Hello,
after building my first version od DC/DC booster as described here I have testes few capacitors 33uF, 45uF and at last 1000uF (electrolytic, low ESR) - it needs about 60ms to get fully charged.
Test circuit with white LED 20mA lightup + fading. Time I got between recharging is about 1m30s.

My question is: Is there any equation to calculate optimal capacity/charge time, to achieve best power efficiency?

Correct me if I'm wrong in my approach to find size of reserve capacitor with charge time equal to time I need to take all measurements, send data with RF module and receive ACK and go to sleep.
So not to draw power from booster only to run uC (atmega328p) while capacitor is already charged?

Right now I have made self running power circuit ( BATT -> MCP1640 -> NCP302 + IRF7410PbF) where time is regulated by capacitor attached to pin 5 (CD) of ncp302, so system is autonomous.

In context of regulating time before measurements I think it's a must to do it on uC site, because I don't see reason to change capacitors for every time change.

So:

• wake up uC
• increase counter
• go to sleep
• every X times wake up all devices take measurement
• reset counter
• go to sleep
  is the best and only approach.

This X should be given to node form Gateway during first start or config.

Any comment about capacity and wake approach?

Hello,
I have finaly build (recreated) what Charles showed...

I have one problem now. How to turn MCP for the first time battery is inserted.
Can someone show me how he did it?

My schema looks like this and is EN by low state from voltage supervisor.

Would be nice if you could write in english...

Hello,
finaly I have all needed parts and I started to build my own ULPNode.

To make as small current as possible I'm looking for best mosfets for this purpose.
I have found: IRLR6225PBF with n-channel (2V - ~10mOhm) or SiA436DJ (1,8V ~12,5mOhm)

and SiSS23DN with p-channel (Rds(on) (-1,8V - ~11,5 mOhm)

Did anyone found something better than this?

@Sergii said in ULPNode schematic:

WDYT Guys from openenergymonitor used a bit more expensive Step-Up regulator
LTC3525 https://github.com/openenergymonitor/Hardware/tree/master/emonTH/emonTH_V1_5 for sensor node. Would it have some benefits to use it in UPNode

• In the doc of LTC3525 there is info, that quiescent current in shutdown mode is 0.1uA, while the same parameter in MCP1640 is 0.7uA - so worse.
• In LTC you have Input Start-Up Voltage 0.85V, in MCP this is 0.65V - so better.
• In MPC you get regulated output voltage from 2.0 to 5.5V (with 6M8 and 3M9 resistors you get 0.28uA loss, or you can use mosfet to disconnect resistors completly), in LTC series you have to use different components.
• Compare price on http://www.digikey.com LTC $44.1/pcs while MCP $0.57/pcs

LPC seams to have better efficiency for 1.2 V, 10mA ~88% while MPC has ~82%,
but for 100mA both have 80%...

So I would pick MPC because it's regulated, and has lover voltage start and significantly lower price

Thanks for answer.
Eeee waiting for other people solution is good for business and this is my fun to find the solution

So first of all I bought those parts NCP and MCP but I had to look for those in china because in Poland no one would like to sell me 3-5 pcs. I could buy 100 or 50 pcs. which is a way to much.

So I have a week to reorganize my schem and think about other "tricks". Of course It would be the best to touch those components and try a few things... but I'll have to wait.

I'll try to update my schema and then we can talk further.

Ok, I tried to draw this in eagle but I'm missing few things...

First I redraw input to get 3.3V as on picture below

According to point 3.4 Enable Pin (EN)
The EN pin is a logic-level input used to enable or
disable device switching and lower quiescent current
while disabled. A logic high (>90% of VIN) will enable
the regulator output. A logic low (<20% of VIN) will
ensure that the regulator is disabled.

So my NCP has to enable booster by applying >90% of Vin - which for 1,5V AA battery = >1,35 V to enable the booster which is easy while we still have 2V limit from the NCP302. But to turn it off I need less then 20% of Vin which is <0,3V.

So my question is how to use NCP302 RST pin 1 to pass those two outputs ? 2V and <0,3V ?
Second thing is what capacitor should i use between MCP1640 and NCP302 ?

Can you point me to the right direction because I stuck on this:

I have read a little bit and... I'm a little bit wiser

As I understand schema from https://hallard.me/ulpnode-low-power-secret/ atmega 328p must be on in deep sleep mode and monitor "Wake" pin. When NCP302 will turn PIN1 (Reset Output) to ground:

• atmega will wake up
• booster will charge capacitor
• atmega will measure battery voltage (make some more operations like send info with NRF or aoter device*) and go to deep sleep again
• meanwhile booster will charge capacitor and after reaching 3,6V NCP302 will disconnect it.

`* this operation can be done after "n" wake-up cycles according to eeprom saved information

Did I get it right?

Second thing about NCP302.
As I understand schematic there are two mosfets N and P just before PIN1. One is turning of booster (P) and the second (N) is touching to the ground the pin from atmega 328?

Hello,
I was reading your article "ULPNode Low Power Secret Revealead" which sounds great, but I watched schematics on gtihub and I cant find your solution MCP1640 + NCP302LSN20T1.

First question is did you stuck with some problems during tests? I would like to build this part of powering the board and test it. Can you provide information about capacitor and resistor which gives you 2V limit ?

According to http://ww1.microchip.com/downloads/en/DeviceDoc/22234B.pdf you picked "MCP1640" (without B/C/D) as it has "true disconnect" and PWM/PFM features. In chapter 5.2 and 5.4 I can get equation to calculate resistor and capacitor but... did you use any inductor to minimize noise and did calculated values fit your live tests?

Sorry being so lazy and for direct asking but I would like to use your experience and make big step forward

Regards
MJ