Vodacom’s true broadband speeds revealed

As a follow on we are trying to get hold of the raw data from a few speedtest.net servers as well which will allow us to do a comprehensive speed comparison. Not easy though :)
 
The questionable variable is really how many users cellular providers have on what tower during peak usage hours. Each tower is a shared pool of bandwidth, and with only 21 Mbps to play with the effects of users coming and going will create a noticeable variance in the throughput each user experiences at a given time on each tower.
 
Aragon said:
The questionable variable is really how many users cellular providers have on what tower during peak usage hours. Each tower is a shared pool of bandwidth, and with only 21 Mbps to play with the effects of users coming and going will create a noticeable variance in the throughput each user experiences at a given time on each tower.
Click to expand...

The tower can handle more than 21Mbps :)
 
Exactly what I get with with Vodacom.

At one point, I used to get 2mbps ... but the many, many months I've received about .8mpbs ... regardless of the 3.6, 7.2, 21 or whatever claims they make. I actually just just shrug everytime I see a 'Vodacom upgrading all clients to xx mbps'

Research in the UK by Broadband Genie in late 2009 revealed that the average speed is only 0.87Mbits/sec a far cry from the reported advertised speeds of between 3.6 Mbps and 14 Mbps.

Another research project in the UK, this time by Epitiro, shows that the average mobile broadband speed in the UK clocked in at 0.9 Mbps (Megabits per second). The fastest 20 test agents recorded an average speed of 1.8 Mbps.

Similar results were achieved in the US. Results from research conducted by PC World in late 2009 showed that the real average speed of mobile broadband services in the US was below 1 Mbps.
 
Is this a sponsored article to mitigate the effects on Vodacom of no longer being able to use the term Broadband? Could we have some real empirical data and real world examples, times and places if this is meant to be at all useful?
 
Yeah well if you charge people R2 per MB or R300 per GB of mobile data, they're simply gonna use as little as possible. This results in less traffic to the base stations, less traffic over the backhaul network, and overall better performance for the select few that can afford to hammer their mobile broadband connections.

Also, remember that in the US and UK, they have very high caps on their mobile broadband pacakages. Some of them are even uncapped. So getting just under 1mbps average[/B ]is a huge feat for these overseas networks, and in fact makes Vodacom's and MTNs slightly better results look paltry at best. I wonder what will happen to Vodacom's and MTN's networks when they are eventually forced to drop prices by competition from Cell C. I reckon they won't even be able to sustain 512kbps AVERAGE throughput. Forget 21mbps.. they will hardly achieve half a megabit.

Bring down the cost per GB, in line with international norms, then do some speed tests, and then you can gloat about your network, if there is reason to.
 
Tommygun said:
Is this a sponsored article to mitigate the effects on Vodacom of no longer being able to use the term Broadband? Could we have some real empirical data and real world examples, times and places if this is meant to be at all useful?
Click to expand...
Nothing of the kind. We gained access to the actual data from their speed test results, and Vodacom in fact did not want us to do an article on it.

I would love to get access to this type of data from all the providers, but as a test you may want to ask them to release it and see what the result is ;)
 
drukkie said:
The tower can handle more than 21Mbps :)
Click to expand...

Well, the HSDPA standard, using 64-QAM and 15 codes per channel achieves 21 Mbps per channel. So how many channels does a typical tower support?

Firstly, let us assume that we are only looking at a single 120 degree sector of the tower. In the 850 MHz band, it appears that the total bandwidth is around 70 MHz. At 2100 MHz, the total bandwidth appears to be around 90 MHz. Since each HSDPA channel is 5 MHz, this gives you between 14 and 18 channels. I am not qualified in radio engineering, but I suspect that you will have to leave some room between those channels, so the total number will probably be slightly lower, maybe around 12 to 14 channels to allow gaps between the channels.

Next, we have to assume that you can only use 1/4 of the available channels per group of 4 neighbouring towers (4-colour problem). This puts the "safe" number of channels per sector at between 3 and 4. Some googling revealed that many operators only use two channels per sector, but let us use an absolute maximum of 4 channels.

This means is 4 * 21 Mbps will be shared amongst all users in that sector, giving a tower an absolute maximum capacity of 252 Mbps. While this seems like a reasonably large number, the typical capacity (2 channels per sector) is likely to be half of that at 126 Mbps. You also have to consider that you will only be serviced by one sector, so the bandwidth in a single sector is between 42 Mbps and 84 Mbps, which is not so great if you have to share with 10 other users.

You also have to consider that the actual rate achievable by users will probably be below the maximum of 21 Mbps per channel, since SNR will not necessarily be good enough to allow 64-QAM to work.

Then you also have to remember that voice calls consume some of this bandwidth. These consume between 8 kbps and 12 kbps per call, so this also adds up quickly in a densely populated area.

Over and above all this, I recall a rumour that many towers in SA were in fact limited by backhaul capacity a year to two ago. This may have been addressed since. Even if you have sufficient backhaul capacity, the above constraints imply that you are always sharing a 42 Mbps theoretical capacity with the other users in the sector, and you can only see half of that at any give time because you are limited to a single channel.
 
Last edited:
I think these speedtests are kind of biased. You would expect people who get faster speeds to be more prone to testing the speed of the connection.
 
SaiyanZ said:
I think these speedtests are kind of biased. You would expect people who get faster speeds to be more prone to testing the speed of the connection.
Click to expand...
Definitely the case - which is why we want to gain access to 'global' stats...comparative analysis evens things out ;)
 
Yeah it's amazing how fast it can be when everybody is trying not to use it because it's so damn expensive per GB.
 
personally i find the vodacom 3G experience getting worse and worse, one is lucky to find a spot with good reception in Joburg and even CPT is starting to deteriorate. if the cell C network lives up to it's promises i'm switching to them.
 
My MTN hsdpa is always on ~950kbps but on weekends it, wait for it, goes down to 40kbps. They all suck for now but i'll give mobile network 3 year to improve. Look on the bright side, with smartphones the improvement is a given.
 
Well in my experience the speed at which one connects varies as we can expect due to the number of users at one time on a tower, backhaul, cell breathing etc.
But things are improving on Vodacom in my neck of the woods. I live in Sunward Park, Boksburg and I have just done a Speedtest to VC, Jhb site.



Not too shabby, but this is in a suburb and most users are at work.
Now just to get Vodacom to re-align to antenna in out area for a better signal at home.(I am on the fringe of their footprint) I live 800m from the tower and need an external Yagi antenna to get a good signal - this has been a request I posted with them about 3 or more years ago.
They confirmed the complaint by installing the Yagi for me, but I would prefer not to be tied to one spot in my home for reception.
 
fvdbergh said:
Well, the HSDPA standard, using 64-QAM and 15 codes per channel achieves 21 Mbps per channel. So how many channels does a typical tower support?

Firstly, let us assume that we are only looking at a single 120 degree sector of the tower. In the 850 MHz band, it appears that the total bandwidth is around 70 MHz. At 2100 MHz, the total bandwidth appears to be around 90 MHz. Since each HSDPA channel is 5 MHz, this gives you between 14 and 18 channels. I am not qualified in radio engineering, but I suspect that you will have to leave some room between those channels, so the total number will probably be slightly lower, maybe around 12 to 14 channels to allow gaps between the channels.

Next, we have to assume that you can only use 1/4 of the available channels per group of 4 neighbouring towers (4-colour problem). This puts the "safe" number of channels per sector at between 3 and 4. Some googling revealed that many operators only use two channels per sector, but let us use an absolute maximum of 4 channels.

This means is 4 * 21 Mbps will be shared amongst all users in that sector, giving a tower an absolute maximum capacity of 252 Mbps. While this seems like a reasonably large number, the typical capacity (2 channels per sector) is likely to be half of that at 126 Mbps. You also have to consider that you will only be serviced by one sector, so the bandwidth in a single sector is between 42 Mbps and 84 Mbps, which is not so great if you have to share with 10 other users.

You also have to consider that the actual rate achievable by users will probably be below the maximum of 21 Mbps per channel, since SNR will not necessarily be good enough to allow 64-QAM to work.

Then you also have to remember that voice calls consume some of this bandwidth. These consume between 8 kbps and 12 kbps per call, so this also adds up quickly in a densely populated area.

Over and above all this, I recall a rumour that many towers in SA were in fact limited by backhaul capacity a year to two ago. This may have been addressed since. Even if you have sufficient backhaul capacity, the above constraints imply that you are always sharing a 42 Mbps theoretical capacity with the other users in the sector, and you can only see half of that at any give time because you are limited to a single channel.
Click to expand...

lol, this is like combining WCDMA and GSM.

WCDMA ( the technology that HSPA was developed from) uses codes not channels. There is 1 channel or carrier and on top of this each sector uses a scrambling code as its "channel" this is then combined with more coding ( spreading factors) to separate data between devices attached to that sector. So what you have on the 5mhz of carrier bandwidth is a mash of data that needs to be deciphered using a combination of codes for each device.

Right so then the limiting factor for each sector is the number of codes that may be used. This is refered to as the code tree, basically this is like a binary tree that starts with one code equal to 3.84M chips per second, as the tree is traversed the chip rate reduces.Each chip refers to the symbols transmitted, and each symbol represents an amount of bits based on modulation (QPSK = 2, 16QAM = 4 etc ).

Each user is a assigned a code for each service the tower might provide. One code is reserved for signalling, and the rest of the tree may be used for whatever services may be required.Thus for 21mbps, 15 codes are needed, this is at SF level 16, with one code reserved for signalling. As the 16th code is further traversed down to about level 128 for signalling, further services may be offered on the children of the 16th code that are not used for signalling.

So in effect we have one user occupying a massive percentage of the code tree if they wanted 21mbps (real user speed of around 17mbps) which leaves only a few code resources left for the rest of the users attached to the tower. As these codes are shared resources the tower will then try to limit each user in order to accomodate as much traffic as it can. So each user is then assigned a lower amount of SF level 16 codes and this combined with the modulation scheme that determines the downlink speed that a user might attain. The modulation scheme will depend on the signal quality over the air.
 
2nd & 3rd paragraph's of the article include the ff :

''Telkom have upgraded their ADSL speeds three times since 2002, with the latest upgrade bumping up a small portion of their users to 10 Mbps.
Telkom’s ADSL speed upgrade brings South Africa closer to international standards in terms of fixed line broadband speeds, ...''

Yes, but for only a small portion of their users.

Does this mean that 384 & 512 users are abandoned for all-time? (to simply getting upgrades of at least 1Mb.)

Seems that Telkom are now only / mainly interested in competing in the 3G / 4G arena, and if Cell C are successful with their new offerings, Telkom will lag further behind there, as well. Instead of concentrating on getting one thing right (fixed-line ADSL), they're trying to cover all areas, and may invariably FAIL.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X