Introduction to the challenges of current GSM and GPRS planning Technical Presentation
Prof. Dr. Fred Wagen Senior Consultant Lausanne, Switzerland wagen@wavecall.ch Prof. in telecommunication at the Univ. of Applied Sciences (HES-SO) in Fribourg, Switzerland (www.eif.ch/telecom) Slide 2
Outline GSM-vs-GPRS = circuit-vs-packet: so what? GSM & GPRS planning: similarities and differences The crucial role of radio propagation prediction: accurate model is a must The challenges ahead: prepare to react quickly Slide 3
The challenges of GSM and GPRS planning GSM-vs-GPRS = circuit-vs-packet: so what? GSM & GPRS planning: similarities and differences The crucial role of radio propagation prediction: accurate model is a must The challenges ahead: prepare to react quickly Slide 4
GSM-vs-GPRS GSM-vs-GPRS : so what? time same TDMA-FDMA air interface could also be traffic slots frequency BCCH Slide 5
GSM and GPRS GSM-vs-GPRS : so what? but flexible multi-slots allocation for GPRS GPRS 3 up 1 down GPRS 1 up 2 down then 3 down GSM Voice BCCH => fewer empty slots Slide 6
GSM-vs-GPRS : so what? GSM Voice: SMS: WAP: data: a few errors can be tolerated but no delay (< 200 ms) store & forward connecting 9.6 kbps CS or HSCSD on the move but not moving Slide 7
GSM-vs-GPRS : so what? GPRS Voice: SMS: WAP: data:??? may be for voice mail services cheaper for operator and users learn from i-mode Packet Data, always on (claim or reality?) (terminals with voice recording) Needs a new phone and a PC (laptop, palmtop), or a PDA (Palm, EPOC, PocketPC, ), or a smart phone Slide 8
GSM-vs-GPRS : Voice-vs-Data bearers 20 ms GSM voice GSM data 13 kbps (Full Rate) 9.6 kbps 12.2 kbps (EFR) 20 ms 20 ms 260 bits (FR) 260 bits (FR) 192 bits+padding 456 bits R=1/2 R=1/2 456 bits 456 bits R=1/2 Interleaving on 8 ½ bursts Interleaving on 4 bursts 2x57 bits Average 22.8 kbps per slot 270 kbps bursts on air interface 2x57 bits Average 22.8 kbps per slot 270 kbps bursts on air interface Slide 9
GSM-vs-GPRS : GPRS Data GPRS 9.05 kbps or 13.4 or 15.6 or 21.4 kbps 20 ms 181 bits or 268 or 312 or 428 bits R=1/2 or 2/3 or 3/4 or no coding 456 bits CS1 or CS2 or CS3 or CS4 Interleaving on 4 bursts 2x57 bits Average 22.8 kbps per slot 270 kbps bursts on air interface Similar to GSM data but 4 Coding Schemes => 4 data rates Slide 10
GSM-vs-GPRS : coverage GSM voice Good enough Voice Quality (e.g.,mos) GSM bad Analog voice Sufficient quality Transmission quality 9 to 12 db C/I Slide 11
GSM-vs-GPRS : quality coverage Good Quality is a subjective opinion Threshold Bad Must be set by operator to quantify quality and simplify the identification of problems. Slide 12
GSM-vs-GPRS : coverage Good enough bad GSM voice Voice Quality (e.g.,mos) Analog voice GSM Transmission quality Sufficient quality GPRS : even more difficult Data Quality (here bit rate) CS1 CS2 CS3 CS4 9 to 12 db C/I 9 13 17 25 db C/I Slide 13
GSM-vs-GPRS : coverage - urban GSM voice GPRS coverage 21 kbps CS4 3 4 No coverage or other cell 16 kbps 13 kbps 9 kbps CS3 CS1 2 3.5 CS2 Urban environment Slide 14
GSM-vs-GPRS : coverage - LOS GSM voice GPRS coverage 21 kbps 16 kbps CS4 CS3 4 8 12 16 13 kbps CS2 9 kbps CS1 No coverage or other cell Free space Slide 15
GSM-vs-GPRS : so what? GSM-vs-GPRS Requirements for planning Slide 16
GSM-vs-GPRS : Voice-vs-Data GSM voice GSM data 13 kbps (Full Rate) 1 slot up / 1 slot down (total of 2) Support ~1% FER Plan coverage for voice service CS: 9.6 kbps (or 14.4 kbps) HSCSD n*9.6 (or 14.4) kbps 1,2 or 3 slots up / 3, 2 or 1 slots down (total of 4) Error must be handled with ARQ For users on the move accepting best effort service and/or non-moving Slide 17
GSM-vs-GPRS : so what? GSM voice GPRS 13 kbps (Full Rate) Packet data n*9 or *13 or *16 or *21 kbps 1 slot up / 1 slot down (total of 2) n = 1,2 or 3 slots up / 3, 2 or 1 slots down (total of 4) Support ~1% FER Errors must be handled with ARQ Plan coverage for voice service Plan multi-coverage for new services Slide 18
GSM-vs-GPRS : so what? GSM voice Traffic characteristics are well known Quality is in fact a best effort service but good planning lead to acceptable quality GPRS Traffic characteristics are unknown Quality depends on service profile: Precedence class: indicates the importance of the packet with regard to discarding it in case of problems and degradation of QoS when necessary. Reliability class: specifies the mode of operation for various error detection and recovery protocols & how securely the data should be delivered. Delay class: specifies the transfer delay including the uplink radio channel access or downlink radio channel scheduling delay, the radio channel transit delay, and GPRS network transit delay. Peak throughput class: defines the maximum transfer rate. Mean throughput class: defines the long term average transfer rate. Slide 19
The challenges of GSM and GPRS planning GSM-vs-GPRS = circuit-vs-packet: so what? GSM & GPRS planning: similarities and differences The crucial role of radio propagation prediction: accurate model is a must The challenges ahead: prepare to react quickly Slide 20
GSM & GPRS planning: similarities & diff. GSM voice Number of subscribers Traffic statistics + Erlang B + Add control channel if SMS traffic increases GPRS Number of subscribers??? Traffic statistics??? +??? simulation + Reserve channels for GPRS if needed Conventional planning Still not easy to plan when exponential growth combined planning With GSM Very difficult to plan Offer best effort service then plan to be flexible and follow the demand Slide 21
GSM & GPRS planning: similarities & diff. GSM voice GPRS handy coverage must work no drop regions (e.g., indoor) Blue = with indoor coverage Red = enhanced capacity Green = GPRS capacity Brown = GPRS high throughput Slide 22
GSM & GPRS planning: similarities & diff. GSM voice GPRS Tools Coverage Interference Matrix Experience Traffic data C/I FP (Frequency Plan) Combine GSM planning and GPRS-CS1 to 4 analysis Many more and faster iterations are required C/I Tools Coverage Interference Matrix Experience Traffic data C/I FP (Frequency Plan) = propagation prediction No OK Yes, load FP in network C/I No OK Yes, load FP in network Slide 23
GSM & GPRS planning: similarities & diff. GSM voice and GPRS Cannot be planned separately ( or consider full load). Introduce GPRS slowly but be prepared for rapid take-off ( i-mode scenario ): 1. If easy user interfaces are developed (e.g., a mobile internet button on every terminals), 2. If appropriate charging is used (e.g., set-up time is either short or not billed, flat rate combination), 3. If the market can decide (open to innovative services). New Key Performance Indicators (KPI) must be developed and used. Tools must be as accurate and as easy to use as possible to concentrate on difficult tasks. Slide 24
The challenges of GSM and GPRS planning GSM-vs-GPRS = circuit-vs-packet: so what? GSM & GPRS planning: similarities and differences The crucial role of radio propagation prediction: accurate model is a must The challenges ahead: prepare to react quickly Slide 25
Radio propagation prediction Tools Coverage Interference Matrix C/I Experience Traffic data C/I FP OK No Since GPRS traffic is unknown, only various scenarios can be investigated. Starting with a best-effort GPRS add-on to an existing GSM network, various growth scenarios must be taken into account or ways to deploy rapidly the required level of GPRS quality must be found. In both cases, tools allowing to quickly and accurately assess the quality of a possible deployment are then critical. Proper coverage is key: => accurate and fast radio propagation prediction model is a must in an integrated planning tool. Yes, load FP in network Slide 26
Radio propagation prediction Conventional models. The smaller the cell the worse the inaccuracy. Need tuning => measurements => time. Need experience to mitigate the drawbacks. OK with GSM: took 10 years to be fully deployed. Fast and accurate WaveSight. Accuracy independent of cell size. No need for tuning. Need measurements to convince first time users. Allows to test various GSM- GPRS scenarios before field tests and deployment. Slide 27
The challenges of GSM and GPRS planning GSM-vs-GPRS = circuit-vs-packet: so what? GSM & GPRS planning: similarities and differences The crucial role of radio propagation prediction: accurate model is a must The challenges ahead: prepare to react quickly Slide 28
The challenges ahead: react quickly!!!!? Plan all possibilities? Or prepare to react quickly? Slide 29
The challenges ahead: react quickly Everything may not work as expected : Suppress the guess work wherever you can accurate propagation predictions will help Slide 30
Conclusion GSM and GPRS have the same FDMA-TDMA airinterface. But different coverage characteristics depending on data rate (Coding Scheme), on unused voice slots, and on expected or required user services. Challenges are coming from the many unknowns. Number of unknowns can be reduced by using more accurate planning tools wherever possible. Gather the proper tools, databases and methodology now! Slide 31