Speaking Engagements & Private Workshops - Get Dean Bubley to present or chair your event

Need an experienced, provocative & influential telecoms keynote speaker, moderator/chair or workshop facilitator?
To discuss Dean Bubley's appearance at a specific event, contact information AT disruptive-analysis DOT com

Sunday, June 22, 2008

HSDPA on the fringes of coverage

There's a lot of talk about whether 3G dongles with flatrate data represent direct like-for-like competition against home DSL or cable broadband. My view is that at some levels this is true - especially for people unable or unwilling to take out long contracts for fixed-line connections, eg students or people moving home regularly. There are also segments of the market that want both fixed & mobile connections for different reasons, and a zillion sub-segments with unique requirements of their own.

That's all well and good, but discussion recently about fixed-line broadband (in the UK at least), has highlighted a lot of unhappiness about headline advertised data rates ("8MB Broadband!", "Superfast 20MB cable"), where the figures quoted (usually with "up to" in small print) don't match the reality. In those cases, it's usually a limitation imposed by the length/quality of the copper, so that if you're living 4km from the local exchange, you get a lousy speed. Although obviously there's a lot more to it around contention rates, backhaul and so on, there is at least a chance that you could be given a "guaranteed" data rate rather than a hypothetical one.

My experience with HSDPA recently has been a little patchy. In particular, I'm becoming increasingly aware of one of the limitations of the technology - the way it dishes out much more capacity to those people in the best radio conditions. It's not just that the signal degrades with distance / indoor penetration (ie the equivalent of DSL line length), but the fact that the system deliberately biases itself towards people with better radio reception.

This exacerbates the difference between the best and worst grades of service - to a greater degree than happens in the fixed world. On ADSL, you'd have to be pretty unlucky not to get 2-3Mbit/s on a supposed 8M service (ie maybe 70% less than headline rate). But on HSDPA, it's quite possible to get a 90% or more reduction in speed when you're on the fringes of coverage, versus when you have 5 bars on the signal meter.

I notice this regularly - I have 2 branches of Starbucks near me on Baker St in London, perhaps only 200-300 metres apart. One has seating in the basement, the other at ground level. Sitting in one of them, I'll regularly get 1Mbit/s+ with high reliability & low latency (this is on Hutchison 3, and I haven't got above 2Mbit/s at all). But in the other, I'm often lucky to get 100kbit/s, and often the network will just seem to 'hang' for 20-30 seconds. I'm not sure if they're both in the same cell, but the difference between 1-bar signal and 5-bars is at least a 10x change in throughput.

Does anyone else experience something similar? Or is this an issue that's going to cause customer service issues, or perhaps drive the need for femtos, or 900MHz 3G? In either case, it strikes me that any regulatory approach to selling "What you pay for is what you get" broadband is really going to struggle with mobile.


wirelessman said...

That's about right. The difference in the minimum and maximum HSDPA data rate (i.e. edge of cell vs sitting under the mast) is about a factor of 25. But wait, it gets better. HSPA+, which uses MIMO to double data rates, increase the top rate (doubles it actually) but does nothing for the bottom rate (leading to a factor of 50). This is the dirty secret and huge design challenge of wireless.

In practice, there are two main solutions to this problem, neither of which require fancy air interface technology, (i) smaller cells and (ii) more spectrum...lots more spectrum.

Anonymous said...

dean, can you comment on which starbucks has the bad coverage. one where you sit in the basement or the other one?


Martyn Davies said...

I don't see that femtocells will solve this problem. The radio coverage would be better, but then it's a backhaul problem. A bit like WiFi services today; you can find radio coverage, but enterprises don't always invest in enough backhaul to make the service useable.

Aditya Kaul said...

HSDPA coverage is definitely an issue indoors and we need a combination of femtos and WiFi to solve that. Although we all keep harping about backhaul, and coverage issues what about an oversubscribed air interface when majority of users in the cell are using data services like video stream, music stream, music downloads? In the longer run we need a combination of peer to peer techniques and multimode devices that connect to the "best interface" to solve this bandwidth bottleneck at the access layer, assuming that we will have many more 3G dongles and embedded HSDPA chips in laptops, MIDs. We all know what happens to an oversubscribed WiFi network at conferences, don't we?

alex said...

UMTS900 will elegantly solve today's problem pointed out by Dean, together with higher capacity backhaul to existing macro sites.

femtos may solve the subsequent capacity problem - that is once the additional radio efficiency included "for free" in UMTS900 and HSPA+ is exhausted.

Dean Bubley said...

Thanks for your comments all.

Anonymous - the poor signal was in the basement cafe.

Anonymous said...

Most places that I have been to are selling HSDPA plans in the same way they do with ADSL. Peak rates...which does not happen very often in ADSL either.

As the 3G cells get dense for capacity, perhaps we would see some statistical leveling. I believe at AT&T they advertise average speeds of 400-700kbps. Average over what period? But that is still much better than the illusion of 7.2Mbps. Under the cell site and at 4am when most of the world is asleep.

I doubt that if the service is delivering 500kbps on the average, there would be much complaints. That is the level that most web pages load at a reasonable rate. I have not heard of mass complain about ADSL throughput. I have a 10Mbps plan. I know it is much faster than my previous 5Mbps plan but I am definitely not getting 10Mbps. I'm lucky to get 5Mbps. But at the rate that I'm getting compared the old exorbitant rate, it seems like my carrier has conditioned me into not complaining.