Low quiescent current regulator or unregulated battery power?

I have a project that doesn't need rock-steady voltage, anything between 3.3V and 5V should work. It will spend 98% of its time idle. Will I get lower idle consumption if I run off unregulated battery power (3 or 4 AA NiMH cells) instead of using an MCP1702-5002?

Interesting question. Some components, like atmega, attiny, draw a little more current at higher voltages, even when sleeping. You can reduce that by regulating the voltage down to 3.3V, but then the regulator isn't perfectly efficient, either. Please do some experiments and let us know what you find.

MCP1702-5002

Isn't that a 5V regulator? 3 or 4 NiMH won't provide enough voltage to allow the regulator to output 5V. I would suggest MCP170x-33xx.

If you use a regulator, then by definition you are wasting power. It's that simple. You only need a regulator if some part of the circuit requires a strictly steady voltage.

Digital electronics specifically does not as long as the voltage remains within the limits set by the device which in the case of the ATmega328 is quite wide. The only limit of significance here is the allowable clock speed for the lowest voltage encountered.

A NiMH AA is 1.2V, yes? You can run with 4 unregulated for a 16 MHz Atmega with no problem. Measure your pack and make sure it's not exceeding 6V. I don't recall if they measure higher fresh off of a charger, has been a while since I used them, and then it was for digital cameras.
3 NiMH AAs could be problematic as the chip needs 3.8V minimum to be fully in spec at 16 MHz; less than that and things may not work properly, Serial being the first thing noticed with some chips.

Connect the battery pack to 5V on the header pin. This is bypassing the reverse polarity protection diode and the voltage regulator, be sure not to get it backwards!

A LDO is able to regulate voltage down only, the MCP1702-5002 won't help you, it is 5V variant (-3302 variant could be more interesting). For some low power applications it may be interesting to use a linear regulator to regulate the voltage down to minimal working voltage because typically devices consume less current with lower supply voltage.

Paul__B:
If you use a regulator, then by definition you are wasting power. It's that simple.

You are "wasting" power but saving current. Some of the power is used by the regulator but even more power is not used in the load - the total power consumption decreases (with the right load ofc).

Smajdalf:
Some of the power is used by the regulator but even more power is not used in the load - the total power consumption decreases (with the right load ofc).

This is the point I was making in post #1. While building low power sensor nodes, I was forced to use 3.3V LDO regulators to protect the RFM95 modules I was using, and to power the atmega328 at 3.3V also, since they are connected. Measuring the current drawn in sleep mode, I noticed the total dropped compared to my earlier tests running the atmega on unregulated battery voltage (3XAAA NiMH), even though the LDO and sleeping RFM95 were now part of the circuit.

(I use low self discharge "ready charged" NiMH cells such as Panasonic "Enerloop". Regular NiMH would be useless for this.)

Aasmund:
Will I get lower idle consumption if I run off unregulated battery power (3 or 4 AA NiMH cells) instead of using an MCP1702-5002?

So what did your measurments tell you ?

The key issue here is how much power the device uses in the 2% of the time the code is running and doing stuff, the sleep current consumption difference between a regulator circuit and non regulator circuit is just not significant.

An ATmega328P with an MCP1700 regulator will consume around 2uA in sleep. Without the regulator about 0.5uA. The running current of a bare ATmega is going to be around 5mA, so 2,500 times the sleep current.

If all the project did was sleep a set of AA batteries would last about 114years and 456years respectivly.

Given that the batteries will likely self discharge in 2 -5 years, the sleep current is insignificant until is gets to a good deal in excess of 25uA or more.

Its the running current that will drain the batteries.

PaulRB:
Isn't that a 5V regulator? 3 or 4 NiMH won't provide enough voltage to allow the regulator to output 5V. I would suggest MCP170x-33xx.

The current prototype runs on 6 AA cells (IKEA NiMH LSD) using the regulator on the Arduino board, and needs new batteries every three-four days. So it runs on 5V, that's why i was thinking in terms of 5V. If I go for regulated power I will probably test both 3.3V and 5V.

srnet:
So what did your measurments tell you ?

I haven't made any yet, I wanted to ask for advice before I embarked on experiments. No need to reinvent a square wheel :slight_smile:

My hunch is that unregulated or low-Iq regulator probably doesn't make much difference in practice, I'll be very happy if I get a month runtime instead of four days.

srnet:
Given that the batteries will likely self discharge in 2 -5 years, the sleep current is insignificant until is gets to a good deal in excess of 25uA or more.

Its the running current that will drain the batteries.

Sure. The regulator is just one of the changes I will need to make to actually get power savings and a proper "idle/sleep" mode. The current prototype is essentially always "on". Other planned changes to the model:

  • Replace Arduino with bare Atmega328, possibly at 8 MHz
  • Switch off power to PCA9685 module and LEDs when not doing anything (relay or NPN/mosfet, further reading and experiments needed)

Maximum current when active is around 80 mA at 5V and 45 mA at 3.3V.

Very interesting math about the battery self-discharge and its implications for Iq, that's one of the things that didn't cross my mind.

Anyway, thanks all. This is very useful info.

Switch off power to PCA9685 module and LEDs when not doing anything (relay or NPN/mosfet, further reading and experiments needed)

Shouldn't be needed. The pca chip has a sleep/standby mode which only draws a couple of uA. You may need to control it's output enable pin with an arduino pin.

Neat! I'm using a PCA9685 module breakout board (SparkFun or Adafruit clone). It doesn't look like it has any power-consuming extras, except for a led or two. Those should be simple enough to disable.

Edit July 2021:
I'm calling this a success. After the following modifications:

  • Replace Arduino clone with Atmega328p running at 8 MHz, internal oscillator
  • Remove power led from PCA9685 breakout board
  • Rewire LEDs so PCA9685 is current source rather than sink
  • 3 NiMH batteries as unregulated power source, nominally 3.6 V but higher when batteries are freshly charged
  • sleep/powerdown PCA9685 and Atmega328 whenever possible

the project has now been running for five months on the same battery pack. Previously I had to replace the batteries every four days.

Rewiring from sink to source was necessary as otherwise the chip would open fully and all LEDs would light up when the chip went to sleep mode. LEDs are somewhat dimmer since max voltage is lower, but it does not matter for this project.
I tested 1MHz too, but that was too slow and caused visible stepping when brightening/dimming LEDs.