Trivia for the day: minimum voltage

Occasionally forum members have asked about the minimum voltage necessary for an AVR processor to run correctly. Atmel states the m328P operating voltage is 1.8 to 5.5.

I can say with confidence that at least one processor stops function correctly at 1.6V.

Don't be a fool. Obey the datasheet. 1.8V really is the minimum.

On a related note, this thing works great. I highly recommend it. http://www.adafruit.com/products/757

[quote author=Coding Badly link=msg=2073344 date=1422954994]

On a related note, this thing works great. I highly recommend it. http://www.adafruit.com/products/757

[/quote]

If I understood right, I can power my atmega328P with 1.8 V and using that little device, I'm able to talk to either 3V3 or 5V devices via SPI or I2C?

mart256: If I understood right, I can power my atmega328P with 1.8 V and using that little device, I'm able to talk to either 3V3 or 5V ...

Exactly.

The one I have has been reliable with the low voltage side at 1.5V 1.4V 1.3560V (the batteries are draining fast) and the high voltage side at 5V. At this point it is a race between that, an ATtiny84, and a pair of AA batteries.

...devices via SPI or I2C?

Or anything else including a simple HIGH / LOW digital output. I'm using Knock-Bang which has a 3µs time quanta with a 1µs toggle and a 2µs pulse. Based on my experience I suspect the device is reliable at 1Mbps. (It probably works at much higher speeds. I just don't have the experience to have an opinion.)

Don't be a fool. Obey the datasheet. 1.8V really is the minimum.

Table 29.3 of the mega328 d/s also says, at 1.8V the "Safe Operating Area" is [u]4 Mhz maximum frequency[/u].

I should have mentioned: the processor is running at 128kHz.

(And the batteries are now down to 1.2760.)

You can always use a boost converter if you must run from a 1.5V battery. Eg. Pololu 3.3V Step-Up Voltage Regulator U1V11F3.

However tests show that for a low-power application the power used by the converter is substantially more than used by the processor in power-down sleep mode.

For example, I measured around 300 µA from the supply to the converter when the processor was asleep. Compare this to 4.5 µA when the processor is powered directly from 3.3V.

Now this is well within spec for the converter, which claims “Good efficiency at light load: <1 mA typical no-load quiescent current”. It was indeed well under 1 mA.

Another similar product JeeLabs AA Power board had somewhat better performance with the processor asleep: 36µA.

The conclusion I draw from this is to stick to straight batteries if power conservation is important, using less batteries and a converter draws a lot more current.

The [u]old[/u] bully way was to use linear regulators having very low quiescent current draw. http://jeelabs.org/2011/06/21/mcp1702-current-draw/

Also, I just checked and Microchip has them for operation down to 1.8V and below, if you really need the regulation, say for measuring analog voltages. http://www.digikey.com/product-search/en/integrated-circuits-ics/pmic-voltage-regulators-linear-ldo/2556290?k=mcp1702

1.2360V is end-of-life. The evidence indicates the batteries are what failed.

The adafruit device I have appears to be reliable down to at least 1.2360V. It is definitely worth $4!