June 16th, 2010
Sorry I’ve been quiet – Life’s hectic!Here are some of the areas I have been thinking about and will blog on shortly. Feel free to add your thoughts before I get there to start the ball rolling:
- Periodic vs Event Driven MRs.
- Is SON in LTE going to remove the ‘need’ for MRs?
- How many MRs do I need to do xxx?
May 3rd, 2010
In all of the networks I have visited and analysed transmit powers are too high and power control is too slow.
Both of these ‘features’ will give a call-in-trouble a better chance to stay up at the cost of reducing the prospects to multiple other calls. Let me explain: ‘conservative’ settings of transmit power and power control mean that the transmit power to that call will be higher this will improve the coverage/ interference immunity to that mobile. Sounds good, but the cost to the network is extra transmit power causing interference to any other calls on the relevant frequency(ies). In the early days of GSM there were few mobiles to be interfered with, coverage was the most important issue while you rolled out your network. Those years are long gone, now networks are interference limited, especially if you want to use EDGE/ Evolved-EDGE or even AMR-HR to increase capacity. Every extra milliwatt of transmit causes a reduction in C/I which either reduces capacity by increasing the likelihood of a codec mode change or even a drop.
Most networks do not like to reduce maximum transmit power on the basis that it may reduce coverage/ in-building coverage. It is very hard to say how much transmit power reduction you can do without losing in-building coverage. MR processing can give you a good hint, but caution would be advised. A ‘little’ extra interference to avoid customers losing coverage, when the friends on different networks have coverage is worth it.
But given that most cells have 3+ carriers then having a BCCH carrier on this extra power is worthwhile, if the other carrier’s transmit power is kept constrained.
If you care about your users, your KPIs and your revenue reduce transmit power whenever possible:
- increase the speed of power control decision – processing power is cheap now, so make a decision after 2 MRs,
- increase the step size of power reductions
- use dynamic power control steps, if your BSS supports it.
- constrain some carriers to lower power – use concentric cells, I.U.O., E.R.P. or whatever you vendor calls it.
- * MOST IMPORTANTLY* if running at full power can’t keep the call going, drop it quickly. Keep RadioLinkTimeout low.
Saving *a* call at all costs is too high-a-cost.
April 26th, 2010
Most optimisers talk about measurement reports, but is that what they really mean? Well not the good ones at least.
The measurement report is the message from the mobile to the BSS. It may appear to contain a lot of information but it only contains half the story. When the BTS receives the measurement report it adds two important extra pieces of information and possibly adds a third. The sum of the two parts is called the measurement results. Without those extra pieces a lot of the potential of the measurement report is lost.
The 3 additional parts are:
- Uplink Measurements – RxLev and RxQual. Most networks are uplink limited, if the mobile had more transmit power then the coverage and C/I could be increased (at the cost of a warm head and reduced battery life). Optimising for downlink coverage/ capacity is not sufficient.
- Downlink transmit characteristics – transmit power and DTX. Without these You cannot interpret RxLev DL and know whether to use the full or sub values. Again vital information to use the Measurement Report.
- [optional, depending on your equipment vendor] look up the reported neighbours in the measurment report to give BCCH ARFCN not just the index into the neighbour list. If optimisers/ planners are constantly adding new sites and changing configurations then this is essential. Without this information you would need accurately timestamped configured neighbour lists for every change. Or you could log all SYS INFO 5′s and friends. The best way is to get the BSS to do the look up for you, after all it knows neighbour list. At least 2 of the big vendors do the lookup for you, if you know where to look. If not then one missed CM change and your neighbour lookups are wrong and you may not know it. That could be adding the wrong ‘missing neighbour’, so even mis-identifying an interferer and causing real issues.
Most people say Measurement Reports because they grew up looking at drive-test logs and seeing the Measurement Reports from the mobile.
So do they really mean Measurement Reports? No, the good ones are been lazy, they just haven’t switched to calling them Measurement Results. The bad one, probably still think the Measurement Report is everything they need, but they won’t be reading this. They will be trying to drive test microcells and in-building systems. So they might mean Measurmenet Reports, but I wouldn’t bother listening to them anyway.
Mr. Measurement Results
April 22nd, 2010
Most mobiles are not location aware and they still work. If I can make and receive calls without the mobile knowing the location then I can gather the data needed to optimise those calls. The MR tells us where the mobile is in ‘radio space‘, i.e.using which serving cell and near to which neighbours/ potential interferers. For the great majority of optimisation tasks that is sufficient.
For years optimisers have lived with maps which show either cell locations or (pretty but less useful) showing predicted coverage. Given OSS statistics of network performance, poor performance issues can be identified. Once identified maps will help one can guess on the likely location of the issues and even possible causes. Then the choice is drive-test or try-it and see what happens. If you have worked that way for years the map is key. But there are 2 problems with this way of working:
- it is manual – issues are tackled one at a time
- it is reactive, tackling the biggest issues first
An automated network-wide approach is possible if you embrace the MR and throw away the map (well at least put it into a drawer for now). Are you employed to fix issues or to optimise the network?
Can I really live without location?
I would love to say yes, but in reality: No. There are a few tasks where location is required:
- Antenna azimuth/ downtilt optimisation
- Identifying hotspots for adding additional capacity
- Identifying coverage holes – no MRs where there is no coverage
- calls from the area around the coverage hole will show poor coverage – that can give you a very useful hint
So the next question is how can I locate all of my calls? And that is too big for this post.
April 21st, 2010
On several occasions in my career I have compared the first few MRs in a call with the remainder. Each time I was looking at different aspects of those MRs. My conclusion is that the first few MRs are:
- more noisy, RxLev has a higher standard deviation
- higher BER/ FER: there is a small but statistically significantly worse RxQual
- fewer neighbours reported
And interestingly this also affect the first few MRs at the start of a TCH, following an ASSIGNMENT COMMAND from an SDCCH.
My guess is that this primarily because of the MS. RxLev and RxQual measurements will be affected by the frequency and phase stability of the mobile’s oscillator. The handling of neighbour list changes when going from Idle (SYSINFO 2) to SDCCH (SYSINFO 5) to TCH (different set of SYSINFO 5) appears to vary from mobile to mobile. The measured reduction of reported neighbours even occurs when the 3 neighbour lists are identical. Sounds like a question for the mobile vendors to me.
My recommendation is don’t make handover decisions too quickly at the start of a call, unless the call is instantly in quality problems, and don’t make important decisions on the basis of the first few MRs, eg DFCA.
Waiting just 5 seconds (10 MRs) will improve the usefulness of the MRs.
April 21st, 2010
Most definitely all technologies!
Not wishing to sound like a politician during this election time, but in a way this is the wrong question. The real question is: ‘What do you have to do to optimise a network?’
To keep it simple:
- 2G – Frequency Planning
- 3G – Antenna downtilt and azimuth
- 4G – sub-channel planning
All of these can be achieved with MR processing. In 2G and 4G, I would propose that MR processing is the best way. I will talk about each of these in the future.
In 3G, there is an element of needing to be able to predict the effect of moving antennas, this cannot be estimated purely from MRs. Some form of propagation model is needed also. This makes it expensive to optimise; so expensive that the cost of the site visit to move the antenna becomes trivial (depending on how many you alter). It has always seemed to me that 3G was a very clever idea, but no-one thought about how do you optimise it when it was designed. Fortunately 4G is not making that mistake. My advice has been and will continue to be – if you data capacity can stand the delay, skip 3G. GSM/ EDGE+ to 4G will be much easier to manage.
If you don’t agree, please sign-up and comment/ put up your own posts!
April 9th, 2010
I have vague memories that Mister Mister was pop band in the late eighties but that is not the relevant answer. Yup, a quick Google later and the details can be found on Wikipedia.
The MR in Mr. MR is the Measurement Report. This message discarded in the billion’s contains the best source of information for radio optimisation. I was in the right place at the right time to be involved with implementation of IOS (Intelligent Optimisation System); recognising the great value of the MR I pushed for it’s development and growth, earning the moniker ‘Mr. IOS’. But that is a Motorola product with its limitations and flaws, the key concept is the processing of MRs, so I prefer the label ‘Mr. MR’.