Menneisyys
04-02-2006, 11:35 AM
Battery life is very important with PDA's, especially if you don't have any chance to charge them or swap batteries while on the move. Then, every additional minute you can squeeze out from the battery may be important. In this article, I show you what you can do in order to optimize battery life.
Also, I show you how much power the backlight, the wireless units, the buzzer/speaker etc. consume. A lot of plain wrong urban legends (for example, the question of disabling the infrared receive or the Bluetooth unit) are still being widely considered thruth - in this article, I explain why they are not true (any more).
You may already have heard that the main processors (Central Processing Units, CPU's) in the Pocket PC consume more power (and, therefore, chew through your battery faster) when used at higher clock frequencies. Many people even underclock (that is, run at a lower frequency - for example, 208 MHz instead of the standard 400/520/624 MHz) their PDA's to achieve better battery life. Also, you may well know if you enable certain (mostly, wireless) functionality on your PDA, battery power will be consumed far faster.
Now that a really excellent application for the exact measuring of the battery load, acbPowerMeter, has been released (http://pdaphonehome.com/forums/attachment.php?attachmentid=8188&d=1142835308) (alternative download here (http://www.winmobiletech.com/sekalaiset/acbPowerMeter.CAB)), it has become much easier to exactly see how much milliampers the battery load at a given time is.
Before acbPowerMeter, measuring the battery load was incomparably harder. The easiest, but slowest way was topping (charging until it's full) the battery and, then, waiting for some hours running the given application / enabling the to-be-tested unit (for example, the Bluetooth / Wi-Fi unit) on the PDA and, then, checking how many percents of the battery charge remained.
As the battery load depends not only on what hardware units are enabled and what the backlight level is but also on the CPU load (that is, how many percents of the CPU time a particular application uses), testing the battery drain of (just) a given program/game has been possible using other means. You could use the CPU usage display in ImmierSoft XCPUScalar (http://www.immiersoft.com/) (if it actually worked on your particular PDA model and didn't crash it). It displays an icon on the taskbar, showing the actual CPU load. Of course, it can't be used if the taskbar is hidden (for example, in full-screen titles like most games) and it delivers unreliable/useless, not updated results if the CPU load is over a certain threshold (about 85-90%). This was the way I've measured the CPU load of the software titles I've tested so far (see for example the article Everything you may need to know about sound recording on the Pocket PC (http://pocketpcmag.com/blogs/menneisyys/112005SoundRecorderApps.asp), the article on voice and remote controllers (http://www.pocketpcmag.com/blogs/index.php?blog=3&title=what_voice_controller_apps_are_there&more=1&c=1&tb=1&pb=1) for some numerical results measured this way).
As has been pointed out, XCPUScalar only displays the CPU load caused by the running programs and not the actual amperage flowing through the battery. This also means it is unable to measure the power consumption of non-software components like the Wi-Fi unit, the backlight or the Bluetooth unit.
The most exact, but hardest-to-do way to do these measurements have so far been connecting an Ampere counter between the battery and the PDA. This is very hard - almost impossible without using a dummy battery (if you wanted to refrain from soldering wires into your PDA). Note that, while it's much easier to do, inserting an Ampere counter in the external power cord is a far worse idea because, then, you would not only measure the real battery load, but also the additional housekeeping functionality load like charging the battery, which would cause some excess power consumption even when the battery is fully topped up.
Now, with the release of acbPowerMeter, the situation has completely changed. You only need to download acbPowerMeter, install and start it (it will only work on WM2003+ devices - don't bother installing it on Pocket PC 2002 ones) and see what happens when you start a program in the foreground, en/disable a, say, wireless unit in your PDA or in/decrease the backlight level. You must make sure your PDA is not connected to any power source (including USB-based recharging). Note that the application will not work (will only display 0 mA's no matter what you do) with several PDA models - for example, the iPAQ 2210 (example screenshot here (http://www.winmobiletech.com/042006BatteryConsumption/0mAWithManyDevices.bmp.png)). With other PDA's (for example, the HTC Wizard (AKA i-mate K-jam, Qtek 9100, MDA Vario, T-Mobile MDA)), the value it displays will be rarely correct and will never change, no matter what you do, unless you soft reset your PDA (then, a new value will be displayed - but it not necessarily will be correct either).
acbPowerMeter may run in the background. It will interfere in very few applications (it was only Georally EX that has crashed in my tests) and will cause only minimal overhead (you may encounter some really short pauses every 10 seconds). After running the to-be-tested application/game for 30-40 seconds at least, you can just terminate it, return to acbPowerMeter and evaluate the results - it will show you the Amperage used when the tested application was active.
For example, in the following screenshot (showing playing The Great Gold Rush, the excellent multiplayer game (http://www.pocketpcmag.com/blogs/menneisyys/102005MPPPCGames.asp) in multiplayer, Wi-Fi P2P mode),
http://www.winmobiletech.com/042006BatteryConsumption/PL720WiFiP2PAndGoldRush.png
there're two distinct phases. The first phase, which (almost constantly) used about 290 mA for about 3 minutes, and the second one, which took slightly over a minute and resulted in taking about 500 mA. Note that the horizontal axis shows the time and the vertical axis shows the Amperage. (Note that you'll need to check out the thin line; the thick only shows an average and, because of the peaks, is pretty useless for our purposes.)
Another example is as follows:
http://www.winmobiletech.com/042006BatteryConsumption/PL720CPUSpeedIncrease.png
This shows the effect of increasing the CPU clock frequency. In the first two phases (separated by a comparatively large, ~110 mA peak) for about one and a half minute, the CPU ran on the lowest frequency and, therefore, consumed the least power (the Amperage was about 78 mA). The second phase (showing the effects of increasing the CPU clock to the middle value - with the Pocket Loox 720, 400 MHz) took slighltly less than a minute and shows that, then, the amperage was slightly over 100 mA (which is considerably higher than the 78 mA of the lowest CPU frequency). Finally, the third phase, showing about 130 mA and lasting from about the third to the fourth minute, shows the effect of increasing the CPU clock to the maximal 520 MHz.
Note that
* there will always be some unavoidable peaks in the graphs, related to task switching. When you switch tasks (for example, open Start menu and start a new program or switch back to acbPowerMeter), the CPU will have a heavy load resulting in a considerably higher battery usage. Don't let these peaks mislead you. To make sure they don't have an influence on your measurements, giving you false results, give at least 30-40 seconds to every program you want to test (as some kind of cool-down time) so that the peaks introduced by task switching become insignificant and easy-to-separate from the rest of the graph.
* the battery load is only measured every 10 seconds (this is a hardware/operating system-based restriction). This is why the graphical representation of the battery load can't have really good resolution, particularly if you only run tests for 1-2 minutes at most. This also means really huge, but short peaks may not be displayed or at least not at their full height.
Some real-world results
Now, for some real-world measurements, comparing the highly popular WM5 device, the Dell Axim x51v to the best Pocket PC of the previous, WM2003SE generation, the Fujitsu-Siemens Pocket Loox 720 (PL720 for short).
I've made measurements of several categories. First, I've measured the power consumption caused by the hardware; that is,
* the different backlight levels
* Bluetooth-related power consumption:
1. enabling the Bluetooth unit (between 20-35 mA); enabling its being discoverable (nothing - this means they can always be discoverable, it won't cause any excess battery load)
2. setting up and maintaining the BT serial port connection (also see this blog entry (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=608&more=1) on this; minimal or, with the PL720, no overload)
3. setting up and maintaining the Bluetooth Personal Area Network (PAN) connection (see this (http://www.pocketpcthoughts.com/forums/viewtopic.php?p=371262) (alternatives: iPAQ HQ (http://www.ipaqhq.com/forums/showthread.php?p=117792), AximSite (http://www.aximsite.com/boards/showthread.php?p=869373), PPC Magazine (http://www.pocketpcmag.com/forum/topic.asp?TOPIC_ID=18904), FirstLoox (http://www.firstloox.org/forums/showthread.php?p=42475)) article for more info on BT PAN)
4. starting and using the ThinkOutside StowAway BT keyboard (driver version 4.3). As can be seen in the screenshot (the first few dozen seconds have been taken with activated BT unit but without an activated keyboard connection; then, the local peak shows I've started typing on the keyboard that, in turn, connected to the Pocket PC - this is why the processing overhead), there is almost no additional battery drain if you use the ThinkOutside StowAway keyboard.
Please note that, with the Dell Axim x51v, I've measured the power consumption of both the official Microsoft BT stack and that of the Widcomm BT stack created by the AximSite community (http://www.aximsite.com/boards/showthread.php?t=114983) (as of version 02). As it has turned out, the self-created Widcomm BT stack has exactly the same power consumption as the official Microsoft one. (Note that the BT PAN measurement on the x51v - as the MS BT stack doesn't support BT PAN - has been made using the Widcomm stack only).
*The additional battery load caused by enabled Infrared (IrDA) receive in Settings/System/Power/Beam (or Settings/System/Connections/Beam) (http://www.winmobiletech.com/042006BatteryConsumption/IrDAAutoRcv.png). As can be seen, you can safely keep it enabled - it doesn't cause any additional (measurable) battery load. This is really-really useful to know if you often receive files/business cards via infrared and, due to the common misconception, always switch off your infrared auto-receive between these sessions. As can be seen, the additional battery load may have been an issue a long time ago, with old PDA's - but not now. One thing is certain: with current, up-to-date PDA's (or, at least not the ones I've tested. Your mileage can vary - please test your IrDA auto-receive to be absolutely sure), you don't need to switch it off to conserve power.
*Wi-Fi: searching and in peer-to-peer (P2P) mode (see this article (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=588&more=1) on the latter). As can be seen, if your Wi-Fi unit is actively searching networks, it will have slightly higher power consumption than when it has already found one and connected to it. (Incidentally, as Wireless Zero Configuration, the built-in Wi-Fi networking utility in all WM2003+ devices, still keeps searching for new networks even if there's already an active connection resulting in (some) additional battery drain and slightly descreased maximal throughput rate, you may want to consider disabling it entirely. See this thread (http://www.pocketpcmag.com/forum/topic.asp?TOPIC_ID=15393) for more info on all this. Then, however, you won't be able to use popular networking tools like Hitchhiker any more.) It can also clearly be seen that enabling power save does not decrease the battery load at all in P2P mode on these two PDA's. (It may decrease the load when connecting to an Access Point in infrastructure mode, though.)
* Finally, I've tested the additional load the maximal volume level causes on the two devices. Also take into account that the PL720 has an exceptionally loud speaker (much louder than that of even the iPAQ 36xx/37xx series, which, in turn, is still much louder than, say, the other iPAQ's and the x50v/x51v - also see this blog entry (http://www.pocketpcmag.com/blogs/index.php?blog=3&title=is_the_x50v_loud_what_about_the_ipaq_hx4&more=1) on these questions), unlike the x51v. This is why the additional battery drain caused by the maximal sound level is considerably higher on the PL720 than on the x51v. The test was done using TapzMania, the excellent multiplayer game (http://www.pocketpcmag.com/blogs/menneisyys/102005MPPPCGames.asp) featuring probably the best music I've ever heard in a Pocket PC game. I've played the first and third phase (between 340-360 mA) with speaker on and the second (about 295 mA) with speaker off. This shows cranking up the volume will only slightly increase the power usage.
Note that these values were measured with the low-impedance (loud) built-in speaker; when you connect a high-impedance (amplified) speaker to your PDA, the additional battery drain will be much lower, even if you fully crank up the volume. That is, the ~50 mA additional battery load shows the worst possible scenario.
The comparison chart of the hardware tests
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t1.html). Sorry guys and gals - the forum engine doesn't make it possible to include HTML tables in here. Therefore, you'll need to click these chart links. Note that the article version posted to my blog (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=706&more=1&c=1&tb=1&pb=1) has in-line tables.
All results are in milliampers. The tests were made using the CPU in automatic mode and, unless testing the additional battery drain of the backlight, the Bluetooth and the Wi-Fi unit, using the minimal backlight level and switching off all wireless units.
These tests show that...
* The "hacked" Widcomm BT stack doesn't at all consume more power than the (default) Microsoft BT stack on the x51v.
* You don't need to disable the IrDA unit in your PDA in Settings/System/Power/Beam (or Connections/Beam) (http://www.winmobiletech.com/042006BatteryConsumption/IrDAAutoRcv.png) - keeping it enabled doesn't consume almost any power in current Pocket PC's.
* The preferred way of local, peer-to-peer networking connections, Bluetooth PAN, increases power consumption a bit, especially if you take into account its periodic power consumption peaks, which may be related to some kind of a networking housekeeping and may be, probably, decreased by tweaking the networking parameters. The additional power consumption of BT PAN is still very low compared to peer-to-peer Wi-Fi connections. The latter never make use of any kind of adaptive power saving mode in P2P connections, even if you enable power saving (a Pocket Loox 720 example screenshot here (http://www.winmobiletech.com/042006BatteryConsumption/PL720WiFiAutoPower.png) and a x51v screenshot here (http://www.winmobiletech.com/042006BatteryConsumption/x51vWiFiAutoPower.png) of doing this).
* Pocket PC PDA's don't reduce their transmit power even when the connected Pocket PC's are adjacent to each other (and, therefore, they all sense the signal power more than sufficient as can be seen for example in this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/GoodSignalLevelWizard.png)). An example of this can be seen in this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/PL720WiFiP2PAndGoldRush.png), taken on the PL720 and (in the first half) showing a(n idle) P2P Wi-Fi connection. During it, I've taken the other PDA (the HTC Wizard) the PL720 was connected to via Wi-Fi p2p from the close vicinity (very good signal level) to far away, behind walls (poor signal level) to test whether there is any power transmission fine-tuning. As can be seen, there isn't - the Wi-Fi unit in the Pocket Loox still used the same 288 mA as in the case of excellent circumstances (that is, the two Pocket PC's being next to one another). Incidentally, the ~500 mA section in the screenshot shows the added ~200 mA battery load of playing the excellent multiplayer game Gold Rush (http://www.pocketpcmag.com/blogs/menneisyys/102005MPPPCGames.asp) in multiplayer mode. As can be seen, playing action games using Wi-Fi P2P will certainly kill the battery even on a device like the Pocket Loox 720, not to speak about much more power-hungry ones like the Dell Axim x51v. With the latter, some 700 mA would have been used at 520 MHz (and even more in automatic/turbo mode), which is really a lot.
Hardware manufacturers/Wi-Fi driver writers, you should add dynamic adaptation into varying environmental conditions - for example, changing the transmit power and, therefore, reducing the battery load whenever possible!
Note that it's only with manually reducable Wi-Fi power (see the Wi-Fi power slider in the Wizard - see this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/HTCWizardWiFIPowerSet.png)) that you can have any effect on the power consumption of built-in Wi-Fi units. It works, therefore, the same way as the Class I / Class II (software) switch in the only Pocket PC PDA that (officially) supports long-distance Class I Bluetooth, the old Fujitsu-Siemens Pocket Loox 600.
Resco Audio Recorder tests
Resco Audio Recorder has an excellent, speech-targeted vocoder, Speex, which makes it possible to record conversations at a much higher quality (at a given bitrate) than with MP3, WMA or OGG, the generic audio codecs widely used today. The only downside of it is the vastly increased power consumption, which I also emphasized in my article on Pocket PC-based sound recorders (http://pocketpcmag.com/blogs/menneisyys/112005SoundRecorderApps.asp).
In these tests, I've examined whether decreasing the CPU speed helps in reducing the battery load while recording with Resco AR. As can be seen, the answer is yes. If operated in power save mode (that is, at 208 MHz on most current devices), the application uses considerably less battery power than in, say, the default automatic mode. Then, however, closing a recording will sometimes take as much as half a minute (as opposed to closing a recording instantly in faster modes), even when using fast cards to write to.
Also note that, while underclocking the test Pocket PC's to 208 MHz clearly reduced the battery drain (and, consequently, increased battery life), this will not be the rule of thumb with all devices. For example, on the (comparatively slow - 400 MHz only) "old" HP iPAQ 2210, underclocking the device to 200 MHz (with XCPUScalar) only resulted in slight battery life improvement: about 1.5% an hour. While recording consumes about 18.5% of the battery life on my test 2210 (its battery is rather aged - bought the unit almost three years ago and am still using the same battery) an hour without underclocking, on an underclocked (and, therefore, really sluggish, particularly as far as closing a recording is concerned) iPAQ 2210, the charge level decrease is about 17% an hour - not much lower than using the default, 400 MHz CPU clock frequency.
Note that I used my standard quality 3, 32 kHz, 18 kbps, disabled screen configuration in the test. The previous remarks (switched off wireless) also apply here.
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t2.html).
Games
As is widely known, most games are really CPU-intensive, which also means huge battery drain. In these tests, I've tested some games (and, in addition, the TCPMP multimedia player (http://tcpmp.corecodec.org/about)) in this respect. I paid special attention to the effects of underclocking the PDA to see whether a given game (or TCPMP) still remains playable/usable on a (heavily) underclocked Pocket PC.
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t3.html).
The previous remarks (switched off wireless, minimal backlight) also apply here. (Note that the very contrasty screen of the PL720 makes it essential, in most cases, to maximize the backlight level when watching videos on the device with TCPMP, unlike with the x51v. The same stands for any device that has (color reproduction and contrast-wise) the same SONY screen as the PL720 - for example, the iPAQ hx4700. Sometimes, the radically reduced contrast and washed-out screen may pay off, battery usage-wise.)
As can be seen in here, Dell Axims' generally consuming more power than really high-end, (much) more expensive Pocket PC's (like the PL720 or the HP iPAQ hx4700): the Axim consumed considerably more power than the Loox, even at running the same 520 MHz, in all gaming tests. This is certainly a bad news for everyone that considers their x51v as, like me, mostly a gaming PDA (or wants to buy one to play games) because of the, in games, really excellent and unmatched Intel 2700g graphics chip.
Dell Axims, on the other hand, excel at everything (mostly - this means most computations are done by the 2700g chip, not the CPU) relying on their 2700g, battery drain-wise. The TCPMP and the Toy Golf test clearly show this.
With TCPMP, if you use the built-in video decoding capabilities of the 2700g (to do this, you won't need to do anything: the i2700g Decoder mode is the default on 2700g-enabled Pocket PC's), you can achieve even better battery consumption results than with otherwise much better Pocket PC's because the latter need to rely on CPU-only, non-2700g-based decoding.
In addition, you can also further reduce the battery load (by some 0-50 mA, depending on the video - with videos that have fast, frequent scene changes and moving view, more power is consumed; with videos that don't do the same, less) by underclocking the x51v to 208 MHz (and, of course, sticking with the default 2700g Decoder mode). Then, unlike with software- or XScale-based decoding non-2700g-equipped PDA's need to use, the videos are still played smoothly. (Now, I wish the screen of the x50v/x51v were a tad easier on the eyes in Landscape mode - see this article (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=578&more=1) for more information on this.)
In general, the following can be stated: When you constantly use an application that really utilizes the CPU and want to increase battery life, you should always try reducing your CPU clock speed, especially on 624 MHz Pocket PC's. Even the 624 -> 520 MHz switch can result in significant battery life savings, while the speed degradation won't be visible in most applications/games.
Idle CPU power consumption
As has already been pointed out, the bigger the CPU load and clock frequency, the higher the battery drain. In these tests, I've measured the battery drain caused by each (officially) settable CPU clock frequency on both test devices while the CPU is in idle mode (that is, you don't run anything on the PDA). Don't forget to check out the screenshots in the first row of the chart for an all-in-one representation of all these measurement values!
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t4.html).
As can clearly be seen, the PL720 has beaten the x51v in this test too. It should also be stated that if you use anything else than the Automatic / Power Save (208 MHz) modes, your battery drain will be higher even when the CPU is idling than in the Automatic / Power Save mode because, then, the CPU won't be automatically run on lower clock speeds when there's nothing to do. Always using any non-automatic/non-208 MHz mode is, therefore, discouraged unless you want to play games or other CPU-intensive stuff at a given clock speed. Then, you may indeed be better off with explicitly switching to, say, 520 MHz (on the x51v) or 400 MHz (on the PL720) to reduce the in-game battery load. If the CPU is often idling (because you mostly use your PDA for light PIM stuff) and you don't always want to manually adjust the CPU clock speed, it's better to leave it in the automatic CPU clock speed mode.
CPU speed control
OK, I understand - now, how can I control the clock frequency of the PDA - you may ask.
With several Pocket PC's, this feature is already built-in. For example, with the PL720, you will need to go to Start/Settings/System/Processor (http://www.winmobiletech.com/042006BatteryConsumption/PL720CPUSpeedSetDialog.png); here, "Turbo mode"stands for (the maximal) 520 MHz, "Standard"to 400 MHz and "Power Saving"to 208 MHz.
With the X51v, you'll need to go to Start/Settings/System/Power/Processor (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-Auto.png); here, "Maximal performance" stands for (the maximal) 624 MHz (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-624.png), "Normal" to 520 MHz (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-520.png) and "Power-Saving" to 208 MHz (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-208.png).
Finally, with XScale devices that don't offer any kind of built-in CPU speed setting capabilities, you may want to turn to the already-mentioned XCPUScalar (http://www.immiersoft.com/download.htm). It's very easy to use - you just move the slider from the starting position (in this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/XCPUScalar-1.bmp.png), for example, 400 MHz) to a lower value (for example, 200 MHz (http://www.winmobiletech.com/042006BatteryConsumption/XCPUScalar-2.bmp.png)).
Note that you'll need to download a separate version for PDA's equipped with 200 MHz XScale CPU's (the old HP iPAQ 1910/1915, for example) and you'll need to get an entirely different, free over/underclocker utility, JS Overclock, for StrongARM-based (a lot of PPC2k2 devices have been shipped with StrongARM CPU's instead of XScale: for example, the iPAQ 31xx/36xx/37xx/38xx series) PDA's available here (http://www.pocketpcfreewares.com/en/index.php?soft=131).
Also note that there's another XScale/Samsung overclock utility, Pocket Hack Master 2005 (http://www.antontomov.com/products/pocket_hack_master.php). As can be seen, it can not only be used with XScale, but also with Samsung CPU's running in a lot of current (and past) PDA's. It also offers a lot of additional tweaking capabilities; please read this forum (http://www.antontomov.com/forum/index.php?showforum=39) for more info/user opinions.
Finally, people having Pocket PC's with the Texas Instruments (TI) CPU (for example, the HTC Wizard (i-mate K-jam, Qtek 9100, MDA Vario, T-Mobile MDA)) will want to cast a glance at the only working solution for their device, OmapClock (http://forum.xda-developers.com/viewtopic.php?t=40284) (also see this review (http://www.pocketpcmag.com/blogs/index.php?blog=3&title=htc_wizard_users_why_is_it_worth_upgradi&more=1&c=1&tb=1&pb=1).)
Recommended links
In addition to the already-linked articles (make sure you click the links in the article above), I really recommend the xv6700 bechmarks thread "Battery Power Consumption / Usage Tests - Comparisons / Observations - Contribute Plz (http://pdaphonehome.com/forums/showthread.php?t=68941)". The results there are pretty much the same as my measurements: for example, enabling Bluetooth doesn't add much additional battery drain (actually, it's 5mA only with the xv6700, that is, considerably lower than that of the PL720 / x51v); this means you can keep the BT unit on in your new-generation (WM5) HTC Phone Edition devices all day long.
Also, I show you how much power the backlight, the wireless units, the buzzer/speaker etc. consume. A lot of plain wrong urban legends (for example, the question of disabling the infrared receive or the Bluetooth unit) are still being widely considered thruth - in this article, I explain why they are not true (any more).
You may already have heard that the main processors (Central Processing Units, CPU's) in the Pocket PC consume more power (and, therefore, chew through your battery faster) when used at higher clock frequencies. Many people even underclock (that is, run at a lower frequency - for example, 208 MHz instead of the standard 400/520/624 MHz) their PDA's to achieve better battery life. Also, you may well know if you enable certain (mostly, wireless) functionality on your PDA, battery power will be consumed far faster.
Now that a really excellent application for the exact measuring of the battery load, acbPowerMeter, has been released (http://pdaphonehome.com/forums/attachment.php?attachmentid=8188&d=1142835308) (alternative download here (http://www.winmobiletech.com/sekalaiset/acbPowerMeter.CAB)), it has become much easier to exactly see how much milliampers the battery load at a given time is.
Before acbPowerMeter, measuring the battery load was incomparably harder. The easiest, but slowest way was topping (charging until it's full) the battery and, then, waiting for some hours running the given application / enabling the to-be-tested unit (for example, the Bluetooth / Wi-Fi unit) on the PDA and, then, checking how many percents of the battery charge remained.
As the battery load depends not only on what hardware units are enabled and what the backlight level is but also on the CPU load (that is, how many percents of the CPU time a particular application uses), testing the battery drain of (just) a given program/game has been possible using other means. You could use the CPU usage display in ImmierSoft XCPUScalar (http://www.immiersoft.com/) (if it actually worked on your particular PDA model and didn't crash it). It displays an icon on the taskbar, showing the actual CPU load. Of course, it can't be used if the taskbar is hidden (for example, in full-screen titles like most games) and it delivers unreliable/useless, not updated results if the CPU load is over a certain threshold (about 85-90%). This was the way I've measured the CPU load of the software titles I've tested so far (see for example the article Everything you may need to know about sound recording on the Pocket PC (http://pocketpcmag.com/blogs/menneisyys/112005SoundRecorderApps.asp), the article on voice and remote controllers (http://www.pocketpcmag.com/blogs/index.php?blog=3&title=what_voice_controller_apps_are_there&more=1&c=1&tb=1&pb=1) for some numerical results measured this way).
As has been pointed out, XCPUScalar only displays the CPU load caused by the running programs and not the actual amperage flowing through the battery. This also means it is unable to measure the power consumption of non-software components like the Wi-Fi unit, the backlight or the Bluetooth unit.
The most exact, but hardest-to-do way to do these measurements have so far been connecting an Ampere counter between the battery and the PDA. This is very hard - almost impossible without using a dummy battery (if you wanted to refrain from soldering wires into your PDA). Note that, while it's much easier to do, inserting an Ampere counter in the external power cord is a far worse idea because, then, you would not only measure the real battery load, but also the additional housekeeping functionality load like charging the battery, which would cause some excess power consumption even when the battery is fully topped up.
Now, with the release of acbPowerMeter, the situation has completely changed. You only need to download acbPowerMeter, install and start it (it will only work on WM2003+ devices - don't bother installing it on Pocket PC 2002 ones) and see what happens when you start a program in the foreground, en/disable a, say, wireless unit in your PDA or in/decrease the backlight level. You must make sure your PDA is not connected to any power source (including USB-based recharging). Note that the application will not work (will only display 0 mA's no matter what you do) with several PDA models - for example, the iPAQ 2210 (example screenshot here (http://www.winmobiletech.com/042006BatteryConsumption/0mAWithManyDevices.bmp.png)). With other PDA's (for example, the HTC Wizard (AKA i-mate K-jam, Qtek 9100, MDA Vario, T-Mobile MDA)), the value it displays will be rarely correct and will never change, no matter what you do, unless you soft reset your PDA (then, a new value will be displayed - but it not necessarily will be correct either).
acbPowerMeter may run in the background. It will interfere in very few applications (it was only Georally EX that has crashed in my tests) and will cause only minimal overhead (you may encounter some really short pauses every 10 seconds). After running the to-be-tested application/game for 30-40 seconds at least, you can just terminate it, return to acbPowerMeter and evaluate the results - it will show you the Amperage used when the tested application was active.
For example, in the following screenshot (showing playing The Great Gold Rush, the excellent multiplayer game (http://www.pocketpcmag.com/blogs/menneisyys/102005MPPPCGames.asp) in multiplayer, Wi-Fi P2P mode),
http://www.winmobiletech.com/042006BatteryConsumption/PL720WiFiP2PAndGoldRush.png
there're two distinct phases. The first phase, which (almost constantly) used about 290 mA for about 3 minutes, and the second one, which took slightly over a minute and resulted in taking about 500 mA. Note that the horizontal axis shows the time and the vertical axis shows the Amperage. (Note that you'll need to check out the thin line; the thick only shows an average and, because of the peaks, is pretty useless for our purposes.)
Another example is as follows:
http://www.winmobiletech.com/042006BatteryConsumption/PL720CPUSpeedIncrease.png
This shows the effect of increasing the CPU clock frequency. In the first two phases (separated by a comparatively large, ~110 mA peak) for about one and a half minute, the CPU ran on the lowest frequency and, therefore, consumed the least power (the Amperage was about 78 mA). The second phase (showing the effects of increasing the CPU clock to the middle value - with the Pocket Loox 720, 400 MHz) took slighltly less than a minute and shows that, then, the amperage was slightly over 100 mA (which is considerably higher than the 78 mA of the lowest CPU frequency). Finally, the third phase, showing about 130 mA and lasting from about the third to the fourth minute, shows the effect of increasing the CPU clock to the maximal 520 MHz.
Note that
* there will always be some unavoidable peaks in the graphs, related to task switching. When you switch tasks (for example, open Start menu and start a new program or switch back to acbPowerMeter), the CPU will have a heavy load resulting in a considerably higher battery usage. Don't let these peaks mislead you. To make sure they don't have an influence on your measurements, giving you false results, give at least 30-40 seconds to every program you want to test (as some kind of cool-down time) so that the peaks introduced by task switching become insignificant and easy-to-separate from the rest of the graph.
* the battery load is only measured every 10 seconds (this is a hardware/operating system-based restriction). This is why the graphical representation of the battery load can't have really good resolution, particularly if you only run tests for 1-2 minutes at most. This also means really huge, but short peaks may not be displayed or at least not at their full height.
Some real-world results
Now, for some real-world measurements, comparing the highly popular WM5 device, the Dell Axim x51v to the best Pocket PC of the previous, WM2003SE generation, the Fujitsu-Siemens Pocket Loox 720 (PL720 for short).
I've made measurements of several categories. First, I've measured the power consumption caused by the hardware; that is,
* the different backlight levels
* Bluetooth-related power consumption:
1. enabling the Bluetooth unit (between 20-35 mA); enabling its being discoverable (nothing - this means they can always be discoverable, it won't cause any excess battery load)
2. setting up and maintaining the BT serial port connection (also see this blog entry (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=608&more=1) on this; minimal or, with the PL720, no overload)
3. setting up and maintaining the Bluetooth Personal Area Network (PAN) connection (see this (http://www.pocketpcthoughts.com/forums/viewtopic.php?p=371262) (alternatives: iPAQ HQ (http://www.ipaqhq.com/forums/showthread.php?p=117792), AximSite (http://www.aximsite.com/boards/showthread.php?p=869373), PPC Magazine (http://www.pocketpcmag.com/forum/topic.asp?TOPIC_ID=18904), FirstLoox (http://www.firstloox.org/forums/showthread.php?p=42475)) article for more info on BT PAN)
4. starting and using the ThinkOutside StowAway BT keyboard (driver version 4.3). As can be seen in the screenshot (the first few dozen seconds have been taken with activated BT unit but without an activated keyboard connection; then, the local peak shows I've started typing on the keyboard that, in turn, connected to the Pocket PC - this is why the processing overhead), there is almost no additional battery drain if you use the ThinkOutside StowAway keyboard.
Please note that, with the Dell Axim x51v, I've measured the power consumption of both the official Microsoft BT stack and that of the Widcomm BT stack created by the AximSite community (http://www.aximsite.com/boards/showthread.php?t=114983) (as of version 02). As it has turned out, the self-created Widcomm BT stack has exactly the same power consumption as the official Microsoft one. (Note that the BT PAN measurement on the x51v - as the MS BT stack doesn't support BT PAN - has been made using the Widcomm stack only).
*The additional battery load caused by enabled Infrared (IrDA) receive in Settings/System/Power/Beam (or Settings/System/Connections/Beam) (http://www.winmobiletech.com/042006BatteryConsumption/IrDAAutoRcv.png). As can be seen, you can safely keep it enabled - it doesn't cause any additional (measurable) battery load. This is really-really useful to know if you often receive files/business cards via infrared and, due to the common misconception, always switch off your infrared auto-receive between these sessions. As can be seen, the additional battery load may have been an issue a long time ago, with old PDA's - but not now. One thing is certain: with current, up-to-date PDA's (or, at least not the ones I've tested. Your mileage can vary - please test your IrDA auto-receive to be absolutely sure), you don't need to switch it off to conserve power.
*Wi-Fi: searching and in peer-to-peer (P2P) mode (see this article (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=588&more=1) on the latter). As can be seen, if your Wi-Fi unit is actively searching networks, it will have slightly higher power consumption than when it has already found one and connected to it. (Incidentally, as Wireless Zero Configuration, the built-in Wi-Fi networking utility in all WM2003+ devices, still keeps searching for new networks even if there's already an active connection resulting in (some) additional battery drain and slightly descreased maximal throughput rate, you may want to consider disabling it entirely. See this thread (http://www.pocketpcmag.com/forum/topic.asp?TOPIC_ID=15393) for more info on all this. Then, however, you won't be able to use popular networking tools like Hitchhiker any more.) It can also clearly be seen that enabling power save does not decrease the battery load at all in P2P mode on these two PDA's. (It may decrease the load when connecting to an Access Point in infrastructure mode, though.)
* Finally, I've tested the additional load the maximal volume level causes on the two devices. Also take into account that the PL720 has an exceptionally loud speaker (much louder than that of even the iPAQ 36xx/37xx series, which, in turn, is still much louder than, say, the other iPAQ's and the x50v/x51v - also see this blog entry (http://www.pocketpcmag.com/blogs/index.php?blog=3&title=is_the_x50v_loud_what_about_the_ipaq_hx4&more=1) on these questions), unlike the x51v. This is why the additional battery drain caused by the maximal sound level is considerably higher on the PL720 than on the x51v. The test was done using TapzMania, the excellent multiplayer game (http://www.pocketpcmag.com/blogs/menneisyys/102005MPPPCGames.asp) featuring probably the best music I've ever heard in a Pocket PC game. I've played the first and third phase (between 340-360 mA) with speaker on and the second (about 295 mA) with speaker off. This shows cranking up the volume will only slightly increase the power usage.
Note that these values were measured with the low-impedance (loud) built-in speaker; when you connect a high-impedance (amplified) speaker to your PDA, the additional battery drain will be much lower, even if you fully crank up the volume. That is, the ~50 mA additional battery load shows the worst possible scenario.
The comparison chart of the hardware tests
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t1.html). Sorry guys and gals - the forum engine doesn't make it possible to include HTML tables in here. Therefore, you'll need to click these chart links. Note that the article version posted to my blog (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=706&more=1&c=1&tb=1&pb=1) has in-line tables.
All results are in milliampers. The tests were made using the CPU in automatic mode and, unless testing the additional battery drain of the backlight, the Bluetooth and the Wi-Fi unit, using the minimal backlight level and switching off all wireless units.
These tests show that...
* The "hacked" Widcomm BT stack doesn't at all consume more power than the (default) Microsoft BT stack on the x51v.
* You don't need to disable the IrDA unit in your PDA in Settings/System/Power/Beam (or Connections/Beam) (http://www.winmobiletech.com/042006BatteryConsumption/IrDAAutoRcv.png) - keeping it enabled doesn't consume almost any power in current Pocket PC's.
* The preferred way of local, peer-to-peer networking connections, Bluetooth PAN, increases power consumption a bit, especially if you take into account its periodic power consumption peaks, which may be related to some kind of a networking housekeeping and may be, probably, decreased by tweaking the networking parameters. The additional power consumption of BT PAN is still very low compared to peer-to-peer Wi-Fi connections. The latter never make use of any kind of adaptive power saving mode in P2P connections, even if you enable power saving (a Pocket Loox 720 example screenshot here (http://www.winmobiletech.com/042006BatteryConsumption/PL720WiFiAutoPower.png) and a x51v screenshot here (http://www.winmobiletech.com/042006BatteryConsumption/x51vWiFiAutoPower.png) of doing this).
* Pocket PC PDA's don't reduce their transmit power even when the connected Pocket PC's are adjacent to each other (and, therefore, they all sense the signal power more than sufficient as can be seen for example in this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/GoodSignalLevelWizard.png)). An example of this can be seen in this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/PL720WiFiP2PAndGoldRush.png), taken on the PL720 and (in the first half) showing a(n idle) P2P Wi-Fi connection. During it, I've taken the other PDA (the HTC Wizard) the PL720 was connected to via Wi-Fi p2p from the close vicinity (very good signal level) to far away, behind walls (poor signal level) to test whether there is any power transmission fine-tuning. As can be seen, there isn't - the Wi-Fi unit in the Pocket Loox still used the same 288 mA as in the case of excellent circumstances (that is, the two Pocket PC's being next to one another). Incidentally, the ~500 mA section in the screenshot shows the added ~200 mA battery load of playing the excellent multiplayer game Gold Rush (http://www.pocketpcmag.com/blogs/menneisyys/102005MPPPCGames.asp) in multiplayer mode. As can be seen, playing action games using Wi-Fi P2P will certainly kill the battery even on a device like the Pocket Loox 720, not to speak about much more power-hungry ones like the Dell Axim x51v. With the latter, some 700 mA would have been used at 520 MHz (and even more in automatic/turbo mode), which is really a lot.
Hardware manufacturers/Wi-Fi driver writers, you should add dynamic adaptation into varying environmental conditions - for example, changing the transmit power and, therefore, reducing the battery load whenever possible!
Note that it's only with manually reducable Wi-Fi power (see the Wi-Fi power slider in the Wizard - see this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/HTCWizardWiFIPowerSet.png)) that you can have any effect on the power consumption of built-in Wi-Fi units. It works, therefore, the same way as the Class I / Class II (software) switch in the only Pocket PC PDA that (officially) supports long-distance Class I Bluetooth, the old Fujitsu-Siemens Pocket Loox 600.
Resco Audio Recorder tests
Resco Audio Recorder has an excellent, speech-targeted vocoder, Speex, which makes it possible to record conversations at a much higher quality (at a given bitrate) than with MP3, WMA or OGG, the generic audio codecs widely used today. The only downside of it is the vastly increased power consumption, which I also emphasized in my article on Pocket PC-based sound recorders (http://pocketpcmag.com/blogs/menneisyys/112005SoundRecorderApps.asp).
In these tests, I've examined whether decreasing the CPU speed helps in reducing the battery load while recording with Resco AR. As can be seen, the answer is yes. If operated in power save mode (that is, at 208 MHz on most current devices), the application uses considerably less battery power than in, say, the default automatic mode. Then, however, closing a recording will sometimes take as much as half a minute (as opposed to closing a recording instantly in faster modes), even when using fast cards to write to.
Also note that, while underclocking the test Pocket PC's to 208 MHz clearly reduced the battery drain (and, consequently, increased battery life), this will not be the rule of thumb with all devices. For example, on the (comparatively slow - 400 MHz only) "old" HP iPAQ 2210, underclocking the device to 200 MHz (with XCPUScalar) only resulted in slight battery life improvement: about 1.5% an hour. While recording consumes about 18.5% of the battery life on my test 2210 (its battery is rather aged - bought the unit almost three years ago and am still using the same battery) an hour without underclocking, on an underclocked (and, therefore, really sluggish, particularly as far as closing a recording is concerned) iPAQ 2210, the charge level decrease is about 17% an hour - not much lower than using the default, 400 MHz CPU clock frequency.
Note that I used my standard quality 3, 32 kHz, 18 kbps, disabled screen configuration in the test. The previous remarks (switched off wireless) also apply here.
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t2.html).
Games
As is widely known, most games are really CPU-intensive, which also means huge battery drain. In these tests, I've tested some games (and, in addition, the TCPMP multimedia player (http://tcpmp.corecodec.org/about)) in this respect. I paid special attention to the effects of underclocking the PDA to see whether a given game (or TCPMP) still remains playable/usable on a (heavily) underclocked Pocket PC.
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t3.html).
The previous remarks (switched off wireless, minimal backlight) also apply here. (Note that the very contrasty screen of the PL720 makes it essential, in most cases, to maximize the backlight level when watching videos on the device with TCPMP, unlike with the x51v. The same stands for any device that has (color reproduction and contrast-wise) the same SONY screen as the PL720 - for example, the iPAQ hx4700. Sometimes, the radically reduced contrast and washed-out screen may pay off, battery usage-wise.)
As can be seen in here, Dell Axims' generally consuming more power than really high-end, (much) more expensive Pocket PC's (like the PL720 or the HP iPAQ hx4700): the Axim consumed considerably more power than the Loox, even at running the same 520 MHz, in all gaming tests. This is certainly a bad news for everyone that considers their x51v as, like me, mostly a gaming PDA (or wants to buy one to play games) because of the, in games, really excellent and unmatched Intel 2700g graphics chip.
Dell Axims, on the other hand, excel at everything (mostly - this means most computations are done by the 2700g chip, not the CPU) relying on their 2700g, battery drain-wise. The TCPMP and the Toy Golf test clearly show this.
With TCPMP, if you use the built-in video decoding capabilities of the 2700g (to do this, you won't need to do anything: the i2700g Decoder mode is the default on 2700g-enabled Pocket PC's), you can achieve even better battery consumption results than with otherwise much better Pocket PC's because the latter need to rely on CPU-only, non-2700g-based decoding.
In addition, you can also further reduce the battery load (by some 0-50 mA, depending on the video - with videos that have fast, frequent scene changes and moving view, more power is consumed; with videos that don't do the same, less) by underclocking the x51v to 208 MHz (and, of course, sticking with the default 2700g Decoder mode). Then, unlike with software- or XScale-based decoding non-2700g-equipped PDA's need to use, the videos are still played smoothly. (Now, I wish the screen of the x50v/x51v were a tad easier on the eyes in Landscape mode - see this article (http://www.pocketpcmag.com/blogs/index.php?blog=3&p=578&more=1) for more information on this.)
In general, the following can be stated: When you constantly use an application that really utilizes the CPU and want to increase battery life, you should always try reducing your CPU clock speed, especially on 624 MHz Pocket PC's. Even the 624 -> 520 MHz switch can result in significant battery life savings, while the speed degradation won't be visible in most applications/games.
Idle CPU power consumption
As has already been pointed out, the bigger the CPU load and clock frequency, the higher the battery drain. In these tests, I've measured the battery drain caused by each (officially) settable CPU clock frequency on both test devices while the CPU is in idle mode (that is, you don't run anything on the PDA). Don't forget to check out the screenshots in the first row of the chart for an all-in-one representation of all these measurement values!
The comparison chart is here (http://www.winmobiletech.com/042006BatteryConsumption/t4.html).
As can clearly be seen, the PL720 has beaten the x51v in this test too. It should also be stated that if you use anything else than the Automatic / Power Save (208 MHz) modes, your battery drain will be higher even when the CPU is idling than in the Automatic / Power Save mode because, then, the CPU won't be automatically run on lower clock speeds when there's nothing to do. Always using any non-automatic/non-208 MHz mode is, therefore, discouraged unless you want to play games or other CPU-intensive stuff at a given clock speed. Then, you may indeed be better off with explicitly switching to, say, 520 MHz (on the x51v) or 400 MHz (on the PL720) to reduce the in-game battery load. If the CPU is often idling (because you mostly use your PDA for light PIM stuff) and you don't always want to manually adjust the CPU clock speed, it's better to leave it in the automatic CPU clock speed mode.
CPU speed control
OK, I understand - now, how can I control the clock frequency of the PDA - you may ask.
With several Pocket PC's, this feature is already built-in. For example, with the PL720, you will need to go to Start/Settings/System/Processor (http://www.winmobiletech.com/042006BatteryConsumption/PL720CPUSpeedSetDialog.png); here, "Turbo mode"stands for (the maximal) 520 MHz, "Standard"to 400 MHz and "Power Saving"to 208 MHz.
With the X51v, you'll need to go to Start/Settings/System/Power/Processor (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-Auto.png); here, "Maximal performance" stands for (the maximal) 624 MHz (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-624.png), "Normal" to 520 MHz (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-520.png) and "Power-Saving" to 208 MHz (http://www.winmobiletech.com/042006BatteryConsumption/x51vCPUSpeedSetDialog-208.png).
Finally, with XScale devices that don't offer any kind of built-in CPU speed setting capabilities, you may want to turn to the already-mentioned XCPUScalar (http://www.immiersoft.com/download.htm). It's very easy to use - you just move the slider from the starting position (in this screenshot (http://www.winmobiletech.com/042006BatteryConsumption/XCPUScalar-1.bmp.png), for example, 400 MHz) to a lower value (for example, 200 MHz (http://www.winmobiletech.com/042006BatteryConsumption/XCPUScalar-2.bmp.png)).
Note that you'll need to download a separate version for PDA's equipped with 200 MHz XScale CPU's (the old HP iPAQ 1910/1915, for example) and you'll need to get an entirely different, free over/underclocker utility, JS Overclock, for StrongARM-based (a lot of PPC2k2 devices have been shipped with StrongARM CPU's instead of XScale: for example, the iPAQ 31xx/36xx/37xx/38xx series) PDA's available here (http://www.pocketpcfreewares.com/en/index.php?soft=131).
Also note that there's another XScale/Samsung overclock utility, Pocket Hack Master 2005 (http://www.antontomov.com/products/pocket_hack_master.php). As can be seen, it can not only be used with XScale, but also with Samsung CPU's running in a lot of current (and past) PDA's. It also offers a lot of additional tweaking capabilities; please read this forum (http://www.antontomov.com/forum/index.php?showforum=39) for more info/user opinions.
Finally, people having Pocket PC's with the Texas Instruments (TI) CPU (for example, the HTC Wizard (i-mate K-jam, Qtek 9100, MDA Vario, T-Mobile MDA)) will want to cast a glance at the only working solution for their device, OmapClock (http://forum.xda-developers.com/viewtopic.php?t=40284) (also see this review (http://www.pocketpcmag.com/blogs/index.php?blog=3&title=htc_wizard_users_why_is_it_worth_upgradi&more=1&c=1&tb=1&pb=1).)
Recommended links
In addition to the already-linked articles (make sure you click the links in the article above), I really recommend the xv6700 bechmarks thread "Battery Power Consumption / Usage Tests - Comparisons / Observations - Contribute Plz (http://pdaphonehome.com/forums/showthread.php?t=68941)". The results there are pretty much the same as my measurements: for example, enabling Bluetooth doesn't add much additional battery drain (actually, it's 5mA only with the xv6700, that is, considerably lower than that of the PL720 / x51v); this means you can keep the BT unit on in your new-generation (WM5) HTC Phone Edition devices all day long.