Typically batteries are measured by this type of circuit, it draw zero current during idle and draw current only when MOSFET conducts which is just enough to make the voltage measurement. Your GPS will draw significant amount of the current, one minute every hour, so resistor divider is a minor problem and will consume like <1mA x 1ms.

 

How about this:
View attachment 248292

This circuit will vary with temperature because of the zener diode.

 

This circuit will vary with temperature because of the zener diode.

By 20 parts per million per degree Celsius. Perhaps that might not be a problem?
https://www.analog.com/media/en/technical-documentation/data-sheets/LT1004.pdf

 

I need to use the smallest packages for all components available i.e. 0201, DFN0606-3 etc. because of the micro PCB size (about 8 x 35 mm).

 

That variation may, or not be a problem. Pet collars are subject to wider temperatures as the pet ventures outside.

 

That variation may, or not be a problem. Pet collars are subject to wider temperatures as the pet ventures outside.

If you design and position the device properly, it will benefit from the body temp of the pet.

 

What am I doing wrong in the simulation? ADC pulses should be decreasing.

spice3_PMH550UPE attached

 

What am I doing wrong in the simulation?

You have made a source-follower. The voltage at the MOSFET source will be the voltage at the gate minus the Vgs(th) the gate threshold voltage. It will no longer depend on the voltage at the drain.

 

That variation may, or not be a problem. Pet collars are subject to wider temperatures as the pet ventures outside.

I hope 20 ppm/°C isn't a problem, because you'd not be able to use ordinary thick-film resistors, because they can be as much as 100 ppm/°C .

 

The problem was in wrong P-MOSFET library. It works now. The ouput CVD voltage is lower by 633 mV because of the P-MOSFET DS Voltage.

 

The MOSFET gate-source voltage is very temperature-dependent and varies considerably from one device to another.

 

The RVD needs two MOSFETs to work correctly.

 

That still suffers from exactly the same problems as the first version .
[edit] no it’s OK- I didn’t see it was P-channel. A P-channel MOSFET symbol would have been helpful! And it’s the same a @Juhahoo suggested in post #61
My circuit in post #58 will actually work and needs fewer components.

 

SIMetrix simulation of Capacitor Voltage Divider with MAX40023 Op Amp. The ADC input voltage is 3.23V, current through CVD is 485 fA, Op Amp 3.3V current 54nA/16uA.

Delete TXT at the end of *.wxsch.txt
SPICE Op Amp model (for SIMetrix) is in the link above.

 

Surface resistivity for FR4 board is between 10^12Ω per square and 10^16Ω per square depending on environment. The distance between pads of any surface mount resistor is about the same as its width, so you should include a parasitic resistance of 10TΩ across each of your two "divider" capacitors. Then run the simulation, and think about a cat rambling through dewy grass and what the level of humidity on your pcb would be.

A quick calculation gives a time constant of about half and hour.

 

"A quick calculation gives a time constant of about half and hour." That's fine I only need to measure the battery voltage every hour. I will get the same values with two parallel resistors 10T to C1 & C2.

 

Typically batteries are measured by this type of circuit, it draw zero current during idle and draw current only when MOSFET conducts which is just enough to make the voltage measurement. Your GPS will draw significant amount of the current, one minute every hour, so resistor divider is a minor problem and will consume like <1mA x 1ms.
View attachment 248304

I simulated it and it draws 1.4/124µA on RVD. Looks good.
With NP-MOSFET's 2.6/120µA. I have a battery 550mAh.

 

"A quick calculation gives a time constant of about half and hour." That's fine I only need to measure the battery voltage every hour. I will get the same values with two parallel resistors 10T to C1 & C2.

You're not following. . . after about 3 time constants (90 minutes) the voltage on the A/D will be dependent ONLY on the two leakage resistances and NOT AT ALL on the capacitor values, REGARDLESS of how often you sample it.
@Juhahoo 's circuit (post #61) and mine (Post #57) will actually work.

 

Finally I have decided to use a proposal from Juhuhu. Resistor voltage divider switched for adc measurement by P-MOSFET and NPN transistor. This solution has virtually no effect on battery life.

 

I think I found the best solution, a single 0.78 x 0.78 mm component in the DSBGA4 package. TPS22916B 5.5-V, 2-A, 60-mΩ, 10-nA leakage load switch with output discharge from Texas Instruments.

SIMetrix simulation.