Battery life of the LG X110 netbook

Introduction to the LG X110 netbook

The X110 netbook from LG is what I purchased to advance my many computer-related projects during vacations, and I consider it a good buy. What I like about it:

• Its keys are not far from the size of a normal desktop (17.5 mm against 19 mm), and the keyboard does not bend when typing hard, as on the ASUS Eee PC 1000 series. It is also slightly more roomy than that of the MSI wind, of which the X110 is supposedly a clone, in that the cursor keys are lowered a little. The keyboard quality is especially important for me, as I work in a terminal most of the time.
• Almost everything worked out of the box with Linux (Ubuntu 9.04, which a guy at the shop installed for me).
• The touchpad's right side works as a scroll wheel, which reportedly is not supported by the MSI's touchpad.
• The display is rather bright when you turn it up; indoors the lowest or next to lowest brightness setting is sufficient, which allows you to save battery life (see next section).

Battery lifetimes under different conditions

My X110 has the small battery, with a nominal 2200 mAh. The manufacturer claims a battery life of 2 hours 20 minutes with that battery. Curious as I am, I wanted to know

1. whether this is realistic,
2. how much it varies depending on what factors,
3. and what you can do to minimise power consumption when you are far from external power.

To measure the battery lifetime, I let the netbook run on battery until it was nearly empty and extrapolated to get the time it would have taken to run empty completely (see below for details). I did this for several scenarios:

Minimum load

Display on standby (xset dpms force standby), no CPU load, no external devices connected

Moderate load

Display set to minimum brightness, xosview running (which takes about 10% CPU on this bitty box), no external devices connected

CPU 100%

Running the raytracer POVRay in an infinite loop without displaying the result, display on standby, no external devices connected

Brightness 100%

Display set to maximum brightness, CPU idle, no external devices connected

Maximum load

Compiling the Qt GUI toolkit after clearing disk cache (giving the hard disk some exercise), display set to maximum brightness, network cable and external mouse connected

Here are the results:

Going by the data, turning the display brightness up takes even more additional power than number crunching. So a good way to prolong your battery life is to keep brightness down – sit in the shadow, not in the sun when using it outdoors, and take care to look at the display perpendicularly, which gives you maximum brightness; set a low default and turn it up only when needed, not the other way round.

The CPU and the display between them seem to account for most of the power consumption of a fully loaded system. When one takes the difference between minimum consumption and 100% CPU and brightness respectively and adds these three components up, the result is 1407 mA, only about 100 mA less than the maximum load result. So the additional power required for increased hard disk activity and an external mouse and network is comparatively low.

What I did not test: None of the scenarios included an external monitor, audio output, and USB devices other than a small laptop mouse. These ports may consume additional power when active. A USB device may consume up to 100 mA at 5 V according to the USB specification (see Wikipedia), which amounts to 45.5 mA for a 11.1 V (nominal) battery (which the average currents above imply).

When the X110 is switched off

Considering you may use your netbook primarily when travelling (I do), how best to store it between trips deserves some mention. Litium ion batteries should best be stored half empty and must not be discharged completely (see Wikipedia). So it is of some interest how fast the battery discharges on its own.

Here I used a simpler approach than for the battery lifetimes of the running netbook. I just left the X110 switched off for some time and checked the remaining charge after the next reboot. The average discharging currents computed from the results were:

These numbers seem low compared to the discharge current when the netbook is running, and indeed they are. But the first row of the table implies that the netbook's battery will be completely discharged after one week unless you remove the battery. The second row says that the battery will discharge on its own within 1½ years. (I had put up a much larger average current earlier, but it was probably due to measuring only a short period after the battery had been fully loaded, when it presumably discharges fastest.) So, to prevent degradation of your battery, you should remove it from the netbook. (If you have the space to store it in your fridge, better still, for at fridge temperatures lithium ion batteryies degrade most slowly.)

Battery charging times

The time it takes to charge the battery fully seems to be independent of the system load; presumably the current from the AC/DC converter increases as needed. Charging increases the reported battery fill level linearly most of the time, only when the battery is very nearly full, the charge current is reduced, and the charging process is completed more slowly, almost asymptotically.

I measured this in the same way as the discharging when the netbook was running, by logging the battery fill level over time. The total charging time including the asymptotic part was 2:28:05, the end of the linear charging (which makes the battery almost full) was at 2 hours, and the interpolated end of the charging process at the linear charging current (974 mA) was at 2:12:03. Translation: if you charge the battery for a good 2 hours, you're good. Only if you want to squeeze the very last electron inside before your trip to the jungle, you have to charge it for 2 1/2 hours.

How I measured these times

As inspiration for those who want to measure their battery lifetime themselves, here's how I did it. All the following requires that you are running Linux; as usual non-UNIX operating systems are unsuitable for anything other than flashing pretty pictures at stupids.

First, I opened a log of the terminal session with the script command to record the output of the following commands. Then I started the following infinite loop:

while true ; do date ; cat /proc/acpi/battery/BAT1/state ; sleep 60 ; done

The commands inside the loop print out the current time and the state (including charge level) of the battery, respectively. The latter uses Linux's all-you-can-eat kernel information shop, the proc file system. After the first output, I disconnected the external power supply. The loop causes the battery state to be printed every minute, allowing to track the charge level. (In those scenarios where the display was blanked, I displayed the screen briefly at interals.)

I then waited until the battery was nearly empty (around 50 mAh), reconnected the power supply, interrupted the loop (Ctrl-C) and terminated the session log. The log now contained times and battery state information at approximately minute intervals. I wrote a small Perl script (download) to convert the log to a tab-separated value table with the times in seconds and shifted so that the first time was zero. (The script uses the Date::Parse module to extract the time; I could have made this easier by using date +%s instead of just date.) I then performed a linear fit to those data using gnuplot. The abscissa (x-axis) offset is the total battery lifetime in seconds, and the slope is the (negative) discarge current in the unit mAh/s = 3.6 A. It did not work for me to fit the x-axis offset directly; this turned out to be numerically ill-behaved, so I computed it from the y-axis offset and the slope.