https

Click for full-size image
Time to Recharge Already?

Research into how Palm Tungstens consume battery power,
and strategies for getting several times more battery life.


Synopsis:
Through a series of careful laboratory measurements, we determine that CPU speed and LCD backlighting are, by far, the top two battery power consumers on present-day Palm Tungsten devices. We then carefully analyze and characterize the relative amount of power each of these consumes to design a workable strategy for significantly extending Palm battery lifetime by several times.


PDAs have always been "battery challenged".
Back at the dawn of PDAs, their 16 Mhz processors were barely powerful enough to drive their screens and perform the minimal processing chores required to search an address book and maintain a calendar. And they could generally run for five or six hours on a set of batteries. If the device was used as a "quick reference" for brief periods of time, those batteries could last for weeks or even months.

As battery technology advanced from replaceable carbon-zinc through nickel-cadmium, nickel-metal-hydride, lithium-ion and finally to today's high-performacne lithium polymer, PDA processor speed/power and display quality advanced just quickly. As a result, despite the fact that battery technology has advanced tremendously, after continuous-use PDA battery life still falls into the 3 to 5 hours range.

If you're just looking stuff up . . .
3 to 5 hours of total battery life is still fine for typical "PDA" use where you're quickly searching your address book or carrying your calendar around. But as processor power has jumped, display quality has increased significantly, flash memory density has skyrocketed, and memory prices have fallen, PDAs have moved solidly into the category of handheld media devices. In many instances, when properly configured, they are able to compete head-to-head with more expensive and much more limited special purpose devices.

Today's large screen, high-resolution (480x320) high-performance (400+ Mhz) PDAs are entirely useful for watching movies, listening to music, and reading books . . .

Reading books and watching movies on my Palm
I love to read. Some people are watchers and others are listeners, but I'm a reader. I read on my Tungsten T5 when I'm eating out alone, travelling, waiting in line, waiting for my name to be called  . . . and sometimes even when waiting for the traffic light to turn green. It's my "time filler", which I use to keep from ever being bored or impatient when things are not moving as quickly as I might like (which is pretty much any time I'm away from my work. :) The main memory of my Palm is mostly filled with Science Fiction eBooks. They provide endless hours of interesting entertainment in just 250 kbytes of compressed text.

So eBooks are the highest density form of entertainment in terms of hours of entertainment per byte of storage. But like any form of media, they require that my Palm stays on more or less continuously, consuming precious battery power minute by minute.

No matter what form of media you want your PDA to deliver, movies, music or books, using any of these means that instead of turning your PDA on, looking something up, and turning it off again, you are inherently leaving it on and running it as long as its batteries will last. So total battery life becomes much more important when our PDAs are operating in "media delivery mode" over the span of a few hours than when they are operating in "quick lookup mode" over the span of many weeks.

What's a "milliamp hour" ??
A battery's total power storage capacity is measured in "milliamp hours". This is milliamps x hours, or "current" multiplied by "time" where a milliamp is one thousandth of an amp. This means that if a battery could supply 1000 milliamp hours of power, it could supply 1000 milliamps for one hour, or 1 milliamp for 1000 hours, or any combination in between, such as 100 milliamps for 10 hours, etc. This simple formula means that any fraction by which we are able to reduce the current our PDA is sucking from its battery is the same fraction by which its remaining battery life, from that point forward, will be increased.

  Palm Model    Battery Capacity  
T T2 T3900 mAh
T51300 mAh
TX1250 mAh

The motivation for this research was to determine exactly how much power our Palm Tungsten LCD backlights consume. I wanted to know what the actual measured power consumption difference was from full brightness to minimum. And what about processor speed? Just how much power can be saved by deliberately "underclocking" the Palm all the way down to 52 Mhz?

Hint: I was able to reduce the Palm T5's current
consumption from 245 milliamps, down to just
105 milliamps  . . . for a total battery lifetime
increase of 233% or more than 2.33 times.




Controlling the Palm Tungsten's processor clock speed:

Overclocking:
Once upon a time PDA processors were clocked at 16 Mhz, which was too slow for some "extreme" applications that some people wanted to run. So the discovery was made that, at the (sometimes extreme) expense of power consumption and battery life, the processor clock rates of those old PDAs could be significantly "juiced up" to give processor-intensive applications a lot more performance.

Before long, "overclocking" utilites were available for virtually all handheld PDA-style devices. While the overclocking utility authors were at it, they added "underclocking" capabilities to their applications to make them more complete. Unfortunately, as the years have gone by, PDAs have become so much more powerful, faster, and capable, that the need to overclock them has largely disappeared. So there is no longer a wide range of clock rate setting utilities to choose among. For some reason the idea of "underclocking" PDAs to extend their battery life hasn't caught on (yet) enough for people to create underclocking utilities for their own sake. So people who want to underclock their Palm devices are currently stuck using older and less heavily supported overclocking utilities that also happen to offer underclocking support.

"LightSpeed" and "PXAClocker"
Although there were once a great many overclocking utilities for the early Palm devices, I have only been able to locate two which operate on the more recent Palm Tungsten family. And also, very sadly, my strong favorite of the two, LightSpeed, has not been updated for the newer Palm T5 and TX devices, on which it simply refuses to run. However, lucky users of older Tungsten T/T2/T3/C, Zire 21/71 and Sony CLIE devices WILL be able to underclock with LightSpeed to extend the running time of those Palm devices.

Click for full-size image

LightSpeed - Click for full size image
The very nice and clean "Lightspeed" Palm clocking utility.

PXAClocker
For newer Palm devices, specifically the Tungsten T5 and TX, PXAClocker is the only compatible processor clocking utility I have found. The program is extensively feature-packed with an amazing array of options (see screen samples below). However, careful measurements of real-time battery current drain revealed that, fancy (and intimidating/confusing) as all pf PXAClocker's features shown below are, NONE of them have ANY effect — not any — on the Palm's resulting current drain and battery life.

Click for full-size images

Click for full-size imageClick for full-size imageClick for full-size image

What's worse (much worse) is that many of these options are extremely dangerous and WILL lockup your Palm totally and permanently, rendering it completely non-functional for days as the battery slowly drains. No form of soft or hard reset will bring it back — everything is ignored.

During my early experiments with PXAClocker on my T5 it locked up totally. After waiting for days for the battery to drain, it never did sufficiently. So I was forced to crack open the unit and unsolder the battery to finally regain the use of the Palm. The gadeteer side of me would love the idea of all those bells and whistles if they were useful for anything at all. But since I have absolutely determined that they do not reduce the Palm's power consumption by even the slightest amount, they are just "booby traps" in the purest sense of the word, and the PXAClocker utility is FAR more dangerous to use than it should be or needs to be. (Especially on Palm devices without user-removeable batteries!)

Still, I'm using PXAClocker . . .
With no other choice than PXAClocker to slow down my T5's processor clock, and since I'm able to use it in combination with minimum display brightness to get 2.28 times more battery life!, it's what I am using. If you choose to follow in my footsteps, it's what you will need to use too. So PLEASE read all of this page carefully before using PXAClocker yourself (unless you have a soldering iron handy!)




PXAClocker — The Good News:

The good news about PXAClocker, is that as long as you steer clear of its pitfalls and landmines, it works well and reliably. The screen above was taken from my T5. It shows that while the system default processor speed is 416 megahertz, the processor clock rate will be reduced to just 26 megahertz whenever the Palm's calculator (Calc) or electronic book reader (eReader) are running in the foreground. PXAClocker automatically changes the Palm's processor speed on a "per-application" basis, allowing a high degree of performance and battery life tuning and customization.

As mentioned above, since battery drain equals "rate x time", any applications where you spend a great deal of time, such as book reading, music listening, or movie watching, will yield large cumulative battery life benefits if the system's processor runs as slowly as possible while it's in each of those high-usage applications.


The chart above may not be surprising. Brigther backlighting consumes more current from the Palm's battery. But I wanted to know exactly how much more, and what percentage of the Palm's total power usage was backlighting compared with the rest of the Palm. I was also wondering whether the current versus brightness relationship might be non-linear and perhaps have a "sweet spot" where it was possible to get some more brightness at less current cost. But as you can see, the experimental results are just about as linear as you could ask from laboratory measurements.

One important fact I was also curious to learn, and determined with certainty, was that the backlight's current consumption is independent of processor speed. The current readings shown in the chart above were taken with the processor running at its minimum stable speed of 52 Mhz. But the difference in backlight current consumption, from maximum to minimum brightness, is always approximately 60 milliamps (0.060 amps) and is completely independent of processor speed.

When the processor is running as slowly as possible, the difference in the Palm's overall battery consumption, between minimum and maximum brightness, is significant. This can be seen clearly in the chart above which is "zero-based" to give a true sense for the proportion of current being consumed by the Palm T5.


The clock rate power consumption chart also has no surprise. Higher processor speed results in greater power consumption. But, again, knowing the relative amount of power consumption is the whole point behind taking precise measurements.

One coincidental fact is that the range of power consumption from minimum to maximum processor clock speed is almost exactly the same as for minimum to maximum display backlight brightness.

And, again, the power consumption due to processor speed is independent of any due to display brightness. The difference from running the processor at 52 Mhz and drawing 100 milliamps and running it at its normal "full speed" of 416 Mhz and drawing 170 milliamps is 70 milliamps, or 0.070 amps.

Important Note: The slowest clock speed shown in the chart above — 26 megahertz — and the three black clock speeds shown in the middle, are not usefully available. The Palm crashes badly when set to any of those speeds. (An example of why PXAClocker is dangerous without instructions.)




Putting it all together:

The following chart assembles this raw data into a simple chart showing the various current consumptions and total theoretical running times for the four combinations of maximum and minimum clock speed and brightness. Since the scales and relationships are all highly linear, you can easily compute the performance of any mid-point between:

  Brightness     Clock  
  Speed
  
  Current  
  Consumption
  
  Running Time     %  
  Normal
  
Max Max 245 mA 5.31 hrs. 100%
Min Max 185 mA 7.03 hrs. 132%
Max Min 175 mA 7.43 hrs. 140%
Min Min 105 mA 12.38 hrs. 233%

As the chart above makes clear, running high-usage applications at the minimum speed they require, rather than always at the Palm's default maximum, can deliver significant saves of up to 400% or 2.12 hours.

And similarly, running the Palm's display as low as possible is always a good idea. When indoors, if any application will be used for a long time, turning the screen's brightness all the way to minimum can result in a savings of 320r 1.72 hours.

And finally, since the savings from clock rate and backlight brightness are additive, the combined savings of minimum clock rate and minimum backlight intensity can result in an overall savings of 133% for a total running time of 12.38 hours . . . more than SEVEN HOURS longer battery life from the Palm T5's standard internal battery.



Jump to top of page
Gibson Research Corporation is owned and operated by Steve Gibson.  The contents
of this page are Copyright (c) 2014 Gibson Research Corporation. SpinRite, ShieldsUP,
NanoProbe, and any other indicated trademarks are registered trademarks of Gibson
Research Corporation, Laguna Hills, CA, USA. GRC's web and customer privacy policy.
Jump to top of page

Last Edit: Mar 28, 2013 at 11:39 (388.03 days ago)Viewed 68 times per day