How exactly does cell phone/tower bandwidth work?

[HibyPrime1]

Googling this didn't find any useful answers, so I thought I'd try here.

Town A has a single cell tower providing HSDPA (or other 3G/4G network type) at 14Mbits/sec. 3 smartphones download a movie, do all 3 users get 14Mbits? or is it split between them? Does it depend on the backend of the tower?

If the bandwidth is split between them, what happens when there are two towers in the same area, with two phones connected to tower A, and zero connected to tower B. When the two phones start downloading and bandwidth is split, do one of the phones auto-reconnect to tower B for full download speed?

Also, since most phones have more than one band they use, does it use more than one at any given time? If not thats not the case: if the 800MHz region is crowded, does the tower send the phone over to the 1900MHz band (or whichever band)?

 

[Mark R]

Googling this didn't find any useful answers, so I thought I'd try here.

Town A has a single cell tower providing HSDPA (or other 3G/4G network type) at 14Mbits/sec. 3 smartphones download a movie, do all 3 users get 14Mbits? or is it split between them? Does it depend on the backend of the tower?

If the bandwidth is split between them, what happens when there are two towers in the same area, with two phones connected to tower A, and zero connected to tower B. When the two phones start downloading and bandwidth is split, do one of the phones auto-reconnect to tower B for full download speed?

Also, since most phones have more than one band they use, does it use more than one at any given time? If not thats not the case: if the 800MHz region is crowded, does the tower send the phone over to the 1900MHz band (or whichever band)?

HSDPA data rates are 'per radio channel'. So, if there are 3 phones in a town, accessing a tower, then the tower will try to allocate each phone a radio channel (as long as the tower has enough channels allocated to it.)

The tower tells each phone a signal telling it which channel to use for the next 2 ms. Once that 2 ms is up, the phone goes idle, and awaits further instructions. If there are enough channels to go around, the tower just keeps immediately sending 'channel available' messages to each phone. If there are more phones then channels, then the tower will allocate time slices in round-robin fashion. Note that the data rate isn't evenly shared - phones will use whatever data rate they estimate will give optimal speed, so if a phone has a weak signal, it will switch to a lower data rate for its time-slice.

How many channels a tower has, depends on how many channels the network operator bought from the government, and how they wish to distribute them (neighboring towers can't use the same channels, or they will interfere - and busy towers need more channels to handle the capacity). Network operators put huge amounts of effort into optimising channel distribution on their towers, so that as many calls and data as possible go through - everytime a call fails to go through because the channels are all used, the operator loses an opportunity to make money.

Similarly, UMTS towers will allocate channels to voice or data as needed. If a tower has 5 channels, and is operating 1 voice UMTS channel and 4 HSDPA data channels. Under heavy voice call load, it can switch HSDPA channels into UMTS voice mode. (One UMTS voice channel can carry dozens of voice calls simultaneously). This ensures voice calls get through in reasonable quality, at the expense of slowing down data access.

Typically, there are about 10-20 channels in a band, with typically 1 or 2 bands available in any particular country. In most cases, a network operator will only have channels in a single band - but very occasionally, there are networks which span multiple bands. A single tower won't use 2 bands for the same network. Instead, a network operator may put neighboring towers on different bands in order to minimize interference and maximize capacity in densely populated areas.

Phones tend to go for the tower with the best signal quality and strength - rather than bandwidth availability. There may be other rules for deciding when a phone should switch from one channel to another, but I doubt available bandwidth has anything to do with it. However, it is worth pointing out that the very latest HSDPA specs (don't think they've got into phones yet) do allow a phone to shotgun data via multiple towers, for very fast data rates.

 

[exdeath]

Same as any RF communication, based on symbol rate, symbol density, signal to noise ratio, and spectrum width, and how narrow the modulation and power management of the modulated signal is to make the most of the spectrum. Tower bandwidth is fixed and determined by these factors, all devices share the bandwidth. There are so many towers, and also the quick burst and streaming nature of mobile device communication it works out ok.

 

[HibyPrime1]

HSDPA data rates are 'per radio channel'. So, if there are 3 phones in a town, accessing a tower, then the tower will try to allocate each phone a radio channel (as long as the tower has enough channels allocated to it.)...

Thanks for the answer, very thorough. I've always noticed phones have ping times that vary widely even from one minute to the next, I suppose that explains why that is too.

I asked this because I was trying to understand why all cell companies seem to basically give everyone the shaft when it comes to data. I knew it couldn't be ONLY that they are 'evil', if that was the case then there would be startup companies all over the place giving far better deals. It would seem that simply adding new towers only helps with the coverage part of the problem, it does nothing for bandwidth when there are other towers near.

Thanks again.

 

[Mark R]

It would seem that simply adding new towers only helps with the coverage part of the problem, it does nothing for bandwidth when there are other towers near.

That's not strictly true. Towers will only interfere on a channel if they are powerful enough and close enough that a phone is able to pick up both tower's signals simultaneously.

The solution is to use very low power cells - these are 'microcells' (short range cells on low towers - e.g. street lights), or 'picocells' (which are ultra-short range cells, that look like a Wifi AP, that are installed inside buildings, or outside at street level, and designed to provide capacity for extremely-dense phone traffic - e.g. inside shopping malls, airports, railway stations, etc.). This way you can have many cells, close enough to provide the requisite capacity.

When extra capacity is needed for short periods, but a new tower isn't warranted, the network operators often have portable microcells - a trailer with a short tower, cell equipment, generator and microwave (or satellite) uplink - they can then haul it to wherever it is needed - e.g. for a big festival.

 

[jersiq]

How many channels a tower has, depends on how many channels the network operator bought from the government, and how they wish to distribute them (neighboring towers can't use the same channels, or they will interfere - and busy towers need more channels to handle the capacity). Network operators put huge amounts of effort into optimising channel distribution on their towers, so that as many calls and data as possible go through - everytime a call fails to go through because the channels are all used, the operator loses an opportunity to make money.
Typically, there are about 10-20 channels in a band, with typically 1 or 2 bands available in any particular country. In most cases, a network operator will only have channels in a single band - but very occasionally, there are networks which span multiple bands. A single tower won't use 2 bands for the same network. Instead, a network operator may put neighboring towers on different bands in order to minimize interference and maximize capacity in densely populated areas.

Phones tend to go for the tower with the best signal quality and strength - rather than bandwidth availability. There may be other rules for deciding when a phone should switch from one channel to another, but I doubt available bandwidth has anything to do with it. However, it is worth pointing out that the very latest HSDPA specs (don't think they've got into phones yet) do allow a phone to shotgun data via multiple towers, for very fast data rates.

Interesting perspective on the UMTS side. CDMA is a little different.

Of course, CDMA uses a shared channel, and it is preferable to have the carriers contiguous. I am going to use CDMA here to also indicate EVDO because I am lazy.

In a typical 850 block you can get 9 channels from the B-Band of the 850 block and ~8 carriers in the A-Block. Generally, all towers within the area will have a matching carrier count up to borders where they will "hand down" to the next highest carrier. So, any new site placed will have the same frequencies being used as their neighbors. It's often referred to as the wedding cake design, where if you looked at a map of an area, it would look like the tiers of a wedding cake from a top-down perspective.

EVDO systems do have the capability to go from the 850 to PCS blocks, but anytime you do an "hyperband" handoff, it's incredibly difficult on the mobile. I have seen multiple clusters with 3 carriers of 850 EVDO and 2 PCS carriers of EVDO interacting. But the general idea is to use the measurements from the phone to indicate which band that call will be handled on. For example if you are within 2 miles of the site, and your Ec/Io for the PCS carrier is > -4.0, you will begin your data session on the PCS carrier. Or the operator could just "let things fly" and use the built in hashing algorithm to place you on a particular carrier.


So, back to the points in the thread, as long as all traffic is treated as "Best Effort" the scheduler in the BTS will schedule them fairly. The scheduler is an algorithm (usually proprietary dependent on the BTS manufacturer) which takes the needs of all users and allocates air interface accordingly. The basis for the rate given is predicated on RF quality between the mobile and the BTS. In EVDO, you have a DRC your moblie states to the BTS. The scheduler then decides how it's going to allocate it's resources to you, keeping in mind it has other users to serve.

 

[-sandro-]

I still didn't get the answer to the first question:

Town A has a single cell tower providing HSDPA (or other 3G/4G network type) at 14Mbits/sec. 3 smartphones download a movie, do all 3 users get 14Mbits? or is it split between them?

 

[jersiq]

I still didn't get the answer to the first question:

Town A has a single cell tower providing HSDPA (or other 3G/4G network type) at 14Mbits/sec. 3 smartphones download a movie, do all 3 users get 14Mbits? or is it split between them?

The answer isn't clear cut. First, we have two different aspects of the BTS to consider:
1) The backhaul bandwidth from the BTS to the MTSO
2) What throughput the RF protocol supports

I am going to neglect addressing point 1, as it's merely academic and becomes a queuing theory problem. We will say that the bandwidth for the backhaul is 100 Mbps.

Diving into point two, there are two more constraints:
a) where the users are relative to each other
b) where the users are relative to the BTS

As to point 2a, first you have to understand how cell sites are built. There is a widely used technique called sectorization. This takes the "circle" that a site would normally cover if it had omni-directional antennas and chops it up into smaller "sectors."

This has numerous benefits, but the most advantageous is that it allows to "split up" the RF interface to the BTS. So if the maximum throughput on an HSPDA channel is 14 Mbps, we now have three discrete sectors each providing 14 Mbps. If we had all three users in the same sector (let's say they are at 12:00 in the picture above) then the scheduler algorithm in the BTS will decide how best to serve each user on the DSCH.

Generally, each BTS manufacturer uses proprietary code to weight how it will serve each user on shared bandwidth. Usually, to gain spectral efficiency, the higher data rates will be given to those users who report a better Channel Quality Index than others. So, in our example if all three users were placed in a radial at 12:00, and they were in gradually "poorer" RF conditions based upon distance from the BTS, the user closest to the BTS would get the higher rates and the user farthest would get the lower rates. If instead all three were right on top of each other and each were reporting the same CQI, the the Fair Queue in the BTS would allocate each the same bandwidth dividing the maximum throughput by three.

However, if we place each user in a different sector (12:00, 4:00, and 8:00) the users are no longer competing for a shared resource. Each device could take full advantage of the available throughput provided that the RF conditions support it.

This brings us to point 2b and understanding how RF conditions affect the throughput. As mentions above, the BTS will receive either periodic or aperiodic CQI reports from the user. In essence, the BTS has a lookup table to determine the maximum coding rate it will apply (QPSK, 16-QAM, 64-QAM) and MIMO selection. If you look in the picture and note there are areas where sectors overlap (darker blue color). These are sometimes referred to as the rolloff areas or even sector Null. Throughput in these regions suffer as you have two competing signals in the same band, your SNR is poorer, which is a measurement that is reported in the CQI. Second, you are in a region that is prone to handoffs between the two sectors. The more frequently you are handing off, the less you are likely to have higher throughputs. The best location relative to the BTS is to be in a centerline with the antenna that is serving you, rather than in those nulls. So the users position relative to the BTS does matter.

So, short winded answer: It depends.

 

[-sandro-]

That is an amazing answer, exactly what I was looking for.