Handover in mobile phone systems 

As you move around when on a call, your mobile phone continuously measures the signal level and quality from nearby cellsites. These measurement reports are streamed to the current active basestation, which determines when and where to initiate a handover sequence. Complex algorithms are used when making these judgements, in order to ensure that best use is made of all available capacity whilst reducing the likelihood of dropping a call during (or by postponing) a handover.

In the case where connection to the current active cellsite is dropped, the system is smart enough to allow the mobile phone to request a new connection on a different cellsite and reconnect the call. This typically causes a short break of up to a few seconds in the conversation. If the call cannot be reconnected, then it drops out.

3G systems (and CDMA) are slightly more complex because it is possible for a mobile to be actively connected to more than one cellsite at the same time. This feature, called soft handover, allows the same signal transmitted by a mobile phone to be picked up by multiple cellsites and the best quality reception selected on a continuous basis.

Femtocell handover

Femtocells do not implement soft handover, regardless of the radio technology used. Instead, all calls are switching instantly to or from the femtocell and the external outdoor cellular network. This is known as “hard handover” and would typically not be audible or noticeable to the caller.

Where 2G and 3G systems from the same mobile network co-exist, as is very common with GSM and UMTS, then handover between 2G and 3G can also occur. Generally speaking, operators prefer to use their 3G systems because of the higher traffic capacity and lower costs. Their systems are therefore configured to automatically select 3G where good reception is available, reverting to 2G when out of coverage – typically either in a rural area or inside buildings which 3G signals can’t so easily penetrate (due to operating at higher frequencies and having fewer 3G cellsites thus being further away).

Many 3G femtocells are also capable of 2G GSM reception. Will Franks, CTO of Ubiquisys, explains that on startup, their Zonegate product scans both 2G and 3G frequencies. Since 2G typically penetrates buildings better than 3G, it allows the femtocell to determine where it is (by reading the cellsite identification on its broadcast channel), derive some timing/clocking reference (as one input to its timing algorithm), and work out which 2G cellsites might be most appropriate to handover to when a mobile phone leaves the femtocell zone. Presumably, these 2G cellsite identities can then be transmitted to the mobile phone as potential handover candidates (known as the neighbour list), and be measured during any active call in case a handover is required.

Femtocell handover/handoff scenarios

The possible handover scenarios are divided between entering and leaving the femtocell zone, and whether switching to/from the external network using 2G or 3G. This is shown in the table below:

The numbers 1-4 indicates the importance of this handover mode from a usability perspective. For example, Ubiquisys has explained that their current implementation includes 3G to 2G call handover when leaving the femtocell zone. Their product hardware is capable of handling all the other modes, and this can/will be added later as a software upgrade.

What is the effect of femtocell handover for the user?

There are two aspects to consider:

a)Usability – does this cause a problem or poor service to the customer
b)Billing – what is the impact on how much the user pays for a call.

Usability

For voice calls, the user is typically unaware whether the phone is using 2G or 3G mode. The call quality is unlikely to vary in good reception areas – other factors present a bigger challenge.

In countries/areas with good 2G coverage, the scenario would be that calls originating outside continue using the 2G service until completion, even when entering the range covered by the femtocell. There is no automatic handover into the femtocell zone, so if you walk into your house during a call it will continue to use the external system for the remainder of the call. From a billing perspective, this is simpler to implement (effectively no change to the current system), although end users might be surprised in some cases. Arguably, this is a relatively unusual case.

Billing Implications of Handover

Operators who have implemented similar schemes, such as dual-mode mobile/WiFi phones, have dealt with the billing aspect as simply as possible. Charges are based on where the call originated (i.e. inside or outside the femtocell coverage), and continue on the same basis regardless of handover to/from the zone. Thus a call started outside would continue to be charged when entering into the femtozone, even though calls made inside are free or included in a bundle. Likewise, calls originating inside the femtocell zone would continue at the same free or discounted rate despite leaving the zone and continuing outside.

Assuming that relatively few calls do require handover to/from the zone, and that the number of minutes balances out, it is probably not worthwhile for the operator to implement any more sophisticated charging solution which wouldn’t net additional revenue. Instead, clarity of the billing mechanism is required in the publicity material/staff training and rates need to ensure that there would be no large loophole for revenue loss to exploit.

Femtocell handover in poor coverage areas

Where users have bought a femtocell to improve coverage in their homes, perhaps a more common case in North America, then this can be a problem. End users know the femtocell provides good coverage indoors, and yet when walking inside the call drops. It has been said that this has been a common customer complaint with Sprint’s Airave service (which uses the CDMA mobile phone standard rather than GSM/UMTS). Perhaps in countries with better all round indoor coverage this may be less of a problem.

When leaving a femtocell, it is easier to switch over to 2G GSM because the mobile phone can scan these different frequencies whilst in a call using 3G. The femtocell can “discover” nearby 2G cellsites by sniffing the operators licensed frequencies and noting their identity. All 2G cellsites also broadcast their own neighbour lists of nearby 2G and 3G cellsites (known as neighbour lists). These can be picked up by the femtocell which can instruct the mobile phone to measure signal level and quality from those as well.

Whilst implementing handover for leaving a femtocell zone is much more under the control of the femtocell itself (because it can instruct the mobile phone what to look out for, and issue commands to force a handover), the case for implementing handover INTO the femtocell zone is a bit more complex.

Radio planning in mobile networks

In the external network, sophisticated radio planning tools are used to determine the best settings and configuration for all cellsites. These tools output many individual parameters (millions of different individual numbers) which are downloaded into the thousands of cellsites. Different values may be determined for transmission power, frequencies used, neighbour lists (for both 2G and 3G), tilt (the angle at which the radio signals are sent/received - some cellsite antenna can change this electronically, others require physical re-alignment), handover algorithm to use etc etc.

In the past, operators would conduct a complex refresh of their entire network perhaps once every two or three months. More sophisticated modern tools have reduced that cycle to days or weeks, and work continues on “Self Optimising Networks” which automate this process still further.

What these complex tools have no awareness of (so far) are the potentially large numbers of low powered femtocells operating in their area. There are potentially three approaches to optimising handover into a femtocell:

1. Adding femtocells to the neighbour lists of the outdoor macrocells. This is unlikely to be a scalable or workable solution. Although neighbour lists can be quite large (I believe it was extended from 32 to 64 a few years ago), the time taken to scan round many different settings increases proportionally. In dense urban areas, there may potentially be some 100’s of femtocells collocated with an outdoor macrocell. It is also questionable whether this would benefit the hundreds or thousands of users served by an outdoor cellsite, where their mobile would be scanning femtocells that they may not have authorised access to. In this case, the mobile phone would not be searching for the most likely cellsite to switch over to, and dropped calls would increase. Additionally, the complex management to download and maintain vast numbers of femtocell candidates adds an overhead to the network operator.
2. Adding some smarts into the mobile phone. One of the key benefits of femtocells is that they work with any standard 3G phone – this is a clear competitive advantage compared with WiFi dual-mode solutions that are restricted to specific (and sometimes more expensive) dual mode devices. However, it could be argued that with some additional functionality in the phone itself, then improved handover into the femtocell zone is enabled. For example, the phone could learn about its femtocell zone and the matching external cellsite used outside. When on a call in the external cellsite it could additionally monitor for the femtocell and switch across to it when in range.
3. Making the femtocell as clever as possible. Ensuring that any calls about to dropout when entering the femtocell zone are quickly restored as soon as the mobile can detect and lockon to the femtocell. Parameters selected by the femtocell, such as the cellid and paging zone, can encourage more rapid identification. Some optimisation may be required in the mobile network too, but the idea would be to avoid any changes to the mobile phone itself. This is one area where femtocell vendors will be able to differentiate themselves. 

Which femtocell handover modes should be implemented

I would say that the benefits of implementing all four modes of femtocell handover outweigh the costs, at least in the short term. For basic voice calls, users do require call continuity, but are much less concerned on whether its carried on 2G or 3G. The priority for vendors is to implement outgoing 3G to 2G handover, on the basis that this would cause the most dropped calls. Some form of inbound call handover would be next, although whether from 2G or 3G probably depends on the type of application – i.e. is it due to poor coverage or for additional services/speed. Perhaps 3G inbound might be higher priority, since this is more likely to dropout due to the poorer in-building penetration of 3G. It’s also more likely the call will be on 3G, because operators seek to push calls onto 3G where the mobile device and network support it.

The above has concentrated on handover for “traditional” voice calls rather than data sessions (or VoIP over a data session). Data handover is handled differently for 2G and 3G – in 2G GPRS, the mobile device has much more autonomy over when to handover, whilst in 3G the data session is carried over the same radio channel as a voice session. Unless using a streaming service such as voice or watching a video, it is generally much more acceptable for a dropout of a few seconds whilst the mobile device switches to or from the femtocell. The mobile device can effectively restart a new data connection if required, yet retain the same end-to-end data session with a server or other party.

This may reduce the effectiveness of some of SIP/IMS femtocell architecture being promoted by some femtocell software vendors and Softbank, the Japanese operator.

The lack of effective handover between 2G and 3G was one of the early problem areas of 3G when it was first deployed. It was an aspect that was left to a later stage of the standardisation process when 3G was first created. Handover from 3G down to 2G was implemented as a priority to address this problem, and improved considerably since then so that it is no longer a problem. Over time, various “tweaks” and improvements were added to cater for common dropout cases and allow calls to be recovered rather than fail.

Summary

We are more likely to see successful handover for traditional Iu based femtocell architectures first, and this will improve over time. I suspect we will see a few additional “tweaks” added into the standard 3G phone specifications, which will improve performance and reduce dropouts in the future. If Voice over IP data calls are to be used, then further optimisation of the standards are likely to be required to avoid poor performance.

Comments (4)

Sam Taylor said:

Location area update a headache in femtos Hello all, One of the technical difficulties that the public knows little about but that has the current triallers fretting is that of location area (LA) updates. The number of location area codes was not a problem in the usual macro/micro deployment but will be with femtos. If someone's femto has the same LA as mine and I am not allowed on theirs when I walk by, my phone sees that base station as "forbidden" and keeps track of the LA in its "not permitted list" when i reach my own femto, I might not be able to get onto my own femto. How about that?

report abuse vote down vote up

May 18, 2008 | url
Votes: +0

Rajesh Pandey said:

... This is my repsone for Sam Taylor's comments- This problem can be avoided by sending proper reject cause to the mobile i.e. network must not use reject cause like 'Location Area not allowed' or 'Roaming not allowed in this location area' becuase on receiving these reject casues the MS will put the LA in its forbidden list. Best reject cause which can be used is 'retry upon entry into a new cell' This will not put the LA into forbidden list and the problem you metioned in your mail can be avoided.

report abuse vote down vote up

June 09, 2008
Votes: +0

3gpp said:

	... .... but still the problem exists that a UE will on each femto perform the LAU and get a LAu reject resulting in several seconds service outage during the LAU/LAU reject phase ... and: "retry upon entry into a new cell" is not a LAU reject cause which triggers a UE mobility procedure ....
	report abuse vote down vote up

May 08, 2009
Votes: +0

Moh Yan said:

... Adding femtocells to the neighbour lists of the outdoor macrocells. This is unlikely to be a scalable or workable solution. Although neighbour lists can be quite large (I believe it was extended from 32 to 64 a few years ago), the time taken to scan round many different settings increases proportionally. In dense urban areas, there may potentially be some 100’s of femtocells collocated with an outdoor macrocell. It is also questionable whether this would benefit the hundreds or thousands of users served by an outdoor cellsite, where their mobile would be scanning femtocells that they may not have authorised access to. In this case, the mobile phone would not be searching for the most likely cellsite to switch over to, and dropped calls would increase. Additionally, the complex management to download and maintain vast numbers of femtocell candidates adds an overhead to the network operator. The "management" is not as complex as you think. I have devised a scheme that uses the tech that is / must be already used in anger in existing UMTS BSCs (for similar/different reasons) . Download traffic is nothing more than file transfer (which is easily optimised) . No broadcasts of massive neighbour cell lists, nor ever changing lists as femto-cells are on and off. The only issue I have is that more storage is required on the BSC to hold the femto/macro cell topology and the UE / femto associations, when all femto-cells are active at the same time, and hand-in is imminent/required for at least one UE per femto-cell. But even then, because hand-in is a low intermittency transient event, tis easy to optimise to allow the above info to be discarded as/when the BSC sees fit to do so.

report abuse vote down vote up

December 22, 2009
Votes: +0