ESP8266 Community Forum

One point worth considering. From what I understand, the most current demanding operation is when the ESP8266 does an analog calibration on startup. And I think this is where it is most likely to fail. So maybe, once it is up and running, it can tolerate a lower supply voltage, like the spec sheet lists.

rudy wrote:One point worth considering. From what I understand, the most current demanding operation is when the ESP8266 does an analog calibration on startup. And I think this is where it is most likely to fail. So maybe, once it is up and running, it can tolerate a lower supply voltage, like the spec sheet lists.

This is true, but you can minimise the power spike at the start by changing the RF calibration settings using system_phy_set_rfoption() with the value 2 (No RF calibration after Deep-sleep wakeup) and system_phy_set_powerup_option() with the value 1 or 2 (1 = RF initialization only calibrate VDD33 and Tx power which will take about 18 ms, 2 = RF initialization only calibrate VDD33 which will take about 2 ms).

I've found that for fixed installations of a sensor unit you can perform RF calibration once when the unit is first set up, and then after the unit is calibrated for the installation site, you can turn it off or only do it periodically (by using system_phy_set_powerup_option() with 0 or 3), reducing the initial current spike by a lot. Since installation is when a battery is most likely to be at its highest charge, this works quite well. After that, a decent sized capacitor will manage the start up spike quite nicely.

Note that these calls should be protected with a corresponding check on existing values as they write to the flash so you only want to call them if the value needs to be changed.

Having said that, you will still need to find a good 2.8V LDO that does at least 400mA peak - but I am going to go through the list in this post and see what I can find. While I am happy with the one I have now, the list has a few that have an even smaller drop out voltage.

PS: @rudy I also really like your idea of using a pair in parallel! I know it was a hack, but it was a brilliant one

davydnorris wrote:

rudy wrote:One point worth considering. From what I understand, the most current demanding operation is when the ESP8266 does an analog calibration on startup. And I think this is where it is most likely to fail. So maybe, once it is up and running, it can tolerate a lower supply voltage, like the spec sheet lists.

This is true, but you can minimise the power spike at the start by changing the RF calibration settings using system_phy_set_rfoption() with the value 2 (No RF calibration after Deep-sleep wakeup) and system_phy_set_powerup_option() with the value 1 or 2 (1 = RF initialization only calibrate VDD33 and Tx power which will take about 18 ms, 2 = RF initialization only calibrate VDD33 which will take about 2 ms).

I've found that for fixed installations of a sensor unit you can perform RF calibration once when the unit is first set up, and then after the unit is calibrated for the installation site, you can turn it off or only do it periodically (by using system_phy_set_powerup_option() with 0 or 3), reducing the initial current spike by a lot. Since installation is when a battery is most likely to be at its highest charge, this works quite well. After that, a decent sized capacitor will manage the start up spike quite nicely.

Note that these calls should be protected with a corresponding check on existing values as they write to the flash so you only want to call them if the value needs to be changed.

Having said that, you will still need to find a good 2.8V LDO that does at least 400mA peak - but I am going to go through the list in this post and see what I can find. While I am happy with the one I have now, the list has a few that have an even smaller drop out voltage.

PS: @rudy I also really like your idea of using a pair in parallel! I know it was a hack, but it was a brilliant one

interesting!
Great strategy!
I am still studying how I will do my system ...
I will use a lithium battery from a cell.
Apparently the LDO regulators will have better performance compared to the switched.
But I have not decided which one to use yet.
Maybe I use two in parallel as quoted earlier. It is an option to consider to make better use of the battery.

davydnorris wrote:I have found that most sensors have a low power or stand by mode - in many it's simply a case of not talking to them, or sending a power down command. In others it's an enable pin of some sort.

I have found a set of sensors that I can control with either a single GPIO pin tied to all the chip enables, or via software - I haven't needed any external components.

My latest design uses the ISL91107IR buck/boost converter, however I am considering going back to just a straight LDO - my last rev used the NCP186A, which is not in your (awesome!) list:

NCP186
Range voltage: 6 V Max
Max current out: 1 A
Quiescent current: 90 uA
Drop out voltage: 100 mV

Several things have made me rethink using the buck/boost:
- the ESP chip will actually operate at lower than 3.3V. I've heard people using as low as 2.8V to power them successfully. Most of my sensors are already using 1.8V and the ones that aren't can also operate with 3.0V or less
- when the input voltage drops below the threshold on the converter, the boost circuit draws more current to make up the voltage difference. On both solar and battery circuits the increased current draw pulls the supply voltage even lower, so you end up in a death spiral, and the increased current drains the power source even faster.

So in the end, it may work out you get better life by dropping your ESP supply voltage to to 3V or even 2.8V if you can, and using a normal high performance LDO instead.

I need to do some more experimenting...

WOW!
Great option!
I'll read the datasheet and search a little bit about it!
It seems to be one with good chain!
Amazing the drop out!
Quiescent current is a bit higher than desired but I'll have to figure out how it will go on a circuit with a lithium battery in a cell.

I will study a bit and decide between this option and an efficient LDO. Maybe 2 in parallel ....