“What is the utility of a mobile security application running on a pocket pc that has run out of battery?”

Wireless infrastructure company, Enterprise Air, asked this question when designing its mission-critical wireless surveillance application, Mobile Mugshot. Through testing different hardware configurations, Enterprise Air found what they believe is the optimal hardware configuration for maximizing a Pocket PC’s useful battery life for a WiFi networked application.

For wireless applications that require a persistent connection to the network, we found that older cards performed as well as newer so -called ‘low power’ WLAN cards. We also found that setting the device and network card driver settings to automatic gave the best battery life/usability combination. Third, and contrary to market reviews of the new PocketPC 2002 operating system, we found the older PocketPC 2000 OS lasted just as long as the newer PocketPC 2002. Last, we were impressed with the HP Jornada 568’s remarkably longer battery life although its expandability is limited when compared to the iPAQ’s system of expansion paks.

Equipment used

At the time of this test, we did not have an iPAQ 3800 series, which is reviewed as having longer battery life than its predecessors. In our test, for devices, we used an iPAQ 3630 running PocketPC 2000, an iPAQ 3765 running PocketPC 2002, and an HP Jornada Model 568 running Pocket PC 2002. For network cards, we used the Low Power WLAN 802.11b CF1 network card from Socket, the Wireless Networker 802.11b CF1 network card from Symbol and the D-Link 802.11b CF2 network card. We also used an iPAQ expansion pak with an extra battery.

For test design, please refer to the section below “How We Tested”.

Application Workflow

Our main concern in designing this test was to measure the battery life of a device under real world conditions as opposed to the types of tests run by manufacturers who set the device to idle and measure how long it lasts before shutting off. Since the Mobile Mugshot application polls the network every 30 seconds, we felt it was a good proxy for PocketPC networked applications that need to be used on a constant basis throughout a shift. Imagine a data collection application where the user is entering data into a form every 30 seconds and hitting submit. This data needs to be sent back to a server in real time. Our test simulates wireless usage under these conditions.

The Mobile Mugshot application interfaces with existing surveillance systems in order to provide alerts and alarms to PocketPC devices. In order to push data to the devices the application must be in constant contact with the server over a wireless network. The server sends a security alert when a security breech occurs. A guard carrying a wireless device running Mobile Mugshot will be alerted by means of a loud audible tone. The guard can then view information on the criminal detected, including their most recent location (presumably in an airport), current image, police photograph on file, arrest/warrant information on file and any other notes and comments on the criminal that are stored in the central server “known criminals” database.

In the application outlined above, there are three main factors that run down the battery:

1)       “always available” network connectivity: since this application requires a wireless LAN to be always available, we could measure the difference between network card types.

2)       continous polling of device to server: the action of polling the server requires the device to wake up from sleep mode, perform a lookup (utilizing the device processor) and then wait.

3)       sound alerts

Results

1. Network Card Type - So called “Low Power” network cards do not always outperform older models

Pocket PC 2000 - iPAQ 3630

In this case, the operating system was the same, the device was the same, only the network cards were different, and the lowest rated power management card, the D-Link, performed best. This is due to the fact that the ‘low power’ network cards will only out outlast the persistent network cards for applications that do not need an ‘always on’ network connection. While the newer ‘low power’ network cards claim their advanced drivers save power, research shows that re-establishing a network connection requires a short surge of power - which indicates a higher drain on battery resources. We know from storing our test devices overnight that if we leave a D-Link card in the device overnight, that device will have no charge left in the morning. The same device, left with a ‘low power’ card will still be charged in the morning. However, in the case of this test, the D-Link card outlasted the ‘low power’ cards since the workflow used in the test required a persistent network connection. The exception to this was when the user set the network and device settings to maximum power usage, the devices with a D-Link card ran out of juice earlier than those with Socket or Symbol cards. This indicates that the drivers for the ‘low power’ cards are indeed smarter or at least better about power management than those for D-Link.

2. Operating System - Pocket PC 2002 shows no significant improvement over Pocket PC 2000

iPAQ - Symbol Card & Extra Battery

In this case, two similar model iPAQs were set to run side by side, running different operating systems. All other variables were the same. There was essentially no difference between the performance of the two operating systems. In the case of “automatic” settings, PocketPC 2000 outlasted the PocketPC 2002 device. On manual power usage settings, PocketPC 2002 outlasted Pocket PC 2000. The difference between the actual times was so small, that we concluded PocketPC 2002 shows no improvement over PocketPC 2000.

3. Network Card Drivers - Socket & Symbol Card Hardware is the same, Drivers different, Similar Performance

In most of our tests, the Symbol & Socket network card performance matched. However, in some cases, there were wide variances, but not in favor of either card. This was probably caused by a feature in these network cards, that they drop the network connection when not required by an application but then use a surge of power to re-establish the network connection when it is needed again.

In the first test above, when we tried to isolate network card impact on battery life, we saw a dramatic difference between Symbol & Socket card drivers set to “automatic”:

The Symbol card outperformed the Socket card by a full 35 minutes, which would imply that the Symbol drivers were smarter than the Socket drivers. However, in subsequent tests, these results were not consistent. Sometimes one outperformed the other and vice versa. The table below shows similar variance in battery life with these cards, this test on iPAQs with an extra battery expansion pak, shows the Socket card to manage power better. What we found significant in this test was the variability in battery life usage with these ‘low power’ cards as opposed to the consistency of the D-Link card’s power usage.

4.        Extra Batteries make a big difference but do not double your battery life

PocketPC2000, iPAQ 3630 & Symbol Card

In the case above, the network card, device & operating system were all equal. The only difference was the extra battery pak. In no case did the extra battery double the life of the device. In our tests, the extra battery could be relied on to increase battery life by 50%.

Pocket PC 2002 -iPAQ 3765 with Extra Battery Pak & Symbol Card

The table above shows the same test done on the PocketPC 2002 operating system and shows no improvement in power management.

5.        Device Manufacturer Really Makes a difference - Jornada Lasts Longer

For our last test, we wanted to compare the Jornada to the iPAQ. . We had read that HP Jornadas had markedly longer battery life but then, we’d also read that PocketPC 2002 was a big improvement over 2000 for battery usage and found no difference. We found that the HP Jornada lasted almost as long as the iPAQ with an extra battery

Pocket PC 2002 - Jornada

Our Recommendation: Choose the Best Value for Your Application Needs

As you saw above, our application requires a persistent network connection, so in our case, we don’t gain anything from spending the extra money on so called ‘low power’ WLAN cards. A D-Link card costs about $100. On the other hand, if your application will be accessing the network only occasionally, choose the Socket card over the Symbol card as it’s slightly less expensive and delivers almost identical performance. For device settings, stick to “automatic” settings whenever you can. While configuring a device to “minimum” power usage settings will slightly outlast “automatic” settings, we found the low brightness and absence of backlight to be a poor user experience.

For your device, consider the HP Jornada 568. The cost of an HP versus the cost of an iPAQ with an expansion pak and extra battery is about equal. However, the Jornada does not have expansion paks. Further, it only has a CF1 slot, which limits the kinds of cards we could use in the device. From our tests, we concluded the D-Link card was the better bet, but we would not be able to use this card with a Jornada since it is too big to fit in the CF1 slot.

Next Steps

Enterprise Air does have the hardware to test other configurations particularly related to Casio ruggedized devices and devices from Symbol technologies. PCMCIA WiFi cards can also be tested. If you have any questions about the tests or particular interest in testing a configuration please let us know by emailing support@enterpriseair.com.

How We Tested

Test Case Workflow

Since we were measuring battery life across devices and configurations, we decided to select a consistent application workflow. And as we were principally interested in the impact of the wireless connectivity on battery life, sounds were left on but set low in volume. The Mobile Mugshot application was altered to insert a time stamped record into a database every 30 seconds. This allowed us to measure the battery life to within 30 seconds of the device shutting down and saved us from having to sit and watch the devices until the screen went blank.

Network Card Types

Since PCMCIA network cards are reputed to use more power than CF cards, and are really meant for laptops, we limited our test to CF cards. There are two main CF network card types, a newer ‘low power’ type that uses specially designed network card drivers to allow the network card to ‘suspend’ itself while not in use. The second kind of card maintains a persistent connection to the network. While the newer ‘low power’ network cards claim their advanced drivers save power, research shows that re-establishing a network connection requires a short surge of power - which indicates a higher drain on battery resources. For this test, we chose two card types: one that allowed the network connection to be suspended (Socket & Symbol cards) and one that maintained a persistent connection (D-Link). The only difference we found between the Socket & Symbol cards was their drivers. Socket & Symbol network cards are basically the same. Symbol has OEM’d the cards from Socket, but written their own drivers. We used both cards in order to measure any difference in power management of the card drivers.

Device & Driver Power Usage Settings

Next, various device power saving settings were isolated and tests defined to highlight the general effect on battery life. The key settings that effect battery life on a device are:

1)       Radio Transmission Power

2)       Power Saving Mode

3)       Device Auto Off

4)       Backlight Auto Off

5)       Brightness

For our test, we defined groups of settings as “automatic” (average), “minimum” and “maximum” power usage settings across card types, device types and card drivers. We used these groups consistently. Device power usage settings were defined as follows:

Device Models & Operating Systems

For this test, three devices were used:

1)       iPAQ 3630 running PocketPC 2000

2)       iPAQ 3765 running PocketPC 2002

3)       HP Jornada Model 568 running Pocket PC 2002

Microsoft claims that PocketPC 2002 has better power saving than PocketPC2000, so we chose two similar iPAQs running each and compared the results of the iPAQ battery usage to that of the HP Jornada running PocketPC 2002, reputed to have longer battery life than an iPAQ.

Device Expansion

The last variable we wanted to measure was the effect of an extra battery on a device. For this, we used an iPAQ PCMCIA expansion pak with an extra battery. But since we only wanted to use CF cards, we put a flash storage PCMCIA card with a CF2 slot into the PCMCIA slot on the expansion pak.

About Enterprise Air

Enterprise Air (www.enterpriseair.com) is an emerging leader in providing end-to-end mobile device software solutions to software applications vendors, systems integrators, and large enterprises. 

Enterprise Air works with vertical industry specialists to extend software application functionality to mobile PDA devices such as Pocket PC and Palm OS devices. Enterprise Air’s unique MobileEdge™ Technology allows any organization to build mobile applications or to mobilize existing applications rapidly and easily.

The Company has built several mobile extensions for law enforcement and public safety applications. For Example the Enterprise Air Mobile MugShot™ application integrates with video surveillance and facial recognition systems, allowing officers in the field to use handheld computing devices to receive alarm information and images from surveillance systems.  Other products of Enterprise Air include solutions for Data Collection and Sales Force Automation.

Enterprise Air has partnerships with large key players such as IBM, Sybase, Compaq and Casio. The company released Version 1.0 of its software in early February and is working with its partners to deploy best of breed mobile application solutions to the market.

Enterprise Air is located in Manhattan, NY. For further information visit www.enterpriseair.com or call Bill Markel or Michael Hejtmanek at (212) 941-1988.