power consumption from 8MHz crystal oscillator

Hello, I've read conflicting information. I'm trying to make my own battery powered Arduino (really just ATMEGA328P-PU). According to Nick Gammon's very detailed analysis, he saved about 4mA by going from external crystal to internal clock, or from 15.15mA to 11.05mA. If you look at his page, it's under the heading "Using the internal clock".

http://www.gammon.com.au/forum/?id=11497

Does that mean that just having an external oscillator consumes 4mA? That seems a lot! Other people say that the external oscillators don't consume that much power, and JeeLab's very power efficient design, which is reported to run for over 6 months, uses an external resonator. I didn't think resonators were that much different from oscillators in terms of power consumption.

Am I misunderstanding Gammon's results? At 4mA, a 2000mAh battery at 3.3V would last less than 1 month.

Looking at that article the 4mA difference was not necessarily caused by the external crystal, the internal clock is 8MHz but the external was probably 16Mhz I.e the normal clock for 5V operation is 16MHz external.

Reducing the clock speed to very low e.g 100khz etc reduces the power usage even more.

Note using the internal clock, doesn't give very accurate time keeping as its frequency can vary with enviromental factors like temperature

A different approach

rather than trying to save 2000 mAh per month by removing a 4mA crystal, how about getting a < £10 silicon solar panel,
setting up one analog input pin to monitor your battery voltage,
and one digital out pin to the gate of fet between solar and battery?

that arrangement, with a 9V rated 180mA small solar panel delivering >20mA for battery topup during daylight might work perpetually.

clever idea, it didn't cross my mind. But for where it's going...might not be very good. People might steal it :)

arusr: Does that mean that just having an external oscillator consumes 4mA? That seems a lot!

No, it means he went from 16MHz to 8Mhz.

The datasheet has graphs of power consumption vs. frequency.

arusr: Hello, I've read conflicting information. I'm trying to make my own battery powered Arduino (really just ATMEGA328P-PU). According to Nick Gammon's very detailed analysis, he saved about 4mA by going from external crystal to internal clock, or from 15.15mA to 11.05mA. If you look at his page, it's under the heading "Using the internal clock".

http://www.gammon.com.au/forum/?id=11497

Does that mean that just having an external oscillator consumes 4mA? That seems a lot! Other people say that the external oscillators don't consume that much power, and JeeLab's very power efficient design, which is reported to run for over 6 months, uses an external resonator. I didn't think resonators were that much different from oscillators in terms of power consumption.

Am I misunderstanding Gammon's results? At 4mA, a 2000mAh battery at 3.3V would last less than 1 month.

For what it's worth, I measured the current draw on a bare Arduino UNO R3 and a MEGA2560 R3 with the oscillator set to both "Low power" and "Full swing". Surprisingly, the "Full swing" setting draws about 1.5 milliamps LESS than "low power".

So, of course, I run mine at "Full swing" now.

Interesting - normally the lower voltage swing on the external circuit would reduce power consumption.

In general external high speed clocks are expensive in power, since the capacitances on the PCB have to be driven (on the order of 20pF for quartz crystal load cap), whereas on-chip capacitances are measured in fF (femtofarads). Ceramic resonators may have much less capacitance perhaps?

The theoretical power consumption of a CMOS chip driving a capacitance C at frequency f is f * C * V^2 where V is the supply voltage. Thus lower supply voltage and clock frequency will reduce power consumption.

BTW current consumption is f * C * V, or 1.6mA for 16MHz into 20pF at 5V

Be careful about the terms “external crystal” and “external oscillator”, because they actually have quite different meanings.

True, I meant the on-chip oscillator for an external crystal, not an external oscillator unit fed into XTAL1 pin.