Welcome to ZenUnwired; a blog dedicated to tracking developments in technology and strategy, and to deciphering the impact of these developments on wired and wireless ISP's, device manufacturers, OS and application developers, and most importantly - you.

Thursday, December 29, 2011

Reign of [the Kindle] Fire

With the Amazon Kindle Fire launch, many are wondering just how its entry will impact the tablet segment. Up until now, the market has essentially been a world of one with the Apple iPad; and Apple continues to reign supreme with the iPad, while competing tablet makers’ attempts to topple the company have failed.

Before its launch, analysts called the Kindle Fire the iPad killer, a "segment defining tablet", etc. But in my opinion, Amazon designed the Kindle Fire not to be any of these things. That said, the Kindle Fire might just turn out to be all of these things. One thing however is clear - Amazon did learn from both Apple's success with the iPad, as well as from the failure of others to gain any traction in the tablet space. The result - the Kindle Fire - a snappy tablet designed around Amazon's customers, and one that is tightly integrated with Amazon's various services.

Consider this: what do buyers really want from their tablets? The most likely answer is going to be - "I want a tablet at a great price; one that is fast, is simple to use, looks good and has tons of content". Of course, the advanced user will want much, much more from their tablets. But the average user's answer is going to be a lot like mine above. The Kindle Fire hits on all of these requirements with a laser-like focus.
  • Price: The Kindle fire, priced at $199 on Amazon.com, and available for a lot lower from various retailers this holiday season, is priced really attractively considering all of the various tablets already available in the market. Sure, you can get some no-name tablets for a lower price, but $199 is a price-point where customers will trade a $20-$30 price differential for brand. In other words, will you buy a no-name tablet for $169 or an "Amazon" tablet for $199? Exactly. No wonder, the Kindle Fire is Amazon's top seller this season.
  • Screen: The Kindle Fire comes with a 7" color touchscreen that delivers 16 million colors in high resolution and the screen uses IPS (in-plane switching) technology (like the iPad). The result: a truly great screen for consuming any type of content - from reading books to watching movies to playing games.
  • Speed: The Kindle Fire is fitted with Texas Instruments' OMAP 4430 1GHz dual-core processor. In other words, this tablet is fast!
  • Simplicity: It is clear that by customizing Android, Amazon has opted out of the Android OS version and upgrade cycle discussions/debacles that plague other tablet makers. Take HTC for instance - The HTC Flyer launched with a promised upgrade to Android Honeycomb 3.2 but almost a year later the Wi-Fi version of the tablet is yet to see that upgrade. The result: a huge backlash against HTC by consumers; one that has set the web/blogosphere on fire (pun unintended). So Amazon doesn't even talk about the version of Android it is running (although its most likely a customized version of Android 2.3). Instead, what customers see is a completely customized UI; built to promote customers own and/or Amazon's content. The result: customers buy the Kindle Fire for the user-experience, not the Android version.
  • Content: The Kindle Fire is clearly designed to promote/sell Amazon's services. The short list of these services include: Kindle books & magazines, Amazon MP3, Amazon Appstore, Amazon Instant Video, Amazon Prime and the Amazon Cloud. If you own a Kindle Fire, it is inevitable that you will end up signing up and using at least a couple of these services.

Bottom Line: Apple, and now Amazon, have made it clear that it’s no longer about “devices.”  Success is about being able to deliver a comprehensive solution of services and applications that is optimized for these devices. The reign of the Kindle Fire has begun, and it looks like Amazon is just getting warmed up.

What is LTE?

One of the best posts out there on what LTE is. You can (and should) read the full post here: http://www.extremetech.com/mobile/110711-what-is-lte

So, what is LTE? Verizon Wireless, in its advertisements, will tell you that it is the fastest, most advanced network. AT&T will say it is the next generation of wireless technology. But that doesn’t say much, now does it? To put it simply, LTE is not just the next generation of wireless technology. LTE is an ongoing, living standard. LTE is a standard that will continuously improve over time. Many expect LTE to be the standard for cellular networks for at least the next decade, possibly even beyond!

LTE stands for Long Term Evolution. Its full name is 3GPP Long Term Evolution for the Universal Mobile Telecommunications System, or 3GPP UMTS LTE for short. Most refer to the standard as UMTS LTE or just plain LTE. While UMTS LTE is the more correct name, it will be called LTE for the remainder of this article for brevity.

But that doesn’t tell us what LTE actually is. LTE is what the 3rd Generation Partnership Project (the group responsible for standardizing and improving the Universal Mobile Telecommunications System, or UMTS) designates as their next step. UMTS is the group of standards that define 3G for GSM networks across the world, including AT&T and T-Mobile’s 3G networks. For those who use CDMA2000 (subscribers of Verizon, Sprint, etc.), then LTE is pretty much the replacement for your mediocre 3G network.

umts logo

LTE is a very good, easily deployable network technology, offering high speeds and low latencies over long distances. For example, Verizon LTE in Dallas, TX was rated with an average download speed of 15.75Mbps and an average upload speed of 1.49Mbps. Verizon’s 3G service was rated with an average download speed of 1.09Mbps and an average upload speed of 0.67Mbps. Its LTE service beat its competitors’ high speed service by a wide margin. Similar ratings followed in other cities as well. Unfortunately, AT&T’s LTE service is too new to take accurate measurements of. When the AT&T LTE network is more loaded with subscribers, then more accurate readings can be taken.

In this article, I will discuss what configurations LTE can be deployed in, why LTE is easily deployable, how LTE works as a radio technology, what types of LTE exist, how LTE affects battery life, what network operators want LTE to do, and the future of 4G as a whole. The most technical parts of the article are LTE can be deployed in, why LTE is easily deployable, how LTE works as a radio technology, and what types of LTE exist. For those who don’t want that information, you can skip to how LTE affects battery life and still get the gist of what I’m saying. But to get the complete picture, reading the whole article is advised.

How LTE is configured for deployment

LTE supports deployment on different frequency bandwidths. The current specification outlines the following bandwidth blocks: 1.4MHz, 3MHz, 5MHz, 10MHz, 15MHz, and 20MHz. Frequency bandwidth blocks are essentially the amount of space a network operator dedicates to a network. An operator may choose to deploy LTE in a smaller bandwidth and grow it to a larger one as it transitions subscribers off of its legacy networks (GSM, CDMA, etc.).

MetroPCS is an example of a network operator that has done this. A majority of its spectrum is still dedicated to CDMA, with 1.4MHz or 3MHz dedicated for LTE depending on the market. Leap Wireless has also done the same thing, except it’s using 3MHz or 5MHz instead of 1.4MHz or 3MHz. Neither of these carriers can afford to cut CDMA capacity by a significant degree just yet, so LTE operates on tiny bandwidths. Additionally, neither operator has enough backhaul (the core network infrastructure and connections to the internet) dedicated to LTE to make larger bandwidths worth it either.
verizon logo

On the other hand, Verizon Wireless has been dedicating 20MHz for LTE all across the board, since it has a nationwide 20MHz block of spectrum available for LTE. Combined with excellent backhaul, Verizon’s LTE service promises to be best in class. AT&T is dedicating 10MHz across the board because that’s all the free space it has, though it makes up for it with much better backhaul, so the performance differential between Verizon and AT&T isn’t noticeable right now. However, when AT&T gets more LTE customers, the difference will become clearer.

Less spectrum means that fewer customers can obtain the same high speeds that Verizon’s LTE customers get when connected to any particular cell. LTE can support up to 200 active data clients (smartphones, tablets, USB modems, mobile hotspots, etc.) at full speed for every 5MHz of spectrum allocated per cell. That means that if a particular tower has 20MHz of spectrum allocated to it, it can support up to 800 data clients at full speed. There are ways of supporting more data clients per 5MHz, but doing so requires sacrificing speed and capacity, as the 200-per-5MHz ratio is the optimal configuration. However, spectrum isn’t everything to LTE quality, as I will discuss later.

Why LTE is easier to deploy

The network architecture for LTE is greatly simplified from its predecessors because LTE is a packet-switched network only. It doesn’t have the capability to handle voice calls and text messages natively (which are typically handled by circuit-switched networks like GSM and CDMA). Anyway, the LTE SAE (System Architecture Evolution) is essentially a simplified version of the one used for UMTS networks today. An LTE network uses an eNodeB (evolved node B, essentially an LTE base station), a MME (mobile management entity), a HSS (home subscriber server), a SGW (serving gateway), and a PGW (a packet data network gateway). With the exception of the eNodeB, everything is considered as part of the EPC (evolved packet core) network. At the tower the eNodeB connects to the EPC.

evolved core packet framework
The MME and the HSS basically handle all duties regarding subscriber access to the network. It handles all the authentication, roaming rules for subscribers, etc. The SGW essentially acts like a giant router for subscribers, passing data back and forth from the subscriber to the network. The PGW provides the connection to external data networks. The most common data network the PGW provides a connection to is the internet. However, if the network operator desires handover with a non-UMTS network like CDMA2000, WiMAX, or a WiFi hotspot network run by the network operator, then an ePDG (evolved packet data gateway) and an ANDSF (Access Network Discovery and Selection Function) for the eNodeB can be installed to support those networks on the EPC.

Most operators around the world will use the basic network design. Verizon Wireless, Sprint-Nextel, Leap Wireless, MetroPCS, C Spire Wireless, and U.S. Cellular have installed or will install the same basic design with one major change: eHRPD will replace the core network connection to traditional UMTS networks.

Palm pre 1xrtt

They won’t be using the proper network design to handover to CDMA2000 because of eHRPD (Enhanced High Rate Packet Data, essentially an enhanced version of the core packet network for EV-DO), which plugs right into the network in a way that is supposed to replace a UMTS network. By its very nature, eHRPD is rather fragile because it attempts to emulate enough of what the LTE network core expects in a UMTS network to communicate and hand over. This is why Verizon’s LTE service has been breaking down at least once every quarter. LTE and CDMA handover wasn’t originally designed to work the way it does now, and the way they’ve implemented it is not officially supported in the standard (well, the 3GPP standard, anyway). Unexpected issues arise every time they do some network tweaking because of this. Sometimes the failure can spread to EV-DO and shut it down, leaving only 1xRTT available.

How LTE actually works

LTE uses two different types of air interfaces (radio links), one for downlink (from tower to device), and one for uplink (from device to tower). By using different types of interfaces for the downlink and uplink, LTE utilizes the optimal way to do wireless connections both ways, which makes a better optimized network and better battery life on LTE devices.

For the downlink, LTE uses an OFDMA (orthogonal frequency division multiple access) air interface as opposed to the CDMA (code division multiple access) and TDMA (time division multiple access) air interfaces we’ve been using since 1990. What does this mean? OFDMA (unlike CDMA and TDMA) mandates that MIMO (multiple in, multiple out) is used. Having MIMO means that devices have multiple connections to a single cell, which increases the stability of the connection and reduces latency tremendously. It also increases the total throughput of a connection. We’re already seeing the real-world benefits of MIMO on WiFi N routers and network adapters. MIMO is what lets 802.11n WiFi reach speeds of up to 600Mbps, though most advertise up to 300-400Mbps. There is a significant disadvantage though. MIMO works better the further apart the individual carrier antennae are. On smaller phones, the noise caused by the antennae being so close to each other will cause LTE performance to drop. WiMAX also mandates the usage of MIMO since it uses OFDMA as well. HSPA+, which uses W-CDMA (an improved wideband version of CDMA) for its air interface, can optionally use MIMO, too.

Verizon Cell Tower

For the uplink (from device to tower), LTE uses the DFTS-OFDMA (discrete Fourier transform spread orthogonal frequency division multiple access) scheme of generating a SC-FDMA (single carrier frequency division multiple access) signal. As opposed to regular OFDMA, SC-FDMA is better for uplink because it has a better peak-to-average power ratio over OFDMA for uplink. LTE-enabled devices, in order to conserve battery life, typically don’t have a strong and powerful signal going back to the tower, so a lot of the benefits of normal OFDMA would be lost with a weak signal. Despite the name, SC-FDMA is still a MIMO system. LTE uses a SC-FDMA 1×2 configuration, which means that for every one antenna on the transmitting device, there’s two antennae on the base station for receiving.

The major difference between the OFDMA signal for downlink and the SC-FDMA signal for uplink is that it uses a discrete Fourier transform function on the data to convert it into a form that can be used to transmit. Discrete Fourier transform functions are often used to convert digital data into analog waveforms for decoding audio and video, but it can be used for outputting the proper radio frequencies too.

The LTE technology itself also comes in two flavors: an FDD (frequency division duplex) variant and a TDD (time division duplex) variant. The most common variant being used is the FDD variant. The FDD variant uses separate frequencies for downlink and uplink in the form of a band pair. That means for every band that a phone supports, it actually uses two frequency ranges. These are known as paired frequency bands. The TDD variant uses one single range of frequencies in a frequency band, but that band is segmented to support transmit and receive signals in a single frequency range.

In the United States, Clearwire is the only network operator deploying LTE in the TDD variant. Everyone else is deploying in the FDD variant. The TDD variant becomes more important in Asia, as China Mobile (the largest network operator in the world in terms of subscriber count) uses TDD frequencies for their 3G network and it plans to upgrade to the TDD variant of LTE. Fortunately, LTE devices can support both variants on a single radio chip without too much trouble.

Enough about specs — what about battery life?

Now we lead to the part that most people care about: how it affects battery life. By itself, LTE devices should last roughly as long as their HSPA+ equivalents because of the optimized radios for both downlink and uplink operations. The reason why LTE devices right now eat batteries for breakfast is because the network operators are forcing these devices into active dual-mode operation.

For Verizon Wireless, this means that all of their LTE devices connect to both CDMA2000 and LTE simultaneously and stay connected to both. This means that you are eating twice the amount of battery for every minute you are connected than you would if you were connected only to CDMA2000 or LTE. Additionally, when you make calls on Verizon Wireless LTE phones, the CDMA2000 radio sucks down more power because you are talking. Sending and receiving text messages causes pulses of CDMA2000 activity, which cuts your battery life more. Arguably, constantly changing radio states could be worse for battery life than a switch into one mode for a period of time and switching back, so text messages may actually kill the batteries faster.

Then there is handover. Handover is the operation in which a device switches from one network to another or from one tower to another. Handover is the critical component that makes any cellular wireless network possible. Without handover, a user would have to manually select a new tower every time the user leaves the range of a tower. (WiFi is an example of a wireless network technology that doesn’t inherently support handover.) When the user travels outside the range of a WiFi network, the WiFi radio will just drop the connection. For cellular networks, this is even more critical because the range of a tower is not very predictable due to factors outside of anyone’s control (like the weather, etc.). LTE supports handover like all other cellular wireless networks, but it improves on it by doing it much faster when handing over to a supported type of network or cell.


However, Verizon is doing handover from LTE to EV-DO and back by plugging in a connection to an enhanced version of the EV-DO data network core called eHRPD. As discussed earlier, this isn’t a great solution by any means. It becomes more problematic when you consider that most LTE signals are very weak. Unfortunately, most customers have no idea because Verizon deceives them into believing it is stronger by using the EV-DO signal strength for the signal bars for LTE for all of their devices except the Galaxy Nexus.

The weak signal and the fragile link-up between EV-DO and LTE make handover occur a lot more than it is supposed to, which eats battery life even more. With AT&T using an HSPA+ network alongside LTE instead of CDMA2000, handover operation is a lot smoother. As far as battery life goes, it should be slightly better than Verizon LTE phones because LTE supports fast handover between UMTS and LTE. AT&T LTE phones are normally not forced into active dual-mode operation because HSPA+ lets you use data and talk at the same time. As a consequence, AT&T has no need to force the device into active dual-mode operation. However, battery life will still be pretty bad because LTE signals are still very weak in most AT&T LTE zones, and AT&T LTE devices default to connecting to LTE signals whenever possible.

C Spire Wireless, MetroPCS, Cricket Wireless, and U.S. Cellular will all have the same problem as Verizon Wireless with LTE battery life because they all plan to do the same thing as Verizon Wireless and force active dual-mode operation. As a result, turning off LTE will significantly improve battery life because the phone switches back to single-mode operation. Or in the case of AT&T phones, passive dual-mode operation (for GSM/HSPA+ handover) since they are typically in passive tri-mode operation for GSM/HSPA+/LTE handover. Passive multi-mode operation means that the device isn’t constantly connected to multiple networks, but will establish a connection and hand over the connection if the signal on the current network is too weak or snaps. This is ideal for multi-mode operation, but it isn’t possible for CDMA/LTE network operators until they make it possible for LTE to handle calls and text messaging.

The panacea

The ultimate goal of the network operators deploying LTE is to replace everything else they have with it. That means that it needs to become possible to handle voice calls, text messages, network alerts, etc. over the data network. However, no one developed the LTE specification with voice and text messaging in mind. It was designed as a data network only. So how do they solve the problem? By developing a VoIP solution that fits their needs. Two main standards came into existence: VoLGA (Voice over LTE via Generic Access) and VoLTE-IMS (Voice over LTE via IMS). VoLGA was based on GAN (Generic Access Network), which is also known as UMA (Unlicensed Mobile Access). Deutsche Telekom was the only network operator that wanted to use this method, as the design for VoLGA was heavily derived from T-Mobile USA’s implementation of UMA for its Wi-Fi Calling feature. No one else wanting to deploy LTE wanted to use it as a final or interim solution, as it would have meant keeping around the legacy GSM core network for this purpose.

Everyone else supported VoLTE-IMS (now referred to as VoLTE), which allowed them to fully discard their older networks and simplify their network design as they decommissioned legacy networks. However, IMS is much more expensive and difficult to deploy than VoLGA, at least for GSM network operators. But IMS also promised more flexibility. IMS could be used to make real-time video calling with all sorts of additional features possible. And so, Deutsche Telekom dropped VoLGA and joined everyone else in supporting VoLTE.

VoLTE uses an extended variant of SIP (Session Initiation Protocol) to handle voice calls and text messages. For voice calls, VoLTE uses the AMR (Adaptive Multi-Rate) codec, with the wideband version used if supported on the network and the device. The AMR codec has long since been used as the standard codec for GSM and UMTS voice calls. The wideband version supports higher quality speech encoding, which would allow for clearer voice calls. Text messages are supported using SIP MESSAGE requests. While video calling has been often discussed as a potential benefit of using VoLTE, no such standard for it exists yet.

t-mobile logo

In a somewhat ironic twist, T-Mobile USA became the first network in the world to commercially deploy IMS-based voice calling and text messaging by using it for an improved WiFi calling solution. An update to the T-Mobile Samsung Galaxy S II and an update to the T-Mobile HTC Amaze 4G both included the new Wi-Fi Calling solution.

Despite not having an LTE network, T-Mobile is the most prepared for deploying LTE to replace its existing networks. Once T-Mobile deploys LTE, it can easily modify the WiFi calling client software to allow it to work over the LTE network as well. Its advanced HSPA+ network architecture also means that it can easily (and relatively cheaply) plug in support for LTE, too. But, T-Mobile has no room in its spectrum nationwide to deploy LTE at the moment. It’ll need to repurpose some of the spectrum it currently uses, which means scaling back the bandwidth for HSPA+ or 2G GSM.

As for Verizon, AT&T, and Sprint deploying VoLTE? Well, Verizon stated at Mobile World Congress 2011 that it plans to begin offering handsets with VoLTE in mid 2012 to early 2013. AT&T has stated that it will deploy VoLTE in 2013. Sprint has not officially said anything about it yet.

The messy future of 4G

We’ve only scratched the surface of what LTE is all about, but this article includes pretty much everything that LTE subscribers would care about. Some of the other aspects of LTE include SON (self-organizing network) capabilities, which allows it to flexibly allocate capacity to parts of the cellular network as it is needed by redistributing connections to an optimal configuration at any given time. Handover to WiFi is another cool feature, too. However, most of the features like the former are pretty much only seen from a network operator’s side of things, and things like the latter may never actually be implemented.

LTE is a significant leap in optimized cellular wireless technology though. If you wish to get the highly-technical details of LTE and its ever-evolving specifications, check out the 3GPP’s specification series for LTE. Specifications for eHRPD and associated CDMA2000 specifications are available on the 3GPP2′s website. The VoLGA specifications are available on the VoLGA Forum’s website. The 3GPP hosts the IMS specifications, with the GSM Association hosting IMS Profile for Voice and SMS specifications on their website. We’ve covered the major highlights in this article, as there is way too much to cover. As the specifications detail, there were many improvements at every level of a cellular network that result in a high-performance, optimized network.

Whether LTE becomes the success story of the mobile industry remains to be seen. Network operators around the world are only now deploying it, and already it is turning into a mess. The 3GPP has already approved over forty frequency bands for LTE. Twenty-five of them are for LTE FDD and the rest are for LTE TDD. Roaming is going to be very difficult on LTE. In the United States and Canada alone, there are ten FDD bands and one TDD band for LTE. In Europe, there are three more bands for FDD LTE. In Asia and Oceania, there are the same three FDD bands for Europe, three more frequency bands for FDD, and the same TDD band as in the United States. The rest of the bands have yet to be used, but they are going to be used. Someone is going to have to figure out how to fit more bands on an LTE device without sacrificing portability.

WiMax is not LTE

And then there’s the 4G mess. Contrary to popular belief, LTE at the current stage was not always considered 4G. The International Telecommunications Union (or ITU) determines what can be considered 4G. Originally, the ITU declared that the collection of requirements known as IMT-Advanced determined what would be considered 4G. LTE did not make the cut (though a future version of it called LTE-Advanced did). Neither did WiMAX or HSPA+. However, the American and Canadian network operators’ collective influence made the ITU revise their specification on what 4G is to include any wireless technology significantly evolved from 3G technologies. Most technophiles are of the opinion that the IMT-Advanced specification determines what can be considered 4G, while most business people prefer the newer definition for 4G. For the purposes of this article, the revised standard is considered 4G. While this is out of the scope of this article (and also not really important either), I’m laying it out now to prevent any arguments. This means that LTE, HSPA+, and WiMAX are all considered 4G technologies, though WiMAX is still officially on the list of 3G technologies too.

I don’t know what the future holds for LTE, but it will certainly be very interesting. This is the most exciting time in the mobile industry since the switchover from analog to digital back in the early 1990s. LTE represents a paradigm shift from hybrid voice and data networks to data-only networks. Going forward, wireless network technology is likely to become more widely used because it will become easier to obtain than wire-based services (cable, DSL, etc.). It is doubtful that it would fully replace wire-based data services though. Hopefully, the issues we face with LTE now will go away over time. At the very least, it might jump-start development in more advanced battery and portable radio technologies that can handle more than what current ones can do

Wednesday, December 28, 2011

Ice Cream Sandwich - From Source Code to the OS on Your Phone [Two Perspectives]

Curious about how Ice Cream Sandwich goes from being source code released by Google to being the OS on your device, be it an upgrade or a fresh release? Here are two perspectives from the Product Management teams at Motorola and Sony Ericsson that I thought were especially insightful.

First, here's Motorola's perspective:

Motorola Update on Ice Cream Sandwich

Like you, we are excited to see that Google released the source code to Android Ice Cream Sandwich (ICS)! We’d like to address the question many of you have now – when can I get my ICS? 

There are many steps and processes that go into Ice Cream Sandwich in a way that works for the carrier and for you.  Once source code is released from Google, it doesn’t automatically update to your device.

Each new version of Android launches with one device partner, in what is called the “Google Experience Device” or GED, the showcase device for a new Android release.  The GED partner for each launch works with Google during the development of the OS so that the device and new Android version are ready for a coordinated simultaneous launch.

Once that GED device ships, the rest of the Android community gains access to the Android source code as its made public shortly after – a critical milestone for device manufacturers and component suppliers, enabling us to start work on integrating the new release into our existing products.

Google has performed the initial public release of the Ice Cream Sandwich source code; additional releases will enable device manufacturers to ship commercial product with ICS.  We are currently assessing this source code, and over the next month we will be determining which devices will get the upgrade and when  — and we will communicate this as information becomes available. From there, the following steps take place:
  1. Merge and adapt the new release for different device hardware architecture(s) and carrier customizations: This means that we take the source code and incorporate it into upgrades for devices on which this can perform well, along with making sure the carrier requirements are met.  Silicon partners such as Qualcomm, TI, and nVidia adapt this to their chipsets in parallel and we incorporate these as they become available. This is also the time when we begin integrating all of the Motorola-specific software enhancements into the source code.  Features like MotoCast, Smart Actions, and our comprehensive enterprise solutions are integral parts of our device experiences, and we want to make sure we continue delivering differentiated experiences for our consumers with these software upgrades.
  2. Stabilize and ‘bake’ the result to drive out bugs: This means that we will prepare the upgrade to meet the quality and stability requirements to enter the wireless carrier’s certification lab.
  3. Submit the upgrade to the carriers for certification: This is the point in the process where the carrier’s lab qualifies and tests the upgrade. Each carrier has different requirements for phases 2 and 3. There may be a two-month preparation cycle to enter a carrier lab cycle of one to three months.
  4. Perform a Customer pre-release: We may perform some customer testing before a final release is delivered publicly to our user base.
  5. Release the upgrade: We are planning on upgrading as many of our phones as possible.  The ability to offer the upgrade depends on a number of factors including the hardware/device capabilities, the underlying chipset software support, the ICS support and then the ability to support the Motorola value add software.
Next, here is Sony Ericsson's perspective:


Ice Cream Sandwich – from source code release to software upgrade

A couple of weeks ago, the source code for Ice Cream Sandwich (Android™ 4.0) was released. This meant the start of an intensive period for the engineers at Sony Ericsson, in order to create a working, stable and certified software release of Ice Cream Sandwich for our 2011 Xperia™ phones.

In our pursuit for greater openness in the Android community, we would like to share some exclusive information of the different phases and activities of Sony Ericsson’s Android development, starting with the release of the Ice Cream Sandwich source code, and leading up to the release of a software upgrade for you to download on your phone. Find out more after the jump!

On the Sony Ericsson Product Blog, we have previously announced that all our 2011 Xperia™ phones will receive software upgrades  to Ice Cream Sandwich (please check the Sony Ericsson Product Blog for upcoming announcements regarding the timing for these software upgrades). However, before we can roll out those software upgrades, there are a lot of activities to first of all get Ice Cream Sandwich to work and become stable on all Sony Ericsson phones. We call this the Bring up phase.

Secondly, and perhaps most important, we must certify and approve the new software release with all the different technologies, networks, and hardware that a modern smartphone should work with. We call this the Certification and approval phase.

Ice Cream Sandwich upgrade process
Illustration showing activities from source code release, to software upgrade.

The Bring up phase: Getting Ice Cream Sandwich to work on our phones
On November 14, the Ice Cream Sandwich source code  was made available. This was also the day when our engineers started their work in order to get Ice Cream Sandwich to run on the 2011 Xperia™ phones, by starting the Bring up phase, as we didn’t have pre-access to the Ice Cream Sandwich source code.

The first thing we do is to integrate the latest available Android source code with the Sony Ericsson development branch, to make sure it compiles. This is done by our engineering teams, who in addition need to confirm that all test and debugging tools are in place. Since we will continue to build more on this basic set of code, the stability of the software is very important at this stage.

In the first Ice Cream Sandwich source code that was released, the Hardware Abstraction Layer (HAL) – the software layer giving applications direct access to the hardware components – was to some extent adapted for a Texas Instruments hardware platform. However, for all 2011 Xperia™ phones, we used a Qualcomm hardware platform. This means we have to replace the default HAL coming with first source code released for Ice Cream Sandwich, with our own HAL.

The HAL changes have impact on several features on a phone, including the camera, different sensors (such as proximity, light, accelerometer and compass), audio, Bluetooth™, Wi-Fi™, GPS, as well as multimedia and graphics components. Thus, we do not only have to modify and configure the HAL according to the Qualcomm hardware platform, but also all the other hardware components used in a phone.

Another layer of complexity to the adaption work is the fact that even though all the 2011 Xperia™ products are based on the same hardware platform and use the Qualcomm 8255 chipset, there are several variations in screen size, memory, peripheral components (such as the camera) and modem constellations. At this stage, we make sure to get all the legacy and basic functionalities (such as phone calls, SMS, MMS, SIM card and access to the SD card) to work.

Integrating Android patches
In the Bring up phase, another task is to integrate a number of patches, to improve and adapt the Android legacy code according to our needs. These are customised patches important to the phone, such as improved error handling. To avoid fragmentation, many of these customised patches are actually contributed back to the Android Open Source Project, so that they are included in the default Android source code for the next software release. This work has made Sony Ericsson one of the main contributors to Android. At this stage, we also integrate our own feature upgrades, such as customised graphics.

As we incorporate these customisations, we also do a basic quality check to confirm that all the integrated changes fulfill the required quality level. Here we run many tests cases, such as confirming the functionality of the music player, video streaming, USB and DRM. When we eventually merge all the functionalities in to one code branch, we run more tests to ensure that all our legacy user scenarios run without problems. This also includes that we ensure that apps installed from Android Market are working properly in general.

Getting the software stable and adding localisation
When all functionalities have been merged into one code branch, we start on the stability tests to make sure the software runs stable in all sorts of use cases. Besides pure lab testing, we rely on a number of live users for testing and reporting defects. And along with pure stability testing, we are continuing with the quality tests. For example, we are performing power consumption tests to improve the power consumption in different user scenarios.

Another important task in this stage is to integrate the localisation support in our applications. In most cases, we are adding more languages to the Android software itself. For example, in the software upgrade we rolled out to our 2011 Xperia™ phones in October, we added support for bi-directional languages such as Arabic.

Once all of these tasks are done, and the software is stable and tested, the software is ready to be submitted for certifications. However, the Bring up phase is overlapped by the Certification and approval phase, as some parts of the software are ready for certification earlier than others. As soon as a specific software module is ready, it’s sent for certification. So while one module can be in the Certification and approval phase, others are still in the Bring up phase.

Android architecture showing position of the HAL.
Illustration showing simplified Android architecture.

The Certification and approval phase: Making sure the software and hardware is compliant
A phone comes with parts and software working with a number of different technologies, such as Bluetooth, Wi-Fi, modem and so on. To use these technologies, we have to certify that the new software is compliant with these technology standards. We therefore need to certify a new software release for all of these different technologies due to operator community requirements (GCF , PTCRB , CCF) and Intellectual Property Rights (IPRs), if we have modified the software module related to such technology.

To a great extent, we try to get global certifications, but in many cases specific countries demand specific tests. For a global, worldwide product, we need local certifications approvals in up to 80 countries. For all these external certifications we are either allowed to test ourselves, and submit evidence that we passed, or we can get external test houses, such as Cetecom, to perform the tests.

Besides the operator and IPR certifications, we must externally comply with certifications for regulatory requirements. For example, we must show evidence that we comply with spurious emissions, SAR, EMC, as well as other environmental and health compliances. But when it comes to software upgrades for existing phone models, not all of the regulatory certifications are needed as they have already been performed once.

On top of this, Sony Ericsson is doing internal self certifications in areas we have identified as extra important to deliver the highest possible quality.

All certifications are based on each phone model, and even if a phone model has been certified with a previous software release, the parts which have been changed must be recertified if we are doing a new software release. So if we, for example, have made changes to the Wi-Fi module in an existing phone model, we need to recertify it. When we are working with a new Android release, such as Ice Cream Sandwich, all the affected areas must be recertified.

Another example is if the Bluetooth stack is changed. Then we have to recertify the Bluetooth part of the new software release. There are many apps the needs to be recertified for each of our phones, if they are changed in a new Android software release, for example Bluetooth, MMS, Wi-Fi, HDMI, DLNA and Adobe FLASH. Besides this, we rerun the Android Compatibility Test Suite (CTS) to make sure that we are compliant.

Contrary to what people may think, it is not the Bring up phase, but the Certification and approval phase that is the most time consuming process when it comes to getting a new software release out on our phones. This is one of the major tasks that are legally required from us as phone manufacturer, but is a task that the custom ROM community doesn’t have to take into consideration. Furthermore, by putting all this efforts into testing and certification, we ensure that quality and conformance is at a top level, in benefit for all consumers worldwide.

Additional approvals might be needed
In some cases, the new and complete software release has to be approved to be used in all the different networks that our phones should work in. Many operators also want to customise the software according to their requirements, which in turn are based on their market, network, differentiation and business model. We implement all of these customisations and create a variety of software packages and releases for each operator.

When all of this is done, we are ready to roll out the software release variants as software upgrades to operators and consumers around the world. The software upgrade can then be downloaded and installed either in the operator’s own update tool (if they have one), from Sony Ericsson’s PC Companion tool, or by over the air updates (FOTA – Firmware Over the Air).