Long Live Your Mobile Battery

I‘ve promised to come back to this topic in my comments to earlier blog postings and now it’s time to do it. Let me start by telling the very recent conversation between my team member and one of the APAC area cellular network operator. It went like this:
                Nokia: Are you planning to deploy this feature X in you cellular network?
                Operator: Why don’t you rather tell me about your display power optimizations? Isn’t it so that displays are consuming most of the power in the smartphones?

This dialog highlighted the need of telling people that how power consumption breaks down in the device between different use cases. Display is one of the biggest battery drainers but cellular radio also belongs to that same category and therefore shouldn't be ignored. By reading further this post you'll notice that display backlights is turned off during phone call and this means 0 mW power consumption for display in that particular use case. That said, no help of display optimisation if you just want to talk longer.

Power consumption breakdown (mW) examples

Above we have four different use cases and their power consumption breakdown. Those pie charts are based on real figures from one of the Nokia Symbian device platforms. Some variation occurs between our HW platforms and especially display share can be much bigger or smaller depending on display technology (LCD, OLED), physical size (larger consume more) and brightness settings (backlight dominate display sub system power consumption).  By looking the pies, following can concluded:

-    Typically three most power hungry units are processor, display and cellular radio. Breakdown is always use case specific and shares differ a lot.
-    During voice call cellular modem is consuming vast majority of the power and other HW parts are most of the time in sleep mode
-    Video call over 3G network and Browsing are good examples of the use cases which load the system heavily and basically all three major units are utilized.
-    3D gaming loads GPU and CPU and display and typically modem is turned off (when not playing online game)

But remember, that in order to understand the full picture, we need to know total power consumption of the use case and also how often it is used. Your application can consume 1W power if it is never used but if it is popular and gets used a lot, then even figure that sounds small starts to make sense to optimise.

In Symbian you get total power consumption of the device by running Nokia Energy Profiler (NEP). Good starting point to get ballpark of application level power consumption is to calculate delta when application is running vs. when it is not installed. Nokia Battery Monitor is another way of getting valuable information of how much your application is contributing to total power consumption. Version 2.0 will be soon available and it provides application level energy monitoring.

How much is daily usage then? People spend lot's of time with popular applications and those will become major shareholder of the device daily energy consumption. At this point energy efficiency starts to show up. 

I decided to highlight advantages of push notifications since synchronizing Internet accesses is such a high priority item in smartphone battery life.  Nokia has introduced Notifications API for Qt/QML applications earlier last year and you can find more info of it by browsing earlier posts in this blog.
In Symbian, Nokia news reader (Nokia Reader) is fetching data by using Notifications API. Let’s see the device battery life impacts when using it.

Without Nokia Reader application using Nokia Notificatications

Each feed is polled separately from the client 

If there are 5 of these feeds and every one of them is polled with 10 minute interval, without synchronizing the time of polling... There will be 30 queries to the network every hour. Http polls cause typically the radio to be active for at least 15 seconds. This means that the radio is on 30*15s = 450 seconds every hour.

With Nokia Reader application using Nokia Notificatications

Each feed is polled separately by a feeder service, data is pushed to client, when needed 

Statistics from more than 7000 feeds show that with 10 minute interval, average feeds have a 11% hit ratio and each hit contains
2 updates on average [about 32 updates / day, not evenly distributed]. 5 feeds lead to about 3.3 updates / hour on average. In addition
the TCP connection needs to be kept open with keep alive messages. Typical cellular network allows about 20 minutes of inactivity (= 3 keep alives / hour are needed). Even if the 5 feeds provide more than that on average, news are typically clustered and thus
additional keep alive messages are probably needed. Let’s assume 1,7 of them in average. This means 5 updates/hour, out of which 3 need 15 seconds of active radio and 2 [keep alive messages] need only 10 seconds. This leads radio being active for 65 seconds / hour.

Summary

We’ll have following results by basing up our calculation in to real data collected from our studies.

-    Without push notifications we have radio active 450 seconds/hour
-    With push notifications 65 seconds/hour

In this example of updating 5 news feeds with 10 minute polling interval, using Notifications API instead of polling brought 86% energy savings.

Got interested in Nokia Reader? Go and download in from Nokia Beta labs!

I was posting earlier in this blog that Nokia offers push notification API for developers. Now I'm happy to tell that a beta release of the Notifications API add-on for the Nokia Qt SDK 1.0 is available.

I recommend to watch Notifications API webinar that gives good understanding what's this all about. There's also informative Notifications API project wiki pages that offers everything you need to get started with application development.

Read more about Qt http://www.forum.nokia.com/Develop/Qt/

Nokia Battery Monitor version 1.0 is now available for free download in Ovi Store. As announced in the previous blog post, Battery Monitor provides a detailed indicator for remaining battery level with usage time and charging time estimates. Battery level monitoring and usage time estimates are supported in all Nokia Symbian devices with touch screen. Widget mini view is available in devices that support homescreen functionality. Remaining charging time estimator is not yet supported in all the latest device models, like Nokia C7-00 and C6-01, but it will be in the future releases. If not supported, the widget will show only “charging” text instead of time estimation. 

In order to estimate remaining charging time, every device model battery charging profile needs to be measured in the lab and fitted to a mathematical model. Benefit of this approach is that it can provide really accurate estimates. Figure 1 shows estimate from a model compared to actual charging time in Nokia N8.

Figure 1: Remaining recharging time model estimate (in seconds) vs. actual charging time for Nokia N8. Source: Nokia Research Center.

Nokia Battery Monitor uses real battery voltage and current measurements for making the estimates. This allows for much more accurate estimates than the information available in the current Symbian HW Resource Manager API, which only provides the same information as the UI battery bars (from 0 to 7). API used for reading voltage and current is the same as used by Nokia Energy Profiler application. Because of this, Energy Profiler cannot be used at the same time with Battery Monitor.

Accuracy of estimations will improve over time as more measurement data is gathered. Especially the overall active usage time estimation requires several days of data. Best accuracy is achieved when there are little variations in daily usage patterns. Accuracy can also depend on external conditions like network strength variations. Data gathering in the background will increase device standby power consumption by less than 0.25 mW. This doesn’t have significant impact on device usage times.

Previous blog post showed screenshots with white background colour. As mentioned in earlier blog posts, this is not energy efficient in devices with OLED displays such as Nokia N8. For this reason, Nokia Battery Monitor will support also black colour theme as shown in Figure 2.

Overview
Nokia has recognised that users of smartphones would like to have better tools for monitoring their battery usage. It can be difficult to understand which applications are draining the battery and how the usage of the device will impact operating times. To increase this understanding Nokia Battery Monitor will be published for Symbian. It will be available soon in Ovi Store for free download.

Today, the remaining energy of Nokia mobile devices is displayed with battery bars in the top right corner of the screen. Granularity of this information is limited and it is difficult to estimate the actual remaining usage time based on this. Charging is indicated only with scrolling battery bars without any estimate of remaining charging time or the level of charge in the battery. Nokia Battery Monitor will fill these gaps by providing more detailed estimates for remaining battery capacity, various activities usage time and charging time. 

Features
Nokia Battery Monitor consists of a home screen widget (Figure 1) and a full screen application (Figure 2). The widget is used for visualization of remaining battery capacity and activity estimates. The full screen application provides more details and statistics of energy usage.
Remaining battery capacity is visualized with a large high resolution battery bar and also shown in percentages. Remaining usage time estimates are provided for voice call, web browsing, music playback and the overall device usage. Remaining charging time of the battery is estimated when the charger is connected. The usage statistics include distribution of overall and daily energy consumption of different applications.

Figure 1: Nokia Battery Monitor home screen widget

Figure 1 shows example views of Nokia Battery Monitor home screen widget. The view on the left shows the remaining usage time for voice calls with the remaining battery capacity. The view on the right shows the remaining charging time until the battery is full.

Figure 2: Nokia Battery Monitor full screen mode

Figure 2 shows the full screen application view of Nokia Battery Monitor. The application consists of three tabs. ‘Estimates’ -view shows a summary of activity estimates and remaining battery capacity. ‘Statistics’ -view shows the distribution of overall and daily energy consumption. 'Info' -view provides instructions and other general information about the application.

Coming soon
More details will be published in blog post part 2/2 when Nokia Battery Monitor is available for download in Ovi Store.

Nokia Battery Monitor has been developed in collaboration between Nokia and VTT Technical Research Centre of Finland. We would really like to hear what you think about it. Please send your feedback to: batterymonitor@nokia.com 

Challenge today

By saying that application is “always on”, we mean that device is registered to IMS system all the time. Connection needs to be kept alive by sending periodic “keep alive” messages. Typical examples are e-mail and different kind of social network services. In addition to keep alive messages, there will be data sent back and forth between client and servers (e.g. e-mails).  Typical smart phone users have multiple services like this. We have all this up and running so what’s the big deal?

Connecting of several always on services has two major impacts,

1. High signaling loads in HSPA network. High signaling load caused by the always-on applications is the hottest topic among operators and network vendors at the moment
2. Smart phone operation time has become unacceptable. Customers are not satisfied with current smart phone operation times. Frequent accessing of WCDMA network consumes too much power

OVI Notifications API provides significant improvement for both challenges

The Ovi Notifications API makes it easy to add push notifications to your applications. With up-to-the-moment, efficient delivery of lightweight notifications, your apps stay at the forefront of users’ attention. At the same time, your mobile app will optimize bandwidth and battery use, giving your customers increased functionality without higher costs.

Key features of Notifications API network in order to reduce network signaling, data transmission and device power consumption

1. One keep alive message for all the services using Notifications API. This is the most important battery saving feature and provides biggest savings when using Notifications API.
2. Optimized keep alive timer period by new, heuristic keep alive messaging solution
3. Send data only when new data is available in the server
4. Zlib compression algorithm enables lightweight notifications (up-to 1,5kB payload)

Taking Notifications API into use

Got interested in? Read more and join us to use Notifications API.

Register to participate in our Notifications API technology preview

http://www.forum.nokia.com/General/API_Notification.xhtml

How to select correct bearer in order to get fastest connection but not forgetting the device battery life? We have several options available in smartphones and each one of those have their own characteristics.

Nokia has made this easy now in new Symbian^3 and feature will be first time introduced in N8. Nokia introduced One Click Connectivity (OCC) that simply does wisely the bearer selection and allows seamless bearer roaming during active use case e.g.  3G HSPA -> WLAN web browsing if authenticated WLAN access point is available.

One Click Connectivity key focuses on three areas

• Centralized connectivity settings
• Connectivity State Soft-Notes (Unified connectivity UI)
• WLAN usability

All these aim at one goal: Consistent connectivity behavior throughout the platform. OCC is not one new feature but several connectivity related improvements gathered under one umbrella

One of the characteristics is that WLAN is prioritized first in bearer selection when it is available. Not just making connection faster, this gives us great savings in energy consumption. See figure below.



Figure: Relative energy consumption comparison between WLAN, 2G data and 3G data connections. Source: Nokia

It is often advised that WLAN should be turned off in standby if you want to save energy in mobile device. True or False? False. Nokia N8 has a new intelligent, autonomous and yet very effective way to do WLAN scanning in device without sacrificing battery life or connectivity performance. Our internal measurements have proven very low figures (in typical user scenarios we have reached <1mA average over 24h measurement period) that won’t have any impact in practice to device standby times.

For developers Nokia has made this easy. Application developer should connect to network just by asking network connection instead of creating own connection management mechanisms.

Tips and hints:

OCC for developers

Are you planning to implement Qt application with network connection?

Top 10 energy saving tips is alive again in FN. I cannot emphasize importance of those tips enough and that's why I wanted to keep repeating myself and link it here again.

Top 10 energy saving tips

We had a webinar on December 2009 in which I highlighted some of the tips via practical examples (in Day 2). I encourage all of you to go and hear those presentations. Other than those tips, there's lot's of other relevant battery life information available.

Mobile Internet Battery Life: Challenges/Solutions

In this session, Nokia experts describe the challenges developers face with mobile internet services’ operating times. This webinar provides practical examples of typical problems related to applications that are always online and performing background data transfers.

Day 1

Day 2

OLED (Organic Light Emitting Diode) displays are here today and people talk of them how brilliant those are. Oh Yes, those are brilliant and now this technology has also entered to mobile devices and therefore it has became relevant to developers who are designing UI SW. When it comes to power consumption, OLED has very unique characteristics vs. TFT(thin-film transistor) LCD’s. Key point being that power consumption varies a lot according to colors of the content in OLED.

We’ve measured 3.2" displays in Nokia with following results:

Figure: 3.2" OLED vs.LCD displays and power consumption in milliwatts [mW].

All white consumes 14x vs. all black screen in OLED! Content makes also difference in power consumption of LCD display but not that much.

Tip: Optimize colors for OLED. Prefer dark colors. Same optimizations apply today for Nokia’s TFT LCD displays. These are technologies used in all Nokia smart phones.

Writing power-efficient code has become more and more important. Todays smartphones are very complex and that makes battery life optimisation challenging. Multiple radios, big displays, always-on usage etc. Marketing forces are demanding more and more impressive graphics, high display refresh rates. List could be endless to explain why our smartphone batteries are draining so fast, and at the same time users are complaining the bad battery life. Why my great device does not last longer than 1-2 days without charging? Pls Mr. Nokia, do something

But as title says, SW is the master. It controls HW resouces and loads CPU and therefore focus on efficiency is essential in order to achieve good device operation time.  In smartphones, multiple simultaneously running use cases is putting the battery beyond its limits. Many of those use cases are such that network interaction is required which adds even more complexity. In practice this means that device is no longer the only party impacting to its battery life.

As an exampe, let's take any always-on mobile application (e.g. IM, e-mail, newsreader etc.)

How does this relate to SW developer?

For those dealing with service business, the following things should be considered:

• Content optimization and filtering for mobile clients (e.g. sending headers instead of downloading the attachments in e-mail apps)
• Synchronization of data connections between device and service (e.g. keep alive messages)

Sending data or just plain signalling has significant impact to device battery life. Especially in WCDMA networks this can be lead bad results if keep alive period has been defined too short.This is one of the key things in e-mail example.

http://www.forum.nokia.com/Design/Power_management/

Device
The place where the battery power  is actually consumed. We can say that the majority of the power is consumed by the CPU, display subsystem and cellular radio. It very much depends on the use case which is the worst one.
Forum Nokia has an excellent “check list” for each one.  You’ve done a lot if you go through top 10 energy saving tips at

http://www.forum.nokia.com/Design/Power_management/Top_10_energy_saving_tips.xhtml

Give options and guide customer to use your SW in the most efficient way! As an example, e-mail again.. let him/her decide how often application downloads new e-mails.