Use this forum to chat about hardware specific topics for the ESP8266 (peripherals, memory, clocks, JTAG, programming)

User avatar
By AdamWu
#40579 TL;DR: https://github.com/Adam5Wu/Micro-power- ... ent-design

I have been searching for efficient power solution for battery operations.

I was aware of linear regulators, but converting excessive voltage as heat doesn't seems to efficient enough;

Then, I found TPS63025x, which seems to be the best switching mode micro-power supply when it comes to <1mA current draw. But still, the 35uA Iq (realistically, about 45uA with everything assembled) seems a bit too high, compared to 2uA Iq of MCP1700.

While at working current draw SMPS is far more efficient than linear regulators, because we must keep the chip sleeping most of the time, so quiescent current is still a very important factor.

Recently I came across MC33464, a power supervisory chip, which seems to be the missing piece for creating the super high efficiency micro-power SMPS!

Basically, it is an under-voltage detector, like an automated switch, when voltage goes above or below threshold, it flips states. What is good about this chip, is the current draw is super low -- ~1uA.

My idea is to make this chip control the SMPS states to significantly reduce the quiescent current. Just like the thermostat for AC system, which regulates the room temperature while saves power by keeping the AC off most of the time, if the room has enough insulation; I could use this chip to regulate the voltage and keeping the SMPS off most of the time, while the esp power draw is very low.

TPS63025x does have a shutdown mode, during which the Iq reduces to 0.1uA. So with MC33464, the Iq of both chips would be around 1.1uA (*) -- as long as the Vout is about threshold; when Vout drops below threshold, MC33464 turns on TPS63025x, which quickly pumps the Vout up; then MC33464 detects the Vout is above threshold, the turn off TPS63025x, the cycle repeats.
(* the system Iq would be higher, due to the need of pull-up/drain)

With this design, it is possible that this SMPS could have comparable Iq with MCP1700, while have 95% efficiency during load, making it a much superior battery power supply! :D
Last edited by AdamWu on Mon Apr 18, 2016 12:36 am, edited 2 times in total.
User avatar
By AdamWu
#40595 There is a bug in previous circuit... :P
When Vin is higher than Vout, Vgs will be negative regardless of the !RST signal, thus renders the control ineffective.

A new version is attached, using N channel instead of P channel FET.
The normally closed drain increases the Iq by another ~0.1uA, but at least it fixes the bug. :P

If only the RST signal from MC33464 was not inverted, the solution would be much more elegant...
User avatar
By AdamWu
#41554 Success! Well sort of.

Here is my journey:

Designed a PCB based on v2 circuit, $1.10 for three boards from OSHPark, they are the best! :D

Assembled together, replacing r2 and r3 with 22K and LED, for easy visual confirmation.
Powered on with no load, got really weird results: with R1 as designed, almost couldn't observe off time (LED constant on), when I replace R1 with very small value, see about 30% off time; add a 0.22f supercap on output, off time increase to about 50%, but input current stables at 1.5mA, 500x greater than what I want.

What can go wrong? Kept checking everything, swapped components... finally, one thing came to mind, shutdown mode is not a simple disconnect!

Dug out the datasheet, bingo! in shutdown mode, the chip effectively discharges the output with 120Ohm resistance. Previously I only noticed their claim of "true shutdown" -- load disconnects from input, which is true. But it appears that they helpfully bundled a discharge circuit...

Thanks, but no thanks.

So then I have to place some "knife hacks" on the PCB board, cutting some traces, rerouting, etc.

Finally, I am able to get a bug-free circuit working, by reducing the output capacitance to 20uf the minimum recommended, and the output goes through a schottky diode before it reaches the load capacitor and R1.

The upside is that when the load is light, the circuit indeed shutsdown for most of the time, and current during shutdown is lower than 10uA (the limit of my meter), much lower than 55uA Iq of my previous assembled model without the power saving design.

The downside is, the diode incurs 0.2v drop, which means that the power saving is reduced, by two factors: 1. discharge of the 20uf 1st level output capacitor at each shutdown, 2. reduced hysteresis, originally 3.3v -> 3v, now 3.1v -> 3v.

Next will see if I can use mosfet to replace the diode.
User avatar
By AdamWu
#41564 V3 circuit! :lol:

Uses MOSFET for cutting backward current during shutdown, so voltage drop should be much smaller;
Use 0.1uf for direct connect output capacitance from TPS63025x, put the larger output capacitor behind the MOSFET. When !shutdown the MOS should allow the larger capacitor be "seen" by the regulator; when shutdown, MOS cut the connection, prevent the energy in the larger capacitor get trashed.
Use a dual P+N MOSFET, able to keep the same PCB footprint.

Just handed in another dollar to OSHPark. Will report back in 3 weeks!