Raspberry Pi Zero power consumption

[geerlingguy]

I just did some testing, since I'm mostly interested in using the Pi Zero in applications where minimal power draw is essential (especially for battery-powered scenarios): http://www.midwesternmac.com/blogs/jeff ... zero-power

The gist of it:

• A+ idle - ~200 mA
• B+ idle - ~240 mA
• model 2 B idle - ~420 mA
• Zero idle - ~65 mA

This is the undisputed power sipping king when it comes to Raspberry Pis (or any clone boards that I've used). My Arduinos still halve that statistic to ~30 mA (even when running code), so they're even better for embedded applications, but this Pi will open up some new opportunities since it's so power-conscious!

[Update: s/mW/mA]

[solar3000]

The next pi should be one that can be powered by a button cell and runs at 8 MHz.

[geerlingguy]

An update: I have been able to hit ~30 mA at idle by doing the following:

• Not have anything plugged into USB
• Run Raspbian Lite (official Jessie image)
• Disable HDMI (/usr/bin/tvservice -o, and put that in /etc/rc.local to make permanent)
• Disable the ACT LED (add lines from this post (http://www.midwesternmac.com/blogs/jeff ... spberry-pi) in your /boot/config.txt and reboot)

See this post for more details: http://www.midwesternmac.com/blogs/jeff ... rve-energy

[solar3000]

I booted my zero with a 32GB card. Plugged in are only composite and keyboard. Also GUI running by default.
I measured 80mA to 140mA.

You disabled everything. So what would you do with it in that state?

[geerlingguy]

It would work as a data logger, or could do a lot of small activities (not networked) via GPIO, a la Arduino. Granted, there are many other reasons a full Linux PC might not be suitable in those applications, but at least power is not going to be the main blocker for me anymore!

For most of my projects, though, networking is a requirement, so I'll be firmly in the 80-120 mA range (using USB WiFi and nothing else).

[solar3000]

My arduino GEMMA2 uses 9mA on full blast.

[rpdom]

My arduino GEMMA2 uses 9mA on full blast.

and how well does it play full HD video?

[alexeames]

• A+ idle - ~200 mW
• B+ idle - ~240 mW
• model 2 B idle - ~420 mW
• Zero idle - ~65 mW

My A+ and Zero idling measurements were exactly the same. 100 mA for both (@ 5.09V) I notice you used a very cheap USB charge doctor for your measurements. I wonder if it's a little "wonky" at the low end? I used a calibrated shunt.

(edit to add that my measurements were all taken with USB keyboard/mouse dongle and HDMI attached and in use)

[mikerr]

I've not managed to go below 66ma even with:

Pi Zero
5.00v via GPIO
Jessie-lite
no USB devices connected
tvservice off
power/ACT LED off

measured from input of the bench power supply pictured:


More usefully though I can get it down to 69ma with wifi dongle connected by sending usb to sleep for as long as needed:

Code: Select all

echo 0 | sudo tee /sys/devices/platform/soc/20980000.usb/buspower >/dev/null
sleep 10
echo 1 | sudo tee /sys/devices/platform/soc/20980000.usb/buspower >/dev/null

with wifi back on and active its 165ma (edimax nano)

[daveake]

I concur with Mikerr ... I get:

Pi Zero with no peripherals (no USB, no monitor), serial terminal only. After boot: 87.9mA

HDMI Off: 70.7mA

HDMI and LED Off: 68.1mA

All above measured on a Fluke 179 with 5V applied via GPIO pins.

I trust the Fluke above the USB tester.

Dave

[alexeames]

I trust the Fluke above the USB tester.

And that's no random accident or unexpected event

[solar3000]

My arduino GEMMA2 uses 9mA on full blast.

and how well does it play full HD video?

At one frame every six minutes probably.
I was messing with 'geerlingguy'
He was talking about low power. How about my digital watch. I haven't change the original battery for ten years. So how many pico amps would that be?

[daveake]

Another way of reducing power consumption is to reduce the voltage that the Pi Zero runs from. With HDMI and the LED off, and voltage measured at the Pi GPIO pins (so after the series current meter) I get:

5V: 68.2mA = 341mW

3.3V: 85.9mA = 283mW

which is a saving of 17%. Woo :p

Remember that if you have a USB peripheral attached, it may need the full 5V; my experience though is that many devices (including all the WiFi adapters I've tried) are happy on 3.3V.

Also, I've been running tests to see how low a voltage supply it will run from. This is useful partly because lower voltages reduce the power consumption a little, but mainly because of the possibility of running the Pi Zero directly from 3 AA cells.

To test, I connected a bench PSU to the 5V/0V GPIO pins. Rather than rely on the PSU voltage (and some rather thin test leads) I connected a Fluke 179 to the same pins to monitor the voltage that the Pi sees.

The lowest PSU voltage that the my Pi Zero reliably boots from is 2.9V, at which level the Fluke reports min/max figures (during the entire boot phase up to login prompt) of 2.68 - 2.81V. The drop from the PSU voltage is down to resistance in the leads from PSU to Pi.

If I drop the PSU voltage to 2.8V then the Zero boots about 80% of the time. Replacing the leads with shorter/thicker ones may well get it to boot more reliably, but this does indicate roughly where the limit is. And remember that it's not wise to run close to a limit.

So the above isn't a statement that "the Pi Zero works at 2.9V", as the actual limit will vary with temperature, from device to device, with the resistance in the leads and internal resistance of the power source, but it does show that a supply of much less than 5V can work. Which is good if you want to run it from 3 AAs which are pretty much dead at 0.9V per cell anyway .

Also remember that no peripherals were connected, so HDMI/USB may well stop working before the supply gets down to 2.9V.

Dave

[geerlingguy]

I've not managed to go below 66ma

[...]

More usefully though I can get it down to 69ma with wifi dongle connected by sending usb to sleep for as long as needed.

Thanks for posting your results (and to others too!); I did use a chintzy little USB 'Charger Doctor', and yes, it's highly likely to be inaccurate at very low readings (just like my Kill-a-Watt meter is highly unreliable below about 50W).

However, it has reliably reproduced a range of 20-40 mA for idle with LED off and TV off, and 40-70 mA for LED on and TV on, so at least relatively speaking, it's been accurate. I'm receiving a slightly nicer (still nothing near Fluke-grade) USB power meter in the mail tomorrow, and will re-test and see how it compares. It's also reliably reported the Zero to consume 30-50% less than my two Pi A+'s, which were the previous low-power champs.

If I get a chance, I'll also bring the Zero to a friend's workshop, where he has a few different highly accurate meters, and we'll do more testing.

[geerlingguy]

My arduino GEMMA2 uses 9mA on full blast.

and how well does it play full HD video?

At one frame every six minutes probably.
I was messing with 'geerlingguy'
He was talking about low power. How about my digital watch. I haven't change the original battery for ten years. So how many pico amps would that be?

You just needed to end with /s

[geerlingguy]

• A+ idle - ~200 mA
• B+ idle - ~240 mA
• model 2 B idle - ~420 mA
• Zero idle - ~65 mA

My A+ and Zero idling measurements were exactly the same. 100 mA for both (@ 5.09V) I notice you used a very cheap USB charge doctor for your measurements. I wonder if it's a little "wonky" at the low end? I used a calibrated shunt.

(edit to add that my measurements were all taken with USB keyboard/mouse dongle and HDMI attached and in use)

My measurements were taking headless with nothing plugged in, so that could definitely make the ranges a lot narrower; when I plug in HDMI, keyboard, wireless trackpad, and WiFi to the Zero, it jumps up to ~500-600 mA!

[daveake]

I've not managed to go below 66ma
... still nothing near Fluke-grade

. Any cheap DMM will be an order of magnitude more accurate and precise than the USB tester; you just need to get it inline between power source and Pi. I only mention the Fluke because, after 30 years of cheaper meters, I finally bought one last month

Another factor is that the 5V going into a meter will be somewhat less than 5V afterwards. This will cause the Pi to use a bit more current. In my tests I measured the voltage at the Pi, and turned the PSU up a bit till it read 5V there.

Dave

[geerlingguy]

I've not managed to go below 66ma
... still nothing near Fluke-grade

. Any cheap DMM will be an order of magnitude more accurate and precise than the USB tester; you just need to get it inline between power source and Pi. I only mention the Fluke because, after 30 years of cheaper meters, I finally bought one last month

I just need to bite the bullet and get one... as a Christmas gift to myself

I have a decent-ish analog multimeter I built myself from a kit eons ago, but that's about it.

[rpdom]

My arduino GEMMA2 uses 9mA on full blast.

and how well does it play full HD video?

At one frame every six minutes probably.
I was messing with 'geerlingguy'
He was talking about low power. How about my digital watch. I haven't change the original battery for ten years. So how many pico amps would that be?

That depends on which battery it takes. Mine uses a CR1620 cell, which has a capacity of around 75mAh. If it lasts 10 years then the average discharge rate would be 0.075(Amps) / 87660(hours) or 0.000000856A which is 0.000856 mA, 0.856 uA, 856 nA or 856000 pico Amps. (If I got my maths anywhere near right)

[FTrevorGowen]

[Slightly O.T.] W.r.t. to "Charger Doctor" current measurements - FWIW I measured the behaviour of three samples cf. an Adafruit current meter over a nominal current range of 9 to 750 mA when the devices first became available. The details are within my webpages here:
http://www.cpmspectrepi.webspace.virgin ... octor.html
Trev.

[alexeames]

[Slightly O.T.] W.r.t. to "Charger Doctor" current measurements - FWIW I measured the behaviour of three samples cf. an Adafruit current meter over a nominal current range of 9 to 750 mA when the devices first became available. The details are within my webpages here:
http://www.cpmspectrepi.webspace.virgin ... octor.html
Trev.

So if I read your pictures right Trevor, there was ~30mA range between the three when measuring ~200mA?

[FTrevorGowen]

[Slightly O.T.] W.r.t. to "Charger Doctor" current measurements - FWIW I measured the behaviour of three samples cf. an Adafruit current meter over a nominal current range of 9 to 750 mA when the devices first became available. The details are within my webpages here:
http://www.cpmspectrepi.webspace.virgin ... octor.html
Trev.

So if I read your pictures right Trevor, there was ~30mA range between the three when measuring ~200mA?

That's not quite the "whole story". Having observed that the C.D.'s were more linear than the (Hall effect based?) Adafruit meter I was checking for variations in the linear slope and offsets. Given that the device's "quantization error" is ~10mA (cf. ~10mV/Rload for the DMM used to measure the voltage across the resistive loads use to determine the "reference currents", Ise3) two of the devices exhibited an offset consistent with that error margin (-10 and +11 mA). The other device's offset was much greater, ie. +38mA. Ironically that device's slope was somewhat closer to 1.0 (0.96 cf. 0.94). However, compared to the Adafruit current meter's values of -120mA and 1.55, the C.D.'s values were all significantly "better".
Trev.

[solar3000]

...........
which battery it takes. Mine uses a CR1620 cell, which has a capacity of around 75mAh. If it lasts 10 years then the average discharge rate would be 0.075(Amps) / 87660(hours) or 0.000000856A which is 0.000856 mA, 0.856 uA, 856 nA or 856000 pico Amps. (If I got my maths anywhere near right)

its a casio databank/calculator. I don't remember if I ever opened it.
Can't be. Its got to be less than that. At full charge it is over 3V. So the watch should start complaining at below 2V. So it probably uses a few micro amps. I mean after more than ten years.
I just checked. I purchased the watch Oct 2003. So that's 12 years. The backlight still shines. So it should still be just below 3V.
Its got to be above 1/2 or 3/4 capacity.

[rpdom]

...........
which battery it takes. Mine uses a CR1620 cell, which has a capacity of around 75mAh. If it lasts 10 years then the average discharge rate would be 0.075(Amps) / 87660(hours) = 856000 pico Amps. (If I got my maths anywhere near right)

its a casio databank/calculator. I don't remember if I ever opened it.
Can't be. Its got to be less than that. At full charge it is over 3V. So the watch should start complaining at below 2V. So it probably uses a few micro amps. I mean after more than ten years.
I just checked. I purchased the watch Oct 2003. So that's 12 years. The backlight still shines. So it should still be just below 3V.
Its got to be above 1/2 or 3/4 capacity.

I expect my watch uses more power than yours. I bought it in 2005 and I've had to change the battery a couple of times (ok, the last one was a cheap battery and only lasted a couple of years). The radio time receiver on it probably uses a fair bit of power. That runs three or four times every morning for around 5-10 minutes.

[geerlingguy]

So, just to give an update (and a mea culpa): I retested everything with the PowerJive USB meter that just arrived today (it's much nicer in every way than the Charger Doctor), and now I'm getting results that jive (see what I did there?) with the others posted in this thread:

USB keyboard, trackpad, WiFi, and monitor plugged in: 310 mA
Nothing plugged in: 100 mA
Disable LED and HDMI: 80 mA

Unfortunately I can't get any better resolution than +/- 10 mA with either of these meters, but it looks like my Charger Doctor is going to go the way of the dodo... it's junk under a few hundred mA!