Paul Brooks pbrooks-link at layer10.com.au
Tue Apr 7 22:59:49 AEST 2009

George Bray wrote:
> This ongoing comparison between fixed and wireless capability has got
> everyone fooled.  Telstra's musings about the capacity of their 3G
> network vs FTTN/FTTP are designed for journalists to replay to the
> punters, quoting the headline "speed" for comparison. After today's
> news, this line of argument should change to comparing 20Mbps 3G
> wireless speed with the current capacity of a fibre (~1500 Mbps).
> Both headline figures are nothing like regular end-users will see, but
> if your aim is to "educate" the mainstream media in comparisons of
> bandwidth capability this obfuscation is to your advantage.
Whenever this comes up, I like to point out the broadband is measured by 
two metrics - capacity* (whether peak or average) and _latency_.
All the debate so far, and Telstra's musings about taking over the 
country with ubiquitous 3G coverage, focusses on link capacity - 100Mbps 
here, 42Mbps wireless there, etc. and the media don't know enough to ask 
about latency and delays.

Every 3G network I've used runs regularly at 80 - 150ms round-trip-time, 
and regularly peaks unpredictably to 400+ ms. This compares to DSL or 
cable, where latency is less than 10ms.

There are several applications where 3G networks simply won't provide 
the same quality user experience as a cabled network, regardless of the 
link capacity, because of the high and variable latency - your VoIP 
session will work much better over a bog standard 1.5Mbps ADSL1 link 
than it will over a 42 Mbps 3G wireless session, and forget about online 
shooter games.

> Forget even 350,000 users.  Take a comparison of 100 concurrent users
> on each technology, downloading (and uploading) as much as they can.
> The wireless system fails at some point because the bandwidth of the
> medium is shared between concurrent users. On fibre, each user should
> get the full capacity of the service they purchase.
Sorta-kinda - FTTH networks are usually based on a PON architecture, so 
the 32 - 64 users sharing the PON link have to share between themselves 
just like the radio access, and HFC networks.
> By declaring that Australia's retail network will be a high-capacity
> piece of glass to every premise, the Govt has opened up the
> possibility of markedly different broadband applications -
> applications that couldn't be considered with a FTTN network, let
> alone some wireless one.
> So what *are* the capacity/distance properties of today's FTTP fibre
> systems?  Paul? Richard?
I'll have a crack...

GPON, the current latest generation FTTP system, shares a 2.5Gbps 
downstream link amongst 32 - 64 users (split ratios smaller and larger 
than this are possible, but not common. 32-way splits are the most 
common implementation). If everybody was madly downloading 
instantaneously then each would get around 78 Mbps max, but stat muxing 
means that the vast majority of the time you should be able to saturate 
your FTTP modem's 100BaseT port.

The distance limit is about a 20km diameter area centred around the 
splitter (governed by need to synchronise timing of upstream 
transmissions), but the splitter itself I believe can be some distance 
from the head-end transmitter (the ONU) - I think the maximum total 
distance of a user from the headend is about 60km, but it may be less.  
Think of a dandelion, with a 40km stalk and a 20km diameter puffball 
with up to 32 - 64 users within the puffball area, and you'd be close to 
the picture.

There are no slow-downs caused by distance - whether you are 40km from 
the ONU or 4km, everyone shares the capacity equally.

Personally, I think one of the biggest improvements of FTTP is not that 
people get a bigger pipe, but rather that the improved service distances 
available with fibre means there should be no more urban blackspots - 
for mine, its much better to provide some level of broadband access to 
those that have none, than to give someone that already has a fat pipe a 
fatter pipe.

*I refuse to use the word 'speed' - the speed is always the same, 300 
meters per microsecond in air (close enough), or 200 metres per 
microsecond in glass fibre.


More information about the Link mailing list