flexible sleep time (depending on the situation and battery power, etc.).

When mass production, of course you also have to calculate programming costs.

If you are satisfied with the higher power of 2 orders of magnitude, you can also put ESP8266 into deep sleep, because this also uses microampere current.

When ESP8266 wakes up for a few seconds every hour, does it really have to be 65nA?

Suppose it takes 5 seconds to start, connect to WiFi and send some data. Approximately 200mA is required during this time. So the average current is about 278µA.

In this case, the 5µA of the wake-up circuit is not important.

If you only wake up once a day, or wake up only on demand, that is another matter.

I totally agree with your comment. With WPA2, the time I see on Wifi is more like 10 to 12 seconds.

Disable dhcp, use optimized firmware and get the number of milliseconds

Or use ESP-now mode instead of standard wifi; up to 20 nodes (using MAC address), 256 bytes per packet, no router or AP, so no (direct) internet connection

Any links to this circuit or hardware? I like to make some for myself.

Picky: I have characterized the TPL51xx timers, using them to drain 10uF film capacitors, using a flying amplifier to buffer capacitor voltage and tracking the output voltage of the amplifier.

The timer does reach the 35nA standard, but it only works when the power supply voltage is lower than 2.5V. At the 4.2V voltage mentioned in the video, they draw about 100nA of current. This is excellent performance (slightly less than 1mAh per year), and well below the 1uA quoted by the project, but not as surprising as believed on the front page of the data sheet.

More picky: Take a look at the TPS3839, which has a typical supply current of 150nA. When disabled, the typical current of the TPS73733 is 20nA. So look at a few hundred amps.

I have been playing similar games, which is very useful to me.

The main difference is that I use a variable resistor to set the timer to different time periods.

I will buy one to have a try, but since it is from the United States and will be sent to the United Kingdom, the postage has doubled the board's cost.

I might not like these types of devices, but if it's about a device that just waits for a button to be pressed, why not make a device that is completely powerless and perform the requested operation on startup?

Then you can use the button to connect the battery to the device;-)

No, according to the article, there is also a timer to wake it up periodically.

This is the purpose of TPL5111.

That requires strength.

It has the added benefit of waking up regularly, so you can check the battery status and make sure it is still working. Otherwise, capacitors and latch circuits can be used to keep the ESP active. If there is no latch circuit, you must press the button until the wifi completes its function.

You can, except that it takes enough time to guide, find a WiFI network, ask to join it, perform an encrypted challenge-response for identity verification, and use DHCP 4 methods to obtain an IP address. In addition to DNS lookup, it may be After 3 TCP handshake and other operations, and someone is unlikely to long press the button... Therefore, you need to "lock" something until the operation is completed and you are told to close it again. Of course, you can design a circuit to do this for you-it is not impossible to do so with a tiny quiescent current, but it is more difficult than you would expect to avoid all kinds of noisy false triggers.

If you also want to add a timer to wake up and regular self-tests/call home, then you obviously also want to tell your ESP (or whatever you wake up) if it is woken up due to a key press or timer-fire. Soon, you will get something much larger than TPL5111 and absorb more quiescent current.

Sometimes it is easier to just add a small chip-someone puts all the thought, design, miniaturization and construction time for you :-)

If only one button is needed, it can be switched off with zero current. You can supply power to the microcomputer through a relay (for high power use) or a transistor, which is one of the first actions of the microcomputer. You can first get power through a button without bypassing the relay/transistor. This is what I did for the children's toys that I made, because young children don't remember to turn things off.

This is not an ultra-low power thing, but it will consume most of the time, so higher current is a good trade-off.

This is why you want to boot. ESP8266 only needs a few milliseconds to start up. The first step should be to lock the power button. Do its stuff and then kill yourself by releasing the latch :)

Hello there

Here is the same architecture.

Really good

Does a value below 1uA really make a substantial difference in battery life? I am pretty sure that the self-loading rate of those lithium-ion battery packs is> 1uA.

I also think that more people should look at LiFePO4. When fully charged, they output 3.3V voltage and have a very flat discharge curve, so they can be used without a regulator, thus avoiding the loss of a static regulator.

It is difficult to say before the current consumption is amortized.

The above project is opened every hour. Suppose it stays on for 100ms, and (knowing ESP8266) uses about 100mA of current during this time. This will be allocated to a stored charge of about 2.8uAh.

Coupled with 1uA of self-leakage current and 1uA of idle current, the total amortized current consumption is about 4.8uA per hour. In this budget, the two terms representing pure loss reach 40%, and the idle current is only half of this loss.

Lowering the idle current to 200nA reduces the total budget to 4uA. The loss clause accounts for about 30% of the total. Idle current accounts for about 5% of the total and about 17% of waste.

This is an improvement worth considering.

Reducing the idle current to 50nA will reduce the total budget to 3.85uA, and the idle current will account for about 1.3% of the total budget, which is a waste of about 5%. By then, you may have reached the point of diminishing returns for reducing idle current.

If you want to connect to wifi and say hello to some server, then sticking to 100 milliseconds is a dream thing... it takes about 10 seconds...

You can save some time by hard-coding the IP instead of using DHCP.

TPS3839 can work as expected in this circuit, which surprised me a bit, because when the input voltage is suddenly eliminated like this, the chip cannot guarantee that the output will be pulled down. It needs at least 0.6V to work. It seems that there must be enough parasitic capacitance on Vdd to keep it long enough to drive its output low. You will see that it needs energy from somewhere to charge the gate of the output pull-down FET. I recommend adding a tiny capacitor between pins 1 and 3 (but only like 0.001uF, so as not to slow down the response time; this capacitor is approximately equivalent to a 6ms delay) to help ensure the expected operation. According to the data sheet:

"8.4.3 Lower than power-on reset (VDD

When the voltage on VDD is lower than the power-on reset voltage (V(POR)), the RESET output is uncertain. In this case, do not rely on the RESET output to achieve normal device functions. "

The circuit is good. Thoughtful!

Connect the trigger to the CH_PD (enable) pin. Find the GPIO that is high at startup and connect it to CH_PD as well. Now, the trigger turns on the microcontroller, and when it finishes its work, it can turn itself off by setting the GPIO low.

I built something similar last year, but it is only used for button press detection, and there is no timing wakeup. It uses a pair of complementary Mosfets, one half is used to isolate the power line from the ESP, and the other half is used to lock the power supply. ESP locks the power immediately after starting. It uses UDP for communication because it is much faster than TCP, which saves a lot of power. It uses a pair of AAA batteries, and tests have shown that they can last for several years for problematic applications. Although ESP8266 is not suitable for battery power supply, the use of UDP protocol (such as CoAP) can keep the average power consumption low enough to achieve battery-powered intermittent signaling.

Smart mines enabled via the network extend battery life

I personally think that it takes a long time for the MCU to find the WiFi network, connect to it, obtain a DHCP address, then send its network message, and then go to sleep. For comparison, the actual reading of the sensor is very fast. These network tasks can easily take 30 seconds or more. This is the main driving force for the required "wake-up time" (when the high-power RF is on), and the "wake-up time" is the main driving force for the system battery life.

Using UDP is very different from the "on" time required to send a message. If I remember correctly, it changed it from a few seconds using TCP/IP to an "on" time of less than one second required to send a message with UDP. If only the hub device is notified on the LAN, UDP can work normally. I used an N-channel mosfet in the ground wire of the ESP to turn it off to eliminate the leakage current when the ESP is off. Press the button to power the ESP and mosfet, and charge the capacitor to keep the mosfet and ESP long enough to enable the ESP to open the P-channel mosfet through the GPIO pin (pulled low) to keep the ESP powered. After sending the message, the GPIO pin is switched to a high level to turn off ESP. If anyone is interested, I will provide a schematic diagram. At that time, the cost of Texas Instruments was higher than that of ESP, so the mosfet solution was also higher.

I am interested in your schematic, if you can send it. I found that the 8266 boot time is about 10 seconds, which is surprisingly long.

Please be kind and respectful to help make the comment section great. (

)

The site uses Akismet to reduce spam.

By using our website and services, you expressly agree to our placement of performance, functionality and advertising cookies.