Edegperf

Conference Call TR45.3/98.04.06.
07R4
(TR45/98.03.03.19R6)
1999 Universal Wireless Communications Consortium Page 177 of 306
ATTACHMENT 1 1
2
3
4
In 136 HS, 8-PSK modulation along with GMSK with the same symbol rate, 270.833 ksps, is used: 5
6
8-PSK 7
8
9
10
Advantages: Very fast link adaptation; No extra complexity 11
(0,0,1)
(1,0,1)
(d(3k),d(3k+1),d(3k+2))=
(0,0,0)
(0,1,0)
(0,1,1)
(1,1,1)
(1,1,0)
(1,0,0)
Conference Call TR45.3/98.04.06.07R4
(TR45/98.03.03.19R6)
136 HS Slot Format for 8-PSK and GMSK 1
2
3
4
5
6
7
8
9
10
11
12
13
14
576.92 Sec.
TS
3
IS
58
GP
8.25
TS
3
IS
58
TSS
26
TS
IS
TSS
GP
Tail Symbols
Information Symbols
Training Sequence Symbols
Guard Period
Gross User Payload (8-PSK) : 348 bits - 2 stealing bits
Gross User Payload (GMSK) : 116 bits - 2 stealing bits
(TR45/98.03.03.19R6)
Symbol and Bit Rates 1
2
3
4
Symbol Rate
Bits/Symbol
Payload/Time Slot
Gross User Rate/Time Slot
Gross User Rate/Carrier
270.833 ksps
3
346 bits
69.2 kbit/s
553.6 kbit/s
270.833 ksps
1
114 bits
22.8 kbit/s
182.4 kbit/s
Modulation Types
8-PSK GMSK
(TR45/98.03.03.19R6)
8-PSK Coding and Puncturing Parameters 1
2
3
4
Puncturing
Interleaving
Convolutional
Coding
BH DATA BCS
DATA Block
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8
Slot
576.92 sec
TDMA FRAME
4.615 msec
,
_
26
120
(TR45/98.03.03.19R6)
1
2
Coding scheme
PCS-6 PCS-5 PCS-4 PCS-3 PCS-2 PCS-1
Radio Interface Rate
(kbps)
69.2 57.35 51.6 41.25 34.3 22.8
Input bits (per 20ms block)
1384 1147 1032 825 686 456
Convolutional coding rate
n.a. 1/3
Polynomials
n.a. G0, G1 G0, G1 G0, G1 G0, G1 G0, G1, G2
Tail bits
n.a. 6 6 6 6 6
Number of encoded bits
n.a. 2422 2076 1662 1384 1386
Remaining bits after
puncturing
n.a. 1384 1384 1384 1384 1384
Output bits
1384 1384 1384 1384 1384 1384
G0 = 1 + D
2
+ D
3
+ D
5
+ D
6
, G1 = 1 + D + D
2
+ D
3
+ D
6
, G2 = 1 + D + D
4
+ D
6
3
Table 1: Coding and Puncturing Parameters 4
5
6
Puncturing 7
8
In this section more detailed descriptions of the different puncturing schemes are given. 9
10
PCS-1 11
12
The puncturing matrix P is defined by: 13
14
P(n)=1, n except for n=467,947 where P(n)=0 15
(TR45/98.03.03.19R6)
1
PCS-2 2
3
No puncturing is used. 4
5
PCS-3 6
7
P(n)=1, n except for n=12k,12k+5 k0,,138 where P(n)=0 8
9
10
PCS-4 11
12
14
P(n)=1, n except for n=12k,12k+5, 12k+7, 12k+8 k0,,172 where P(n)=0 15
16
PCS-5 17
18
20
P(n)=1, n except for n=10k+3,10k+5, 10k+7, 10k+9 k0,,230 21
and n=2303,2305 where P(n)=0 22
23
PCS-6 24
25
This is an uncoded scheme, and hence no puncturing is applied. 26
27
28
29
30
(TR45/98.03.03.19R6)
Channel Coding Schemes for Packet Data 1
2
3
Service name Code rate Modulation Gross rate Radio interface rate*
per Time slot
PCS-1 0.329 8-PSK 69.2 kbps 22.8 kbps
PCS-2 0.496 8-PSK 69.2 kbps 34.3 kbps
PCS-3 0.596 8-PSK 69.2 kbps 41.25 kbps
PCS-4 0.746 8-PSK 69.2 kbps 51.6 kbps
PCS-5 0.829 8-PSK 69.2 kbps 57.35 kbps
PCS-6 1.0 8-PSK 69.2 69.2 kbps
CS-1 0.49 GMSK 22.8 kbps 11.2 kbps
CS-4 1 GMSK 22.8 kbps 22.8 kbps
4
* The radio interface rate includes the signaling overhead for the RLC/MAC layer. 5
6
7
8
(TR45/98.03.03.19R6)
1
ATTACHMENT 2 2
3
Equalizer Simulations 4
5
Performance for High Velocities 6
7
While the physical layer performance shown in Attachment 8 looks very good for the 8
Pedestrian environment (3 km/h), the performance with the same simple equalizer degrades 9
at vehicular speeds. Figure 1 shows the BLER performance for the Vehicular A120 10
interference limited environment using the simple DFSE equalizer. These results show that 11
an error floor develops for the lighter coded schemes, while the uncoded PCS-6 coding 12
scheme is totally unusable. 13
14
When the higher dispersion Vehicular channel B environment is considered, the performance 15
is even worse as shown in Figure 3. 16
17
To solve this problem an improved equalizer for 8-PSK was developed based upon the IRW 18
channel tracking scheme developed for the GMSK very high vehicular speed equalizer. This 19
improved 8-PSK equalizer, developed for the harshest Vehicular B120 environment is an 8 20
tap DFSE design using the IRW channel tracker. 21
22
The BLER performance of this improved 8-PSK equalizer for the original Vehicular A120 23
environment is shown in Figure 2. These results show a significant improvement in 24
performance with no error floor for the coded schemes. 25
26
27
28
29
30
(TR45/98.03.03.19R6)
1
2
Figure 1. BLER Performance for 8-PSK simple equalizer in ITU Vehicular A120 3
Interference Limited environment 4
5
6
Figure 2. BLER Performance for 8-PSK improved equalizer in ITU Vehicular A120 7
9
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
vehicular A, 120km/h, 2br ,no FH
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
0 5 10 15 20 25 30 35
10
2
10
1
10
0
vehicular A, 120km/h, 2br
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
Performance in Time Dispersive Environments 1
2
In the Pedestrian environment, there is essentially no performance difference between the 3
Channel A and Channel B environments, as shown in Attachment 8. The simple 8-PSK 4
equalizer can handle the time dispersion of up to 3.7 usec in the Pedestrian Channel B. 5
However, in the ITU Vehicular Channel B model, time dispersion of up to 20 usec is seen 6
which results in considerably worse performance in the Vehicular Channel B environment as 7
compared to the Vehicular Channel A environment when using the simple equalizer. The 8
Vehicular B120 BLER performance using the simple equalizer is shown in Figure 3. 9
10
The BLER performance for the Vehicular B120 environment using the improved 8 tap DFSE 11
equalizer is shown in Figure 4. These results show that the performance for Vehicular 12
Channel B with the improved equalizer is better than the Vehicular A120 results for the 13
simple equalizer. The dynamic system simulation results given in Attachment 6 for 14
Vehicular A120 are using the simple equalizer and should therefore be taken as worse case 15
for the Vehicular environment when using the improved equalizer. 16
17
18
Figure 3. BLER Performance for 8-PSK simple equalizer in ITU Vehicular B120 19
21
The complexity of the improved DFSE equalizer used for the Vehicular environment is 22
approximately 4 times that of the simple DFSE equalizer. 23
24
25
26
27
28
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
vehicular B, 120km/h, 2br ,no FH
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
1
Figure 4. BLER Performance for 8-PSK improved equalizer in ITU Vehicular B120 2
Interference Limited environment. 3
4
The corresponding throughput performance based on the above BLER curves using the 5
improved equalizer is shown in Figure 5 for Vehicular A120 and Figure 6 for Vehicular B120 6
respectively. 7
8
Figure 5. Throughput Performance of 8-PSK improved equalizer for Vehicular A120 9
0 5 10 15 20 25 30 35
10
2
10
1
10
0
vehicular B, 120km/h, 2br ,Tracking DFSE
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
0 5 10 15 20 25 30 35
0
10
20
30
40
50
60
70
C/I [dB]
T
h
r
o
u
g
h
p
u
t

[
k
b
i
t
/
s
]
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
1
2
Figure 6. Throughput Performance of 8-PSK improved equalizer for Vehicular B120 3
Additional Equalizer Improvements for Very High Velocities 5
Additional improvements have been made in the equalizer for very high mobile speeds. A 6
new full adaptive MLSE equalizer with a 6-tap channel model is evaluated. The channel 7
impulse response is broken into two parts, the unknown response of the medium (3taps), and 8
the known response of the transmit filter and the receive filter. Only the three taps of the 9
medium are tracked. The tracker is based on an Integrated Random Walk (IRW) model for 10
each tap of the medium response. A least-squares type synchronization algorithm was used 11
to determine the best sampling point for each burst 12
13
Simulation results for 500 km/hr mobile speed in a ITU Vehicular A channel environment for 14
1900 MHz are presented below. GMSK is used resulting in a peak radio interface rate of 15
182.4 kbps (uncoded). No receiver diversity is assumed. 16
17
Block error rate performance for 500 km/hr mobile speed in a ITU Vehicular A environment 18
for 1900 MHz for coding schemes CS-1 through CS-4 as well as one additional coding 19
scheme (R=0.79) are presented in Figure 7 for Eb/No and Figure 8 for C/I. No receiver 20
diversity is assumed. 21
22
0 5 10 15 20 25 30 35
0
10
20
30
40
50
60
70
C/I [dB]
T
h
r
o
u
g
h
p
u
t

[
k
b
i
t
/
s
]
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
1
Figure 7. BLER vs Eb/No Performance of GMSK improved channel tracking 2
equalizer 3
4
5
Figure 8. BLER vs C/I Performance of GMSK improved channel tracking equalizer 6
7
The corresponding fractionally loaded throughput performance expressed as a CDF for 8
GMSK using the above BLER results for the interference limited case is shown in Figure 9. 9
0 2 4 6 8 10 12 14 16 18 20
10
4
10
3
10
2
10
1
10
0
GMSK ITU Channel A Vehicular 500km/h; no FH ; no diversity ; 1900MHz
EbN0[dB] (Eb = Energy per modulated bit)
B
L
E
R
(
B
l
o
c
k

E
r
r
o
r

R
a
t
e
)
R=0.5
R=0.64
R=0.73
R=0.79
R=1.0
0 5 10 15 20 25
10
4
10
3
10
2
10
1
10
0
GMSK ITU Channel A Vehicular 500km/h; no FH ; no diversity ; 1900MHz
C/I[dB]
B
L
E
R
(
B
l
o
c
k

E
r
r
o
r

R
a
t
e
)
R=0.5
R=0.64
R=0.73
R=0.79
R=1.0
(TR45/98.03.03.19R6)
The aggregate average throughput for these results is 149 kbit/s. These results are again 1
without mobile station antenna diversity. 2
3
4
5
Figure 9. Throughput Performance for GMSK with improved channel tracking 6
equalizer 7
8
9
10
11
12
0 20 40 60 80 100 120 140 160 180 200
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C
.
D
.
F
.
Vehicular A 500 km/h no diversity
Throughput S [kbps]
(TR45/98.03.03.19R6)
Attachment 3 1
Peak to Average Ratio and Backoff 2
3
In order to decrease the Peak to Average Ratio a linearized GMSK pulse is used as transmitter filter. The pulse 4
shape is depicted in Figure 1. 5
6
7
2 1.5 1 0.5 0 0.5 1 1.5 2
symbols
8
Figure 1: Pulse shape of linearized GMSK filter 9
10
11
The resulting Peak to Average Ratio, PAR, using the linearized GMSK filter is therefore: 12
13
PAR
8-PSK
= 2.5 dB 14
15
PAR
GMSK
= 0 dB 16
17
The peak value for the PAR-value calculation is defined as the amplitude value which is exceeded by 0.1 % of 18
the amplitude values. 19
20
Power Considerations 21
22
Considering the peak-to-average information presented above, the mobile transmit power can be analyzed. 23
Table 1 shows the power considerations for a mobile for8-PSK, and Table 2 for GMSK. 24
(TR45/98.03.03.19R6)
1
Table 1 8-PSK 2
3
Number of
Active Uplink
Timeslots
PA Back-off due
to Linear
Modulation
Resulting MS Tx
Power per
Timeslot after
PA Back-off
from 1 Watt
Max Allowed
MS Tx Power
per Timeslot not
to exceed total
power = 28 dBm
Resulting Power
Considering PA
Back-off and
Total power
1 2 28 37 28
2 2 28 34 28
3 2 28 32.3 28
4 2 28 31 28
5 2 28 30 28
6 2 28 29.2 28
7 2 28 28.6 28
8 2 28 28 28
4
5
6
7
8
Table 2 GMSK 9
10
Number of
Active Uplink
Timeslots
PA Back-off due
to Linear
Modulation
Resulting MS Tx
Power per
Timeslot after
PA Back-off
from 1 Watt
Max Allowed
MS Tx Power
per Timeslot not
to exceed total
power = 28 dBm
Resulting Power
Considering PA
Back-off and
Total power
1 0 30 37 30
2 0 30 34 30
3 0 30 32.3 30
4 0 30 31 30
5 0 30 30 30
6 0 30 29.2 29.2
7 0 30 28.6 28.6
8 0 30 28 28
11
12
13
14
15
16
PA Back-off impacts 17
18
(TR45/98.03.03.19R6)
Figure 2 shows the link performance for 8-PSK with different output back off values 1
assuming an RF2108 power amplifier from RF Micro Devices. The performance for an ideal 2
power amplifier is also showed as reference. The results show that the impact on performance 3
from non-linearity in the PA is negligible for an output back off larger than 1 dB. 4
5
Non-linearity in the PA will also affect the spectrum of the transmitted signal, especially the 6
spectrum leakage into adjacent channels. The spectrum in Figure 3 is obtained using a 7
linearized GMSK pulse. With a 2dB output back off value, this leakage is acceptable 8
considering that the implementation of the pulse shaping is not optimised. Further, the chosen 9
amplifier acts only as an example and larger margins to the GSM spectrum mask can be 10
achieved with other alternatives. 11
12
6 8 10 12 14 16 18 20 22 24 26
10
3
10
2
10
1
10
0
R
a
w
B
E
R
C/I [dB
ideal PA
1dB output back off
2dB output back off
13
Figure 2: Link Performance Degradation Due to Non linear PA. 14
15
100 0 100 200 300 400 500 600
80
70
60
50
40
30
20
10
0
frequency [kHz]
P
.
S
.
D
.
GSM spectrum mask
Ideal PA
Output back off 3 dB
16
Figure 3: Spectrum of Transmitted Signal 17
18
(TR45/98.03.03.19R6)
ATTACHMENT 4 1
2
Radio Link Adaptation 3
4
Link adaptation of modulation and channel coding is used in order to achieve the maximum 5
performance (optimize the throughput) under varying conditions. The modulation is adapted 6
between 8-PSK and GMSK while the channel coding is adapted between six different coding 7
cases for 8-PSK and four different cases for GMSK. These cases are detailed in Attachment 8
1 and are labeled PCS-1 through PCS-6 for 8-PSK and CS-1 through CS-4 for GMSK. For 9
the 8-PSK modulation, PCS-6 has the highest throughput while PCS-1 has the lowest. 10
Correspondingly, for GMSK the highest throughput case is CS-4 and the lowest CS-1. 11
Figure 1 shows the calculated throughput performance per time slot for 8-PSK without 12
diversity, using BLER curves assuming TU3 channel and frequency hopping using T=R*(1- 13
P
B
) where R= Maximum rate and P
B
=BLER (non-frequency hopping results are shifted ~1.5 14
dB higher in C/I). The heavy line shows the throughput performance using link adaptation 15
between the various modes. 16
17
The system simulation results given in Attachment 6 show the achieved user data rates using 18
this link adaptation technique. 19
20
21
22
0 5 10 15 20 25 30 35 40
0
10
20
30
40
50
60
70
80
C/I [dB])
T
h
r
o
u
g
h
p
u
t

S

[
k
b
p
s
]
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
23
Figure 1. Calculated Ideal Link Adaptation Throughput per Timeslot 24
25
The actual results achieved using both modulations in the link level simulations with ITU 26
channel models and diversity are shown in figure 2 (GMSK shown with dotted lines). 27
(TR45/98.03.03.19R6)
1
Figure 2. Link Adaptation Throughput per Timeslot for Ped A 2
3
4
5
0 5 10 15 20 25 30 35 40
0
10
20
30
40
50
60
70
Pedestrian A 3 km/h with diversity
C/I [dB]
T
h
r
o
u
g
h
p
u
t

S

[
k
b
p
s
]
PCS6
PCS5
PCS4
PCS3
PCS2
PCS1
CS4
CS3
CS2
CS1
(TR45/98.03.03.19R6)
Attachment 5 1
Guard Bands 2
3
The co-existence of 136 HS carriers and 136 carriers in the same cell or adjacent cells has 4
been investigated. Figure 1 shows the definition of Guard Band with respect to 136 HS and 5
136 carriers. Note that the Guard band definition is the frequency between the channel 6
spacing of 136 HS to the channel spacing of 136; i.e. the center channel to channel spacing 7
is defined as 115 kHz plus the Guard Band as shown in figure 1. The cell planning for co- 8
existence is shown in Figure 2. Note that the worse case is in those cells where carrier E3 is 9
adjacent to 136 carrier D1. 10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Figure 1. Definition of Guard Band Between 136 HS carriers and 136 Carriers. 32
33
136 HS carriers Guard Band 136 carriers
(TR45/98.03.03.19R6)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Figure 2. Cell Planning Diagram for 136 HS carriers and 136 Carriers. 25
(TR45/98.03.03.19R6)
The interference simulation results for the above co-existence case is shown in Figure 3. The 1
interference from 136 HS into the most interfered 136 channel (D1) is recorded. The solid 2
lines show 136 co-channel interference 3
(right-most curve) and 136 to 136 adjacent channel interference (left). The dashed lines show 4
the interference from 136 HS into 136 as a function of the guard band. Thus, if the guard 5
band is 50 kHz, the amount of interference from 136 HS into 136 is less than the co-channel 6
interference between 136 channels but greater than the adjacent channel interference between 7
136 channels. For a guard band of 100 kHz, the amount of interference from 136 HS into 136 8
is less than both the co-channel and adjacent channel interference between 136 channels. 9
10
11
12
Figure 3. Interference with 136 HS and 136 Carriers 13
14
15
180 175 170 165 160 155 150 145 140 135 130
0
10
20
30
40
50
60
70
80
90
100
solid: co and adjacent interf, dashed: EDGE interf
Interference power [dBW]
c
.
d
.
f

[
%
]
200 kHz
150 kHz
100 kHz 50 kHz
(TR45/98.03.03.19R6)
Attachment 6 1
System Level Simulations for 136 HS Outdoor 2
Static Capacity Simulations 3
4
In order to evaluate the system performance of the 136 HS concept, static packet data 5
simulations have been performed. The following assumptions have been made: 6
1/3 reuse 7
No frequency hopping 8
Log-normal fading with standard deviation of 10 dB 9
Path loss according to M.1225 Models 10
Pedestrian A, Vehicular A50, and Vehicular A120 environments 11
Exponentially distributed packet size with a mean of 1600 bytes 12
2 Branch MRC diversity is used. 13
No power control is assumed 14
Ideal Link adaptation is performed every 100 ms, i.e. 5 coding blocks. 15
The time step of the simulator is 20 ms, corresponding to one coding block. This 16
means that the interference situation can change 5 times during one link adaptation 17
interval. 18
The throughput is measured after channel decoding. 19
The offered load was fixed for all cases. 20
The throughput per carrier distribution for ideal link adaptation is analyzed with fractional 21
loading from which the aggregate average throughput for users in the system is determined. 22
The spectrum efficiency is then evaluated (see attachment 7 for errors in link adaptation 23
signaling). The spectrum efficiency is calculated in the system simulations as: 24
[ ] cell MHz s Mbit
MW
S
N
i
i
/ / /
1
25
where S
i
is the throughput of user i, N is the number of served users, M is the number of cells 26
and W is the total available spectrum. 27
28
Figure 1 shows the throughput distribution for ideal link adaptation, with 8-PSK coding 29
schemes PCS-1 to PCS-6 for link adaptation. This static simulation uses 40% load and 30
mobile station antenna diversity. Table 1 summarizes aggregate average throughput/user and 31
spectrum efficiency for this Pedestrian A simulation. 32
33
(TR45/98.03.03.19R6)
1
Figure 1. Throughput distribution obtained from static simulation with PCS-6 through 2
PCS-1 for Pedestrian A Environment. 3
4
5
Mobile antenna
Diversity
Aggregate Average throughput/
timeslot
50.75 kbit/s
carrier
406 kbit/s
Spectrum efficiency 0.813 Mbit/s/MHz/site
Table 1 Summary of static system simulation results for Pedestrian A. 6
7
Figure 2 shows the throughput distribution the Vehicular A50 environment using the 8-PSK 8
simple equalizer. This static simulation uses 30% load and mobile station antenna diversity. 9
Table 2 summarizes aggregate average throughput/user and spectrum efficiency for this 10
Vehicular A50 simulation. 11
12
0 100 200 300 400 500 600
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C
.
D
.
F
.
Throughput S [kbps]
(TR45/98.03.03.19R6)
1
Figure 2. Throughput distribution obtained from static simulation with PCS-6 through 2
PCS-1 for Vehicular A50 Environment. 3
4
5
Mobile antenna
Diversity
timeslot
48.75 kbit/s
carrier
390 kbit/s
Table 2 Summary of static system simulation results for Vehicular A50 Environment. 6
7
These results indicate that in the low speed vehicular environment an aggregate average 8
throughput greater than 384 kbit/s is obtained even with the simple equalizer. These results 9
would improve with the more complex equalizer. 10
11
Figure 3 shows the throughput distribution the Vehicular A120 environment using the 12
improved channel tracking equalizer as described in Attachment 2. This static simulation 13
uses 10% load and mobile station antenna diversity. Table 3 summarizes aggregate average 14
throughput/user and spectrum efficiency for this Vehicular A120 simulation. 15
16
0 50 100 150 200 250 300 350 400 450 500
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C
.
D
.
F
.
Vehicular A 50 km/h with diversity
Throughput S [kbps]
(TR45/98.03.03.19R6)
1
2
3
Figure3. Throughput distribution obtained from static simulation with PCS-6 through 4
PCS-1 for Vehicular A120 Environment using improved channel tracking equalizer. 5
6
7
Mobile antenna
Diversity
timeslot
47.75 kbit/s
carrier
382 kbit/s
Table 3 Summary of static system simulation results for Vehicular A120 Environment 8
with improved channel tracking equalizer. 9
These results indicate that with the improved channel tracking equalizer, essentially 384 10
kbit/s aggregate average throughput can be provided even at 120 km/hr. Therefore the 11
objective of providing 384 kbit/s data service up to 100 km/hr is realized. 12
13
Load Effects 14
15
There will be a trade-off between the spectrum efficiency and the quality in the network. 16
System simulations show that by increasing the load (even up to 100%), the spectrum 17
efficiency will increase. However, the throughput for individual users will drop and may 18
result in low throughput. Figure 4 shows the sepctrum efficiency for the Pedestrian A 19
environment as a function of Load. 20
21
0 50 100 150 200 250 300 350 400 450 500
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C
.
D
.
F
.
C/I [dB]
(TR45/98.03.03.19R6)
1
2
3
4
Figure 4. Spectral Efficiency as a Function of Offered Load for Pedestrian A 5
Environment. 6
7
Pedestrian A Spectral Efficiency Vs Load
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
40% 50%
Percent Load
100%
M
b
i
t
/
s
/
M
H
z
/
s
i
t
e
(TR45/98.03.03.19R6)
1
Coverage Simulations 2
3
4
Figure 5 shows how the throughput varies over the cell area for the Pedestrian environment. 5
Note that smart antennas are not assumed. Smart antennas will of course further increase the 6
coverage. The results are shown using an E
b
/N
0
distribution corresponding to 95% 136 7
speech coverage. The distance attenuation is calculated according to the ITU Channel 8
Models. The results include mobile antenna diversity (2 branch, maximum ratio combining). 9
10
Figure 5 Coverage with mobile receiver diversity for Pedestrian A Environment 11
12
13
14
15
Figure 6 shows the results for the low speed Vehicular A50 Environment using the same 16
criteria as used in Figure 5. Figure 7 shows the results for the higher speed Vehicular A120 17
Environment using the same criteria as used in Figure 5. These results are for the simple 18
equalizer. 19
20
0 100 200 300 400 500 600
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C
.
D
.
F
.
Throughput S [kbps]
(TR45/98.03.03.19R6)
1
Figure 6 Coverage with mobile receiver diversity for Vehicular A50 Environment 2
3
4
5
Figure 7 Coverage with mobile receiver diversity for Vehicular A120 Environment 6
7
100 150 200 250 300 350 400 450 500
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C
.
D
.
F
.
Throughput S [kbps]
50 100 150 200 250 300 350 400 450
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C
.
D
.
F
.
Throughput S [kbps]
(TR45/98.03.03.19R6)
Dynamic System Simulations 1
2
In order to further evaluate the system performance of 136 HS using ITU Models, additional 3
dynamic packet data simulations have been performed. The following assumptions have been 4
made: 5
Interference Limited 6
1/3 reuse 7
No frequency hopping 8
Log-normal fading with standard deviation of 10 dB 9
Path loss according to M.1225 ITU Models 10
Pedestrian Channel A and B, Vehicular Channel A and B 50 km/hr, Vehicular 11
Channel A and B 120 km/hr environments simple equalizer 12
Power control is not used 13
2 Branch MRC diversity is used. 14
Ideal Link adaptation 15
The time step of the simulator is 20 ms, corresponding to one coding block. Each 16
radio block is explicitly simulated. 17
27 cell sites are used. 18
19
These additional simulations provided more detailed analysis of the system. 20
21
Traffic Model 22
A session-based traffic model is used. In-session users may be both active and idle. The 23
number of packets sent in a session is geometrically distributed with a mean of 10. A packet 24
is defined as for example a web page rather than an IP packet. Sessions arrive according to a 25
Poisson process. Packets are sent to user at the rate of 0.3 packets/sec. 26
27
The bursty behavior of a packet data system is modeled using a truncated Pareto distribution 28
for the packet interarrival times. 29
30
The packet (file) sizes are lognormally distributed with a mean of 12 kbytes. 31
32
Simulation 33
Each radio block was explicitly simulated included queuing, retransmission, etc. The time 34
step of the simulator was 20 msec. The total simulation time is 300 sec. Ideal link 35
adaptation is used. Packets are scheduled using an ideal G-based scheduling algorithm. 36
37
Dropping Criteria 38
A leaky bucket algorithm is used for user dropping. Each user is assigned a counter 39
initialized at 32. The counter is decreased by one for a NACK and increased by 2 (up to a 40
maximum of 32) for an ACK. The user is dropped if the counter reaches zero. 41
42
(TR45/98.03.03.19R6)
Performance Criteria 1
Since users in a packet data system will have different user throughputs, a quality measure is 2
defined. The quality measure for the 384 kbit/s service is that 95% of the users should have a 3
session throughput exceeding 10% of 384 kbit/s (or 38.4 kbit/s). The same quality measure 4
is used for the 64 kbit/s service which means that the system can be 100% loaded. Therefore 5
the spectral efficiency is obtained at 100% load for the 64 kbit/s service. 6
7
The session throughput is defined as the total number of bits a user transmitted in a session 8
divided by the total time for the transmissions. Dropped users are given a session throughput 9
of zero even if they transmitted some data before being dropped. 10
11
Simulations were performed with increasing load until the quality measure was reached. At 12
this load, the spectral efficiency is calculated. 13
(TR45/98.03.03.19R6)
384 kbit/s Service Results 1
For the 384 kbit/s packet data service, Figure 8 shows the spectrum efficiency results (per 2
sector) versus the lowest 5% percentile session throughput with the quality measure indicated 3
by the horizontal dashed line. Results for Pedestrian A, Vehicular A50 and Vehicular A120 4
using the simple equalizer are shown. Note that the spectral efficiency per site would be 5
obtained by multiplying by 3. 6
7
8
9
10
Figure 8 Spectral Efficiency Performance for 384 kbit/s Packet Data Service 11
Environment A. 12
250 300 350 400 450
10
20
30
40
50
60
70
80
384 kbps
Spectral efficiency [kbps/MHz/sector]
5
%

p
e
r
c
e
n
t
i
l
e

o
f

s
e
s
s
i
o
n

t
h
r
o
u
g
h
p
u
t

[
k
b
p
s
]
pedA3
vehA50
vehA120
(TR45/98.03.03.19R6)
1
Figure 9 shows the loading results (in number of users per sector) for the 384 kbit/s packet 2
data service, versus the lowest 5% percentile session throughput with the quality measure 3
indicated by the horizontal dashed line. Results for Pedestrian A, Vehicular A50 and 4
Vehicular A120 are shown. 5
6
7
8
9
Figure 9 Loading Performance for 384 kbit/s Packet Data Service Environment A. 10
11
12
70 75 80 85 90 95 100 105 110 115
10
20
30
40
50
60
70
80
384 kbps
Average number of users per sector
5
%

p
e
r
c
e
n
t
i
l
e

o
f

s
e
s
s
i
o
n

t
h
r
o
u
g
h
p
u
t

[
k
b
p
s
]
pedA3
vehA50
vehA120
(TR45/98.03.03.19R6)
1
Figure 10 shows the spectrum efficiency results for the 384 kbit/s packet data service (per 2
sector) for the Pedestrian B, Vehicular B50 and Vehicular B120 environments using the same 3
criteria as in Figure 7 again using the simple equalizer. As indicated in Attachment 2, the 4
Vehicular A results should be taken as worse case for both Vehicular A and B when the 5
improved channel tracking equalizer is used. The results are shown versus the lowest 5% 6
percentile session throughput with the quality measure indicated by the horizontal dashed 7
line. Again it should be noted that the spectral efficiency per site would be obtained by 8
multiplying by 3. 9
10
11
12
13
Figure 10 Spectral Efficiency Performance for 384 kbit/s Packet Data Service using 14
simple equalizer for Environment B. 15
200 250 300 350
10
20
30
40
50
60
70
384 kbps
Spectral efficiency [kbps/MHz/sector]
5
%

p
e
r
c
e
n
t
i
l
e

o
f

s
e
s
s
i
o
n

t
h
r
o
u
g
h
p
u
t

[
k
b
p
s
]
pedB3
vehB50
vehB120
(TR45/98.03.03.19R6)
1
Figure 11 shows the loading results (in number of users per sector) for the 384 kbit/s packet 2
data service for the Pedestrain B, Vehicular B50 and Vehicular B120 environments using the 3
same criteria as in Figure 10 again using the simple equalizer. The results are shown versus 4
the lowest 5% percentile session throughput with the quality measure indicated by the 5
horizontal dashed line. 6
7
8
9
Figure 11 Loading Performance for 384 kbit/s Packet Data Service using simple 10
equalizer for Environment B. 11
12
Additional spectral efficiency results are given in the Deployment Matrix in Attachment 16. 13
14
50 55 60 65 70 75 80 85 90 95 100
10
20
30
40
50
60
70
384 kbps
Average number of users per sector
5
%

p
e
r
c
e
n
t
i
l
e

o
f

s
e
s
s
i
o
n

t
h
r
o
u
g
h
p
u
t

[
k
b
p
s
]
pedB3
vehB50
vehB120
(TR45/98.03.03.19R6)
Attachment 7 1
Control Signaling Overhead and Performance 2
In the following, it is assumed that the physical and RLC/MAC protocol layer structure of 136 HS is similar 3
to the structure of Enhanced General Packet Radio Service (EGPRS). The RLC/MAC protocol provides fast 4
medium access via a reservation based medium access scheme, supplemented by selective ARQ for efficient re- 5
transmission of erroneous data blocks. There are three groups of control channels: broadcast information 6
(PBCCH), medium access signaling (PCCCH) and fast associated signaling such as ACK/NACK messages 7
(PACCH). A comprehensive overview of the GPRS control channels and their use is provided in [2]. 8
As for any HSD system overlaid on 136, the 136 HS broadcast and common control signaling (PBCCH and 9
PCCCH) can be placed on either an 136 HS carrier or on the 136 DCCH. For worse case analysis calculation 10
purposes, the PBCCH and PCCCH are considered to be on the 136 HS carrier itself. 11
For this scenario the PBCCH and PCCCH channels are multiplexed on time slot 0 in each TDMA frame. Out 12
of the blocks in the multiframe structure, 1/12 of the blocks are reserved for PBCCH while the remaining blocks 13
are shared between PCCCH and data. The number of blocks reserved for PCCCH is changed dynamically on a 14
per block basis, according to the need of random access and paging capacity. 15
16
Worse Case Maximal user bit rate accounting for signaling overhead 17
The maximal radio interface bit rate for an 136 HS 200 kHz 8-slot mobile station is 8*69.2=553.6 kbps. The 18
actual user bit rate is lower due to broadcast and common control signaling, associated signaling and RLC/MAC 19
header overhead. 20
The control channel needed for broadcast and common control signaling (PBCCH and PCCCH) is 21
approximately 50% of time slot 0, which yields a remaining data rate of 553.6*7.5/8=519 kbps. 22
Associated control channels used for ACK/NACK signaling (also including forward link channel quality 23
reports sent on the reverse link) also reduces the user data rate. Assuming, that an acknowledgment block is sent 24
after every fifth block, the signaling overhead for a full duplex 8-slot service is 1/(8*5)=2.5%, since one 25
acknowledgment block is sufficient for all 8 time slots. Hence, the resulting data rate is 519*(1-0.025)=506 26
kbps. 27
Finally, to arrive at the actual user bit rate, the RLC/MAC overhead must be taken into account. The size of 28
the RLC/MAC headers are 24 information bits per block plus 16 CRC bits, yielding (24+16)/1304=3.1% 29
RLC/MAC overhead. Hence, the maximal 136 HS user bit rate in both forward and reverse links is 30
506*(1-0.031)=490.3 kbps. This figure is valid for the worst case when all signaling takes place on the 136 HS 31
carrier. 32
33
Fast control signaling robustness 34
During packet transfer, fast and robust control signaling is essential for the delay performance. 136 HS uses 35
two main control messages, both of which are well protected by means of extensive channel coding. 36
1. ACK/NACK messages are sent on the PACCH logical channel, which always uses the most robust 37
modulation (GMSK) and the most robust channel coding, regardless what schemes are used for the data transfer. 38
The information is also protected by a 40 bit FIRE code, which provides excellent error detection and which 39
can also be used for error correction. Furthermore, if a PACCH block including acknowledgments for e.g. five 40
blocks is not received correctly, the next PACCH block will automatically include the old acknowledgments and 41
new acknowledgments. The data transfer can thus continue uninterrupted. Figure 1 depicts PACCH block error 42
distribution for a 45% loaded 1/3 system without antenna diversity, using the same models as in Attachment 6. 43
The performance can be further improved by using the 40 bit FIRE code for error correction, which was not 44
done in the simulation. 45
2. To inform the receiver what channel coding is used, a two bit message is included in each block. It is 46
block coded to 8 modulated bits, which makes this signaling considerably more robust than the data 47
transmission. 48
49
Link adaptation signaling errors 50
(TR45/98.03.03.19R6)
In the reverse link, a Channel Quality Report is included in each ACK/NACK block for use by link 1
adaptation and power control. Since the PACCH logical channel used for this signaling is very robust (see 2
above), errors are uncommon. However, should a block error occur, the error is detected and the modulation and 3
coding scheme used for the previous block is used also for the next. Therefore, the performance in case of 4
control channel errors can be approximated by the performance when using longer update intervals. In an 5
extreme case of 50% BLER on the PACCH (which only occurs for 1% of the mobiles according to Figure 1), a 6
nominal link adaptation update interval of 5 blocks corresponds to an actual update interval of 10 blocks. The 7
degradation caused by this is not significant, as Figure 2 clearly shows. 8
Further, the information telling what coding scheme is used is sent using in-band signaling, for which 9
performance is considerably better than for data (see previous question). The very few errors in these messages 10
do not affect performance significantly. 11
The conclusion is thus that in the unlikely case of errors in the link adaptation control signaling loop, that is 12
handled very well by the link adaptation mechanism. 13
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
10
20
30
40
50
60
70
80
90
100
C
.
D
.
F
.

[
%
]
BLER
14
Figure 1. Block error rate distribution for PACCH in a 1/3 reuse system with 45% load 15
0 10 20 30 40 50 60 70
0
10
20
30
40
50
60
70
80
90
100
C
.
D
.
F
.

[
%
]
Throughput [kbps]
20 RLC blocks update interval
16
Figure 2. User throughput distribution per timeslot for different link adaptation update 17
intervals in a 1/3 reuse system with 45% load 18
19
(TR45/98.03.03.19R6)
Admission Control 1
136 HS as well as all other data systems need admission control and congestion control algorithms to prevent 2
the system from producing unacceptable packet delays. Admission control prevents a new user from entering the 3
system if the system is already fully loaded, whereas congestion control drops users that cannot be served 4
without causing unacceptable packet delay. 5
6
136 HS admission control scheme 7
The 136 HS protocol includes means for continuously monitoring system characteristics such as queue 8
lengths, number of re-transmissions of a certain packet, number of users in different cells etc. Such 9
characteristics are the preferred information on which admission control is based. The actual algorithm is vendor 10
specific and is not limited by the RTT. 11
The admission control algorithm is preferably supplemented by a congestion control algorithm that drops a 12
user for which the link quality is low enough to cause excessive block re-transmissions. 13
14
References 15
[1]. ETSI, GSM 03.64, Overall description of the General Packet Radio Service (GPRS) Radio Interface, 16
1997. 17
[2]. J. Cai and D. Goodman, General Packet Radio Service in GSM, IEEE Communications Magazine, 18
October 1997, Vol. 35, No. 10. 19
[3]. H. Olofsson, J. Nslund, J. Skld, Interference Diversity Gain in Frequency Hopping GSM, in 20
Proceedings of the 45
th
IEEE Vehicular Technology Conference (VTC95), 1995, pp. 102-106. 21
[4]. H. Olofsson et al, Improved Interface Between Link Level and System Level Simulations Applied to 22
GSM, in Proceedings of the 6
th
IEEE International Conference on Universal Personal Communications 23
(ICUPC97), 1997, pp. 79-83. 24
(TR45/98.03.03.19R6)
Attachment 8 1
Physical Layer Simulation Results 2
Performance without Frequency Hopping TU3 Environment 3
4
Performance for PCS-1 to PCS-6 without frequency hopping in a Typical Urban environment 5
is depicted in Figure 1 and Figure 2 (data rates shown per time slot). The velocity is 3 km/h 6
(900 MHz). No diversity is assumed. 7
0 5 10 15 20 25 30 35
10
3
10
2
10
1
10
0
B
L
E
R
C/I [dB]
PCS1 (22.8 kbps)
PCS2 (34.3 kbps)
PCS3 (41.25 kbps)
PCS4 (51.6 kbps)
PCS5 (57.35 kbps)
PCS6 (69.2 kbps)
8
Figure 1. BLER Performance for PCS-1 to PCS-6 without frequency hopping (C/I). 9
0 5 10 15 20 25 30 35
10
3
10
2
10
1
10
0
B
L
E
R
E
b
/N
0
[dB] (E
b
= Energy per modulated bit)
PCS1 (22.8 kbps)
PCS2 (34.3 kbps)
PCS3 (41.25 kbps)
PCS4 (51.6 kbps)
PCS5 (57.35 kbps)
PCS6 (69.2 kbps)
10
Figure 2. BLER Performance for PCS-1 to PCS-6 without frequency hopping 11
(Eb/No). 12
(TR45/98.03.03.19R6)
Performance with Frequency Hopping TU3 Environment 1
2
Performance for PCS-1 to PCS-6 with frequency hopping in a Typical Urban environment is 3
depicted in Figure 3 and Figure 5 (data rates are given per time slot), while the performance 4
for CS-1 to CS-4 (which replaces ECS-5 to ECS-8) is depicted in Figure 4 and Figure 6. The 5
velocity is 3 km/h (900 MHz). No diversity is assumed. 6
7
0 5 10 15 20 25 30 35
10
3
10
2
10
1
10
0
B
L
E
R
C/I [dB]
PCS1 (22.8 kbps)
PCS2 (34.3 kbps)
PCS3 (41.25 kbps)
PCS4 (51.6 kbps)
PCS5 (57.35 kbps)
PCS6 (69.2 kbps)
8
Figure 3. BLER Performance for PCS-1 to PCS-6 with frequency hopping (C/I). 9
10
0 5 10 15 20 25 30 35
10
4
10
3
10
2
10
1
10
0
C/I [dB]
B
L
E
R
CS1 (11.2 kbps)
CS2 (14.5 kbps)
CS4 (22.8 kbps)
CS3 (16.7 kbps)
11
Figure 4. BLER Performance for CS-1 to CS-4 with frequency hopping (C/I). 12
(TR45/98.03.03.19R6)
0 5 10 15 20 25 30 35
10
3
10
2
10
1
10
0
B
L
E
R
E
b
/N
0
[dB] (E
b
= Energy per modulated bit)
PCS1 (22.8 kbps)
PCS2 (34.3 kbps)
PCS3 (41.25 kbps)
PCS4 (51.6 kbps)
PCS5 (57.35 kbps)
PCS6 (69.2 kbps)
1
Figure 5. BLER Performance for PCS-1 to PCS-6 with frequency hopping (Eb/No). 2
0 5 10 15 20 25 30 35
10
4
10
3
10
2
10
1
10
0
Eb/No [dB] (Eb=Energy per modulated bit)
B
L
E
R
CS1 (11.2 kbps)
CS2 (14.5 kbps)
CS4 (22.8 kbps)
CS3 (16.7 kbps)
3
4
Figure 6. BLER Performance for CS-1 to CS-4 with frequency hopping (Eb/No). 5
6
(TR45/98.03.03.19R6)
Performance of 8-PSK with ITU Pedestrian A and B channels 1
2
Performance for 8-PSK without frequency hopping but with antenna diversity in an ITU 3
Outdoor to Indoor and Pedestrian A coverage limited environment is depicted in Figure 7 4
(note that the BLER is plotted versus energy per symbol (Es/No) therefore Eb/No would be 5
obtained by reducing these numbers by 4.77 dB). The interference limited performance is 6
shown in Figure 8. The velocity is 3 km/h (1900 MHz). These results were obtained using a 7
simple equalizer. 8
9
10
11
Figure 7. BLER Performance for PCS-1 to PCS-6 in ITU Outdoor to Indoor and 12
Pedestrian A Coverage Limited Environment . 13
14
15
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
outindoor pedest A, 3km/h, 2br ,no FH
Es/No [dB] (Es=energy per modulated symbol)
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
1
Pedestrian A Interference Limited Environment. 3
4
The performance for Pedestrian Channel B under the same conditions is shown in Figure 9 5
and Figure 10. 6
7
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
1
Pedestrian B Coverage Limited Environment . 3
4
5
Pedestrian B Interference Limited Environment. 7
8
9
10
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
outindoor pedest B, 3km/h, 2br ,no FH
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
Performance of 8-PSK with ITU Vehicular A50 channel 1
Vehicular A50 coverage limited environment using the simple 8-PSK equalizer is depicted 3
in Figure 11 (note that the BLER is plotted versus energy per symbol (Es/No) therefore 4
Eb/No would be obtained by reducing these numbers by 4.77 dB). The interference limited 5
performance is shown in Figure 12. The velocity is 50 km/h (1900 MHz). 6
7
Coverage Limited Environment 9
Interference Limited Environment. 11
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
Performance of 8-PSK with ITU Vehicular A120 channel 1
Vehicular A120 coverage limited environment is depicted in Figure 13 (note that the BLER 3
is plotted versus energy per symbol (Es/No) therefore Eb/No would be obtained by reducing 4
these numbers by 4.77 dB). The interference limited performance is shown in Figure 14. 5
The velocity is 120 km/h (1900 MHz). Note that with using the the improved channel 6
tracking equalizer described in Attachment 2, all the coded schemes can be used. The 7
performance of the simple equalizer for this environment is given in Attachment 2. 8
9
Coverage Limited environment 11
12
0 5 10 15 20 25 30 35
10
2
10
1
10
0
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
Performance of 8-PSK with ITU Vehicular B120 channel 1
Vehicular B120 coverage limited environment is depicted in Figure 15 (note that the BLER is 3
plotted versus energy per symbol (Es/No) therefore Eb/No would be obtained by reducing 4
these numbers by 4.77 dB). The interference limited performance is shown in Figure 16. 5
The velocity is 120 km/h (1900 MHz). Note that with using the the improved channel 6
tracking equalizer described in Attachment 2, all the coded schemes can be used. The 7
performance of the simple equalizer for this environment is given in Attachment 2. 8
9
Coverage Limited environment 11
12
0 5 10 15 20 25 30 35
10
2
10
1
10
0
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
(TR45/98.03.03.19R6)
Performance of GMSK with ITU Pedestrian A and B channels 1
Performance for GMSK without frequency hopping but with antenna diversity in an ITU 2
Outdoor to Indoor and Pedestrian A coverage limited environment is depicted in Figure 17 3
(note that the BLER is plotted versus energy per symbol (Es/No) however with GMSK there 4
is 1 bit per symbol and therefore is equal to Eb/No). The interference limited performance is 5
shown in Figure 18. The equivalent performance for the Pedestrian B channel is shown in 6
Figure 19 and Figure 20. The velocity is 3 km/h (1900 MHz). These results were obtained 7
using a simple equalizer. 8
Figure 17. BLER Performance for GMSK simple equalizer in ITU Pedestrian A 9
11
Figure 18. BLER Performance for GMSK simple equalizer in ITU Pedestrian A 12
Interference Limited Environment 13
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
0 5 10 15 20 25 30 35
10
2
10
1
10
0
outindoor pedest A, 3km/h, 2br, no FH
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
(TR45/98.03.03.19R6)
1
Figure 19. BLER Performance for GMSK simple equalizer in ITU Pedestrian B 2
4
Figure 20. BLER Performance for GMSK simple equalizer in ITU Pedestrian B 5
0 5 10 15 20 25 30 35
10
2
10
1
10
0
outindoor pedest B, 3km/h, 2br, no FH
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
(TR45/98.03.03.19R6)
Performance of GMSK with ITU Vehicular A channels 1
Vehicular A50 coverage limited environment is depicted in Figure 21 (note that the BLER is 3
plotted versus energy per symbol (Es/No) however with GMSK there is 1 bit per symbol and 4
therefore is equal to Eb/No). The interference limited performance is shown in Figure 22. 5
The velocity is 50 km/h (1900 MHz). Likewise, performance of GMSK for Vehicular A120 6
channels is shown in Figures 23 and 24 where the velocity is 120 km/h (1900 MHz). These 7
results were obtained using a simple equalizer. 8
9
(TR45/98.03.03.19R6)
1
Figure 21. BLER Performance for GMSK simple equalizer in ITU Vehicular A50 2
4
0 5 10 15 20 25 30 35
10
2
10
1
10
0
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
(TR45/98.03.03.19R6)
1
4
0 5 10 15 20 25 30 35
10
2
10
1
10
0
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
(TR45/98.03.03.19R6)
Performance of GMSK with ITU Vehicular B channels 1
Vehicular B50 coverage limited environment is depicted in Figure 25 (note that the BLER is 3
plotted versus energy per symbol (Es/No) however with GMSK there is 1 bit per symbol and 4
therefore is equal to Eb/No). The interference limited performance is shown in Figure 26. 5
The velocity is 50 km/h (1900 MHz). Likewise, performance of GMSK for Vehicular B120 6
channels is shown in Figures 27 and 28 where the velocity is 120 km/h (1900 MHz). These 7
results were obtained using a simple equalizer. 8
9
(TR45/98.03.03.19R6)
1
Figure 25. BLER Performance for GMSK simple equalizer in ITU Vehicular B50 2
4
0 5 10 15 20 25 30 35
10
2
10
1
10
0
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
(TR45/98.03.03.19R6)
1
4
7
8
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
CS1
CS2
CS3
CS4
(TR45/98.03.03.19R6)
Attachment 9 1
136 HS Coverage with respect to 136 voice (from link budgets) 2
3
Table 1. General assumptions 4
Vehicular
Maximum peak power 28 dBm
Path loss 128.1 + 37.6*log10(d), d in
kilometers
Log-normal fade margin 11.3 dB
Handoff gain 4.7 dB
5
Table 2. Assumptions for IS-136 6
Required Eb/(N0+I0) 17.0 dB
Antenna diversity gain in uplink 6.0 dB
7
Table 3. Assumptions for 136 HS 8
Frequency hopping No frequency hopping
BLER level in EGPRS 10 %
Antenna diversity gain Included in Eb/No
9
Table 4. Additional assumptions for data terminals 10
Mobile antenna gain (=lower body loss) 2.0 dBi (0.0 dBi for speech
terminals)
Antenna diversity in the mobile, gain Included in Eb/No
11
Mobile antenna gain of 2.0 dBi is used because data terminals will not be used very close to 12
the users head and therefore the body loss can be assumed to be lower than for speech 13
terminals. Receiver antenna diversity has been assumed for data terminals which is included 14
in Eb/No. 15
16
136 HS Link Budget Appendix 4 shows the link budgets for the vehicular environment for 17
both the 384 kbit/s and 144 kbit/s packet data services, EGPRS/PCS-4 (384 kbps) and 18
EGPRS/PCS-1 (144 kbit/s). 136 HS Link Budget Appendix 5 shows the link budget for an 19
asymmetric case, vehicular environment, EGPRS/PCS-4 (downlink, 384 kbit/s) & CS-3 20
(uplink, 134 kbit/s). 136 HS Link Budget Appendix 6 shows the link budget for vehicular 21
environment, EGPRS/PCS-3 (330 kbps). Table 5 shown below is a link budget for IS-136 22
speech which is used for comparison. 23
(TR45/98.03.03.19R6)
Table 5. Link Budget Template for IS-136 Speech. Use for comparison. 1
Item Forward Link Reverse Link
Test environment
Test service
Multipath channel class
(a0) Average transmitter power per traffic channel
(a1) Maximum transmitter power per traffic channel
(a2) Maximum total transmitter power 35.00 dBm 28.00 dBm
(b) Cable, connector, and combiner losses (enumerate sources) 2.00 dB 0.00 dB
(c) Transmitter antenna gain 13.00 dBi 0.00 dBi
(d1) Transmitter e.i.r.p. per traffic channel 46.00 dBm 28.00 dBm
(d2) Total Transmitter e.i.r.p. = (a2-b+c) 46.00 dBm 28.00 dBm
(e) Receiver antenna gain 0.00 dBi 13.00 dBi
(f) Cable and connector losses 0.00 dB 2.00 dB
(g) Receiver noise figure 5.00 dB 5.00 dB
(h) Thermal noise density -174.00 dBm/Hz -174.00 dBm/Hz
(i) Receiver interference density (NOTE 1) 0 mW/Hz 0 mW/Hz
(j) Total effective noise plus interference density
= 10 log ( 10
((h+g)/10)
+ i )
-169.00 dBm/Hz -169.00 dBm/Hz
(k) Information rate (10 log (R
b
)) 46.87 dB(Hz) 46.87 dB(Hz)
(l) Required E
b
/(N
0
+ I
0
) 17.00 dB 17.00 dB
(m) Receiver sensitivity = (j+k+l) -105.13 dB -105.13 dB
(n) Hand-off gain 4.70 dB 4.70 dB
(o) Explicit diversity gain 0.00 dB 6.00 dB
(o) Other gain 0.00 dB 0.00 dB
(p) Log-normal fade margin (90% cell edge reliability) 11.30 dB 11.30 dB
(q) Maximum path loss
= ( d1 - m + (e - f) + o + n + o - p )
144.53 dB 143.53 dB
Maximum Range 2727 m 2565 m
Range for data terminals
Data terminal antenna gain (lower body loss)
Mobile Station antenna diversity gain
Maximum path loss for data terminals
Maximum range
(TR45/98.03.03.19R6)
1
According to the vehicular A50 link budgets, the downlink range of bit rate 384 kbit/s with 2
136 HS is about 82.5 % of IS-136 uplink speech range. It has been assumed that downlink is 3
the limiting factor for 136 HS range. If asymmetric connection is considered, more coding 4
can be applied on the uplink to increase its range. It should also be noticed that the service 5
area of 384 kbit/s is more than 0.825*0.825 of IS-136 cell area because the coverage areas of 6
adjacent IS-136 cells must be overlapping to guarantee service at cell border with log normal 7
fading. The downlink range of 136 HS PCS-3 330 kbit/s is approximately 96.6 % of IS-136 8
uplink speech range. Therefore, 136 HS PCS-3 packet data services with 330 kbit/s can be 9
offered essentially with the same coverage as IS-136 speech. 10
11
For the 144 kbit/s data service, EGPRS/PCS-1 can be used to provide the capability. 12
According to the A120 Vehicular link budgets in Appendix 4, the downlink range for the 144 13
kbit/s service for 136 HS is about 126% of the IS-136 uplink speech range. 14
15
The link budgets show maximum path loss and maximum range figures with terminal 16
diversity. 17
18
Summary of 136 HS packet data ranges and coverage areas is shown in Table 6. The 136 HS 19
Outdoor/Vehicular analysis in Table 6 includes terminal diversity gains. 20
21
Table 6. Maximum range of 136 HS compared to IS-136 speech uplink range (from link 22
budgets) 23
24
Range / m Coverage area / m
2
136 HS /CS-3 134 kbit/s (uplink) 101 % x IS-136 102 % x IS-136 cell
136 HS /PCS-3 330 kbit/s (downlink) 96.6 % x IS-136 more than 93 % of IS-136 cell
136 HS /PCS-4 384 kbit/s (downlink) 82.5 % x IS-136 more than 68 % of IS-136 cell
136 HS /PCS-1 144 kbit/s (downlink) 126 % x IS-136 more than 158 % of IS-136
25
26
(TR45/98.03.03.19R6)
Attachment 10 1
Deployment Model Results Matrix, Spectrum Efficiency Models 2
3
It is necessary to use the value for the Total Number of Cell Sites for Coverage 4
Efficiency for the Spectrum Efficiency computations to meet the coverage 5
requirements specified in the Deployment Model for each Test Environment. 6
7
The Deployment Model Results Matrix for 136 and 136+ are presented in Tables 1 8
and 2, respectively. 9
Table 1. Deployment Model Results Matrix for 136 10
11
Market
Requirement
Total
Number of
Cell Sites
Total Number
of RF
Channels
Number of Voice
Channels/RF
Channel
Coverage
Efficiency
(km
2
/site)
Spectrum
Efficiency
(Mbits/s/MHz/site)
Spectrum
Efficiency
(E/MHz/site)
Coverage
Efficiency
4.33 26 3 34.6 - -
Vehicular
Spectrum
Efficiency
- 520 3 - 0.228 20.25
Coverage
Efficiency
369 439 3 0.075
(1)
0.424
(2)
- - Pedestrian-
Outdoor to
Indoor Spectrum
Efficiency
- 2090 3 - 0.010 0.47
Coverage
Efficiency
4 32 3 0.197 - - Indoor
Spectrum
Efficiency
- 196 3 - .09344 8.33
12
(1)
Coverage Efficiency for Pedestrian Environment for Outdoor to Indoor Coverage 13
(2)
Coverage Efficiency for Pedestrian Environment for Outdoor Coverage 14
15
16
Table 2. Deployment Model Results Matrix for 136+ 17
18
Market
Requirement
Total
Number of
Cell Sites
Total Number
of RF
Channels
Number of Voice
Channels/RF
Channel
Coverage
Efficiency
(km
2
/site)
Spectrum
Efficiency
(Mbits/s/MHz/site)
Spectrum
Efficiency
(E/MHz/cell)
Coverage
Efficiency
6.03 36 3 24.89 - - Vehicular
Spectrum
Efficiency
- 540 3 - 0.256 14.52
Coverage
Efficiency
564 672 3 0.049
(1)
0.276
(2)
- - Pedestrian-
Outdoor to
Indoor Spectrum
Efficiency
- 2688 3 - 0.0118 0.30
Coverage
Efficiency
6 36 3 0.110 - - Indoor
Spectrum
Efficiency
- 204 3 - .09696 5.56
19
(1)
Coverage Efficiency for Pedestrian Environment for Outdoor to Indoor Coverage 20
(2)
Coverage Efficiency for Pedestrian Environment for Outdoor Coverage 21
22
(TR45/98.03.03.19R6)
General Assumptions: 1
2
The spectrum efficiency results for voice capacity, E/MHz/Site, are bounded by the 3
capacity calculations presented in Method 1 of Attachment 17. 4
5
6
The information rate per RF channel for 136 is 28800 bits/s, i.e. 9600 bits/s per 7
server. 8
The information rate per RF channel for 136+ is 43200 bits/s, i.e. 14400 bits/s per 9
server. 10
11
Examples of Information Rate Calculations for each Environment 12
13
Vehicular 14
15
For the Vehicular Test Environment, it is assumed that 3-sector site configurations 16
are used for the coverage area. 17
18
For 136: 19
Erlangs per Sector = 1312.5 (System Erlangs) / (4.33 (Total Number of Cell Sites) * 20
3 (Sectors per Site)) = 101.04 Erlangs per Sector 21
Total Number of Servers Required per Sector for 101.04 Erlangs @ 1% Blocking = 22
118 23
Total Number of RF Channels per Sector = 40 (119 Servers and 1 DCCH) 24
Information Rate per Sector = 119 (Servers) * 9600 = 1.1424 Mbits/s 25
Information Rate per Site = 1.1424 * 3 = 3.4272 Mbits/s 26
Spectrum Efficiency = 3.4272 Mbits/s / (15 MHz) = 0.228 Mbits/s/MHz/site 27
28
For 136+: 29
Erlangs per Sector = 1312.5 (System Erlangs) / (6.03 (Total Number of Cell Sites) * 30
3 (Sectors per Site)) = 72.55 Erlangs per Sector 31
Total Number of Servers Required per Sector for 72.55 Erlangs @ 1% Blocking = 88 32
(TR45/98.03.03.19R6)
Pedestrian 1
2
For the Pedestrian-Outdoor to Indoor Test Environment, it is assumed that the 3
indoor coverage area (25 km
2
) is wholly contained within the outdoor coverage area 4
(40 km
2
). It is assumed that the Number of Required System Erlangs for this 25 km
2
5
includes both the indoor and outdoor market requirements. It is assumed that 6
omnidirectional site configurations are used for the indoor coverage area (25 km
2
) 7
contained within the outdoor coverage area and that 3-sector configurations are 8
used for the remaining (15 km
2
) outdoor coverage area. 9
10
For 136, the Pedestrian indoor coverage area requires 334 sites and the outdoor 11
coverage area requires 35 sites for a Total Number of Cell Sites of 369. For 136+, 12
the indoor coverage area requires 510 sites and the outdoor coverage area requires 13
54 sites for a Total Number of Cell Sites of 564. 14
For 136: 15
For 25 km
2
(Outdoor and Indoor): 16
Erlangs per Sector = 2190 (System Erlangs) / (334 (Total Number of Cell Sites)) = 17
6.56 Erlangs per Site 18
Total Number of Servers Required per Site for 6.56 Erlangs @ 1% Blocking = 13 19
Total Number of RF Channels per Site = 5 (14 Servers and 1 DCCH) 20
Information Rate per Site = 14 (Servers) * 9600 = 0.1344 Mbits/s 21
23
For 15 km
2
(Outdoor) 24
Erlangs per Sector = 450 (System Erlangs) / (35 (Total Number of Cell Sites) * 3 25
(Sectors per Site)) = 4.29 Erlangs per Sector 26
32
Resultant Spectrum Efficiency for 136: 33
((0.00896 Mbits/s/MHz/site * 334 Sites) + (0.021 Mbits/s/MHz/site * 35 Sites)) / 369 34
Sites = 35
0.010 Mbits/s/MHz/site 36
37
For 136+: 38
For 25 km
2
(Outdoor and Indoor): 39
Erlangs per Sector = 2190 (System Erlangs) / (510 (Total Number of Cell Sites)) = 40
4.29 Erlangs per Site 41
Information Rate per Site = 11 (Servers) * 14400 = 0.1584 Mbits/s 44
(TR45/98.03.03.19R6)
1
For 15 km
2
(Outdoor) 2
Erlangs per Sector = 450 (System Erlangs) / (54 (Total Number of Cell Sites) * 3 3
(Sectors per Site)) = 2.78 Erlangs per Sector 4
10
Resultant Spectrum Efficiency for 136+: 11
((0.01056 Mbits/s/MHz/site * 510 Sites) + (0.02304 Mbits/s/MHz/site * 54 Sites)) / 12
564 Sites = 13
0.0118 Mbits/s/MHz/site 14
15
(TR45/98.03.03.19R6)
Indoor 1
2
For the Indoor Test Environment, it is assumed that omnidirectional site 3
configurations are used and that the base stations (radiating elements) are not 4
deployed on partial floors. 5
6
For 136: 7
Erlangs per Site = 500 (System Erlangs) / (4 (Total Number of Cell Sites)) = 125 8
Erlangs per Site 9
Total Number of Servers Required per Site for 125 Erlangs @ 1% Blocking = 144 10
Information Rate per Site = 1.4016 Mbits/s 13
15
For 136+: 16
Erlangs per Sector = 500 (System Erlangs) / (6 (Total Number of Cell Sites)) = 83.33 17
Erlangs per Sector 18
Information Rate per Site = 1.4544 Mbits/s 22
(TR45/98.03.03.19R6)
Attachment 11 1
Answers to A1.2.20 for 136 and 136+ 2
3
136 4
Class Bit Rate Delay BER/FER Examples
A1.2.20.1 A 7.4 kb/s
12.2 kb/s
7.4 kb/s
12.2 kb/s
9.6 kb/s
51.7 ms
46.7 ms
31.7 ms
26.7 ms
< 30 ms
1% BER
3% BER
1% BER
3% BER
variable
Voice - IS-641-A, interleave 2
Voice - US1, interleave 2
Voice - IS-641-A, interleave 1
Voice - US, interleave 1
IS-130 asynch data
A1.2.20.2 B variable (max 9.6 kb/s) < 30 ms < 3% uncorrected BER IS-130+WAP
136+ Packet (Mango)
A1.2.20.3 C variable -
half,full,double,triple rate
(full rate = 9.6 kb/s max)
variable error-free with ARQ
(discounting the
undetected error rate of
the CRC)
IS-130/135 asynch data/fax
A1.2.20.4 D variable variable error-free with ARQ
(discounting the
undetected error rate of
the CRC)
136+ Packet (Mango),
SMS (IS-136-A), WAP, GUTS
5
(TR45/98.03.03.19R6)
1
136+ 2
Clas
s
Bit Rate Delay BER/FER Examples
A1.2.20.
1
A 8 kb/s
9.6 kb/s
~ 50 ms
< 30 ms
< 3% FER
variable
Voice - IS-641A
IS130/135 asynch data/fax
A1.2.20.
2
B variable (avg 9.6
kb/s)
< 30 ms < 3% uncorrected
BER
IS130/135+WAP
IS136 Packet (Mango)
A1.2.20.
3
C variable -
half,full,double,triple
rate
(full rate = 9.6 kb/s
typ.)
variable error-free with
ARQ
IS130/135 asynch data/fax
A1.2.20.
4
D variable variable error-free with
ARQ
IS136 Packet (Mango),
SMS (IS136A), WAP,
GUTS
3
(TR45/98.03.03.19R6)
Attachment 12 1
Layer 1 Description 2
Indoor Environment, 1.6 MHz carrier 3
4
Glossary of abbreviations: 5
6
ARQ Automatic Repeat reQuest 7
BS Base station 8
DL Downlink (forward link) 9
TDMA Time Division Multiple Access 10
FDD Frequency Division Duplexing 11
FEC Forward Error Correction 12
FH Frequency Hopping 13
JD Joint Detection 14
LA Link Adaptation 15
LC Load Control 16
Mbps Mega bits per second 17
MS Mobile Station 18
NRT Non Real Time 19
PC Power Control 20
RT Real Time 21
TDD Time Division Duplexing 22
TH Time Hopping 23
TX transmission 24
UL Uplink (reverse link) 25
Physical channels 26
136 HS Indoor can operate in FDD mode and in TDD mode. The channel spacing of the 27
wideband mode is 1.6 MHz both in FDD and in TDD mode. The basic physical channel is a 28
certain time slot on a certain carrier frequency. In the following, an overview about the 29
multiframe, unit frame and time slot structure is given. The last subsection defines the 30
modulation method. 31
Multiframe 32
Multiframe structure is presented in Figure 1. The contents of the control channel cluster 33
indicated in the figure is dependent on the higher layer protocols, too. 34
35
(TR45/98.03.03.19R6)
0 1 2 3 24 25
0 1 2 3 24 25 0 1 2 3 49 50
multiframe with 26 frames
multiframe with 51 frames
superframe
51 (26-frame) multiframes or 26 (51-frame) multiframes)
hyperframe
2048 superframes
0 1 2 3 49 50
0 1 2 3 2046 2047
Frame (2k) Frame (2k+1)
Frame 2k
TS3:0-7
Frame 2k+1
TS3:0-7
Control channels Control channels
1
Figure 1 The multiframe structure for 136 HS Indoor. 2
FDD and TDD frames 3
In the following sections, a unit frame structure is presented separately for FDD and TDD 4
modes. 5
FDD frame 6
The unit FDD frame is presented in Figure 2. The length of the FDD frame is 4.615 ms 7
which is 12000 symbol periods. 8
9
72 s
288 s
4.615 ms
TDMA Frame
1/64 Slot
1/16 Slot
1.6 MHz
10
Figure 2 The unit FDD frame structure of 136 HS Indoor. 11
(TR45/98.03.03.19R6)
TDD frame 1
The TDD frame is of the same length as the FDD frame but it is divided into downlink and 2
uplink parts (Figure 3). The switching point between uplink and downlink can be moved in 3
the TDD frame to adopt asymmetric traffic. The minimum length of uplink and downlink 4
parts is one eighth of the frame length (577 s). 5
6
72 s
288 s
4.615 ms
TDMA Frame
1/64 Slot
1/16 Slot
1.6 MHz
Downlink Uplink
Switching point between
uplink and downlink
7
Figure 3 The unit TDD frame structure of 136 HS Indoor. 8
9
In the TDD frame structure, it is assumed that the same mobile station is not receiving in the 10
last slot of the downlink part and transmitting in the first slot of the uplink part. 11
Time slots 12
The TDMA frame is subdivided into time slots. Two different types of time slots are 13
presented for 136 HS Indoor in Figure 2 and in Table 1 : 1/64 time slot and 1/16 time slot. In 14
a 1/64 time slot there are 187.5 symbol periods and in a 1/16 time slot 750 symbol periods. 15
136 HS Outdoor/Vehicular uses 1/8 time slot. 16
17
Table 1 : Time slot lengths in seconds and in symbol periods. 18
Time slot type Length in seconds Length in symbol periods (SP)
1/64 time slot 72 s 187.5 SP
1/16 time slot 288 s 750 SP
1/8 time slot 577 s 208.33 SP (136 HS
Outdoor/Vehicular)
19
A 136 HS Indoor TDMA frame of length 4.615 ms can consist of 20
64 1/64 time slots of length 72 s (15/208 ms) or 21
16 1/16 time slots of length 288 s (15/52 ms) or 22
any mix of these time slots of different lengths fitting together in the TDMA frame. 23
(TR45/98.03.03.19R6)
The 1/64 slot can be used for all services from low rate speech and data to high rate data 1
services. The 1/16 slot is to be used for medium to high rate data services. 2
The physical content of the time slots are the bursts of corresponding length as described in 3
Section 1.3. 4
Bursts 5
Traffic bursts for the 1.6 MHz carrier 6
Three types of traffic bursts are defined for 136 HS Indoor: the Short burst 1 (SB1), the Short 7
burst 2 (SB2) and the Long burst (DB). 8
Long burst 9
The Long burst uses 1/16 time slot. It consists of two tail symbol fields, two data symbol 10
fields, training sequence field and guard period (Figure 4). The use of individual symbols is 11
defined in Table 2. 12
The payload of the Long burst (number of data symbols) is 684 symbols. The training 13
sequence is located in the middle of the burst so that none of the data symbols is too far from 14
it. This improves the channel estimate compared to the use of the preamble type training 15
sequence. Burst tail symbols are defined to have fixed value (e.g. 0s) in order to make 16
different detection methods possible. In the end of the burst there is the guard period of 11 17
symbol periods. The guard period is a protection interval between bursts for time alignment 18
uncertainty, time dispersion and power ramping. 19
Long burst can be used for all services from medium rate data (64 kbps) to high rate data up 20
to 2 Mbit/s. 21
Data symbols
342
TS
49
Data symbols
342
TB
3
288 s
GP
11
TB
3
22
Figure 4 Burst structure of the Long burst. TB stands for burst tail symbol, TS for training 23
sequence and GP for guard period. 24
Table 2 : The contents of the Long burst fields and the use of individual symbols 25
Symbol number
(SN)
Length of field Contents of field
0-2 3 Tail symbols
3-344 342 Data symbols
345-393 49 Training sequence
736-738 3 Tail symbols
739-749 11 Guard period
26
(TR45/98.03.03.19R6)
Short bursts 1 and 2 1
The Short bursts use 1/64 time slot. There are two types of Short bursts that differ in the 2
length of the training sequence. The training sequence of the Short burst 1 is 49 symbol 3
periods long whereas the training sequence of the Short burst 2 is 27 symbol periods long. 4
Thus the Short burst 1 provides better channel estimate and allows for longer multipath 5
delays than the Short burst 2 by sacrificing some of the burst payload. 6
The Short bursts 1 and 2 consist of two tail symbol fields, two data symbol fields, training 7
sequence field and guard period (Figure 5 and Figure 6). The payloads (number of data 8
symbols) of the Short bursts 1 and 2 are 122 symbols and 144 symbols, respectively. The use 9
of the individual symbols is defined in Table 3 and Table 4. 10
Both burst formats (Short 1 and 2) can be used for all services from speech to high rate data 11
up to 2 Mbit/s. 12
Data symbols
61
Training sequence
49
TB
3
72 s
GP
10.5
TB
3
Data symbols
61
13
Figure 5 Burst structure of the Short burst 1. TB stands for burst tail bit and GP for guard 14
period. 15
Table 3 : The contents of the Short burst 1 fields and the use of individual symbols. 16
Symbol number
(SN)
0-2 3 Tail symbols
177-187 10.5 Guard period
17
Data symbols
72
TS
27
TB
3
72 s
GP
10.5
TB
3
Data symbols
72
18
Figure 6 Burst structure of the Short burst 2. TB stands for burst tail bit, TS for training 19
sequence and GP for guard period. 20
(TR45/98.03.03.19R6)
Table 4 : The contents of the Short burst 2 fields and the use of individual symbols. 1
Symbol number
(SN)
0-2 3 Tail symbols
177-187 10.5 Guard period
2
Modulation 3
The basic modulation parameters including pulse shaping are summarized in Table 5. 4
Table 5 : Basic modulation parameters 5
Carrier symbol rate 2.6 MSymbol/s
Carrier spacing 1.6 MHz
Data modulation Binary Offset QAM
Quaternary Offset QAM
Pulse shaping Root Raised Cosine
(roll-off 0.35)
6
Modulation 7
In this section, symbol rates and durations are defined, the mapping of bits onto signal point 8
constellation is shown and the pulse shaping is defined. 9
Symbol rate 10
The symbol rate and symbol duration are shown in Table 6. 11
Table 6 : Summary of symbol rates and durations 12
Symbol rate Symbol duration
136 HS Indoor 2.6 Msymbol/s 0.384 s
13
Mapping of bits onto signal point constellation 14
In 136 HS Indoor the data modulation is either Binary Offset QAM (BOQAM), which is 15
sometimes also referred to as Offset QPSK (OQPSK), or Quaternary Offset QAM 16
(QOQAM), which is sometimes also referred to as Offset 16QAM. 17
Offset QAM may in general be expressed as: 18
( ) ( ) ( ) ( ) ( ) ( ) s t a h t kT t a h t k T t
k
k
c k
k
c

1
]
1
+
1
]
1 + 2 2 1
2 2 1 cos sin
, (3-3) 19
(TR45/98.03.03.19R6)
where
c
=2f
c
, f
c
is the carrier frequency, 1/T is the symbol rate (T=T
b
for Binary Offset 1
QAM and T=2T
b
for Quaternary Offset QAM), a
k
is the kth data symbol taking on values of 2
t1 for Binary Offset QAM and t1 and t3 for Quaternary Offset QAM and h(t) is the impulse 3
response of the shaping filter. The difference between Offset QAM and conventional QAM is 4
the delay of T (half a symbol period for QAM) in the quadrature branch. This time shift 5
prevents zero-crossing signal transitions, as shown in Figure 7 and Figure 8. This improves 6
the Peak-to-Average Power Ratio, which makes Offset QAM more suitable for using with 7
non-linear amplifiers. 8
The complex envelope of an Offset QAM signal is 9
( ) ( ) ( ) ( ) [ ]
( )
u t a h t kT j a h t k T
j a h t kT
k k
k
k
k
k
+ +

+
2 2 1
2
2 2 1
( mod )
(3-4) 10
1
1
0.5
0.5
0
0
-1
-1 -0.5
-0.5
O O
O O
1.5
-1.5
-1.5 1.5
11
Figure 7 Signal point constellation for BOQAM with rectangular pulse shaping (O,
_______
) 12
and GMSK (----). 13

1
1
0.5
0.5
0
0
-1
-1 -0.5
-0.5
O O
O O
O O
O O
O O
O O
O O
O O
14
Figure 8 Signal point constellation for QOQAM with rectangular pulse shaping (O,
________
) 15
Pulse shape filtering 16
In 136 HS Indoor, the pulse shaping filter has square root raised cosine spectrum with 17
impulse response given by: 18

( )
( ) ( )
( )
h t
E
T t T
t T t T t T
t T
+ +
1
]
1
1
2
1
2
1 2 4 2 1 2
1 4 2
2

/
sin / / cos /
/
, (3-3) 19
which is uniquely defined by the roll-off factor . Here, the value 0.35 is chosen for the roll- 20
off factor . E is the energy of the pulse h(t) (usually normalized to 1). The impulse response 21
(TR45/98.03.03.19R6)
h(t) and the energy density spectrum of h(t) with the roll-off factor =0.35 are depicted in 1
Figure 9. 2
0 0.5 1 1.5 2 2.5
-0.2
0
0.2
0.4
0.6
0.8
t/us
h
(
t
)
-5 0 5
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
f/MHz
d
B
3
Figure 9 Binary Offset QAM basic impulse h(t) and the corresponding energy density 4
spectrum of h(t) with the roll-off factor =0.35 for a symbol duration 0.38 s (1/2.6 5
Msymbol/s). 6
Examples of gross bit rates and service mappings 7
This chapter presents the gross bit rates of 136 HS Indoor, some examples how the burst can 8
be used to provide different data rates. 9
Gross bit rates of 136 HS Indoor bursts 10
Table 7: Gross bit rates of different burst types of 136 HS Indoor 11
Slot type Burst type Modulation Gross bit rate
per single slot
(kbit/s)
Total gross bit
rate
(using all slots)
(Mbit/s)
1/64 Short 1 BOQAM 26.4 1.69
1/64 Short 1 QOQAM 52.8 3.38
1/64 Short 2 BOQAM 31.2 2.00
1/64 Short 2 QOQAM 62.4 4.00
1/16 Long BOQAM 148.2 2.37
1/16 Long QOQAM 296.4 4.74
(TR45/98.03.03.19R6)
1
Service mappings with 136 HS Indoor bursts 2
Table 8: Examples of service mappings for 136 HS Indoor 3
Required
user bit
rate
(kbits/s)
Code
rate
Slot type Burst type Modulatio
n
Number of
basic
physical
channels per
frame
8 0.5 1/64 Short 2 BOQAM 0.5
64 0.5 1/64 Short 2 BOQAM 4
144 0.5 1/16 Long BOQAM 2
384 0.5 1/16 Long BOQAM 5
1024 0.5 1/16 Long BOQAM 14
2048 0.5 1/16 Long QOQAM 14
4
(TR45/98.03.03.19R6)
1
Enhancements 2
Utilization of the reciprocal channel for TDD operation 3
In TDD operation the fast fading is the same both in uplink and downlink if the Doppler 4
frequency is sufficiently lower than the frame frequency. This reciprocal channel can be 5
utilized for 6
open loop control 7
transmission diversity 8
Adaptive antennas (AA) 9
The well-known advantages of adaptive antennas system are following 10
improved range 11
improved spectrum efficiency 12
improved quality 13
Without doubt, the amount improvement which AA techniques provide depends on the 14
antenna array complexity. As AA techniques are often more powerful when used in the 15
receiver side, the possibility of implementing a small antenna array in the mobile receiver is 16
an interesting alternative in the 136 HS Indoor system to balance the up- and downlink 17
performances. 18
The 136 HS Indoor system is designed to well support the usage of adaptive antennas. For 19
example, appropriate training sequences with good cross-correlation are selected to optimize 20
the usage of AA techniques for interference cancellation. 21
Interference cancellation by using digital antenna array implemented in the receiver is a very 22
powerful technique against co-channel interference as well as other spatially distributed 23
interferences such as adjacent channel interference. The system requirements for AA-IC- 24
receiver are not as stringent as for the JD-receiver. For example, the AA-IC-receiver does not 25
strictly require synchronous system. A drawback of asynchronism is that the interference is 26
not the same over a burst which leads to gradual performance loss if the interference cannot 27
be estimated separately for the both ends of a burst. In the synchronous system, the 28
optimization of training sequence cross-correlation properties is of key importance. 29
Antenna diversity 30
All capacity results shown in this document are for downlink without antenna diversity. Since 31
antenna diversity could be applied in the uplink, a considerable improvement could be 32
expected compared to these downlink results. 33
Inter-cell interference suppression 34
At least two types of interference cancellation techniques are applicable in 136 HS Indoor 35
system: joint demodulation of co-channel signals and adaptive antennas. Interference 36
cancellation (IC) is the preferred technique which can be implemented in the receiver to 37
suppress the interference which is present in the receiver. This can be done in real time by 38
post-processing the burst in the receiver memory. 39
(TR45/98.03.03.19R6)
The advantages of co-channel inter-cell interference suppression in 136 HS Indoor are as 1
follows 2
improved spectrum efficiency (lower reuse, or saved air interface time) 3
easier cell deployment 4
improved quality 5
make the receivers operation more reliable in the case of sudden interference changes 6
The first bullet refers to the capacity increase resulting from the receivers improved 7
susceptibility with co-channel interference. A way to gain the capacity is to design the 8
network with lower reuse when IC-receivers are used. Another way not requiring lowering 9
the reuse factor, is to take advantage of the saved air time as the IC-receiver copes with less 10
channel coding and/or retransmission. 11
The second bullet implies the possibility to rely on IC as a method to equalize the location 12
dependent interference level changes. An example of such case is a street micro-cellular 13
system where street crossings suffer from a higher interference level. Already, fast link 14
adaptation is specified in 136 HS Indoor system to tackle the same problem, but at the 15
expense of extra radio resources. 16
The third bullet, implies the possibility to provide a better service quality for a user having 17
receiver with IC-capability. For example, a user with IC-receiver may be granted with a 18
higher service bit rate as it can cope with less channel coding. 19
The fourth bullet emphasizes the fact that interference cancellation can be thought as real- 20
time adaptation as it removes the interference which already were present in the receiver. If 21
the receiver can adapt to the sudden interference level changes, it helps avoid call drop outs 22
or allows extra time to perform handover to a new base station or channel. 23
Joint demodulation of co-channel signals 24
In TDMA systems, as the number of nearby co-channel signals is few and the signals have 25
independent propagation paths, there is a high probability for the existence of a dominant 26
interfering signal (DI) in the receiver. The probability of DI is further increased by DTX, 27
fractional loading and cell sectorization. Joint demodulation of desired signal and DI 28
provides substantial interference suppression gain and, moreover, makes it feasible to 29
implement a receiver with reasonable complexity. 30
The receiver can be divided into two parts: joint channel estimator and detector. From the 31
detector complexity and structure point of view, detection of two independent Bin OQAM 32
signals simultaneously is identical to the detection of a single Quat-OQAM signal. Therefore, 33
as Quat-OQAM is supported anyway by a 136 HS Indoor receiver, the support of IC for Bin- 34
OQAM signals does not require major changes in the demodulator. The practical limitations 35
of receiver complexity limits the operational environment of the JD-receiver as for Quat- 36
OQAM modulation 37
To support joint channel estimation and DI identification, base stations need to be 38
synchronized to make training sequences from co-channels overlap with each other and 39
training sequences with good cross-correlation properties need to be used. Base station 40
synchronization is an implementation option, not a requirement. 41
(TR45/98.03.03.19R6)
Improved channel coding 1
The use of different channel coding algorithms need not be restricted. By optimized the 2
channel coding the link levels performance and system capacities can be improved. The 3
following items are examples for improving the channel coding. 4
optimized puncturing 5
Turbo codes 6
longer constraint lengths, now K=5 for convolutional codes and K=3 for Turbo 7
codes 8
Interleaving depth can be adjusted to optimize trade off between bearer C/I 9
requirement and delay 10
Fast power control (frame-by-frame) 11
Fast power control (frame-by-frame) could be used to improve the performance in case where 12
frequency hopping cannot be applied. Such case is e.g. if the operator has only one carrier 13
available. 14
15
(TR45/98.03.03.19R6)
Attachment 13 1
Figures and Tables 2
4
Figure 1. See A1.2.5 5
6
7
8
9
ACP
- 28 dB (back off of 1,5 dB)
- 32 dB (back off of 3 dB)
- 36 dB (ideal PA)
C/I e.g. 5 dB
Co-channel interference Adjacent channel interference
11
12
13
14
Figure 3. See A1.2.18.1 15
16
17
18
19
(TR45/98.03.03.19R6)
1
Table 1. See A1.2.20.1 2
Link adaptation options
Case 1 Case 2 Case 3 Case 4
User bit rate 390 kbit/s
Number of slots used /
frame
3 slots / frame 4 slots / frame 6 slots / frame 8 slots / frame
Slot size 1/16
Modulation Quat-O-QAM Bin-O-QAM
Coding block 1800 for RS encoder/2250 for convolutional encoder
Coding rate 0.44 0.33 0.44 0.33
Outer code Rate 4/5 Reed-
Solomon code
Rate 4/5 Reed-
Solomon code
Rate 4/5 Reed-
Solomon code
Rate 4/5 Reed-
Solomon code
Inner code 1/2-rate CC with
puncturing
1/3-rate CC with
puncturing
1/2-rate CC with
repetition
1/3-rate CC with
repetition
Interleaving depth 30 frames (138 ms)
Type of interleaving block interleaving
Frequency hopping slot-by-slot
Mobile speed 3 km/h
Channel model ITU Indoor A
3
4
5
Figure 4. Table 1. See A1.2.20.1 6
7
8
9
10
11
(TR45/98.03.03.19R6)
Table 2. See A1.2.20.3 1
frame
Slot size 1/16
Modulation Bin-O-QAM
(Quat-O-QAM results are found in figures.)
Coding rate Variable
Basic code 1/2-rate convolutional code
Interleaving depth 4 or 8
Frequency hopping slot-by-slot (need not be consecutive)
Mobile speed 3 km/h
E
b
/N
0
UL/DL - / - 2.1 / 5.1 - / - 2.4 / 4.9
C/I UL/DL - / - -4.6 / -1.8 - / - - / -
2
3
5
6
7
8
(TR45/98.03.03.19R6)
2
3
4
5
Table 3. See A1.2.20.3 6
Case 1 Case 2
frame
8 slots / frame 16 slots / frame
Slot size 1/16
Modulation Quat-O-QAM
(Bin-O-QAM results are found in figures.)
Interleaving depth 8 slots
Mobile speed 3 km/h
E
b
/N
0
UL/DL - / - 4.6 / 7.0
C/I UL/DL - / - 4.2 / 7.6
7
8
9
10
(TR45/98.03.03.19R6)
2
3
4
5
7
8
9
10
(TR45/98.03.03.19R6)
See A1.4.2 1
Table 4.1 Adjacent channel power attenuation with SSPA1
PA
Input backoff (IBO)/dB
0 2 4 6
1. adjacent
channel
-24.8 -30.4 -33.5 -33.0
2. adjacent
channel
-46.8 -50.3 -54.6 -56.0
3. adjacent
channel
-62.0 -68.7 -69.8 -69.8
4. adjacent
channel
-74.1 -75.9 -76.0 -75.8
Table 4.2 Adjacent channel power attenuation with SSPA2
PA
Input backoff (IBO)/dB
0 2 4 6
1. adjacent
channel
-25.0 -30.0 -33.6 -34.1
2. adjacent
channel
-49.7 -53.5 -58.4 -61.5
3. adjacent
channel
-64.8 -70.1 -71.5 -71.3
4. adjacent
channel
-75.7 -76.5 -76.5 -76.3
2
3
4
60 %
40%
85%
70% 40%
60%
65%
75%
80%
70%
60%
30%
Re-use 3 Network
Network with re-use 1 and
fractional loading
6
7
(TR45/98.03.03.19R6)
Attachment 14 1
Performance 2
4
Introduction 5
136-HS consists of a 200 kHz narrow carrier and a 1600 kHz wide carrier. Known as 136 HS 6
Outdoor/Vehicular and 136 HS Indoor, respectively, these components of 136 HS provide a 7
harmonized path for UWC-136 High Speed Data transmission for bit rates up to 2 Mbit/s. 8
The carrier spacing of the 136 and 136+ 30 kHz carrier is optimized for speech and low bit 9
rate transmission and for higher data rates a wider bandwidth is needed. Even if it is possible 10
to support bit rates up to 64 kbit/s with 30 kHz carrier, a more cost efficient solution is to 11
increase the bandwidth. Widening the bandwidth up to 200 kHz makes it possible to support 12
bit rates up to about 384 kbit/s with full coverage and with moderate equalizer complexity in 13
136 HS Outdoor/Vehicular. When higher bit rates up to 2 Mbit/s are needed, 136 HS Indoor 14
is the solution. 136 HS Indoor is targeted especially for micro cellular and indoor 15
environments where high bit rates are required. It is possible to support also low bit rate 16
services in long delay spread environments with 136 HS Indoor but for those environments 17
and bit rates 136 HS Outdoor/Vehicular provides a more cost and spectrum efficient solution. 18
19
This attachment presents link performance and system capacities of 136 HS Indoor. Link 20
budget calculations are presented in appendices 6, 7 and 9. 21
22
Table 1 UWC-136 High speed data 23
136 HS Outdoor/Vehicular 136 HS Indoor
Carrier spacing 200 kHz 1.6 MHz
Bit rates Up to 384 kbit/s Up to 2 Mbit/s
(4 Mbit/s with high C/I)
Target
environments
Vehicular, micro Indoor
Coverage Full coverage Limited coverage
24
136 HS Outdoor/Vehicular and 136 HS Indoor are harmonized to support fast and easy 25
deployment of 136-HS. The following aspects have been considered for co-existence of 136 26
HS Outdoor/Vehicular and 136 HS Indoor. 27
Spectrum co-existence, carrier spacing of 136 HS Indoor a multiple of 200 kHz. Adjacent 28
channel interference between 136 HS Indoor and 136 HS Outdoor/Vehicular allows a 29
narrow guard band. 30
Easy handover between 136 HS Outdoor/Vehicular and 136 HS Indoor due to the same 31
frame length which makes it easy to perform inter-frequency measurement. 32
Easy implementation of dual mode 136 HS Outdoor/Vehicular and 136 HS Indoor 33
terminals with one reference clock and with same modulation schemes. 34
(TR45/98.03.03.19R6)
Link Level Results 1
This document presents link level simulation results of 136 HS Indoor. The results are 2
obtained with COSSAP. Both non-real time service results with ARQ and real time service 3
results with FEC are presented. For Real time (RT) services the interference averaging 4
concept relies heavily on the link adaptation. Therefore a few link adaptation options are 5
presented for RT services. For providing input to system simulations the link level 6
simulations are run against one co-channel interferer and for range calculations against 7
Gaussian noise. These results are valid both for FDD and TDD operation. In TDD operation, 8
however, these results could be improved by utilizing the reciprocal channel for e.g. open 9
loop control and transmission diversity. 10
The results in this document are shown against average(C)/average(I). The interface between 11
link and system level simulations does not directly utilize these curves but the burst-by-burst 12
collected information. For more information about the interface see the following chapter. 13
The basic assumptions and technical choices of the link level simulations are shown in the 14
Table 2 below. 15
136 HS Indoor is proposed for the Indoor Office environment, data services only. However, 16
in order to fully evaluate 136 HS Indoor, simulations have been performed also for some 17
other combinations of environment and services. 18
Table 2 Link level assumptions and technical choices 19
Channel estimator Correlator, delay search window of 3 symbols,
Independent estimation from burst-to-burst
Equalizer Decision Feedback Equalizer (DFE) or
Soft Output Viterbi Algorithm (SOVA)
Number of equalizer taps ITU Indoor A: 3 taps
ITU Outdoor to indoor A: 3 taps
Quat-O-QAM
Channel coding Convolutional codes, K=9, puncturing / repetition for rate matching
Concatenated code for LCD 144 and LCD 384: Reed-Solomon
(500,400) or Reed-Solomon (210,168) + Convolutional code (K=9)
Power control Slow power control, not modelled in link level
Interference modeling One co-channel interferer (for capacity)
Gaussian noise (for range)
Antenna diversity Not used in C/I simulations
(C/I simulations are for downlink capacity)
Used in E
b
/N
0
simulations for uplink range
Frequency hopping Frame-by-frame hopping or slot-by-slot hopping
Uncorrelated frequencies
Time hopping Included in link level frequency hopping for interference diversity, not
separately modelled in link level
20
The simulation results for each service and environment are shown in the following chapters. 21
(TR45/98.03.03.19R6)
Validation of simulation chains 1
The link level COSSAP simulation chains of 136 HS Indoor has been validated by comparing 2
the simulation results in non-fading AWGN channel (Figure 1) and in 1-path Rayleigh fading 3
channel (Figure 2) to the theoretical BER-curves. 4
5
Figure 1 Simulated results compared to
theoretical value in AWGN channel
(overhead due to training sequence and
tail bits is not taken into account in
E
c
/N
0
)
Figure 2 Simulated results compared to
theoretical value in 1-path Rayleigh fading
channel
The simulated results are a little worse than theoretical values as expected. This is due to the 6
non-ideal channel estimation. 7
LCD 384 in ITU Indoor A 8
Four different link adaptation options have been simulated for LCD 384 service in Indoor A 9
environment. Two of them use Bin-O-QAM modulation and 6 and 8 slots per frame, 10
respectively, whereas two other options use Quat-O-QAM modulation and 3 and 4 slots per 11
frame, respectively. The rate 4/5 Reed-Solomon code (500, 400) has been used as outer code 12
and punctured convolutional code as inner code. The bit interleaving is done after the inner 13
code over 30 TDMA frames (138 ms). 14
The C/I curves for LCD 384 in Indoor A are presented in Figure 3. With case 4 (Bin-O- 15
QAM, 8 slots per frame), C/I requirement for BER 10
-6
is 2 dB. In the other end, C/I of 15.5 16
dB is required when 3 slots and Quat-O-QAM modulation are used. 17
18
User bit rate 390 kbit/s
Number of slots
used / frame
Slot size 1/16
(TR45/98.03.03.19R6)
Coding rate 0.44 0.33 0.44 0.33
Solomon code
Rate 4/5 Reed-
Solomon code
Rate 4/5 Reed-
Solomon code
Rate 4/5 Reed-
Solomon code
Inner code 1/2-rate CC
with puncturing
1/3-rate CC with
puncturing
-rate CC with
repetition
1/3-rate CC with
repetition
Interleaving depth 30 frames (138 ms)
Frequency hopping slot-by-slot
Mobile speed 3 km/h
E
b
/N
0
UL/DL
(BER=10
-6
)
- / - - / - - / - - / -
C/I UL/DL (BER=10
-
6
)
- / 15.5 - / 11.5 - / 6.0 - / 2.0
1
2
3
Figure 3 LCD 384 in ITU Indoor A, C/I 4
5
LCD 144 in ITU Outdoor to Indoor A 6
Four different link adaptation cases have been simulated for LCD 144 service. Two of them 7
use Bin-O-QAM modulation and 3 and 4 slots per frame, respectively, whereas two other 8
ones use Quat-O-QAM modulation and 1 and 2 slots per frame, respectively. The rate 4/5 9
Reed-Solomon code (210, 168) has been used as outer code and punctured convolutional 10
code as inner code. Between the outer and the inner code, symbol interleaving has been used. 11
The bit interleaving is done after the inner code over 30 TDMA frames (138 ms). 12
(TR45/98.03.03.19R6)
The C/I curves for LCD 144 are presented in Figure 4 and the E
b
/N
0
curves in Figure 5. 1
Those figures suggest that the tight BER requirement for LCD services (10
-6
) can be achieved 2
with the proper code design. C/I ratios near 0 dB are achievable with the total code rate of 3
0.24. The LCD 144 service can be provided even with one 1/16th slot (Quat-O-QAM 4
modulation, code rate 0.49). E
b
/N
0
values have been derived for cases 2 and 4 that use 2 slots 5
(Bin-O-QAM) and 4 slots (Quat-O-QAM) per frame, respectively. 6
User bit rate 145.6 kbit/s
Number of slots
used / frame
1 slot / frame 2 slots / frame 3 slots / frame 4 slots / frame
Slot size 1/16
Coding rate 0.49 0.24 0.33 0.24
Solomon code
Rate 4/5 Reed-
Solomon code
Rate 4/5 Reed-
Solomon code
Rate 4/5 Reed-
Solomon code
Inner code 1/2-rate CC with
puncturing
1/4-rate CC
with
puncturing
1/3-rate CC
with repetition
1/4-rate CC
with repetition
Interleaving
depth
30 frames (138 ms)
Type of
interleaving
Block interleaving
Frequency
hopping
slot-by-slot
Mobile speed 3 km/h
Channel model ITU Outdoor to Indoor A
E
b
/N
0
UL/DL
(BER=10
-6
)
- / - - / 11.4 - / - - / 6.4
C/I UL/DL
(BER=10
-6
)
- / 22.0 - / 10.0 - / 4.0 - / 1.0
7
(TR45/98.03.03.19R6)
1
Figure 4 LCD 144 BER vs. C/I in ITU
Outdoor to Indoor A, DL
Figure 5 LCD 144 BER vs. E
b
/N
0
in ITU
Outdoor to Indoor A, UL
2
UDD 144 in ITU Indoor A 3
Case 1 Case 2 Case 3
Number of slots used
/ frame
2 slots / frame 4 slots / frame 8 slots / frame
Slot size
1/16
Mobile speed 3 km/h
E
b
/N
0
UL/DL 2.7 / - 2.3 / - - / -
C/I UL/DL - / 2.1 - / -3.8 - / -
4
(TR45/98.03.03.19R6)
1
Figure 6 UDD 144 throughput (kbit/s) vs.
E
b
/N
0
in ITU Indoor A, UL
E
b
/N
0
in ITU Indoor A, DL
2
3
C/I in ITU Indoor A, UL
C/I in ITU Indoor A, DL
4
UDD 144 in ITU Outdoor to Indoor A 5
The simulation parameters of UDD 144 are shown in the table below. The packet services of 6
136 HS Indoor are based on Type II Hybrid ARQ. Two different modulations, Bin-O-QAM 7
and Quat-O-QAM, are used. The basic code has been CC(1,2,9), i.e. 1/2-rate convolutional 8
code with constraint length of 9. Interleaving was over 8 frames and the slot size of 1/16 is 9
used. 10
Case 1 Case 2 Case 3
/ frame
2 slots / frame 4 slots / frame 8 slots / frame
Slot size 1/16
(TR45/98.03.03.19R6)
Mobile speed 3 km/h
E
b
/N
0
UL/DL 2.8 / - 2.1 / - - / -
C/I UL/DL -1.5 / 2.3 -6.2 / -3.7 - / -8.0
1
2
E
b
/N
0
in ITU Outdoor to Indoor A, DL
E
b
/N
0
in ITU Outdoor to Indoor A, UL
3
4
C/I in ITU Outdoor to Indoor A, DL
C/I in ITU Outdoor to Indoor A, UL
5
(TR45/98.03.03.19R6)
/ frame
4 slots / frame 8 slots / frame 12 slots / frame 16 slots /
frame
Slot size 1/16
Interleaving depth 4 or 8
Mobile speed 3 km/h
E
b
/N
0
UL/DL - / - 2.1 / 5.1 - / - 2.4 / 4.9
C/I UL/DL - / - -4.6 / -1.8 - / - - / -
2
3
E
b
/N
0
in ITU Indoor A, DL
E
b
/N
0
in ITU Indoor A, UL
4
(TR45/98.03.03.19R6)
1
C/I in ITU Indoor A, DL
C/I in ITU Indoor A, UL
2
UDD 384 in ITU Outdoor to Indoor A 3
/ frame
4 slots / frame 8 slots / frame 12 slots / frame 16 slots /
frame
Slot size 1/16
Interleaving depth 4 or 8 slots
Mobile speed 3 km/h
E
b
/N
0
UL/DL - / - 2.0 / - - / - 2.3 / -
C/I UL/DL - / - -4.6 / -1.5 - / - - / -6.5
4
(TR45/98.03.03.19R6)
1
E
b
/N
0
in ITU Outdoor to Indoor A, DL
E
b
/N
0
in ITU Outdoor to Indoor A, UL
2
3
C/I in ITU Outdoor to Indoor A, DL
C/I in ITU Outdoor to Indoor A, UL
4
Case 1 Case 2
/ frame
8 slots / frame 16 slots / frame
Slot size 1/16
Modulation Quat-O-QAM
(Bin-O-QAM results are found in figures.)
(TR45/98.03.03.19R6)
Mobile speed 3 km/h
E
b
/N
0
UL/DL - / - 4.6 / 7.0
C/I UL/DL - / - 4.2 / 7.6
1
2
Figure 22 UDD 2048 throughput (kbit/s)
vs. E
b
/N
0
in ITU Indoor A, DL
vs. E
b
/N
0
in ITU Indoor A, UL
3
4
vs. C/I in ITU Indoor A, DL
vs. C/I in ITU Indoor A, UL
In Figure 24 and Figure 25 it is shown how bit rate of 2.048 Mbit/s can be achieved with 136 5
HS Indoor. It is possible to transmit 2 Mbit/s with Bin-O-QAM modulation if at least 14 6
slots/frame are allocated for the bearer and the C/I is high enough. If Quat-O-QAM 7
modulation is used, the 2 Mbit/s can be achieved with 8 slots and by allocating all 16 slots 8
the maximum bit rates exceeds 4 Mbit/s. 9
(TR45/98.03.03.19R6)
1
Effect of frequency hopping on packet services 2
The effect of frequency hopping can be illustrated by the following simulation results. In the 3
sense of the throughput frequency hopping decreases the performance about 0.7 dB. 4
5
Figure 26 ARQ with (FH) and without (FF) frequency hopping. 6
However, throughput is not the only measure for packet service quality. It is also beneficial 7
that ARQ-scheme does not produce very high occasional delay. In this sense, frequency 8
hopping improves the performance and it also helps to average the interference caused to 9
other users. 10
Without frequency hopping there are occasionally longer delays for individual packets. 11
However, because the average delays in these two cases are about the same, there are also 12
low delays in the non-hopping case. 13
In indoor_A channel delays over 10 bursts (~46 ms) were not present if C/I was over 10 dB. 14
With C/I = 0 dB the maximum delay is measured to be 20 bursts (less than 100 ms). Without 15
frequency hopping maximum delays at C/I of 10 dB can exceed 20 bursts and with C/I = 0 16
dB maximum delays can exceed 60 bursts (~ 280 ms). In the non-hopping case the delay 17
distribution has longer tail than in the hopping case. 18
0 20 40 60 80 100 120
10
3
10
2
10
1
10
0
Delay distribution for nonhopping IndoorA
interleaving over 8 burst, FF, CC(1,2,9)
C/I = 5dB(x), 0dB(+), 5dB(o), 10dB(square), 20dB(diamond), 25dB(v)
packet transmission delay / bursts
n
u
m
b
e
r

o
f

e
v
e
n
t
s

/

t
o
t
a
l
n
o
m
b
e
r

o
f

e
v
e
n
t
s
0 10 20 30 40 50 60
10
3
10
2
10
1
10
0
Delay distribution for hopping IndoorA
interleaving over 8 burst, FH, CC(1,2,9)
C/I = 5dB(x), 0dB(+), 5dB(o), 10dB(square), 20dB(diamond), 25dB(v)
packet transmission delay / bursts
n
u
m
b
e
r

o
f

e
v
e
n
t
s

/

t
o
t
a
l
n
o
m
b
e
r

o
f

e
v
e
n
t
s
19
Figure 27 Delay distribution for non- Figure 28 Delay distribution for hopping
(TR45/98.03.03.19R6)
hopping UDD UDD
System Level Results 1
This chapter describes the executed system level simulations and the assumptions used in them. 136 2
HS Indoor is simulated utilizing Interference Averaging (IA) concept. The main features of the IA- 3
concept are frequency hopping on the whole operator bandwidth, time hopping, link adaptation, type 4
II hybrid ARQ and quality based Handover (HO). Frequency Hopping (FH) and Time Hopping (TH) 5
provide interference averaging and Link Adaptation (LA) provides interference diversity which are 6
the cornerstones of the scheme. 7
Basic assumptions 8
Only the downlink direction is considered. It is expected that downlink will be the capacity 9
limiting direction. Advanced receiver techniques such as antenna diversity can be used in the 10
uplink to make its capacity exceed that of the downlink. Additionally, the locations of the 11
interfering transmitters (mobile stations) will also change providing more interference 12
diversity gain through FH and TH in the uplink. Hence the lower limit for performance is 13
presented. 14
In the following the major assumptions that have effect on the performance are presented. 15
General models 16
Test environments, propagation models, mobility models and quality of service (QoS) criteria 17
for Real Time (RT) users is modeled according to [1]. In addition, for packet services, if a 18
user receives all requested bits within 150 ms the user is regarded as a satisfied user. 19
Channel allocation 20
Channels are allocated frame by frame independently in each BTS. RT services are 21
prioritized over NRT services. Channel allocation for RT services is based on first come first 22
served principle. Within NRT services users having difficulties to maintain the minimum bit 23
rate and NRT users with lowest transmitted power per slot are prioritized over the rest of the 24
NRT users. 25
For channel allocation, a channel matrix for each BTS is defined. It contains f t slots, where 26
f is the amount of frequencies available for the BTS and t is the amount of slots in one 136 27
HS Indoor frame. A channel separation of 1.6 MHz is used, thus the maximum size for the 28
channel matrix is nine frequencies and 64 or 16 time slots for 1/64 and 1/16 slots structures 29
respectively. However, to speed up simulations with some simulation cases less frequencies 30
and/or time slots have been used. The amount of used frequencies and time slots in each 31
simulation case are indicated together with the simulation results. Use of less frequencies and 32
time slots (smaller channel matrix) in the simulation than the maximum number available on 33
the band causes less diversity and more blocking, thus this is a pessimistic assumption and 34
worse performance is obtained. 35
Frequency and time hopping 36
Random uncorrelated memoryless frequency hopping is applied. The random hopping 37
introduces interference diversity, uncorrelated means that the channel fading of different 38
frequencies are uncorrelated, memoryless means that each frequency has the same probability 39
(TR45/98.03.03.19R6)
to be chosen regardless of the previously used frequencies. As an example, if nine 1
frequencies are used in a BTS, there is 1/9 probability that two consecutive slots of one 2
particular connection use same frequency. For time hopping there are no restrictions which 3
time slots a terminal can use. User carrier and interfering carrier are slot synchronized but this 4
is not utilized in the simulations, e.g. for interference suppression. 5
Power control 6
Slow pathloss based power control is applied. The dynamic range of the power control is 30 7
dB or less. 8
Handover 9
Simple pathloss based handover with hand over margin of 3 dB is used, i.e., handover is 10
performed to a new base station if the pathloss to the new base station is 3 dB lower than to 11
the serving base station. In indoor simulations, HO to a BTS in a different floor is prohibited. 12
Taking into account very high standard deviation (12 dB) for indoor slow fading, this 13
probably has a negative effect on the capacity. 14
Link Adaptation 15
For RT users quality based link adaptation is used. The channel coding rate is increased when 16
the connection experiences a bad frame and if the serving BTS has extra radio resources 17
available. Amount of channel coding will be decreased if less channel coding would have 18
continuously been sufficient for certain amount of frames. If needed, LA can be done at the 19
beginning of an interleaving period. 20
For NRT the link adaptation is performed with type II hybrid ARQ, thus the coding rate 21
depends on the amount of retransmission. The modulation is kept constant over the whole 22
simulation. 23
Type II hybrid ARQ 24
In the ARQ scheme used, user data is coded with -rate convolution code and interleaved 25
over 4 bursts. The interleaving is done in a such way that decoding is possible after two of 26
four bursts have been received. Thus the effective coding rate is one. If the decoding is not 27
successful, a third burst is sent and decoding is redone. After third burst the coding rate is 28
2/3. If the decoding is still not successful fourth burst is sent and decoding is done again, now 29
with the coding rate of . If the decoding is still not successful, the burst with the lowest 30
C/(I+N) value is resent and the original burst and the retransmitted burst are combined by 31
adding their C/(I+N) values (maximal ratio combining). This repetition coding is repeated 32
until the decoding is successful. 33
DTX 34
Radio resources of speech user are released during the DTX period. If there are no resources 35
available after the DTX period bit error rate of 0.5 is applied until resources are allocated or 36
call is dropped. 37
Modulation adaptation 38
Simple modulation adaptation algorithm is used for UDD services. Preferred modulation is 39
Q-O-QAM but B-O-QAM is used for small packets. 40
(TR45/98.03.03.19R6)
Frequency Planning 1
A characteristic feature of the simulated IA concept is very simple frequency planning with 2
frequency re-use 1. Thus all the frequencies can be used at each base station. Through 3
fractional loading the load in each cell is always less than 100%. 4
Interface between system and link level simulation 5
Link level simulation and system level simulations are connected to each other by using the 6
actual value interface. The actual value interface makes it possible to simulate fast radio 7
resource algorithms on the system level. The actual value interface also increases the 8
simulation realism considerably. 9
The actual value interface is a novel way to connect link and system level simulations. In the 10
actual value approach the link level simulation data e.g. bit errors are measured for every 11
burst or frame. This means however, that possible de-interleaving and decoding has to be 12
considered on the system level. The performance of receiver algorithms, such as decoding, 13
are not measured or analyzed in the system level simulations, but their performance is 14
considered on the link level. The effect of receiver algorithms is seen in the link level results 15
that are given as inputs to the system level simulation. 16
If the system level simulation is done with an actual value interface, fast fading has to be 17
taken into account in addition to slow fading and pathloss. With the average value interface 18
fast fading was neglected, since the input from the link level and correspondingly the channel 19
was averaged over a long period (and the fast fading characteristics were included on the link 20
level). With the actual value interface, non-averaged link level simulations are used. Thus 21
fast fading has to be considered on the system level. The same kind of correlated fading 22
process is used both in link- and system level simulations, modeled as Rayleigh, with a 23
certain Doppler frequency. 24
The strength of an actual value interface, compared to the average value interface, is that all 25
the radio resource management algorithms can be simulated on the system level accurately 26
since the simulation resolution is as accurate as the resolution of the mentioned algorithms. 27
The actual value interface enables possible gain or loss of frequency hopping, ARQ and link 28
adaptation algorithms. If frequency hopping was simulated on the link level and the average 29
value interface was used, interference diversity gain could not be modeled. This is because in 30
the link level simulations only interfering user(s) operating with the same frequency as the 31
observed user can be considered. Correspondingly, if ARQ was simulated on the link level, 32
the varying interference conditions that have their impacts to ARQ performance could not be 33
taken into account. Link adaptation can be simulated thoroughly only on the system level 34
since adaptation decisions depend only on the changing interference conditions. If the link 35
adaptation algorithm was fast the average value interface could not be used. 36
37
(TR45/98.03.03.19R6)
For each slot
interleaving
Measure
C/I vs. raw
Take average
raw
Raw BER vs. code
1
Figure 29 Block diagram of interface for RT bearers 2
In the method presented here, link level simulation results are collected on a burst-by-bust 3
basis, i.e. C/I and BER values are collected for each burst and coded BER/FER values for 4
each interleaving period. E.g. if the target service has interleaving over 4 bursts, then C/I 5
values for 4 bursts are observed directly from the fading channel and BER/FER of the coding 6
block is measured over the interleaving period. In the link level simulation raw BER (BER 7
before decoding and de-interleaving) versus C/I ratio is measured for each burst within the 8
interleaving block. In the system level the C/I ratio is measured for each burst within the 9
interleaving block and is mapped to raw BER by using a raw BER vs. C/I curve from the link 10
level. De-interleaving is modeled so that the average raw BER within the interleaving block 11
is calculated. Further, decoding is modeled by mapping the de-interleaved raw BER to the 12
coded BER/FER by using the measured mean raw BER vs. coded BER/FER curve. The actual 13
value interface for 136 HS Indoor is depicted in Figure 61. 14
System level performance results 15
Simulation results are collected to Table 3. Detailed description of the simulations carried out 16
are presented in the following chapters. 17
Table 3 Summary of the system simulation results 18
Deployment model Service Capacity
[Mbit/s/MHz/cell]
Notes
Indoor UDD2048 .332
Indoor with walls UDD2048 .743
Wall attenuation 5 dB
Outdoor to indoor
and pedestrian
UDD384 .811
Isolated cell (high
C/I)
UDD > 2.000 Single cell capacity
19
Performance with Indoor deployment model 20
UDD2048 21
The simulated spectrum efficiency is .332 Mbps/MHz/cell. Of all sessions 98.4 % fulfill 22
quality criteria. If all bits user requests during a session are received correctly within 150 ms, 23
the user is regarded as satisfied user independent of active bit rate. Insufficient active session 24
throughput is the limiting reason for most sessions not fulfilling the performance criteria. 25
(TR45/98.03.03.19R6)
The average active session throughput is 713 kbps and 0.4 % of the sessions have active 1
session throughput 2048 kbps or higher. If offered load is reduced the average active session 2
throughput and ratio of the users having active session throughput 2048 kbps or higher 3
increases. Reuse 1 is used with fractional load of 48%. The distribution of active session 4
throughputs can be seen in Figure 30. In Figure 31 histogram for needed transmissions per a 5
hybrid II ARQ packet is presented. 6
With Indoor UDD2048 simulations the channel matrix was 16 time slots x 9 frequencies. The 7
preferred modulation is Q-O-QAM, but the modulation is switched to B-O-QAM if offered 8
traffic does not require Q-O-QAM. If the BTS buffer has less bits that can be transmitted in a 9
one radio packet, the radio packet is filled with dummy bits. These overhead dummy bits are 10
not included into spectrum efficiency figure nor into active session throughput. The dynamic 11
range of the power control is 30 dB. 12
0 500 1000 1500 2000 2500 3000
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
000 bit/s
p
d
f
13
Figure 30. Histogram for session throughputs. 14
(TR45/98.03.03.19R6)
2 4 6 8 10 12 14 16 18 20
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
1
Figure 31. Histogram for needed transmissions per hybrid II ARQ packet. Both Q-O-QAM 2
and B-O-QAM packet are included. 3
Performance with Indoor deployment model with walls 4
UDD2048 5
Indoor model defined in [1] has hardly any isolation between cells on the same floor. 6
Therefore additional simulations with walls were carried out. UDD2048 simulation with 7
walls is otherwise identical to UDD2048 simulation without walls presented in this paper, 8
expect it has additional wall attenuation of 5 dB and standard deviation of slow fading is 9
reduced from 12 dB to 4 dB. 10
The simulated spectrum efficiency is .743 Mbps/MHz/cell. Of all sessions 98.7 % fulfill 11
quality criteria. If all bits user requests during a session are received correctly within 150 ms, 12
the user is regarded as satisfied user independent of active bit rate. Insufficient active session 13
throughput is the limiting reason for most sessions not fulfilling the performance criteria. 14
The average active session throughput is 692 kbps and 0.4 % of the sessions have active 15
session throughput 2048 kbps or higher. If offered load is reduced the average active session 16
throughput and ratio of the users having active session throughput 2048 kbps or higher 17
increases. Reuse 1 is used with fractional load of 86 %. The distribution of active session 18
throughputs can be seen in Figure 32. In Figure 33 histogram for needed transmissions per a 19
hybrid II ARQ packet is presented. 20
With pico cell UDD2048 simulations the channel matrix was 16 time slots x 8 frequencies. 21
The preferred modulation is Q-O-QAM, but the modulation is switched to B-O-QAM if offer 22
traffic does not require Q-O-QAM. If the BTS buffer has less bits that can be transmitted in a 23
one radio packet, the radio packet is filled with dummy bits. These overhead dummy bits are 24
not included into spectrum efficiency nor into active session throughput. The dynamic range 25
of the power control is 30 dB. 26
(TR45/98.03.03.19R6)
0 500 1000 1500 2000 2500 3000
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
000 bit/s
p
d
f
1
Figure 32. Histogram for session throughputs. 2
3
2 4 6 8 10 12 14 16 18 20
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
4
Performance with Outdoor to Indoor and Pedestrian deployment model 7
UDD384 8
The simulated Spectrum efficiency is .811 Mbps/MHz/cell. Of all sessions 99.7% fulfill all 9
quality criteria. Dropping is the major reason for not fulfilling all quality criteria, the bad 10
quality criterion has not any practical effects. 11
(TR45/98.03.03.19R6)
The average active session throughput is 405 kbps and 46 % of the sessions have active 1
session throughput 384 kbps or higher. The respective ratios for 512 kbps, 1Mbps and 2Mbps 2
are 24%, 3.1% and 0.25%. Thus higher bit rates are also possible in outdoor to indoor / 3
pedestrian environment. The distribution of active session throughputs can be seen in the 4
figure Figure 34. Reuse 1 is used with fractional load of 79%. In the Figure 35 a histogram 5
for needed transmissions per a hybrid II ARQ packet is presented. 6
With the micro cell UDD384 model simulations the channel matrix is 16 time slots x 5 7
frequencies due to simulation complexity reasons (with assumption of 15 MHz spectrum 9 8
frequencies could have been used). Thus channel diversity gain and specially the statistical 9
multiplexing gain is reduced. The preferred modulation is Q-O-QAM, but the modulation is 10
switched to B-O-QAM if offer traffic does not require Q-O-QAM. If less bits that can be 11
transmitted in one radio packet are sent, the radio packet is filled with dummy bits. These 12
overhead dummy bits are not included into spectrum efficiency nor into active session 13
throughput. The dynamic range for the power control is 30 dB. 14
0 200 400 600 800 1000 1200 1400 1600 1800 2000
0
0.005
0.01
0.015
0.02
0.025
0.03
000 bit/s
p
d
f
15
Figure 34. Histogram for session throughputs 16
(TR45/98.03.03.19R6)
2 4 6 8 10 12 14
0
0.1
0.2
0.3
0.4
0.5
1
Discussion 4
System simulation results for 136 HS Indoor are shown. The radio resource management 5
scheme used in these simulations is based on the interference averaging principle. In order to 6
be able to study the effects of fast algorithms such as ARQ and Frequency hopping, the actual 7
value interface between link and system level is implemented. 8
The results represent a lower bound case in several aspects: only downlink results are 9
presented, all the possible enhancements by the interference averaging radio resource 10
management algorithms are not implemented and network parameter optimization is not 11
completed. 12
Issues for further studies include several items. Among these are interference cancellation, 13
antenna diversity, application of ARQ to LCD data, better coding schemes (e.g. increased 14
constraint length, optimal puncturing and concatenated codes), optimized mapping of user 15
data into packets, fast power control and channel allocation to cells. 16
Future work will include the estimation of signaling overhead and degradation of the 17
obtained capacity due to signaling. The estimated signaling load is less than 10 % of the 18
capacity. This load has not been subtracted from the presented capacity figures. 19
References 20
21
[1] ITU-R Doc.8-1/INFO/2, Seattle, January 19 23, 1998
22
(TR45/98.03.03.19R6)
ATTACHMENT 15 1
2
Description of the connection oriented mode of operation 3
4
Connection oriented mode, or circuit switched service as it is also called, is mainly used to 5
provide constant predefined bit rate services. Bit error rate is let to vary according to channel 6
and network load conditions. This attachment shows performance of a mode that is used to 7
provide 144 kbit/s low delay service. In this case, service is provided with 8 PSK modulation 8
and by using five out of eight time slots per frame. Single slot bit rate is 29 kbit/s 9
(approximating 28.8 kbit/s), thus by sending 5 timeslots per each TDMA frame a data rate of 10
145 kbit/s (144 kbit/s) can be achieved. Figure 1 shows one possible time slot arrangement. 11
12
Figure 1 144 kbit/s service using five 28.8 kbt/s slots per one TDMA frame 13
14
Connection oriented mode can be used as such to provide transparent constant bit rate 15
variable bit error rate service, or a higher layer link protocol with error checking and 16
retransmission mechcanisms can be used to provide connection oriented non-transparent 17
constant bit error rate variable delay service. These modes should not be confused with 18
packet radio, which can provide connectionless data services with or without retransmission. 19
Description of the channel coding scheme 20
21
Transparent mode relies on forward error correction (FEC) and therefore a robust channel 22
coding is typically required. By using a set of concatenated codes, a Reed-Solomon code and 23
a convolutional code, one can improve FEC performance especially at good signal to noise 24
ratios. 25
26
How the code rate for 144 kbit/s service is selected is described in the table below 27
28
29
30
Time Slots Uncoded bit rate Channel coding User data rate
1 68.4 kbit/s 29 kbit/s
5 342 kbit/s 0.42 145 kbit/s
8 547.2 kbit/s 232 kbit/s
31
28.8 k 28.8 k 28.8 k 28.8 k 28.8 k
slot
TDMA frame
(TR45/98.03.03.19R6)
A detailed description for the channel coding is provided in scheme format in figure 2 below. 1
It shows the selected code rates and channel coding process with its related functions. 2
Convolutional
Encoder (1/2 rate)
Source
Reed Solomon
Encoder (36,32)
Puncturing
(84 bits)
Interleaving
19 diag
8-PSK
Modulator
Channel
8-PSK
Demodulator
580
Repetition bits
60
640 720
Flush bits
6
726 1452 1368
De-interleaving
19 diag
De-puncturing
(84 bits)
Convolutional
Decoder (1/2 rate)
Reed Solomon
Decoder (36,32)
Framing + +
Reconstruction
from repetition
bits
-
Deframing
1F = 148s = 444 bits
456s
1368
580
640 720 726 1452 1368 1F = 148s = 444 bits
456s
1368
Recovered data Flush bits
6
Tail Data Flag Flag Data Tail Training Sequence
3s 57s 1s 26s 1s 57s 3s
148s
+
-
Note:
- multiplexer
- demultiplexer
s - symbols
All numbers indicate bits except as noted
Burst structure
F - frame
3
Figure 2 Forward error correction scheme for the connection oriented mode 4
5
6
7
Performance of the connection oriented mode 8
9
The connection oriented mode performance was studied with following parameters: 10
11
Carrier frequency: 1900 MHz 12
Mobile speed: 3 kph or 120 kph 13
Frequency hopping: Ideal or none 14
Interference: Noise (SNR) or single co-channel user (C/I) 15
Channel models: Vehicular A and Vehicular B 16
No Mobile Station Diversity 17
18
Ideal frequency hopping means that each TDMA burst is sent over channel that is 19
uncorrelated with the channel of any previous or following burst, i.e., a new channel is 20
initilaized for each burst. Multipath propagation is modelled as given by the channel model 21
Vehicular A or Vehicular B. Non-hopping case means that fading is modelled so that 22
correlation between consecutive frames is taken into account. Frequency hopping is one 23
method to introduce diversity into transmission path. 24
25
(TR45/98.03.03.19R6)
Carrier to interference ratio is calculated as average carrier power over average interference 1
power. Interference is modelled as a single user and the interference multipath channel and 2
frequency hopping modelling is the same as for the carrier. In case of hopping, new channel 3
is initilized for each burst, but the interferer is present for each burst. 4
5
6
7
Figure 3 Performance as function of signal to noise 8
9
10
11
Figure 4. Performance as function of carrier to interference 12
13
Case I: BER performance for Vehicular A
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
6 8 10 12 14 16 18 20
Eb/No dB
B
E
R
120 km/h, FH on
120 km/h, FH of f
3 km/h, FH on
3 km/h, FH of f
Case II: BER performance for Vehicular A
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
8 10 12 14 16 18 20 22 24
C/I dB
B
E
R
120 km/h, FH on
120 km/h, FH of f
3 km/h, FH on
3 km/h, FH of f
(TR45/98.03.03.19R6)
Figure 5: Performance as a function of signal to noise 1
2
Case III: BER performance for Vehicular B
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
6 8 10 12 14 16 18 20
Eb/No dB
B
E
R
120 Km/h, FH on
120 Km/h, FH of f
3 Km/h, FH on
3 Km/h, FH of f
(TR45/98.03.03.19R6)
ATTACHMENT 16
Deployment model result matrix
Table 1
Input assumptions
Test environment Pedestrian/Outdoor 3 km/hr
Test service 384 kbit/s Packet Data
Base station antenna height (m) 15 Meters above average rooftop
Any other assumptions made by the
proponent (e.g. antenna pattern,
sectorization etc.)
Three sector sites, Directional antennas (pd1132 antenna), Simple Equalizer
Deployment results
Total number
of cell sites
Total number
of RF channels
Number of voice
channels per RF
channel
Coverage efficiency
(km
2
/site)
Spectrum efficiency (Mbit/s/MHz/site) for
data
27 75 N/A 0.0899 M.1225
1.347 Optimized
0.9495 (0.3165 per sector) A
0.9318 (0.3106 per sector) B
Table 2
Input assumptions
Test environment Vehicular 50 km/hr
sectorization etc.)
Three sector sites, Directional antennas (pd1132 antenna) Simple Equalizer
Deployment results
Total number
of cell sites
Total number
of RF channels
Number of voice
channels per RF
channel
Coverage efficiency
(km
2
/site)
data
27 75 N/A 11.63 M.1225
32.51 Optimized
1.176 (0.392 per sector) A
0.820 (0.2733 per sector) B
(TR45/98.03.03.19R6)
Table 3
Input assumptions
sectorization etc.)
Deployment results
Total number
of cell sites
Total number
of RF channels
Number of voice
channels per RF
channel
Coverage efficiency
(km
2
/site)
data
27 75 N/A 6.619 M.1225
17.566 Optimized
1.038 (0.346 per sector) A
0.7682 (0.256 per sector) B
Table 4
Input assumptions
Test environment Pedestrian/Outdoor 3 km/hr
sectorization etc.)
Deployment results
Total number
of cell sites
Total number
of RF channels
Number of voice
channels per RF
channel
Coverage efficiency
(km
2
/site)
1
data
27 75 N/A 0.1797 M.1225
2.687 Optimized
1.554 (0.5147 per sector) A
1.533 (0.511 per sector) B

1
Uses 144 kbit/s PCS-1
(TR45/98.03.03.19R6)
Table 5
Input assumptions
sectorization etc.)
Deployment results
Total number
of cell sites
Total number
of RF channels
Number of voice
channels per RF
channel
Coverage efficiency
(km
2
/site)
1
data
27 75 N/A 28.555 M.1225
75.8 Optimized
1.422 (0.474 per sector) A
1.017 (0.339 per sector) B
Table 6
Input assumptions
sectorization etc.)
Deployment results
Total number
of cell sites
Total number
of RF channels
Number of voice
channels per RF
channel
Coverage efficiency
(km
2
/site)
1
data
27 75 N/A 27.295 M.1225
72.44 Optimized
1.176 (0.392 per sector) A
0.894 (0.298 per sector) B
(TR45/98.03.03.19R6)
Attachment: 17
UWC-136 Voice Capacity Enhancements
This section outlines the voice capacity per site of UWC-136 using a number of link quality
enhancements. It is important to note however, that comparing systems using only the capacity per
site (Erl/MHz/site) measure does not provide the most economical and deployable solution for
delivering high capacity in most cases. In this respect, an alternative and equally valid measure that
should be considered is (Erl/MHz/km
2
), coupled with the cost to deploy a system over a given area.
In this respect, the hierarchical cell nature of UWC-136 provides an unlimited capacity in the
Erl/MHz/km
2
measure via macrocells and underlying microcells which can meet any high capacity
needs.
Nevertheless, this section provides the techniques and calculations to illustrate how large individual
cell voice capacity for UWC-136 can also be obtained. To accomplish this, the Erlang-B tables are
first used to benchmark the capacity of the system without any link quality enhancements.
Afterwards, the individual link quality enhancements are presented with respect to their individual
gain and benefit. Finally, two techniques are described to estimate the capacity. The first method
simulates a very simple non-optimized channel assignment algorithm, and thus represents a lower
bound capacity calculation. The second method, views the link quality gains in terms of their
incremental benefit, and considers which methods can be assumed additive to obtain an overall
system benefit and capacity. This technique represents an upper bound on capacity. The actual
capacity lies somewhere between these two capacity figures (Eq.s (2) & (4)) depending on the
particular infrastructures implementation.
1. Baseline Capacity
Typical RF planning for IS-136 is to utilize a 7/21 reuse pattern which results in a C/I>17 dB 90% of the
time. The capacity with no link quality improvements is thus:
#RF channels/sector = 15000 kHz/(30*21) = 23.8
#voice paths/sector = 23.8*3=71.4
Assuming one uses one voice path for the DCCH yields 70.4 (~70) voice paths/sector, so that at 1%
blocking this gives 56.1 Erlangs, or
3.74 Erlangs/MHz/sector or 11.22 Erl/MHz/site
At 2% blocking this gives 59.1 Erlangs, or
(TR45/98.03.03.19R6)
3.94 Erlangs/MHz/sector, or 11.82 Erl/MHz/site
2. Link Quality Improvement Techniques
2.1 Capacity Simulations Adaptive Channel Allocation and Power Control

The first set of simulations was presented in the paper Capacity Enhancements in a
TDMA System, VTC 93, pp 277 280. In these simulations, a standard 7/21 Fixed
Channel Allocation (FCA) system is compared with one that utilizes sophisticated Adaptive
Channel Allocation (ACA) as well as power control. The following assumptions were made
in the simulations:

Log-normal fading with standard deviation of 6 dB
Path loss according to L=C+35log(d)
Traffic distribution: homogeneous over simulated area
Site-site distance 5.2 km (3.2 miles).
A large number of cells (153) were simulated. Information was only collected from
the central part to avoid border effects.
Three-sector sites were used.
Number of voice channels in system: 189
Available voice channels per cell:
FCA system: 9 (3 carriers)
ACA system: 189 (all 63 carriers)
Offered traffic (Erlang/sector/MHz):
FCA system -- 2.9
ACA system -- 5.8
Power Control in ACA system: Quality based, on a per carrier level, quantized in
steps of 1dB for the forward link and 4 dB for the reverse link.
The channel selection in the ACA system was made by combining three different
criteria; courtesy to other users, creation of individual and stable channel reuse and
maintenance of sufficient quality. See the paper mentioned above for details.

The results of the simulations are summarized in Table 1.

System Blocking Lost Calls Outage at 17dB
C/I
FCA 5% 0.8% 11%
ACA 4% 0.9% 9%
Table 1 Summary of system simulation results with fast ACA.

Blocking:
FCA system: 5%
ACA system: 4%
(TR45/98.03.03.19R6)
Lost calls:
FCA system: 0.8%
ACA system: 0.9%
Outage at 17dB C/I:
FCA system: 11%
ACA system: 9%

Figure 1 and 2 show the CDFs of the C/I distributions for the FCA and ACA systems
respectively. Note that although the ACA systems carries twice as much traffic, quality
equal to that of the FCA system is obtained by the efficient channel allocation and power
control. The steeper CDF for the ACA system signifies a more fair distribution of the
resources, which enables a larger number of users without increasing the amount of
outage or lost calls. It should also be noted that in these simulations the power control was
on a per carrier level. One should thus expect that even higher capacity gains can be
achieved if power control on a per time-slot level was introduced.

Figure 1. FCA system

(TR45/98.03.03.19R6)

Figure 2. ACA system

For the second set of simulations, base station power control was applied on a per time-
slot basis, whereas the ACA employed was simpler than that used above. The following
assumptions were made in the simulations:

Log-normal fading with standard deviation of 8 dB
Path loss according to L=C+35log(d)
Site-site distance 3 km
Three-sector sites
Mobile speed 50 km/h
Available frequency band 126 duplex frequencies
No quality degradation by timeslot power control when equalizing
Adaptive Channel Allocation: The channel with lowest uplink interference is
selected.
Base Station Power Control: The power level of each timeslot is set each second
based on measured downlink signal strength and BER-class.
Since the downlink is the limiting from a quality and capacity point of view, results on
uplink C/I are not included below. A 7/21 planned network without the new functions is
assumed to have sufficient quality and serves as a reference.
When adding the new functions the frequency reuse can be decreased from 21 to 12,
which together with the increased trunking efficiency provides roughly twice the original
capacity. This is achieved without compromising downlink quality as can be seen in the
table and figure below. It should be noted that the ACA resource management is
considerably simpler than in the above simulation, and thus that a further increase in
capacity should be possible by combining these two approaches.
(TR45/98.03.03.19R6)
Functionality Fixed base station power
and
FIFO channel selection
Timeslot base station power
control and adaptive channel
allocation
Frequency reuse 7/21 4/12
Frequencies/cell 6 10
Blocking 0.8 % 1.9 %
Carried traffic 10.1 Erlang/cell 20.7 Erlang/Cell
Downlink C/I 10%
level
15.9 dB 16.1 dB
Table 2 Summary of system simulation results with fast ACA.
6 8 10 12 14 16 18 20 22 24 26
1

3

10

30

100
C
.
D
.
F
.

[
%
]
Downlink C/I [dB]
Reuse
7/21
4/12
Figure 3. CDF for Downlink C/I
2.2 Transmit Diversity
In this form of diversity, the base station transmits multiple versions of the signal from multiple antennas. By
using antennas of different polarizations at the base station, the fading on the signals from two antennas can
be made relatively uncorrelated without significant spatial separation. Various forms of transmit diversity
include:
a. Delay diversity, wherein the antenna signals are delayed with respect to each other
and the terminal uses an equalizer to recover the information
b. Intentional frequency offset diversity, wherein an intentional frequency offset is applied
to the second antenna to emulate the effect of fast fading so as to improve the
performance of channel coding
(TR45/98.03.03.19R6)
c. The use of codes across antennas (Space-Time Codes) that are designed to provide
significant performance advantage with multiple antenna transmission.
In the table below, we show the performance gain that can be obtained using delay diversity in IS-136. The
results shown are for a practical receiver in a non-dispersive Rayleigh fading channel. The gains shown are
with respect to a system that does not employ transmit diversity. The fading on the two transmit antennas is
assumed uncorrelated, and the relative delay between the two antenna signals is 41.2 s.
Doppler freq. Diversity Gain at 1%
BER
7 Hz 4.5 dB
184 Hz 4.9 dB
Table: Performance of Transmit Diversity
2.3 Receive Antenna Diversity at Terminal
By using multiple receive antennas at the terminal and associated signal processing algorithms, a significant
performance advantage can be obtained. Antenna diversity at the terminal can be implemented in two forms:
2.3.1 Pre-Selection Diversity:
In this form, there are two receive antennas at the terminal, but only one receive chain. The terminal
measures the signal on both antennas before the timeslot of interest and makes a decision on the best
antenna to receive the slot. The transmit/receive timing defined in IS-136 allows sufficient time for such
measurement before the slot of interest. Since the assumption is that the fading does not change much over
the slot, pre-selection diversity provides noticeable gains only at low vehicle speeds, however, it is under
these channel conditions that channel coding does not work well and the diversity gain is needed. At high
vehicle speeds, the interleaving makes the channel coding present in IS-136 perform quite well. Also, in
interference-limited environments, the performance gain of pre-selection diversity can be made relatively
insensitive to the gain differential exists between the two antennas, which might be the case in a practical
implementation.
The gains of pre-selection diveristy with respect to the case with no diversity are shown below. The gains
were obtained by simulating a practical receiver in a non-dispersive Rayleigh fading environment. The raw
BER, as shown in the table is the BER of the modem before error correction has been applied.
Doppler freq. Antennna
Correlation
Diversity gain at 1% Raw
BER
7 Hz 0.7 5.8 dB
184 Hz 0.7 1.2 dB
Table: Performance of Pre-Selection Diversity
2.3.2 Combining Diversity:
In this form, the terminal has two receive antennas and two receive chains. The signals from both antennas
are received and combined at baseband. Such baseband combining allows the use of powerful interference
rejection techniques that can cancel a dominant co-channel interferer. Such processing can also be made
insensitive to the gain differential between the antennas in an interference-limited environment. The
performance gain of a practical diversity receiver employing interference rejection with respect to a receiver
without diversity is shown in the tables below. The correlation coefficient between the two antennas is 0.7.
(TR45/98.03.03.19R6)
When only one co-channel interferer is present, it is seen that the performance is quite good, but there is a
degradation when more interferers are present. Similar gains are obtained when combining diversity with
interference rejection is used at the base station.
Doppler freq.. Gain wrt 1-antenna receiver at 3% Raw BER
7 Hz > 10 dB
184 Hz 8 dB
Table 1: Performance of combining diversity with interference rejection and one co-channel
interferer
Doppler freq. Gain wrt 1-antenna receiver at 3% Raw BER
7 Hz 8 dB
184 Hz 5 dB
Table 2: Performance of combining diversity with interference rejection and two co-channel
interferers
2.4 Improved Error Correction (ACELP/ Channel Coding 2 (CC2))
TIA TR45.3 has recently adopted improved forward error protection to improve the DQPSK
downlink with the IS-641 ACELP vocoder. This methods adds 2 dB improvement in the 1% FER.
2.5 Voice Activity Detection
Voice activity detection is the process in which the mobile station turns off its transmitter in
between talk spurts. On average this benefits about a 2 dB less interference on the
uplink.
2.6 8kbps coder over 8PSK (IS-641 over 8PSK)
When the 7.4 kbps IS-641 ACELP vocoder is implemented on 8PSK such that two users can share a time
slot on the downlink, the 1% FER can be reached at 21 dB C/N or 16.23 dB Eb/No. As such, this can be
implemented as an underlay when deployed in conjunction with cells planned for the 4-ary modulations.
This doubles the capacity for the users in the 8PSK underlay. In this implementation, a 3 dB handoff
hysteresis is assumed.
3. Voice Capacity in UWC-136, Method 1:
To model the capacity using the above link quality or system improvements, a simulation was run
with the following assumptions:
Log-normal shadowing with standard deviation of 8 dB
Path-loss according to L=C+35*log(d)
Two component lognormal fading model added together for each mobile station (with equal
weight); one base station specific and one common for all base stations
Traffic distribution: homogeneous over simulated area
Site-site distance: 3 km (1.9 miles)
(TR45/98.03.03.19R6)
Three-sector sites were used.
Number of frequencies in the system: 498 i.e., 1494 voice channels (Roughly 15 MHz)
Channel allocation: A 1/3 frequency plan was used together with ACA (least interfered
channel). The reason for the underlying 1/3 frequency reuse plan was to help the ACA
algorithm. (The alternative would have been to have every channel available in every sector of
the system and let ACA alone handle the channel allocation).
Available voice channels per cell: 498
Power control: Quality based on a time-slot basis for the forward link, and signal strength based
for the reverse link.
Link quality improvement assumption: Downlink improvement assumptions (+10 dB): Tx
diversity + 2 dB
7
, MS antenna diversity + 6 dB, ACELP/CC2 + 2 dB. Uplink (+10 dB):
Interference rejection combing + 8 dB, voice activity detection + 2 dB. Provided the original
design of 17 dB 90% of the time is assumed, the link quality improvements allow the system to
operate at 17-10 dB or 7 dB C/I.
Quality target for 10% level of C/I for up- and downlink: 7 dB.

Resulting capacity of the system:

12.1 Erlangs/MHz/sector, or 36.3 Erl/MHz/cell (3 sectors) (1)

Due to the very large number of channels in each sector, the simulated system was not blocking
limited. (The traffic per sector was roughly 182 Erlang, and the number of channels available, 498,
resulted in zero blocking). Thus, the 7 dB quality criterion mentioned above determined the
capacity of the system. The limiting link in these simulations was the uplink, due to the two-
component lognormal fading which favors the downlink quality. The resulting performance curves
are shown in Fig. 1.

7
In conjunction with MS antenna diversity, Tx diversity gains are less ~2 dB.
(TR45/98.03.03.19R6)

1
10
100
-10 0 10 20 30 40
C/I (dB)
C
D
F

(
%
)
C/I uplink
C/I downlink

Figure 36: CDF C/I probability for ACA + power control with quality margin of 7 dB

Underlay gains with IS-641 over 8PSK:

The link quality improvement benefits for IS-641 over 8PSK are +6 dB due to mobile station
antenna diversity (Tx diversity is not assumed to be used with 8PSK). Therefore, the
operating margin with handoff of 21+3=24 dB is reduced to 24-6 = 18 dB. From Fig. 1, the C/I
is above 18 dB ~62% of the time so that the underlay results in a 1.62X increase in capacity or

12.1* 1.62= 19.6 Erl/MHz/sector, 58.8 Erl/Mhz/cell (3 sectors) (2)

It is important to note that the ACA simulator used in this simulation was one originally designed for a
quality criterion of 17 dB in a 4/12 reuse pattern, and hence was not optimized for this scenario. Significant
improvements are expected if quality based power control for the uplink and more optimized ACA
algorithms are utilized. As such, we believe the results in Eq.s (1) and (2) to represent a lower bound on
voice capacity with UWC-136. An upper bound using the same link quality improvements is given in the
next section.
4. Voice Capacity in UWC-136, Method 2:
The spectrum efficiency calculations are based upon the load conditions specified for each test
environment in M.1225 Section 1.3.2. The spectrum efficiency numbers in this section have been
calculated based upon the following assumptions:

The number of 30 kHz channels in the system = 15 MHz/30 kHz = 500
(TR45/98.03.03.19R6)
Each site is configured with 3 sectors
Each sector per site has been optimally configured with the maximum number of RF carriers
and traffic channels
Each 30 KHz RF carrier has 3 voice traffic channels
Each sector has been configured to include 1 DCCH
The offered traffic per sector is based upon Erlang B, 1% Blocking

Case 1, capacity gains with no antenna diversity:
Enhanced link capacity techniques assumed in the forward (downlink) path; transmit diversity,
ACELP/CC2 improved forward error correction, and in the reverse path; adaptive interference
rejection and voice activity detection. The link performance gains associated with these
techniques are

Table 1. Downlink quality gains that lower the required C/N

Technique Performance Gain
(dB)
Downlink Transmit Diversity 4
ACELP/CC2 Improved Forward Error
Correction
2

Table 2. Uplink quality gains that lower the required C/N

(dB)
Adaptive Interference Rejection 8
Voice Activity Detection and
Discontinuous Transmission
2

Adaptive Channel Allocation (ACA) and power control are used to assign the channels for the
forward and reverse path. The system performance gain associated with ACA and power
control is 6 dB for the downlink and 2 dB for the uplink (no quality based power control on the
uplink is assumed.)
The cumulative link and system performance gain is 12 dB and is referenced to a 7/21 frequency
reuse plan. Using the preceding enhanced techniques, the following can be calculated:
For N=7, 37.6log (SQRT 3N) = 24.85 dB
Effective N (Enhanced techniques) = ( 10
((24.85-12)/37.6)
)
2
/3 = 1.61
Effective frequency groups used per site is 1.61*3 = 4.83
Effective RF channels per sector = 500/4.83 = 103.5
Effective traffic channels per sector = 103.5 *3-1 =309
Effective Erlangs per sector = 286.1
Effective Erlangs per cell (3 sectors at a site) = 286.1 *3 = 858.3

(TR45/98.03.03.19R6)
The effective Erlangs/MHz/cell (3 sectors at a site) is calculated by dividing the effective Erlangs
per cell of 858.3 Erlangs by 15 MHz. Therefore, the effective spectrum efficiency for each test
environment is

19.1 Erl/Mhz/sector, 57.2 Erl/MHz/cell (3)

Case 2, capacity gains with mobile station antenna diversity

Enhanced link capacity techniques assumed in the forward (downlink) path; transmit diversity,
Mobile station receive diversity, ACELP/CC2 improved forward error correction, and in the
reverse path; adaptive interference rejection and voice activity detection, and quality based
uplink power control. The link performance gains associated with these techniques are

Table 3. Downlink quality gains that lower the required C/N

(dB)
Downlink Transmit Diversity
8
2
ACELP/CC2 Improved Forward Error
Correction
2
Mobile Station Receive Diversity 6

Table 4. Uplink quality gains that lower the required C/N

(dB)
Adaptive Interference Rejection 8
Voice Activity Detection and
Discontinuous Transmission
2

Adaptive Channel Allocation (ACA) is used to assign the channels for the forward and reverse
path. The system performance gain associated with ACA and power control is assumed 6 dB on
the downlink and 4 dB on the uplink using quality based power control.
The cumulative link and system performance gain is 14 dB and is referenced to a 7/21 frequency
reuse plan. Using the preceding enhanced techniques, the following can be calculated:
For N=7, 37.6log (SQRT 3N) = 24.85 dB
Effective N (Enhanced techniques) = ( 10
((24.85-14)/37.6)
)
2
/3 =1.26
Effective frequency groups used per site is 1.26*3 = 3.78
Effective RF channels per sector = 500/3.78 = 132.4
Effective traffic channels per sector = 132.4*3-1 = 396.1
Effective Erlangs per sector = 370.3

8
Tx diversity only has 2 dB gain in conjunction with MS antenna diversity.
(TR45/98.03.03.19R6)
Effective Erlangs per cell (3 sectors at a site) = 370.3*3 = 1111

The gains due to IS-641 over 8PSK
Since mobile station antenna diversity is used in this scenario it is possible to deploy IS-641 with
8PSK over a significant portion of the cell. The percentage of the cell over which this concept can
be used, can be approximated from the C/I cdf distribution from section 2.1. On the downlink, the
DQPSK (or QPSK) users benefit from 10 dB link quality improvements, and 8PSK users benefit
from only 6 dB (MS antenna diversity, Tx diversity is not assumed, and ACELP/CC2) which
implies that the percentage of users in 8PSK mode are equivalent to those with C/I above 24+10 6
= 28 dB from the graph in section 2.1. On the uplink, 8PSK users benefit from both uplink
improvements (i.e. 10 dB), so that the percentage of users in 8PSK mode are equivalent to those
with C/I above 24+10-10= 24 dB. Assuming balanced link operation, the downlink becomes the
limiting factor. Thus, from the graph in section 2.1 the downlink C/I is above 28 dB approximately
30% of the time, implying that IS-641 over 8PSK can be utilized 30% of the time, resulting in a
1.3X capacity multiplier.

Therefore, the effective Erlangs/MHz/cell (3 sectors at a site) is calculated by dividing the effective
Erlangs per cell of 1111 Erlangs by 15 MHz and multiplying by 1.3 as

370.3/15*1.3 = 32.1 Erl/Mhz/sector, 96.3 Erl/MHz/cell (3 sectors) (4)
This number represents an approximate upper bound on the capacity per site in comparison with (2).
(TR45/98.03.03.19R6)
Attachment 18
Additional Capacity Enhancements
This section presents results which show the significant increase in IS-136 based cellular/PCS
system spectral efficiencies by using hierarchical cell structures with a 4/12 macrocell re-use pattern
and street level underlays.
Although the matrics specified by ITU to obtain spectral efficiencies seem to preclude the use of
hierarchical cell structure, real world IS-136 system deployment experiences have proven that
hierarchical network architecture provides the highest capacity potential and facilitates the widest
range of deployment flexibility.
Additional spectral efficiency enhancing schemes such as those specified in Attachment 17 can be
used to further enhance the capacity of hierarchical cell systems.
1. Hierarchical Cell Structure With 4/12 Macro Re-Use Pattern and Street Level Underlays
Network capacity enhancements using IS-136 based cellular/PCS systems can be obtained using
street level hierarchical cells. The most conservative deployment segregates channels into (1)
groups that may be used by the macrocells and (2) groups that may be used by the hierarchical cells.
Hierarchical cells are placed on street corner intersections to capture pedestrian and in-building
traffic while the macrocell handles high mobility traffic.
Scalable macrocell capacity through tighter re-use along with scalable underlay capacity deployment
scenarios underscore a key competitive advantage of IS-136 based technology over competing
CDMA based technology in terms of "complexity" and voice carrying capacity.
In a typical deployment scenario, omni hierarchical cells with an N=9 re-use are placed at
intersections on diagonals throughout a single macrocell coverage area (see Figure 1). If the
macrocell radius is 500 meters and the city blocks are between 100 and 150 meters, then 18
underlays can be placed within the coverage area of a single macrocell. Each hierarchical cell has 8
other small cells which are at most 2 blocks away. The traffic captured by each of the underlays is
included in the overall traffic captured by the macrocell.
Figure 1
(TR45/98.03.03.19R6)
In general, the larger the percentage of channels designated to the street level hierarchical cells, the
greater the spectral efficiency as shown in Table 1 below.
Segregation
(% of Channels Allocated to
Macrocell
Spectral Efficiency
(E/MHz/Cell) at 1%
Blocking
Spectral Efficiency
(E/MHz/Cell) at 2%
Blocking
0 21.54 22.40
10 29.74 31.72
20 44.30 47.12
30 59.86 63.26
40 76.22 80.34
50 91.60 95.80
Table 1
(TR45/98.03.03.19R6)
Attachment 19
Table of Minimum Performance Capabilities for IMT-2000 Candidate
UWC-136 Radio Transmission Technologies
1 Terrestrial Test Environments
Minimum Performance Capabilities for IMT-2000 Candidate
Radio Transmission Technologies
Test
environments
Indoor Office Outdoor to Indoor
and Pedestrian
Vehicular
Mobility
considerations
mobility type
(low)
mobility type
(medium)
mobility type
(high)
Handover Supported Supported Supported
Support of
general service
capabilities
Required / Not
required
Required / Not
required
Required / Not
required
Packet data Supported Supported Supported
Asymmetric
services
Supported Supported Supported
Multimedia Supported Supported Supported
Variable bit
rate
Supported Supported Supported
Data services
key
capabilities
user bit rates
BER
user bit rates
BER
user bit rates
BER
Circuit-
switched low
and long delay
at least 2 048
kbit/s
10
-6
Exceeded
at least 384 kbit/s*
3
10
-6
Exceeded
at least 144 kbit/s
10
-6
Exceeded
Packet at least 2 048
kbit/s
10
-6
Exceeded
at least 384 kbit/s*
3
10
-6
Exceeded
at least 144 kbit/s
10
-6
Exceeded
(TR45/98.03.03.19R6)
ATTACHMENT 20
MRC Diversity Gains at 10% BLER for 136 HS Outdoor Using Simple Equalizer
Table 1. 8-PSK Vehicular A120
Coding Scheme C/I Gain (dB) Eb/No Gain (dB)
PCS-1 6.6 5.45
PCS-2 10.36 8.64
PCS-3 24.75 ___
PCS-4 ___ ___
Table 2. 8-PSK Vehicular A50
PCS-1 4.57 3.39
PCS-2 5.39 4.82
PCS-3 6.43 6.5
PCS-4 8.93 8.57
Table 3. 8-PSK Pedestrian A
PCS-1 6.43 6.0
PCS-2 6.61 5.64
PCS-3 6.46 5.89
PCS-4 6.63 6.5
PCS-5 6.71 6.68
PCS-6 6.22 6.43
(TR45/98.03.03.19R6)
Table 4. GMSK Vehicular A120
CS-1 5.71 4.82
CS-2 7.32 6.43
CS-3 10.3 7.93
CS-4 ___ ___
Table 5. GMSK Vehicular A50
CS-1 4.88 4.55
CS-2 6.01 5.60
CS-3 6.55 6.13
CS-4 14.94 15.59
Table 6. GMSK Pedestrian A
CS-1 6.61 6.43
CS-2 6.79 6.79
CS-3 6.96 6.96
CS-4 7.86 7.86
The data above is obtained by examining the link level BLER curves obtained from simulations
using the simple DFSE equalizer with and without MRC diversity. An example of the two is
given in Figures 1 and 2 below.
(TR45/98.03.03.19R6)
Figure 1 BLER Performance for 8-PSK Vehicular A50 without Diversity.
Figure 2 BLER Performance for 8-PSK Vehicular A50 with Diversity.
0 5 10 15 20 25 30 35 40
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6
0 5 10 15 20 25 30 35
10
2
10
1
10
0
C/I [dB]
B
L
E
R

(
b
l
o
c
k

e
r
r
o
r

r
a
t
e
)
PCS1
PCS2
PCS3
PCS4
PCS5
PCS6

Edegperf

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Edegperf

Uploaded by

Copyright:

Available Formats

Conference Call TR45.3/98.04.06.

You might also like