ESP8266 power supply issues

[Dajgoro]

Hi.

I've made a low power circuit that uses the esp wifi modules among other things.
The way it works it sleeps wakes up, powers up all the peripherlas with mosfets (pmos) takes the readings and sends them via the esp.
Now the issue is that the ESP is having problems booting up through the pmos and regulator(LD39050). It works fine if I plug it directly to the bench power supply,
but just adding an ampermeter in series is enough to make it glitch or just powering it in parallel with the rest of the circuit.
What sort of power requirements does it have to boot properly, once powered the voltage is 3.3V even with the current circuit, but the slower power up slew rate seems to mess it up. 

[iMo]

What sort of power requirements does it have to boot properly,..

The best source of information in this matter is usually a datasheet or reference manual (we do not read usually)   

[Richard Crowley]

ESP8266 is infamous for being very sensitive to power droop, especially on power-up.  A popular remedy is to put a large (100uF or even more) capacitor as close to the power connections as possible.  Some of the power-up routines apparently draw unusually large current spikes (compared to long-term "normal" current draw)

[Dajgoro]

What sort of power requirements does it have to boot properly,..

The best source of information in this matter is usually a datasheet or reference manual (we do not read usually)   

All it says is: "Operating Current Average value: 80 mA" and the regulator can provide 500.
As for the module itself, I don't think there is a datasheet.

A popular remedy is to put a large (100uF or even more) capacitor as close to the power connections as possible.

I tried rigging caps, did not help.

[Richard Crowley]

I tried rigging caps, did not help.

How big?
Where? (How close to the board/chip)?

Do you have an oscilloscope?  Can you monitor the power voltage directly at the connection to the ESP8266 board?

[Dajgoro]

I tried rigging caps, did not help.

How big?
Where? (How close to the board/chip)?

Do you have an oscilloscope?  Can you monitor the power voltage directly at the connection to the ESP8266 board?

The pcb itself has a ceramic 0603 decoupling cap, and then I bodged a 470uF directly to the power pins.
On the scope I didn't see anything (crt), will borrow a digital scope and test again.

[BocaDev]

I had a similar issue with an ESP8266 and was able to correct the problem by playing a large cap on the ESP8266 module supply input as Richard had stated. Using a scope the power rail was glitching when other peripherals where drawing current. In my case, the supply at the ESP8266 was dropping down around 2.8 volts. I tried a 470uf cap (rather large being 25V, but I had one laying around) and that fixed the problem. I was thinking about using a second regulator (3.3V) to supply the ESP8266 on it's own, but I never got around to it because of other fires to fix.

I hope you get your project fixed and up and running soon.

[Dajgoro]

I had a similar issue with an ESP8266 and was able to correct the problem by playing a large cap on the ESP8266 module supply input as Richard had stated. Using a scope the power rail was glitching when other peripherals where drawing current. In my case, the supply at the ESP8266 was dropping down around 2.8 volts. I tried a 470uf cap (rather large being 25V, but I had one laying around) and that fixed the problem. I was thinking about using a second regulator (3.3V) to supply the ESP8266 on it's own, but I never got around to it because of other fires to fix.

I hope you get your project fixed and up and running soon.

I tried with 470uF as well, tho I read that the cheapo ESP module are so cheap because they failed QA tests, so maybe its partially defective.
Tho I've made another project before with the ESP that was also supposed to be low power, but I ended up wiring the ESP directly to the power source to get it working.

Btw, welcome to the forum.

[metrologist]

I tried with 470uF as well, tho I read that the cheapo ESP module are so cheap because they failed QA tests, so maybe its partially defective.

Might be as the few cheapies I bought were about the same.

[iMo]

The ESP8266 takes up to 170mA peak current when TXing according to the datasheet. With a sub-optimally matched antenna it could be even more.

Also your "ESP module" - does it include a voltage regulator on the module's pcb?
It could happen you create 3.3V with your "external" voltage regulator and you feeding with it the on-board 3.3V voltage regulator (not good, of course).

[EmmanuelFaure]

For your viewing pleasure, here's attached a graph showing the current consumption of an ESP8266. It used DHCP then sent a short message using HTTP POST. Horizontal axis sample number (Sampled at 100kHz. Total time ~5 seconds), vertical axis in miliamps. Measured with a 0.1 ohm shunt and a fast 16 bits ADC.

[iMo]

To handle 350mA peaks you have to use several smd ceramics multilayer capacitors 10uF-22uF wired in parallel (or a good low ESR tantalum ones). Use beefy and short wires. A single 100uF-470uF aluminum would not help much with such fast peaks, imho.

[EmmanuelFaure]

Depending on the current limiting mechanism of the regulator, it won't be enough. In attached file : The raw current consumption measurement file. Number in Amps, 10µs between each line. Picture : Detail of one of the 300mA-isch peak. Measured with a 10µF ceramic capacitor wired close to the ESP.

[iMo]

Try to measure the voltage drop at the 10uF ceramics during that 350mA peak..
As I wrote above I would use several ceramics in parallel.
Also mind the points you wire the capacitor(s) in has to be carefully chosen as the 0.35A peak may create a large drop with thin pcb tracks.

[Dajgoro]

It could happen you create 3.3V with your "external" voltage regulator and you feeding with it the on-board 3.3V voltage regulator (not good, of course).

Not sure what you mean, but the board is battery powered, and the regulator is on the same pcb as the esp.

Try to measure the voltage drop at the 10uF ceramics during that 350mA peak..

Can't, it won't even boot.

[Dajgoro]

I probed it now with a digital scope, and the thig is such rubbish.
Putting a 470uF cap is a joke!
I needed to put 2000uF to get it somewhere decent.
I got it to the point that the power is stable, but it still glitches, and I can't find where the spike is.

[iMo]

470uF..2000uF will not help you much when the glitches/peaks are 350mA (or more) and dozens of useconds long.

Those aluminum capacitors >100uF are good for filtering 50/60Hz in a linear power supply or as a speaker DC decoupling.

For fast transients you must use ultra low ESR capacitors, with low internal inductance and low internal resistance, otherwise the decoupling will not work.

The best are smd ceramics multilayers, then tantalum ones.

A 22uF ceramic multilayer makes better decoupling as a 470uF aluminum capacitor when talking narrow and fast peaks.

PS: Mind the ceramic multilayers suffer a significant decrease of its capacitance when biased close to their nominal voltage (up to 70% down). Therefore use rather large 1207 smd types for 25Volts DC.

[Dajgoro]

470uF..2000uF will not help you much when the glitches/peaks are 350mA (or more) and dozens of useconds long.

Those aluminum capacitors >100uF are good for filtering 50/60Hz in a linear power supply or as a speaker DC decoupling.

For fast transients you must use ultra low ESR capacitors, with low internal inductance and low internal resistance, otherwise the decoupling will not work.

The best are smd ceramics multilayers, then tantalum ones.

A 22uF ceramic multilayer makes better decoupling as a 470uF aluminum capacitor when talking narrow and fast peaks.

Yes I know.

Now I gave it its own 1117 regulator as most arduino schematic usually show, still nothing.
I added back the 2000 uF because even the 1117 could not cope with the huge power surges.
Ceramic caps, I have some, but now I will try to make it a skyscraper.

[iMo]

PS: Mind the ceramic multilayers suffer a significant decrease of its capacitance when biased close to their nominal voltage (up to 70% down). Therefore use rather large 1207 smd types for 25Volts DC.

1206

[Dajgoro]

PS: Mind the ceramic multilayers suffer a significant decrease of its capacitance when biased close to their nominal voltage (up to 70% down). Therefore use rather large 1207 smd types for 25Volts DC.

1206

Yes, I've got a tower of 1206 now stacked up making about 250ish uF in total of ceramic caps, also added few 0603 100nF but nothing.
So I took a 18650, an old diode and plugged it right to the output of the 1117 and then it works.

... And then I realized. 

The RX line was being pulled down by the mcu! 

Ah, well... But still, it won't work without bodging extra caps to it, but now at least I know how much power it needs.

[Rolo]

Makes me wonder how they do it on the Wemos D1 Mini, it has an ESP-12, own (small form factor) regulator and a Serial to USB chip. I used several with no problems.

[Twoflower]

You mentioned you have a scope. Even if not booting. Can you post a picture of the supply during powering on (not necessarily booting) and one with increased sensitivity (e.g. 100mV/div with AC-coupling)? Please measure as close to the CPU as possible with the shorted GND and signal pins as possible (see EEVblog #1085). Please repeat this measurement with different time-basis and see if there's something hidden if you run too slow. Maybe this will show us what's going on here.

Also please measure the exact supply voltage with a decent multimeter. Also as close to the CPU as possible.

[iMo]

Makes me wonder how they do it on the Wemos D1 Mini, it has an ESP-12, own (small form factor) regulator and a Serial to USB chip. I used several with no problems.

It could be because:
1. their antenna is matched better (+/- 100mA)
2. 2x10uF smd multilayer, 1x1uF smd multilayer
3. ME6211 voltage regulator - high speed LDO, optimized for low ESR ceramic decoupling, with fast transient response.

[BocaDev]

I had a similar issue with an ESP8266 and was able to correct the problem by playing a large cap on the ESP8266 module supply input as Richard had stated. Using a scope the power rail was glitching when other peripherals where drawing current. In my case, the supply at the ESP8266 was dropping down around 2.8 volts. I tried a 470uf cap (rather large being 25V, but I had one laying around) and that fixed the problem. I was thinking about using a second regulator (3.3V) to supply the ESP8266 on it's own, but I never got around to it because of other fires to fix.

I hope you get your project fixed and up and running soon.

I tried with 470uF as well, tho I read that the cheapo ESP module are so cheap because they failed QA tests, so maybe its partially defective.
Tho I've made another project before with the ESP that was also supposed to be low power, but I ended up wiring the ESP directly to the power source to get it working.

Btw, welcome to the forum.

Thank you Dajgoro, glad to be on here.

[C]

YouTube video

Andreas Spiess
#47 Power Saving with ESP8266 (Sleep Mode) Tutorial with some Tricks