Battery Life: A Two Year Study

[Jason Dunn]

With such a pretensious title, you might expect nicely polished results, but that never quite happened. Here's the story: about two years ago, I started working on some battery tests using some testing methods I had worked up. The Pocket PC OEMs were a little generous in their battery testing estimates, and I wanted to get down to the truth about battery life. I started doing tests with every Pocket PC I could find, and the result was a big spreadsheet with a lot of numbers.<br /><br />About six months ago, I made a plea to the community to see if anyone had the skills to help me build an engine to host and manipulate the data I had collected. A few people offered, but ultimately none of them followed through on their comittments. I'm working on an application now that will completely re-define how we measure Pocket PC performance and battery life, and once it's realeased my data will quickly become overshadowed by this tool. And no, I can't tell you what it is yet. :wink:<br /><br />So, rather than sit on this data for any longer, I thought I'd just toss it out to the Pocket PC community. So, for better or worse, here's my data. If someone wants to take this data and create some pretty graphics and charts, I'm happy to attach them to this article (400 pixels wide max) - I just didn't have the time neccesary to analyze this data properly. It would also be cool to see someone do up some Q&A's using this data. ie: "Will leaving a memory card in my device affect battery life"? There's so much to learn from this information, so I hope that someone takes up the task of crunching it.<br /><!><br /><b><span>Battery Life Tests</span></b><br />The battery life on most Pocket PCs is still an area where people are disappointed. While the newest Lithium Polymer batteries are getting better and better with power, they still don't last long enough. If battery life is important to you, these tests I've done should help you gauge how long battery life really lasts. These tests are by no means exhaustive, but by testing a constant activity like listening to music and a passive one like idle runtime, you should get a pretty good idea of how long your Pocket PC will last in most cases.<br /><br />The standard runtime test was accomplished by using <a href="http://www.scottandmichelle.net/scott/cestuff.html">KeepAlive</a> by Scott Seligman. KeepAlive sends phantom taps to the screen, making the Pocket PC think it's being used. The backlight settings were changed to never turn off as long as the Pocket PC was in use, and the runtime was written to a text file on the device. Because the Pocket PC is essentially doing "nothing", this test is only a guideline - even with moderate use, the CPU will chew up power and drive this number down. Actually using the device is where the audio test comes in.<br /><br /><b><span>How the Audio Test Worked</span></b><br />With the audio playback test, I wanted to isolate the Pocket PC and truly test it without any complications. I took a 64 kbps WMA file (<a href="http://www.amazon.com/exec/obidos/tg/detail/-/B000055ZY0/qid=1051644845/sr=8-6/ref=sr_8_6/002-2633618-2394468?v=glance&s=music&n=507846">"Smell the Color 9" by Chris Rice</a>), put it on repeat, and turned off the screen. I plugged in a set of headphones, and placed a computer microphone inside the headphones.<br /><br />I then started an audio recording stream on the PC using SoundForge, and let it run. When the Pocket PC is low on battery life, it will stop audio playback and make an warning sound (beep). Since I could see this audio spike in the audio stream, it was easy to spot how long the battery would last. Trying to play music beyond that point or even use the device results in repeated warnings about the battery life - you might get another twenty minutes of use if you're lucky, and beyond that you run the risk of losing your data depending on the backup battery technology used in your Pocket PC.<br /><br />The issue of staggered testing has come up. I've heard people say that if they listen to music, stop for a while, then start again, their battery lasts long. This seemed to make sense, but when I did staggered audio testing on the Casio E-125 with a Microdrive (49 minutes, 43 minutes, and 87 minutes of playback), the end result was approximately 3 hours, the same as the regular test.<br /><br /><b><span>And the Winner is...</span></b><br />The data is available in two formats: <a href="http://www.pocketpcthoughts.com/files/ppcbatterytest-jasondunn.pdf">Adobe PDF</a> and <a href="http://www.pocketpcthoughts.com/files/ppcbatterytest-jasondunn.xls">Excel 2002</a>. Feel free to download, distribute, and generally do whatever you want with it. It would be nice of you to reference this site though, because I'd like to see the discussion of the results focused here.<br /><br />Take a look and tell me what you think!

[bdegroodt]

Nice work Jason. I'm a little confused about the Dell Axim X5 on Auto Speed. You have 2 entries. One that says 14.0 hours and another that says 6.6 hours. What's the difference in the entries?

15 hours out of a PPC from Dell! WOW! That's really impressive.

[Jason Dunn]

Nice work Jason. I'm a little confused about the Dell Axim X5 on Auto Speed. You have 2 entries. One that says 14.0 hours and another that says 6.6 hours. What's the difference in the entries?

That sounds like the lowest vs. highest backlight settings...but my data may have some errors in it too. It's pretty raw. :mrgreen:

[bdegroodt]

Nice work Jason. I'm a little confused about the Dell Axim X5 on Auto Speed. You have 2 entries. One that says 14.0 hours and another that says 6.6 hours. What's the difference in the entries?

That sounds like the lowest vs. highest backlight settings...but my data may have some errors in it too. It's pretty raw. :mrgreen:

My bad. You're right. Add to that the potential to swap batteries on the Axim and that's pretty impressive. Would surely get a full days use without issue or a week of occasional use. Plus I believe the Axim allows you to disable the Quick Start application buttons temporarily (Like a Palm from what I've heard.) so no accidental 2 hour rogue power use. I wonder how the 5450 compares.

This should be a nice tool for comparison! A very important topic as all the wiz bang cool features won't mean a thing when you try to power on your dead PPC.

[Jhokur2k]

I think I might take a crack at running through this - first thing is for me to organize it in a mysql table :D

[chuckers]

Yea i'm incredibly happy with dell battery life, i usually get a weeks worth of typing notes in class or more on one charge. And great comparison chart, too.

[JonnoB]

Illustrates how far ahead of it's time the HP 56x device was... without the benefit of an XScale CPU - it compares well.

[fireflyrsmr]

:D an engineer is in that body somewhere Jason!

as an iPAQ owner I glanced at that data first. The runs with the different size cf cards was a surprise. I wouldn't have thought it would make a difference. Do any of the EEs out there understand that? I'm just an ME who thinks this is all wonderful magic.

the other surprise is the 200% drop between the 3650 and the 3870 with the light on low. While there was only a 40% drop when the music is playing. I understood that the newer iPAQs had a better cpu relative to battery useage but I didn't understand that the lighting technology had that much advantage. Without looking it up i thought the screens were the same. That the new screens came in with those newest fangled ones - 54somethingorother.

[R K]

Of course, the lowest backlight setting can mean different things on different Pocket PCs. For instance, the earlier iPAQs only had four backlight settings, while the H38xx had about fifty.

[ethancaine]

One peice of data I think would be interesting would be dates (month/year) that the models you used were released so as to build a trend chart in Excel.

[khalberg]

I did a little bit of pivot-table manipulation. Interesting to see that the iPAQ 3650 will only survive a few hours with the backlight on, but will go 10+ hours without. I really like the ability to use the iPAQ all day with the PCMCIA sleeve and an 802.11 WiFi PC Card. Don't see that I can post the HTML version of the Excel Pivot table easily on this forum.

Kevin

[klinux]

Very impressive work/data collection Jason!

[Jason Dunn]

Illustrates how far ahead of it's time the HP 56x device was... without the benefit of an XScale CPU - it compares well.

Indeed, it was a very, very impressive device for it's time (even now). Part of the great battery life was the lighting technology they used - it was quite different from the iPAQ. The down side is that the screen just didn't look that great...but in terms of battery life, yeah, it took the iPAQ to school and back again. :mrgreen:

[Jason Dunn]

:D an engineer is in that body somewhere Jason!

I secretly love benchmarks and data crunching...I'm the kind of guy who'd run a benchmark, overclock my CPU by 10 Mhz, then benchmark it again just to see the difference. :mrgreen:

as an iPAQ owner I glanced at that data first. The runs with the different size cf cards was a surprise. I wouldn't have thought it would make a difference. Do any of the EEs out there understand that? I'm just an ME who thinks this is all wonderful magic. the other surprise is the 200% drop between the 3650 and the 3870 with the light on low. While there was only a 40% drop when the music is playing. I understood that the newer iPAQs had a better cpu relative to battery useage but I didn't understand that the lighting technology had that much advantage. Without looking it up i thought the screens were the same.

Yeah, the memory card difference is very interesting, isn't it? I was very surprised. I'm hoping to do a test with a single Pocket PC and a variety of memory cards, brands, etc for performance and battery life. The screen technology advances from iPAQ 36xx to 37xx was shocking as it related to battery life.

One last thing to consider is that not all of these devices were tested immediately out of the box, so some of the batteries may have been outside of their optimal charge range. Without being in a lab with utterly strict controls, this is about the best I could do. :fro:

[Jason Dunn]

One peice of data I think would be interesting would be dates (month/year) that the models you used were released so as to build a trend chart in Excel.

Indeed, that would be cool! Unfortunately, I have zero data on that - I wouldn't even know where to begin. Every OEM would store that data differently - the iPAQs have it in their serial number, but I don't know about HP, Casio, etc...

[heliod]

First of all, great work, an awesome study.

The only thing that puzzles me is how the Axim at 400MHz got a better result than in Auto mode (in which it was supposed to switch to 200 MHz).

[PapaSmurfDan]

Jason, Do you have a PXA250 Axim? If so, how long have you had it?

I'm starting to find mine (from the first batches in December), is starting to wear down thin with the battery. Then again, I did have a bunch of battery drains due to my case. It would awake from an alarm and not turn off because the touchscreen had pressure on it :( Also, I am a heavy wifi user and its not uncommon for me to have my wifi card in for a few hours in and draining it past the 30% mark. Overall, the battery in my device has taken a beating.

I just had a recent trip where I used MS Reader to listen to Audible books and after about 5 hours (at about 50% volume over a car adapter, 200mhz, screen off) I had a very low battery warning :? The books were on a the 64mb Sandisk SD Card. I also use a 256mb SanDisk CF card for ogg storage.

But I am thinking about getting a second battery. When I do get one, would you like an independent test comparing a brand new battery to an worn out one? I am assuming you would like this table to continue to gather data.

-Dan

[Jason Dunn]

The only thing that puzzles me is how the Axim at 400MHz got a better result than in Auto mode (in which it was supposed to switch to 200 MHz).

I thought that was a little strange too. I wonder if the CPU throttling uses up more battery power going up and down vs. just locking it in at 400 Mhz? I haven't truly done enough XScale battery testing to come up with any theories though, so for now it will remain a mystery. :mrgreen:

[Jason Dunn]

Jason, Do you have a PXA250 Axim? If so, how long have you had it? I'm starting to find mine (from the first batches in December), is starting to wear down thin with the battery...

I've had my Axim since...hmm...January? I haven't noticed any battery drain problems yet.

But I am thinking about getting a second battery. When I do get one, would you like an independent test comparing a brand new battery to an worn out one? I am assuming you would like this table to continue to gather data.

Actually, no, this project is pretty much dead. As I indicated, there's a new project underway that will offer a much more comprehensive testing solution, so I want to start over with that. But if you'd like to do your own tests and compare them to mine, go for it! :-) When I can talk about this new system, I will, and we can call have fun with it. :D

[rlobrecht]

The runs with the different size cf cards was a surprise. I wouldn't have thought it would make a difference. Do any of the EEs out there understand that? I'm just an ME who thinks this is all wonderful magic.


I'm not an EE, but here's a guess. More memory means more transistors. Each transistor would probably use a similar amount of power.

[nosaturn]

You'll need to find out the threshhold level for the battery warnings. Some of the different device makers may be tweaking that. Just think, if the standard threshold is say, 40% of charge, the TMobile phone has way more hours of use at that point then say the 3600. That's one reason the OEM/ODM's make their own power controls and might tweak the notification/shutoff levels.
:werenotworthy: your now the Battery God!

[fireflyrsmr]

You'll need to find out the threshhold level for the battery warnings.
That's probably good insight. I know at GM they leave 2 gallons of gas in the tank when the meter reads empty.

More memory means more transistors.

that sounds like a good guess. thanks.

[Abba Zabba]

Hey Jason that was a pretty sweet test :clap: .....sooo what is coming soon come on you can tell me, I don't think anyone is watching :wink:

[ethancaine]

That's probably good insight. I know at GM they leave 2 gallons of gas in the tank when the meter reads empty.

This is a bit off topic, but according to the History channel, the tradition of extra gas at the end of the fuel tank is a tradition stemming from the blatant inaccuracies of the early guages.

Oh, and :way to go: Hoorah History Channel![/quote]

[BubbaJonBoy]

Actually, no, this project is pretty much dead. As I indicated, there's a new project underway that will offer a much more comprehensive testing solution, so I want to start over with that. But if you'd like to do your own tests and compare them to mine, go for it! :-) When I can talk about this new system, I will, and we can call have fun with it. :D
First off - excellent test, fits in well with what I have observed on my Dell Axim. One thought to be aware of for future tests, the newer Dells are coming out with the 255 processor which is rumored to have 20% better battery life.
FWIW
Jon

[Jason Dunn]

[quote="BubbaJonBoy"]First off - excellent test, fits in well with what I have observed on my Dell Axim. One thought to be aware of for future tests, the newer Dells are coming out with the 255 processor which is rumored to have 20% better battery life./quote]

Indeed! Dell is a little too sneaky for my liking slipping in a new processor without changing the model number...kind of screws up the results. :-) That said, it's cool they're giving consumers better products now for the same price. Go Dell go!

[heov]

First off - excellent test, fits in well with what I have observed on my Dell Axim. One thought to be aware of for future tests, the newer Dells are coming out with the 255 processor which is rumored to have 20% better battery life.

Indeed! Dell is a little too sneaky for my liking slipping in a new processor without changing the model number...kind of screws up the results. :-) That said, it's cool they're giving consumers better products now for the same price. Go Dell go!

but it sucks for those who bought the unit say a week earlier and got a pxa250... oh well, i don't have a dell, but if I had an old one, i'd be a little peeved if I bought one now and got an older pxa250 device.

Anyway, about the battery testing thing, is this going to be one of those battery benchmark tests where it runs an operation (same on every device) and measures the battery drain over a period of time? That'd be pretty cool :)

[timmy]

Hi,

I am just wondering if there is an error in the Excel sheet for the iPaq 3650. :?:


Compaq iPAQ 3650 Lowest none 207 29 118,0 2,0

Is actually the first test 207 mins and the second 29 mins ? that explains the low average of 2 hours. Shouldn't it be 209 mins ? That would give 3,5 hours average. more in the neighbourhood I have experienced with mine.

regs Timmy

[Jason Dunn]

I am just wondering if there is an error in the Excel sheet for the iPaq 3650. :?:
Compaq iPAQ 3650 Lowest none 207 29 118,0 2,0
Is actually the first test 207 mins and the second 29 mins ? that explains the low average of 2 hours. Shouldn't it be 209 mins ? That would give 3,5 hours average. more in the neighbourhood I have experienced with mine.

Yes, you're right, that's probably supposed to be 209 - making it 3.5 hours on average, which is more realistic than 2.0 hours.

The data is very raw, which is why I didn't use it for anything fancy. :D