LiSOCl2 battery EOL monitor review

Question

I'm working with a device which will use Lithium thionyl chloride (non-rechargeable) batteries as a backup source; their discharge curve is quite flat and AFAIK you can't tell how is the battery in terms of whether it's good or about to die (EOL) unless you apply a dummy load. There are other more complex methods but my application requires me to save every penny and the client just wants to know if the battery is good or dead. Now the circuit is working but I would like to make sure If I'm actually doing it correctly or outright far from the right path.

My concerns:

1. If this method of applying a dummy resistor when needed and measuring from the ADC at that time a legit method?
2. If I should be concerned that the voltage divider and the ADC might eventually drain the battery, I added a MOSFET to cut off or connect the path between the battery and the voltage divider and ADC.

Answer 1

That solution is good. A couple of points.

1. When on backup the P-FET will be biased on slightly by the voltage drop across D27. The threshold voltage of Q5 needs to be significantly higher than that to ensure it can be switched off properly.

2. When powered normally the VDDSYS should be no lower than the voltage when on backup or again Q5 may turn on.

3. I would add a capacitor across R90. The input of the ADC of the MCU will probably be equivalent to a capacitor of about 20pF charged to zero volts when the voltage is sampled. This has to charge through R190 in parallel with R106. The voltage will drop for a short time when sampling. There is normally a maximum source resistance of 5-10k ohms defined in the MCU data sheet. You are just within acceptable range but capacitor of a 1000pF or so will speed up the sampling and also provide a bit of filtering that may be desirable. Make sure the software does not attempt to read the voltage too quickly after Q5 is enabled.

4. If you have a GPIO available it would be useful to provide separate control for Q5 and the load resistor R92. You can then read the voltage with a light load and a heavy load to better determine the battery condition.

Edit:

The recommendations in the answer from Jonathan S. are also worth including. If the ability to check the battery at light load as well as heavy load is required T5 can be left in the circuit and R56 connected to the battery positive instead of VDDSYS. That will reduce any sensitivity to Q5 threshold voltage.

Answer 2

You can actually save quite a few more components in the circuit while simultaneously improving it: Remove T5, R56 and R58. Then connect the gate of Q5 directly to the collector of T7. You already have signal there that's perfectly suitable for driving Q5's gate. Additionally, this avoids weird interactions between the level of VDDSYS and Q5's threshold voltage, making the circuit more robust.

You should also decrease the value of R95 - you're overestimating T7's beta in saturation. A value of 2k2 would be more appropriate. This then asks T7 to deliver a current gain of 20, which is realistic for a saturated small-signal NPN transistor.

Other than that, the circuit is perfectly fine, including your choice of transistor for Q5 (DMP1555).

Including Q5 to disconnect the voltage divider is also very good idea.