The basic steps are

1. Estimate your target market in terms of the number of customers in your target area that are likely to adopt your network. Work on a 5-year (financing) plan, with capacity being reached in 5 years. So if (!) you can get the site and install the equipment by the end of this year, you estimate the potential number of customers at the end of 2018.

2. Use standard measures of the calling rate plus terminating rate of the customers and add them to get erlang in the busy period. (Use Mother's Day busy hour measures.)

3. Multiply 1. and 2. to get the traffic per target area.

4. Look up the traffic per base station, as supplied by the base station manufacturer.

5. Divide 3. by 4. to get the number of base stations in your target area.

6. Situate the base stations strategically at the intersections of roads and railways, and at airports, schools, business areas and shopping malls to get a 95% coverage.

7. Double up on your base stations that have reached capacity in 2018, based on your measurements. The radius will become a half of its previous value.

8. Remember to calculate your total income from a flat figure in 2014 to its maximum at the end of 2018 to see your project is viable. Do a worst-case scenario of too little marketing and a slow uptake of customers.

The basic steps are

1. Estimate your target market in terms of the number of customers in your target area that are likely to adopt your network. Work on a 5-year (financing) plan, with capacity being reached in 5 years. So if (!) you can get the site and install the equipment by the end of this year, you estimate the potential number of customers at the end of 2018.

2. Use standard measures of the calling rate plus terminating rate of the customers and add them to get erlang in the busy period. (Use Mother's Day busy hour measures.)

3. Multiply 1. and 2. to get the traffic per target area.

4. Look up the traffic per base station, as supplied by the base station manufacturer.

5. Divide 3. by 4. to get the number of base stations in your target area.

6. Situate the base stations strategically at the intersections of roads and railways, and at airports, schools, business areas and shopping malls to get a 95% coverage.

7. Double up on your base stations that have reached capacity in 2018, based on your measurements. The radius will become a half of its previous value.

8. Remember to calculate your total income from a flat figure in 2014 to its maximum at the end of 2018 to see your project is viable. Do a worst-case scenario of too little marketing and a slow uptake of customers.

In line with Ian's responses, The important calculation is what will be the capcity required in terms of Speech & Data for a target area. This gives you the hardware requirement for the TRX units for the designated Spectrum & Technology, then how you want to place these resources across the coverage area (Inddor/Outdoor/Micro/Macro/Pico/Repeaters etc) then you can finlise the amount of Base Staations. Als the 5 year plan is important, because of the space limitation you should run in to trouble of shifting equipment to another location in a year/two time frame. Also you need to make sure the backbone (i.e. backhaul) connectiovity is available and required equipments also to be equipped in the same base stations.

Each base station can have 3 sectors to max...then you can play with Micro/Pico solutions to add up the capacities provided you ahve enough spectrum.

Thanks Ian and Sunil :)

Sunil correctly pointed out the importance of including data requirements (fast growing!), the range of base stations available, the backhaul and rest-of-the world connectivity, and fine tuning of the sectors. Have fun!

The no of base stations are decided based on the coverage requirements. Once the coverage is met then we need to see the Capacity requirements are met sothat the Grade of Service is within acceptable limits. It is an iterative cyclic process which is done round the year. meaning Planning of new base stations and optimising the network happen parallelly. if u want still an indepth answer pl feel free to email me.

Prem is right. Initial plans are always wrong, and need tweaking from field experience.

Can any one comment on the problem and possible simple answer I present here?When  a  user  makes  three  calls  per  hour  and  the  average  call  duration is  two  minutes,  the number of Trunked channels or Capacity is 20 and the Blocking Probability is 2%.  If  the  total  population is  1000,000  in  a  suburban  area  and  cellular  service  penetration  is  70%,  a) how many  sectors  would  be required  to cover the  suburban area ? If one antenna can reach two sectors in distance, b) how many base stations does one need for one to four frequency re-use ? Ans: When  a  user  makes  three  calls  per  hour  and  the  average  call  duration is  two  minutes,  μ  is  3/60  calls  per  minute  =  0.05  calls  per  minute,  and  H   =   2 minutes.  The  user  traffic  intensity  =   μH  Erlang  =   0.05  * 2  =   0.1 Erlang. Then  the  case  of  N   =  20  and  Pb,  =   0.02,  the  Erlang  capacity  is  E   =  13.18 from the Erlang Table. The  number  of  subscribers  supported  in  the sector  becomes  U   =   13.18 Erlang/0.1 Erlang  =   131.8  70%,  a  total  of 700000/U  =  700000/131.8    = 5311.07 rounded up to 5312 sectors.  In  summary,  when  cellular  subscribers  are  uniformly  distributed  among  5312 sectors and each  subscriber generates 0.1 Erlang of traffic during the busy hour. Looking at Figure 5.24 in Ref. N. D. Tripathi and J. H. Reed, Cellular Communications-A Comprehensive and Practical Guide. John Wiley & Sons, Inc., 2014., need 3 for 12 sectors or ¼ x 5312 = 1328 OR 1328 BST (rather big number, I may be wrong). In other words, if one gets more possible users and want better penetration rate then change the population and penetration rate numbers. Please correct (perhaps wrong calculation) or comment. Thank you in advance.

Yes, 3 calls (incoming plus outgoing) per busy-hour, times an average duration of 2 minutes does give 6 erlang-minutes per busy-hour = 6/60 erlang = 0.10 erlang for the bothway traffic intensity.

Yes, a blocking probability of 2% is a common standard for a cellular provider, at the end of the 5-year planning period.

70% cellular penetration sounds initially like an ambiguous statement. Does this mean that you can choose to give service to only 70% of the customers (e.g. omit those just outside the hexagon?), or must you give coverage to all customers, and expect only 70% to subscribe? The first is more economic, but would not be acceptable to the communications regulator. For reality, I would choose the latter.

{So for a million people, you can expect 700 000 to subscribe. Multiplying by the bothway usage 'rate' of 0.1E gives 70 000 Erlang bothway or 35 000 Erlang originating traffic.}, assuming that the busyhour time is unchanging over the whole network.

You can assume for the exercise that you are able to purchase or lease ideal base station sites and you are able to beam evenly to all sectors.

Next we need to calculate the capacity of a sector. I have not got my traffic table handy, but A  =  13.18 sounds about right. So the number of customers  supported  in  one sector is correct at 13.18 Erlang/0.1 Erlang  =   131.8 customers. Yes, 700 000 actual customers divided evenly over the sectors gives 700000/131.8= 1312 sectors rounded up.

Although GSM uses 3, you may be able to squeeze in 4 sectors per 360 degrees, which correctly means you need to divide the number of sectors by 4 to get the number of base stations. 5312/4= 1328 base stations. Yes, this is a lot of base stations. If a base station costs USD 100 000, the network will cost more than $132M and the plant will only be fully used in 5 years time. Not counting running costs such as interest, site rental, switches, interconnect costs, staff, maintenance and energy costs (which could be covered by call charges), each member of the full population would have to be taxed by the government(?) to the extent of more than $132 over 5 years to finance the capital required to install the network.

The only redeeming factor is that if one base station is running quiet because the population have gone to work, then it can take some overflow traffic from adjacent cells during the business peak.

To properly calculate this saving you would have to have some knowledge of the diversity factors for adjacent cells. In practice, most times, adjacent cells serve similar populations, and so this saving is elusive.

I think that the strange statement that “one antenna can reach two sectors in distance” means only that frequency reuse is employed.

Part b) is tricky, because certain practical steps are assumed always to be done in order to prevent overspill from causing interference. For this reason {for bonus points, please?} it is important the base station antennas be sited as low as possible, and the beams be tilted downwards somewhat. In practice then there should be no problem with two cells being sufficiently far apart operating on the same frequency. Wikipedia gives a frequency reuse formula.

Thanks Ian. Sorry just to correct your calculation slightly, 6/60 is 0.1 not 0.01. Please re-evaluate the rest. 

@Benny, Thanks, I fixed my 2 typos in the first paragraph. 

Now 3 calls in the busyhour seems rather high, unless these are business customers only. How many cellular calls did you make between 9 am and 10 am today ?

Thank you for sharing your theoretical & practical knowledge. 3 calls per hour is only an assumption. The interesting part for me is an engineer can now help an investor/a government body, by working backwards. Say one can only provide 500 BST then one can work backward to get approximate number of sectors covered for a certain % of population usage /penetration, and frequency reuse. 

@Benny. Yes, and I feel the Erlang equation is an excellent example of how mathematics can help the engineer. My famed definition of an engineer is "someone who can do for a shilling what any fool can do for a pound".

What a wise investor will do is initially enter the "cream" market, and provide service to the business users, business areas, mobile businesses and major intersections, then progress to the less lucrative markets once capital interest has been mainly paid off.

https://opencourses.emu.edu.tr/mod/resource/view.php?id=9606 follow this