Problems with ESP12f without WiFi on 50mA power source

[tuxedo0801]

Basic Infos

Hardware

Hardware: ESP-12F
Core Version: 2.2.0

Description

I want to drive my ESP12F module from a 50mA@3.3V source. For this application I don't need WiFi at all. So I disabled it.
If I measure the current consumption with my multimeter, I get about 14mA@3.3V. That would be okay for my power source.
But as soon as I connect the ESP to the meant power source, the 3.3V line drops to 1.8V and I can measure a current consumption of about 30..35mA. The blue LED on the ESP12F starts to flicker continuously.
My current theory is:
The ESP8266 has a too high starup current consumption my power source cannot handle. I added an 100µF capacitor, but this does not solve the issue.
I don't have a scope to measure the exact startup current.

Question is:

• Is my code okay (wrt. disabling WiFi completely)? Of course the consumtion I can measure is just about 14mA, but maybe there is a "faster way" of disabling it to reduce startup current consumption?!
• Does anybody know about the startup power consumption without wifi or can do a measurement with a scope?!
• Are there known issues with this scenario (no wifi, low current power source)?

Settings in IDE

Module: Generic ESP8266 Module
Flash Size: 4MB
CPU Frequency: 80Mhz
Flash Mode: qio
Flash Frequency: 40Mhz
Upload Using: SERIAL
Reset Method: ck

Sketch

#include <ESP8266WiFi.h>

void setup() {
    // turn off wifi completely   
    WiFi.disconnect(); 
    WiFi.mode(WIFI_OFF);
    WiFi.forceSleepBegin();
    delay(10);

    Serial.begin(19200);  

}

void loop() {

  Serial.println("HIGH");

  delay(3000);

  Serial.println("LOW");

  delay(3000);

} 

[igrr]

Can you try adding RF_MODE(RF_DISABLED); line between include and void setup?
After that calling WiFi.xxx functions should not be needed.
Note that this requires a pretty recent version, like staging 2.3.0-rc1 or git version.

[tuxedo0801]

Thanks.I will test it this evening.

But I'm still interested in a scope-screenshot showing the current consumption on startup. So If someone can help ...

[igrr]

Bummer, RF_MODE(RF_DISABLED); works only after deep sleep wakeup, not after normal power-up.
I've got some scope pics, will post a bit later.

[igrr]

Some scope pics taken using a 2.5 Ohm current shunt. Scale is therefore 80 mA / 200 mV.

Test 1: Your sketch. You can see that RF calibration is causing large current spikes (300mA peak)

Test 2: Your sketch, and also disabling RF calibration

extern "C" {
#include <user_interface.h>
}

RF_PRE_INIT()
{
  system_phy_set_powerup_option(2);
}

// rest of the sketch

Same trace, closeup of the remaining spike (approx 320mA, 0.8ms):

Test 3: attempt to disable WiFi early. Added a hook to user_init in core_esp8266_main.c:

extern void early_init();

extern "C" void user_init(void) {
  // all the existing code
  early_init();
}

And move WiFi configuration from setup into this new function (inside the sketch):

void early_init() {
  WiFi.persistent(false);
  WiFi.mode(WIFI_OFF);
  WiFi.forceSleepBegin();
}

Same trace, different vertical scale:

The remaining spike is about 100mA, 0.8ms. Can probably be filtered out by a capacitor.

[krzychb]

@igrr,
Do you mean one division on the screen = 40mA and max spikes = 80mA?

[igrr]

I have updated my previous comment with some numbers.
The mentioned hack with an extra hook in user_init can make its way into the core, i guess. Would be useful to turn off wifi depending on certain conditions.

[tuxedo0801]

Thanks for the screenshots. I will test it and report back.

can make its way into the core, i guess

What does this mean? How would one disable wifi in future with this change?

[update]

Ah, now I got it. There might be a new "void early_init()" method which you can use in the sketch to do "very early things"..

[tuxedo0801]

100µF capacitor and code changes as suggested in Test3 give me flickering blue led and console tells me repeatedly:

ets Jan 8 2013,rst cause:2, boot mode:(3,7)

case 2 = reset pin
mode 3 = flash ...

is this normal if power source does not provide enough current?

same setup with mit current on power source boots as expected.

Will repear test with more µF ...

[tuxedo0801]

200µF gives me on console:

ets Jan 8 2013,rst cause:1, boot mode:(3,6)

load 0x4010f000, len 1264, room 16

To compare: with optimal power supply (more than up to 2,5A instead of just 50mA max), I get:

ets Jan 8 2013,rst cause:2, boot mode:(3,7)

load 0x4010f000, len 1264, room 16
tail 0
chksum 0x0f
csum 0x0f
~ld

[igrr]

Looks like you aren't reaching the point when ROM bootloader finishes reading second stage bootloader from flash. How much of a voltage drop does your power supply have at 50mA?

[tuxedo0801]

It's not a "real" power supply.
The ESP12 is attached to a "KNX Transceiver" which is connected to the (wired) KNX home automation bus. The KNX transceiver IC (www.onsemi.com/pub_link/Collateral/NCN5120-D.PDF) is able to provide about 40..50mA at 3.3V (DC-DC converted bus power)
There is no spec/figure/graph showing the voltage drop over current consumption. The current estimated current limit is about 40mA, 50mA should be also possible. I can try to measure the voltage drop when applying a big enough resistor to simulate constant 50mA current consumption.

P.S. I just saw that Test3 might include changes from Test2... Sketch is changed accordingly, has to be tested this evening...

[tuxedo0801]

Good news... with all your suggestes changes (test1+2+3) in addition to an 470µF capacitor on Vcc, ESP now now able to start with my luw current power source...

Thanks a lot.

[tuxedo0801]

btw:

Would be great if we could add those code-changes to the next release?!

@igrr
Should I make a PR or do you add those changes yourself?

[igrr]

Awesome, glad that this worked for you.
Created #2133 to track changes necessary in the core.

[wb8wka]

Is there a reason RF_MODE(RF_NO_CAL); would not be working? 2.3.0 with a Wemos D1 mini. All I can do is either turn the RF off or RF_CAL is on. with with the call to RF_NO_CAL

[dioguerra]

This reduction of power consumption doesn't seem to work on 2.4.0-rc2

using user_init()