NF, here is the dinkum oil on how Qualcomm will handle at-capacity systems.
From a Senior VP for Technology with QUALCOMM, a co-founder and one of
the principal inventors of the IS-95 system.
-----------------------------------------------------------------
Basically, with CDMA there are two regimes of cell design, coverage and capacity. In the coverage regime, which you have in the startup phase of a wireless business, you want to cover as many pops as possible for a given level of investment. One of CDMA's neat features is that it allows you to tradeoff some capacity to gain additional coverage. This is a consequence of the cell receive noise vs. loading curve which rises asymptotically as you approach "pole capacity".
So.......what happens is that as you start to load a cell from one call, 2 calls, 3 calls, etc, is that the total cell noise gradually rises from its low value determined by thermal noise. Consider the impact of increased loading on that first call. At first, he will be using the minimum power that could ever be used over that link. So for a given maximum transmitter power in the mobile, he could also be at the greatest possible distance in this condition. I.e., the cell radius is greatest under this ondition of no other user interference. Is this surprising? I should think not.
Then, as additional users come on, the other user noise increases, slowly at first, then more quickly as you approach the asymptote. An increase in other user noise means you must increase your power to maintain the desired Eb/No. At 50% loading, the total noise increases (and you must increase your power by) 3 dB relative to the unloaded case. The noise increase is simply 1/(1-%load). So at 75% load, it's a 6 dB increase in noise, and at 90% load its a 10 dB increase. Also, the power control loops tend to get a little squirrely as you approach the asymptote causing the average power increase even faster. We recommend running nominal cell loading at around 50% of the asymptotic capacity.
As the noise increases the cell radius shrinks, relative to some nominal design radius. Note that one never should design a cell radius to support only one user. Relative to the design point, you can also say that the cell expands when the loading is less than nominal. In my opinion, this is a feature, not a problem. It allows cells that are experiencing higher than nominal load to borrow capacity from neighbor cells with lower than nominal load by having their own radius shrink and neighbor cells expand. Calls in the border area will find them selves being handled by the lightly loaded cell rather than the heavily loaded cell. The other technologies cannot do this, they just block and drop calls when design load is exceeded. (Note that GSM people have touted dynamic channel assignment, but name me one city with a GSM system that actually does this.)
Of course, if a group of neighboring cells in a certain area are all loaded more than nominal, they will all shrink and not be able to borrow capacity from an expanding neighbor cell because all of them are shrinking too.
This will result in coverage gaps between the cells where calls are more likely to drop or not be made in the first place. What's the cure? Add more capacity in the afflicted areas. How? More frequencies, more sectors or more cells. In that order. (It's least expensive to do in this order.)**
Can this, has this happened?? Yep, yep. If a growing system fails to add capacity as fast as it adds minutes of use, it will certainly happen. Is this a fault of the technology? Absurd! What happens when an AMPS cell gets more than about 20 calls in a sector? It blocks. It has no choice because there are no more channels available. What happens to call attempts in this sector? They are blocked. What happens to handoff attempts for a call moving into this cell? They are dropped. Same answer for IS-136 and GSM.
Based on the number of gripes I hear on the internet, I'd guess that Sprint used fairly aggressive cell spacing and Primeco was a little more conservative. The cell design criteria also has a strong impact on in-building coverage. It costs something like 10-20 dB (depending on construction materials, etc.) to make a PCS signal go through a wall. So a cell design that maximizes outdoor range to minimize total cell count is not going to penetrate inside buildings very well unless the buildings are in the central part of the cell.
Is any of this news? Nope. We've been explaining it to everyone in the industry who would listen since the earliest days.
** More on growing capacity.
Adding frequencies is the cheapest way to grow capacity. If you started at with only a single RF channel then even the D,E,F licences have room for more RF channels. This adds only additional radios and modems to already existing cells. Adding sectors is slightly more expensive because in addition to additional radios and modems, you also need to add the new antennas and RF cable. A side benefit of sectorization is that the higher gain of the narrower sectorized antennas will improve in-building coverage and range a small amount (3 dB for six vs three sectors). Finally, one can add additional cells.
This is most expensive because you incur all the additional costs associated with acquiring and paying for an additional site, additional back haul costs, etc. Although with the new mini-cells we've recently announced (the QCell(TM) 519e and the QCell 508e, for
cellular and PCS networks) will mitigate this somewhat.
---------------------------------------------------------------
Now is that orthogonal or what Surfer Mike who really really should be at Piha? surf.co.nz
That surf looks little, but it is rolling, smooth, clean, clear and plenty big enough. Gentle off-shore breeze, warm.
Now, if you go back up there a little, you'll see that they aim to have only 50% of the erlangs per bushell nominal capacity due to squirrelly orthogonality and depleted batteries and voice quality as noise and decibels increase. In non-technical terms, they suggest running systems at half speed coz it flies off the rails if they get up too much head of steam and the passengers don't like getting jostled, crowded and dropped off the train.
See, nowhere does it mention a dollar. Now why are they and we in business? That's right! For lots of orthogonality [with due genuflections to the greedy little prat from Hong Kong who cares only for money and couldn't care less about orthogonality].
So there!! MightyQ wants orthogonality, not dollars.
As Dave Barry said - in better words: "I just love writing orthogonality in public! Orthogonality, orthogonality, orthogonality."
Mqurice
PS: Better sell Qcom everyone. No disclaimers or boiler-plate. Don't do your own research. Trust me!
To serve me with a summons to sue for damages if I'm not right, my address is:
31416 Taumarunuinikauaroha Track
Waikikamukau
Queenstown [no, I'm not a poofter though some of my best friends are]
New Zealand |
