Low power circuit based on TPL5111

[AKstudios]

I'm designing a circuit based on what I've learned from some Adafruit products, general stuff in the forums and just tinkering around on my own. The idea is to make a low power circuit based on the TPL5111 timer. I'm not using the TPL5110 to completely power the circuit off because the initial boot time and configuration time for all components in my circuit would have a noticeable effect on battery life.

Edit: image removed.

The general idea is that the circuit should run on battery power only until the timer wakes the MCU up. Let's say the battery source is either a 3V lithium battery or two AA's in series, producing ~3V.

The TPL5111 would periodically (let's say every 5 mins) send a signal over the DRVN pin to the MCU. The MCU will be programmed to wake up on external interrupts, so it would be in sleep mode until it receives this signal. Once the signal is received, the MCU then sends a LOW signal over its GPIO pin, which would open the P-MOSFET (Q3) and enable the SHDN pin on the step up DC-DC converter. This lets 3.3V flow through, isolating the battery via the P-MOSFET (Q2) and letting the MCU (and all sensors which were powered down until now) run on 3.3V instead. Now the circuit can do what it needs to do, let's say, take a temperature and humidity reading. Once the readings are done, the sensors are switched off, the MCU sends a HIGH over the GPIO pin so the SHDN pin is disabled, which disables the step up via Q3. This also enables the P-MOSFET (Q2) and the MCU is now powered only by the battery. Ideally, this setup would limit current draw to only the TPL5111 and the MCU in deep sleep mode, which I expect to be around 0.2uA.

The other part of the circuit is - when USB power is present, the battery is completely isolated via Q1, the step up is disabled via Q3 and the circuit is now powered by a LDO voltage regulator that supplies 3.3V to everything. There may be a better way to do this, but this is what I came up with so far.

Ultimately, the questions I have about this are:
1.) Since diode D2 is there for reverse current protection from VBUS into the GPIO, I'm concerned that when the MCU is running on VBAT only, let's assume that's 3V, the voltage drop across the diode would drop it down close to 2.5V and that's getting close to Vgs of the P-MOSFET (at least, the ones that I have). I'm not sure if the FET would fully open at that voltage. I'm assuming the solution to this would be to use a FET with a lower Vgs.

2.) Would the 100K pull-down (R2) consume significant current? If we assume 3V from VBAT, that's around 30uA just burning through the resistor and that kinda defeats the purpose of using the timer in the first place. Not sure if there's another way to make sure the gate is pulled down to 0V without consuming current.

3.) When VBUS is not present, the step-up provides 3.3V. Not sure how the LDO would react to having 3.3V on its output pin, when there is 0V on input. I suppose it depends on the LDO in question, so let's say it's an AP2112-3.3V. Would it be better to switch the ENABLE pin on the LDO using a MOSFET as well and that makes the output line okay?

4.) Would this circuit even work? Just need some advice before I spend money on buying the parts and such.

[adafruit_support_mike]

D2 is backwards if you want Q3 to be a P-mosfet. It will pull the gate high, but won't pull it low. That isn't hard to fix though.. just move the gate connection to the other side of D2 (directly connected to the MCU). That will also solve the loss of Vgs.

The 100k pull-down on Q2 would waste current.. assuming the microcontroller can idle at 100uA, the pull-down current would be 1/3 of what the MCU is using.. and the P-mosfet doesn't care about the resistor's value all that much. 10M should work, and will take the leakage current down to 300nA. You can put a 10M pull-up on Q3's gate too, to hold it off until the MCU pulls it low.

The pull-down on Q2's gate shouldn't connect to the 3.3V rail.. that will just hold Vgs at the forward voltage of the internal protection diode, but never allow the channel to turn on. Just make it a straight shot to GND through the pull-down.

For the LDO, connect a diode from the output back to the input. The diode will be reverse-biased when VBUS is higher than 3.3V (when it's actually regulating the voltage), but the diode will keep the Vin from dropping to 0v when there's no 5V supply. Any current flowing backwards through the regulator would be going through the equivalent of a forward-biased diode, so putting an external one in parallel routes current around the LDO instead of through it. For extra protection, use a Schottky diode there, since the lower forward voltage will keep the LDO's current path from reaching a voltage where it can become forward biased.

Overall the circuit looks good though. Your reasoning is sound, and the incorrect points in the schematic above are basically typos. I go back over each node in a schematic (the idea 0-ohm connections between components) with colored markers, and write a quick description of how each one should behave to catch the places where I've crossed wires.

[AKstudios]

Mike -- thanks for the reply.

move the gate connection to the other side of D2 (directly connected to the MCU).

the gate should also be pulled high when VBUS is available. Should I make the connection directly to MCU and to VBUS on the other side of D2?

the P-mosfet doesn't care about the resistor's value all that much. 10M should work, and will take the leakage current down to 300nA.

That's better, can we go higher to save more current, like 50M or 100M? Or would those be extremely weak pull downs?

For the LDO, connect a diode from the output back to the input. The diode will be reverse-biased when VBUS is higher than 3.3V (when it's actually regulating the voltage), but the diode will keep the Vin from dropping to 0v when there's no 5V supply. Any current flowing backwards through the regulator would be going through the equivalent of a forward-biased diode, so putting an external one in parallel routes current around the LDO instead of through it. For extra protection, use a Schottky diode there, since the lower forward voltage will keep the LDO's current path from reaching a voltage where it can become forward biased.

This makes sense, but wouldn't the LDO be powered through whatever is running from output to the input when there is 3.3V available at output? It would be better to have the LDO run only when VBUS is available and turn off when battery power is available to save the most current.

[adafruit_support_mike]

the gate should also be pulled high when VBUS is available. Should I make the connection directly to MCU and to VBUS on the other side of D2?

A 10M pull-up between the gate and VBAT will keep the mosfet shut off whether there's power to VBAT or not. It's also kind of a philosophical question whether there's any difference between Q3 being open or closed when there's nothing on the VBAT side.

That's better, can we go higher to save more current, like 50M or 100M? Or would those be extremely weak pull downs?

Anything above 10M starts to get iffy. That's about where parasitic leakage current through the components and across the surface of the PCB starts to be noticeable.

This makes sense, but wouldn't the LDO be powered through whatever is running from output to the input when there is 3.3V available at output? It would be better to have the LDO run only when VBUS is available and turn off when battery power is available to save the most current.

There will be a diode drop worth of voltage loss from the output back to the input. The LDO can't do anything if its input voltage is lower than its output voltage.

[AKstudios]

Thanks for all the suggestions. Here's the updated diagram:

Edit: image removed.

The 3.3V on the MCU is actually 'VCC' and switches between VBAT and 3.3V, depending on the source available.

The only thing I'm unsure about is how to close Q2 when 3.3V is present. When the step up is shut down, 3.3V is not present, so the gate is pulled down through R2, letting VBAT flow through and powering the MCU. But when the step is enabled, 3.3V flows through to power MCU but Q2 should should close, isolating VBAT from powering the MCU, which is its main purpose.

[adafruit_support_mike]

Is Q2 strictly necessary? Could you connect the LDO's VBUS line to VIN on the TPL5110?

[AKstudios]

I'm using the TPL5111 as an external watch dog timer for the MCU, so it needs to run on VBAT mostly.

My intention for Q2 and Q3 in the initial circuit is to run the MCU directly from battery when it's in deep sleep mode, bypassing any step up or voltage regulation when not needed to save the most power. Q3 turns on, letting the step up provide 3.3V when measurements need to be taken, and turns off when everything is in deep sleep mode.

[adafruit_support_mike]

It sounds like you have some ambiguity in the signals labeled 3.3V.

In the schematic above, the LDO will power every point labeled 3.3V if VBUS has positive voltage. If you want a circuit that uses VBUS when it's available and falls back to VBAT when VBUS has no power, you've already built it with Q1, D1, and R1.

[AKstudios]

Yes, Q1, D1 and R1 switch between VBUS and VBAT. That part is fine. Im specifically talking about when VBUS is not available.

I want the MCU to switch between VBAT and 3.3V. Not sure if I drew that part correctly, but I see the ambiguity, that is my bad.

The 3.3V on the MCU should actually be 'VCC', which should switch between current directly from the battery when asleep, and to 3.3V from LTC3525 stepup when needed. That's why I tried to add Q2.

[AKstudios]

I think there is too much ambiguity in my diagram and contents of this thread; I need to rethink this setup through.

Having said that, I don't think leaving this post here would help others. I request a mod to delete it.

Thanks!

[adafruit_support_mike]

No worries.. sometimes you develop an idea until it's clear enough for you to see a problem that was hidden before. Most interesting projects run into that at least once.

This is a good example of that happening, and is a valuable reference for that alone. The individual pieces of the circuit we discussed are also useful. I'd suggest keeping it as-is.