10Mbps TV white spaces trial network planned for South Africa

Wimbo

Well-Known Member
Joined
Aug 4, 2004
Messages
119
bla bla bla throughput bla bla bla. Whats the LATENCY like???
Short answer: I have no idea :D

Longer answer: TVWS technology has nothing to do with latency (in the IP packet sense)

TVWS space "technology" is all about finding out if a particular frequency is available for use, on a dynamic basis, without causing interference to the primary use of the spectrum, namely TV broadcasting. Once a usable frequency has been identified, then existing radio-based technology can be used, such as Wi-Fi, LTE, or whatever.

Perhaps it's worth refreshing our memories about the causes of latency?

If you have a direct connection between two Hosts, then the latency is essentially a function of the distance between them. Signals travel at an appreciable fraction of the speed of light in a copper or fibre cable. The longer the cable, the higher the latency. That's why we're talking about say 40 mS for Paris to Washington and 75 mS for RSA to Europe over SEACOM, compared to over 300 mS for a satellite connection.

However, this unfortunately isn't the only factor. If there're one or more routers between the two Hosts, then the latency is increased by two more factors: the time it takes a router to process a packet, and the time the packet sits in a queue on the router waiting to be processed.

If a router is too small for the job, or it's overloaded, then the latency due to the router can be far more significant.

Wireless connectivity adds further latency, for three reasons: the data is often TDD (Time Division Duplexed), which means that sending and receiving take turns; the radio transceiver also has a queue, and most significant of all, the available spectrum is shared between all users. Too many users on a base station can send the latency sky high.

So, the answer to your question is: It doesn't depend on the technology, it depends on how it's implemented, and how many users are using the service, compared to its design capacity.

It's more expensive to "overdesign", which would reduce the latency to something close to the theroretical minimum due to physical distances.
 
  • Like
Reactions: rpm

UnUnOctium

Expert Member
Joined
Mar 6, 2005
Messages
3,127
The article may have implied that I am leading this initiative. I'm not. I just sit on steering committee of the project, that's all.

I should also have said that the project is entirely dependent on the good will of ICASA with whom we have not yet applied for a test license for this project. They have been very positive about TVWS so far and very ready to engage but it is still early days.

If you are looking for more information about TVWS, here is a link to a good background article.

http://www.dailywireless.org/2011/08/12/white-space-war/

-Steve
Of course, the bottle neck (and it's more of a stop-valve in this scenario) is ICASA and whether they will grant a license for this network. I'm guessing soon enough (if not already) that you guys might have it since it's very easy to blame a test if something goes wrong :)

As for the network, do you guys intend to use infrastructure or ad-hoc? And will you guys be down-mixing IEEE 802.11 radios or will you do a proprietary transmit/receive signal chain (AFAIK there's no IEEE 802.22 devices out there you guys can use)?

More technical details please!!! Are they using cognitive network technology?
More than likely not. They'll be using dynamic spectrum access but I highly doubt they'll propagate the intelligence upwards to the network level. I think what they'll be doing is like the FCC did in the states and that's using a database containing spectrum information for a geographic location. In other words, the device connects to the database (probably through the mobile network) and obtains information about which channels are available for use in its location.

If they are using some form of spectrum sensing, I'll be quite interested to have a look at the details but I doubt they would be using anything other than a very rudimentary form (think FFT + energy detection + hard decision). Would definitely love to speak to some of the designers/engineers.

I was about to ask about licensing issues, I guess this answers my main question - we still need the cooperation of ICASA, a major hurdle, as positive as you may want to be.

My other question is how does this affect the migration to DTT and the digital dividend. The way I understand it, TVWS will be making partial use of (future) digital dividend spectrum which will eventually sold/auctioned off in chunks, making TVWS a wasted exercise. Right?
TVWS is actually a precursor to a future technology called cognitive radio. Currently TVWS serves as a quick fix for the problem of spectrum crowding. As it sits, the licensed spectrum is unused a large percentage of the time by its licensees (figures can be as high as 85%, or 100% with the case of Sentech in SA). TVWS technology aims to exploit the unused TV channels (due to interference mitigation requirements as mentioned by Wimbo) since they're large and easy to exploit. Once the digital dividend is fully open though, we'll still likely license it in the old fashioned, exclusive way (regulatory authorities do not like thinking ahead) and we'll back at the problem of "companies have spectrum they don't use and the ones that need it aren't allowed to use it"

it all sounds good in theory, but what about the uplink? I imagine everything needs to go through Telkom in the end? Unless you have some kind of infrastructure dedicated to this project.

10mbps for users is great, but as we currently see with Telkom... sharing a 155mbit ATM uplink over 5000 users doesnt work
1) depends on how it's duplexed, purely a system design parameter (unless they love to repurpose 802.11 transceivers, like all WAPA members do :))
2) depends on what EIRP ICASA allows them to use. If they're allowed to transmit at 'high' power, they'll more than likely implement the CPE as transceivers rather than just receiver. I'm vouching for the latter given the propagation characteristics of the spectrum they aim to exploit.
 

ambo

Expert Member
Joined
Jun 9, 2005
Messages
2,680
The longer the cable, the higher the latency. That's why we're talking about say 40 mS for Paris to Washington and 75 mS for RSA to Europe over SEACOM, compared to over 300 mS for a satellite connection.
Those numbers look closer to the propagation delay (one direction) rather than the round trip time (RTT) that is usually considered when looking at latency. And even then... those numbers are optimistic ;)

Currently the lowest achievable RTT is about 150-odd ms from Cape Town to London via the West coast. That's the router to router latency and excludes any kind of local access network. From JHB you add 25+ ms. The lowest latency you'd see on the East coast cables (Seacom and EASSy) would be from Durban but most of that traffic goes via JHB first. JHB to London via the East coast is around 220ms.

However, this unfortunately isn't the only factor. If there're one or more routers between the two Hosts, then the latency is increased by two more factors: the time it takes a router to process a packet, and the time the packet sits in a queue on the router waiting to be processed.
This is actually a myth as I was discussing a few days ago. The only time that you will see significant forwarding delay on a router is when its congested and the buffers are filling up.
 

Saajid

Expert Member
Joined
Aug 8, 2008
Messages
4,528
Thanks guys for the technical explanations. Very interesting and insightful. Much appreciated.
 
Top