[ Mac Tips ] [ Next Tip ]

On September 20, 2003 I received a new 15" PowerBook from Apple. This is the new model with a 1.25 GHz G4 CPU and a Superdrive. However, the battery seems pitifully small, rated at 46 watts. This is smaller than the most recent TiBooks (61 watts) and about the same size as an iBook battery (45 watts). I originally figured that I was getting about 2 hours battery life in a fairly conservative operating mode as opposed to the 4.5 hours claimed by Apple. Actually I never expected to get 4 hours, but less than 2 is not acceptable. During a call to AppleCare, the technician claimed that a battery life of between 2 to 5 hours is in spec, the advertising notwithstanding.

The whole driving force for the work I did on this page was driven by the fact that the battery, IMHO, is undersized for this particular computer. Therefore I needed to know how to squeeze as much out of the battery as possible. It is one of the more power hungry models, yet it has one of the smallest batteries. This was probably done in an effort by Apple to cram all that stuff into a 1" thick case. I would have preferred a little thicker computer and a MUCH bigger battery.

It is possible that a 1 GHz PowerBook running at Reduced performance, with minimum RAM installed, no Bluetooth or Airport, minimal screen backlighting, no keyboard backlighting and doing very light duty work (so that the graphic processor has nothing to do) MAY touch 4 hours. However, this isn't anything close to a real world case. Remember that "truth in advertising" is one of the world greatest oxymorons.

Update. In Dec 2005, I ran a different series of power consumption tests and I actually did get 4 hours and 17 minutes from this PowerBook, but only in a condition where the computer was not very useable, see Apple Laptop Power Consumption Tests as well as the information in this page.

A few days after the Powerbook arrived OS 10.2.8 came out and I upgraded immediately because there was a claim that an SSH security bug had been patched. After the upgrade, I noticed that the indicated battery life had decreased from 2 hours to about 1.5 hours. However, the actual run time of the computer seemed to have become a little better. This got me to wondering what was going on.

I used a freeware program called X-Charge to monitor actual battery run time in some tests and found that the battery life predictor in the OS is wrong, at least for this particular computer. It consistently underestimates the time remaining by about 1/3 all through a linear discharge. Computers are normally pretty good at doing linear regressions, but this part of the OS isn't doing so well in my case. Others have reported that the lifetime estimator is fairly accurate, some getting 2+ hours of estimated and real battery life, some much less.

Another shareware program, XBattery, reports the current drawn while running from the battery along with battery voltage. I imagine that the CPU inside the battery that keeps track of battery charge state is providing this reporting and the program just plots the result. In any event, this current meter can be used to judge the draw of various modes of operation. It reports current in 1 mA increments, but it also drifts around by 10 to 50 mA for no apparent reason so many readings are required before a consistent result can be obtained. The integration time for the current reporting is also quite long and the response to step load changes is underdamped. It takes about 5 minutes to get a "stable" reading and even that reading may be inconsistent. The readings have to be taken several times after several step load changes to find a consistent value for a step change in current. I figure that even if the absolute accuracy of the current indication isn't very good, it should be ok for relative readings.

The overall result is that the current drawn by this Powerbook can vary by nearly 2.5 times depending on use and settings and I don't believe that I've explored the high end enough yet to find the top. For battery powered operation, the highest loads are of not very great interest. The load is so high that the battery life would be an hour or less. Since I'll never run that way on a battery, knowing the worst battery life is of little interest. The value in the tests is in determining which of the possible operating configurations are the heavy hitters in battery life and in optimizing them as much as possible.

I've done two types of tests. The first series of tests is the measurement of current draw while turning on or off some feature. I picked sets of features that made some sense for battery powered operation, mostly these tests were run at the Reduced performance setting. It is no surprise that the backlight intensity is the big swing factor at a 60% increase between no backlight and full backlight. If you want to get maximum run time, turn the backlight down as far as you can stand it. There is also a significant current draw related to the CPU settings between Highest and Reduced, 24%. The Airport is next at 14%. The rest of the changes all together, don't add up to the load of the Airport card.

The other set of tests are actual run time tests where I set the machine in some state, and then run it that way until the thing shuts down. I then record the time. Since these take longer to run (like 4-6 hours counting the time to recharge for another test) I have fewer of these but I will probably add to them later.

To make the results of these tests more easily interpreted, I have elected to represent them in percent. This gives a relative importance to each power saving (or consuming) feature. These percentages will roughly relate to run time changes on a single battery charge.

Display Backlight Intensity is evaluated relative to the backlight off. Since running with the backlight off is generally impractical the numbers are a little misleading, but there are times when you can get the ambient light just right to work with text even with the backlight off. Maybe this is how Apple gets 4.5 hours.

Bluetooth should be turned off if you have no Bluetooth enabled devices. It doesn't consume a lot of power, but if you never use it, why run it?

Airport power consumption is significant. If you are nowhere near a network, turn it off too. You can elect to leave Bluetooth and Airport icons on your menu bar so that it is easy to activate or deactivate them.

The Internal Speakers do draw some power. If you want to listen to music, consider headphones or external speakers. Headphones won't bother people nearby (such as in an airplane) and at a fixed location, external speakers sound MUCH better than the internal ones. I choose different volume levels for the headphone and speaker tests just to set the volume to a reasonable listening level. The volume with the headphones was at 50% and with the speakers it was 100%.

CPU Performance is a very significant factor in power draw. Use the reduced setting when running on battery power. Your PowerBook will also run cooler.

Keyboard Backlighting draws some power but having it turned up is good for working in a dark environment. The test was done with the keyboard lit all the time. You can have a bright keyboard and still reduce power consumption by setting the keyboard backlight to extinguish after a few seconds of inactivity.

USB Devices draw power from the computer unless they are running from a powered USB hub. I have a powered hub and the test shows that it indeed has no impact. Depending of your USB configuration, your milage may differ.

Note that hard disk power consumption isn't on this list. This is simply because I have not control over the disk spindown (or lack thereof). During all my testing, the "Put the hard disk to sleep when possible" box was checked. However, this option doesn't actually seem to do anything. In almost 3 years, I've never caught my iBook or PowerBook disk spinning down in OS X. The iBook disk would spin down in OS 9. My son's beige G3 will indeed spin down the disk (including the external SCSI disk) while running OS X but the laptops don't. Therefore, its not part of the tests. The internal disks are rated at a little more than 2 watts which is roughly 20% of the computer's power consumption in an otherwise conservative setting. Getting it to spin down when not needed could add up to a half hour to the battery life.

The current draw also depends on what you are doing. Simple text entry doesn't tax the CPU very much. Running iTunes does. Running a graphically intensive game will make the computer work even harder. If you want to run through a whole airplane flight, consider the type of work that you will do. Save the computer's battery for the stuff that you really need a computer for. If you just want to listen to music, get an iPod instead of running your computer down.

By the way, for listening to music on an airplane, you might want to consider a set of noise cancelling headphones. These work great however you might not be able to hear those annoying announcements or the guy next to you snoring....

In these tests I set the computer into the state that I wanted to evaluate and then let it run the battery all the way down using X-charge as a monitor. It plots a graph of the battery capacity vs time so that after the machine is placed back on the charger and awakened, the data can be read off the graph to within a minute or so. For most of these tests, the discharge rate was quite linear. During the tests with Highest performance, there is a bump at the bottom which appears to be the computer being forced to Reduced performance at the end of the battery life.

The computer was set not to sleep at all and the display effects were turned off so that the computer could be started and left to finish the test. Whatever I wanted it to be doing was set up so that it would do it until the bitter end.

Unless otherwise noted, the conditions are as I would use the computer around my own house when on battery power.

Conditions

The results are not surprising. The more you do and the faster you want to do it, the less time that you'll have to do it in. I've also recorded the run time that Apple calculates for reference. The real result and Apple's calculation stay reasonably consistent over these constant load tests. If you change your operating mode during a session (such as by starting a DVD or a graphics intensive game) you may see the prediction change abruptly.

I'll be adding more data later as I have time to gather it.

I noted an interesting effect during the Pocket Tanks run with Highest Performance enabled. At about the time that the computer reminded me that I was running in reserve power, the current draw decreased from 3.1 amps to about 1.6 amps and the battery voltage abruptly increased by 100 mV or so before starting to decay again. I will speculate that the system was forced to Reduced Performance mode in an effort to conserve power. Note that for this computer, the "normal" light load current is about 1.2 amps so that this game demanded over twice the "normal" load.

When fully charged, the battery runs at about 12.58 volts. At full discharge, the voltage gets down to 10.37 volts. During discharge, the battery voltage decreases more or less linearly and the current increases proportionally. This is probably due to the action of the power regulators in the computer. They want to supply constant power to the computer so that when the input voltage decreases, the regulators compensate by drawing more current to keep the delivered power constant. This is the normal response of a regulated DC to DC converter.

Just to set an upper bound, I set the computer up for light duty with virtually everything either off or at minimum usable conditions and let it sit to simulate a very light load. For this computer and this battery, this condition results in a maximum achievable duration of 3 hours and 40 minutes, still 25% short of Apple's claim.

On Oct 3, 2003 Apple re-released 10.2.8 with a couple of bug fixes. One of them was to the battery life indicator. It did something. It behaves completely differently and at least in the last half of the battery life is fairly close. However, during a single trial, it was wildly inaccurate at the very beginning of the discharge.

This is the result of a single test. The computer was set in my normal mode and I was doing web surfing, email, Solitaire Until Dawn, and iChat for most of the test. The discharge was pretty linear until the last few minutes when the power manager abruptly changed the state of charge. Also, the first indication provided was over 4 hours and the indicator changed wildly for the first minute or so. It didn't actually settle down for an hour.

The actual battery life was nearly identical to the previous test in the same conditions. Although I didn't time it accurately under 10.2.7, the ultimate battery life has stayed roughly the same under the same load conditions. Those that are having battery life problem where the life is significantly shorter than the data presented here probably actually have a bad battery and should get it replaced. It is unlikely that the initial release of 10.2.8 did their batteries in. It is more likely that the battery died on its own at roughly the same time or was already on the way out. New batteries with manufacturing defects are more likely to die in the first few charge cycles than later and 10.2.8 came out just a few days after the new PowerBooks started reaching customers.

Those that are having serious battery problems should get a copy of X-charge or XBattery and run their own tests. Presenting Apple with printed evidence of a malformed discharge curve may go a long way in convincing them to replace a bad battery.

I received Panther (10.3) on about October 27, 2003 and in a couple of days I had rerun my standard usage test. Panther did not change my actual battery life at all. I did notice that after periods of inactivity, the hard disk would actually spin down under Panther, but not during actual usage so that didn't help. If your computer is set to sleep in just a few minutes of inactivity, this is a much more effective way of preventing waste of the battery's precious energy.

After a period of usage with Xbattery running, it has recorded the reported capacity of my battery which has settled at 3.7 ah or very nearly at the spec for the battery. Others are reporting up to 4.5 ah and yet other are reporting much less.

Run this command in Terminal:

ioreg -l -w 0 | grep IOBatteryInfo

to get a report of the battery performance direct from the OS. It'll look like this:

{"Capacity"=3711,"Amperage"=1200,"Current"=3702,"Voltage"=12487,"Flags"=1090519045}

The numbers are in "milli" units. 3711 = 3.711 and such.

• "Capacity" is the total battery capacity reported by the battery when fully charged.
• "Amperage" will show the current draw to or from the battery. If the number is 1200, this indicates that the charger is connected. On older versions of Panther, this number is bogus unless the computer is actually running on the battery. Later versions of Panther have fixed this bug. If the computer is running from a power adaptor, the battery current will be zero once the battery is charged. Positive numbers mean charge current, negative numbers mean discharge current.
• "Current" is the capacity left in the battery in its current state of charge.
• "Voltage" is the current battery voltage.

There is a command script available at Angry Kitchen Appliances that packages this information more neatly. Xbattery packages it even more neatly yet.

Apple released an update to the battery manager about the same time as 10.3.2. I reran my standard test and I got exactly the same battery life as before. The difference is that the battery life indicator was improved again and the indication of the total capacity of the battery changed back to the value that I got the first time that I ran a battery calibration. XBattery has kept a running log of the indicated capacity of my battery ever since I installed it. My battery capacity is now indicated at 3.944 Ah, just under the spec for the battery. A new 15" AlBook battery should have about 4.4 Ah of capacity

Apple, as well as most other laptop computer manufacturers, has selected the Lithium Ion technology for its batteries. There are several implications of this selection, some good, some not so good.

Lithium Ion batteries currently have the highest energy density figures of merit for any common battery technology. Smaller and lighter batteries allow the design of smaller and lighter portable equipment. This is a good thing. However, this does not come without cost. Lithium Ion batteries also have the highest life cycle cost of the common battery technologies.

Battery University provides an excellent tutorial of the details for those that are interested, but the bottom line is that Lithium Ion batteries:

1. Cost more to initially purchase
2. Are more touchy about being properly charged (resulting in a more expensive battery charger)
3. Have limited life. They start to die from the day that they are made.

There have been lots of consumer complaints about the lifetime of Lithium Ion batteries. There is a reason for it. Nobody wants to replace a battery that costs more than $100 every couple of years, but with this technology, its just a fact of life. Maybe in the future when the manufacturing technology improves, the lifetime will improve also, but the simple fact is that for now, 3 years or 300 cycles (whichever comes first) is about it from a good battery. Less can be expected from one that was poorly manufactured. Further, even some of them made with a good overall manufacturing process will die much sooner than they should due to latent and undetected manufacturing defects.

There doesn't seem to be a way around the 3 year limit. The batteries will degrade if they are used or not used. They can be made to degrade faster if they are discharged completely flat more than once a month or so. There is some evidence that the lifetime in storage can be optimized a little if the battery is left at 40% charge but even this is not a large effect.

This is an iBook battery on its very last legs as shown on an X-Charge capacity plot. When one of the four cells in this battery starts to degrade faster than the others, the voltage on that cell will drop rapidly when that cell discharges and the whole battery will seem to die suddenly. In this case, you may not even get the 10 minute warning, the voltage collapses so fast that the only response the computer has is to suddenly go to sleep. You can see the same thing on the recharge curve. When the one dead cell rises in voltage, the other cells are still mostly charged and the recharge curve terminates suddenly too.

This same event was recorded by XBattery. XBattery plots actual voltage in addition to calculated capacity. In this case, the rapid decrease (and increase) of the voltage can be detected. When a whole cell dies, the battery voltage should drop by 3 volts, but you can't ever see that because the computer cuts it off at a reduction of about 2 volts.

When a battery gets into this shape, it is history. There is nothing that you can do except to buy another battery or be tethered to your power adapter.

This is an X-Charge plot of a still ok PowerBook battery that came with the computer. However, this one has degraded to about 75% of its former self. This is pretty normal degradation for a year of service. The battery discharge curve is mostly linear except for the first half hour were the load was higher. However, at the very end, it got to about 5% capacity and abruptly died. This trend will continue to higher and higher percentages until it looks like the iBook battery shown above.

The is the same plot from XBattery. It shows exactly the same data, but it isn't quite as easy to read.

However, XBattery also plots current. Note that a the current is higher (the curve is lower because discharge current is defined as negative current) during the first half hour when I was playing Pocket Tanks. This game demands more graphics processor performance and results in higher discharge rates.

The uptick at the very end is when I plugged the computer back in to capture the screen shots. The voltage had increased abruptly because the battery was being charged.

When fully charged, the battery voltage is about 12.3 volts, but during the very first part of the discharge, it decreases by over a volt. Then the voltage on the battery decreases pretty linearly (except for the kink when I quit Pocket Tanks and the load decreased) until about 10.3 volts when the computer sleeps and no more data is taken. As time goes on and this battery degrades more, the voltage will drop below 10.3 volts sooner and sooner until eventually it becomes unusable.

Because the batteries die on their own, it is not a good plan to buy a spare battery unless you actually use it often. It will die almost as fast as the primary battery. When you do buy a new battery, try to determine the date of manufacture and pick a fresh one, just as you would look at the expiration dates on milk cartons at the supermarket. Batteries sold at discount have probably already expired and the vendor knows it. That's why he sells them cheap. Batteries for computers that are currently in production are probably fairly fresh as the stock will be consumed. This assumes that the computer manufacturer properly allocates batteries to production and aftermarket sale. Batteries for computers long out of production will almost certainly be stale. If you buy a laptop computer at the end of its production run and the battery model is not being reused in current production, then you can be pretty sure that you'll only be able to find stale batteries as replacements when you need one. If you do go for the "end of product life" computer at closeout discount, try to pick one that uses a battery that has been carried over to the newest model. That way, when you need a battery, there might still be some in current production.

I've had the PowerBook for a little more than a year and I find that, overall, the battery life is not adequate for most laptop usage. The computer just draws too much power due to the higher performance processor and the larger screen. The extra power also means extra heat. This computer tends to run too warm to comfortably keep on my lap.

As a portable desktop computer however, it does great. The screen is large, bright and sharp. The processor performance is good enough for almost every usage (at least with today's software) and the keyboard is good.

When I travel now, I tend to take my old 500 MHz dual USB iBook. There are several reasons I do this.

1. The iBook is smaller and packs easier
2. The iBook has less replacement cost so that if I lose or damage it in travel, the loss is not as severe
3. The life on the battery is easily double that of the Powerbook. I typically get 4+ hours on the iBook vs 2+ hours on the PowerBook. Further, the battery in the PowerBook is dying now barely makes 90 min. I also have a 2nd battery for the iBook already.

My wife now has a 12" iBook G4 so I ran some tests on current draw on the three machines. These were all done with an aggressive power conservation strategy. The Airport is off and the processor performance is set to reduced. I ran the tests with 3 screen brightness, off, one bar and max. As expected, the faster computers draw more current and therefore won't last as long on a battery.

As expected, the slower computers draw less current. Further, the screen size matters. The Powerbook has almost twice the screen area to update and illuminate. Set with a dim screen, the iBook G3 draws half the current as the PowerBook G4. For the same battery capacity, it will run twice as long. This matters a lot on an airplane. Further, the smaller computer is easier to use in a cramped airplane coach seat.

I'm beginning to believe that the best arrangement for desktop and portable utility would be TWO computers. The smallest possible laptop and a smoking desktop (like an iMac G5) won't cost that much more than a really nice laptop and provide better utility in both situations.

After 14 months, the original Apple battery was beginning to show its age. It's capacity was declining rapidly as can be seen in the graph. Under normal loads, it was providing only a little more than an hour of service. As the XBattery graph shows the early days, the reported capacity of the battery was jumping all over the place, mostly because of system software updates that changed how the battery was used and reported. After the first few months however, the downward trend was obvious. This particular battery never reported as much as 4 ah, if it met spec it should have at least been about 4.5 ah. At last report it was under 3 ah and declining quickly.

Instead of buying an Apple battery, I decided to try the Newer Technology battery. This one was claimed to have 10% or so higher capacity than the Apple battery at the same price. However, it's downside is that when it runs down, it WON'T put the computer to sleep. It'll just go on until the computer shuts down without warning causing a total loss of any unsaved work. This battery started at 4 ah out of the box, improved to 4.45 ah after one discharge/charge cycle, improved again to 4.9 ah after the second cycle. This probably where it will stay for awhile before it starts its inevitable death spiral. It also runs the computer for 3 full hours with the airport running and 4 bars of screen brightness. The best I could ever do with the old battery was 2 hr 40 min with no airport and one bar of screen brightness.

While I haven't actually run the battery down in a full power conservation mode, when I set the screen brightness to one bar and turn the airport off, the projected battery life is just over 4 hours, far better than it has ever been before. Based on the difference in current consumption (1.2 vs 1.5 amps), this is just about right.

However, to avoid beating up on the battery by discharging it too deeply, I have to keep an eye on the state of charge and plug back in when the battery voltage reaches 10 volts (as reported by XBattery). The computer will shut down at 9 volts but since the battery voltage fades quickly below 10 volts, this is less than 5 minutes difference. The Apple battery will put the computer to sleep at 10 volts.

I ran a similar discharge test on the Newer Technology battery and I find that not only does it have considerably more capacity that my old and tired Apple battery, it's discharge characteristics are markedly different. While the XBattery capacity plot looks pretty linear (because the load was nearly constant, the following curves tell a different story.

Just for reference, the load current is similar to the previous test except no Pocket Tanks this time.

However, the voltage characteristic is very different. The voltage stays higher longer at the beginning and folds over more gradually at the end. Since this battery doesn't activate the sleep mode by itself, I had to do it. The Apple battery cuts off at 10.3 volts because there isn't much life in the battery after that. I let this discharge go to 10 volts but it is clear that the battery is done.

X-Charge thinks that the battery isn't quite done at about 5% capacity. At the very end of the discharge, the capacity curve starts heading down again but never reaches zero as the Apple battery does. This probably indicates that the characteristics of the battery monitor computer in this battery are different that the one that is in the Apple battery. What's more is that it lasted 3 hours and 30 minutes, something that the Apple battery NEVER did.

Yet another year has gone by and the Newer battery is starting to show its age. This is the capacity plot over a full year of usage. The capacity is down something near 10% which is pretty good for a year. I would have expected more like 20%.

A little over 300 days back, the battery was changed. There are several small steps on the large up-slope of the new battery that were points were the battery was cycled. After 3 cycles, it had attained it's full capacity. Each subsequent downstep was a deep battery cycle that forced a recalibration and each time the battery lost a little capacity due to aging. The original Apple battery, starting at over 800 days back also improved after a few cycles before it started it's death spiral.

After another 120 days or so, the Newer battery is continuing to degrade. The big blip about 100 days back was short stint on the original Apple battery. Each step in the curve represents a "calibration" of the battery.

This particular Newer battery was an early one that won't force the computer to sleep when it becomes discharged. Apparently, the more recent batteries from Newer now work properly in this regard.

I bought the NewerTech battery sometime in late 2004. It's now late 2007 and the battery is showing its age. I hadn't been using the PowerBook much on the battery simply because it gets to hot to use on a lap and I was traveling with an older iBook. The iBook got much better battery life and, because it was so slow, it didn't get hot. The PowerBook was being used as a portable desktop computer where it did very well. Therefore, the battery didn't get exercised very much, it only had 103 full cycles on it in 3 years.

However, yesterday, I was using it without the charger. I noticed that is said that it had an hour to go which was reasonable for the amount of usage that it had just received. Then a minute later, I got the reserve power warning. Then it went to sleep in a couple of minutes. What had happened?

First, the sleep issue. When the battery was new, NewerTech claimed that the battery would not put the computer to sleep when it ran out of gas, now it is doing that. All through Panther and Tiger, it had acted as NewerTech claimed. However, I have Leopard now and maybe that changed the behavior of the computer some. I really don't know, but something has changed.

Second, the apparent power crash. The graph at the left shows the computed capacity of the battery over four complete discharge/charge cycles as recorded by MyBattery. The first discharge started at the very left edge of the graph. Note that the capacity was decaying normally for a little while to about 66% and then just dropped. On the following charge cycle the battery apparently recovered to about 40% and then the computer said it was good. This is a symptom of a seriously wounded battery. On the next three discharge/charge cycles, the battery apparently IMPROVED. On the last discharge cycle, it got to 20% before crashing and the improvements were getting incrementally smaller. This battery is clearly still wounded, but cycling it apparently increased it's capability as the discharge time was improving too.

This curve is the same event. It shows the battery voltage. At the left is an abrupt drop in battery voltage that occurred when the charger was unplugged. This drop is due to the internal resistance of the battery, about a volt at a load current of about 1.5 amps. Then the discharge occurred, but it was quite noisy. The noise is occurring because the internal resistance of the battery is unstable. When the voltage got almost to 10.4 volts, the computer went to sleep. The rapid rise following is the very first part of the charge cycle. Then the battery would charge until the voltage reaches about 12.56 volts and the charge shuts off. The battery voltage decays slowly for about 15 minutes. When it had stabilized, I started another discharge cycle.

Note that at the beginning of each discharge, the battery voltage dropped a little less each time as the internal resistance of the battery was apparently becoming less. Also note that on subsequent discharges, there was less noise on the discharge. If you look back at the voltage plot when the battery was new, there is not nearly as much noise. Note that this data is recorded only once every 15 seconds so that there is a very good chance that there is a lot going on that is undersampled and doesn't show up in this data.

This kind of instability can only be bad. The battery is clearly wounded and nearing the end of it's life although it might go a little longer before it dies completely.

This is the history of this battery. It's total capacity continues to degrade as all Li-ion batteries will. After 5 or 6 complete charge/discharge cycles, the battery began to behave "normally" although with reduced total capacity. This can never be recovered.

After a couple of weeks floating on charge in the computer, the NewerTech battery began to return to it flakey ways, dropping out somewhere in the discharge. A couple more power cycles brought it back again but it is clear that the battery will fail completely fairly soon.

[ Mac Tips ] [ Next Tip ]