BROADCASTING DIVISION

Application Note

Bit Error Ratio BER

in DVB
as a Function of S/N

Products:

TV Test Transmitter SFQ

Spectrum Analyzer FSE
Spectrum Analyzer FSP

7BM03_3E

BROADCASTING DIVISION
Contents

1 Bit Error Ratio BER in DVB as a Function of S/N respectively C/N........ 3

2 Conversion of S/N (C/N) to Eb/N0 .......................................................... 10

ANNEX 1
Short form discription of SFQ and FSE settings for verifying a C/N of 6.8dB (example) 14

Annex 2
Note: Transmitter Output Power .................................................................... 15

We would like to express our thanks to

PHILIPS/Eindhoven for the support given us
in the preparation of this Application Note.

BROADCASTING DIVISION
1 Bit Error Ratio BER in DVB
as a Function of S/N
respectively C/N block may be up to 0.8 dB per unit referred
to the C/N of the DVB signal as defined by
At what C/N ratio does a set top box still Standard ETR 290.
operate properly? What system margin is .
available in the reception of DVB-C or
DVB-S signals? How can the bit error ratio
as a function of these parameters be
determined exactly?
These questions have top priority in the
development and production of equipment
with DVB capability. In many cases, there is
a defined BER margin for DVB equipment or
chip sets, and the task is to find out to the
limit up to which signal quality may
deteriorate with the DVB system still
operating properly. Different values are to
be expected for DVB-S with QPSK
modulation on the one hand and QAM on
the other hand, because satellite
transmission (QPSK) uses double forward
error correction (FEC), ie Viterbi and Reed
Solomon (RS), whereas for QAM simple
error correction (RS) is used only. Fig. 1 BER as a function of S/N
Determining the bit error ratio is, therefore,
one of the most important measurements in For this reason, this application note
DVB (Digital Video Broadcasting) on cable describes a second, viable approach to
and satellite links. The difficulty is to generating a defined BER, taking into
generate an exactly defined BER. account S /N deterioration:
One approach is to introduce, in FEC White noise of a defined power is
according to Reed Solomon, a known superimposed on the DVB signal. From the
number of bit errors directly after the noise and the modulated DVB signal, the
calculated error protection for the error-free S/N respectively the C/N ratio in dB (with
MPEG2 transport stream (TS). If this consideration of the "roll off" factor) can be
approach is taken, it must be ensured that calculated. After conversion, the
no further bit errors occur on the corresponding BER is obtained for each S/N
transmission path (modulation, frequency value.
conversion, demodulation), caused by noise But there are some constraints using this
in the transmission channel or modulation method. Figure 1 shows the theoretically
errors in the data stream. This condition based restrictions as they occur with QAM
cannot however be met in practice: transmission:
Each unit of a digital TV transmission chain At BER values of about 10-4 to 10-6 - the real
has inherent errors. These errors are range of interest - the graphs for each QAM
explicitly defined by the standard as mode are very steep. This is also shown in
”implementation loss” (IL) or ”equivalent figure 2 "BER in the range of 10-4 to 10-6 as
noise degradation” (END). The additional a function of S/N" which is a zoomed part of
degradation of signal quality from figure 1.
transmission block to transmission

BROADCASTING DIVISION
 ∆ = 1 dB 0.1

BER
BER 0.01
1 x 10-4 4QAM
uncoded
0.001
0.5 dB

4
10
256 QAM
64 QAM ca. 0.9 decades
∆ ca.1 decade
16 QAM
10
5
4 QAM
1 x 10-5 1/2
6
10

2/3
7
10

3/4
8
10

5/6
1 x 10-6 9
10
8 9 10 11 12 . 13 14 15 16. 17 18 19 20 . 21 22 23 24 25
. 26 27 28 29 30 31 32
7/8
S/N dB Fi 10
10

g. 2 Code Rate

BER in the range of 10-4 to 10-6 as a 10

11

function of S/N in QAM

12
10
2 3 4 5 6 7 8 9 10 11 12

Changing the S/N value by 1 dB the BER S/N dB

alters about one decade. A precise noise

source should have an accuracy of about Fig. 3
0.5 dB and as the diagram shows in this BER as a function of S/N in QPSK
case the variation of the BER is again at Modulation with Viterbi correction for
least half a decade. This is too much in the normally used Code Rates
QAM mode.
It is obvious that the sensitivity of BER is
Regarding the QPSK modulation where two nearly twice as in QAM. Changing the S/N
FECs in series correct occuring errors the value by only 0.5 dB the BER alters again
situation is worse. The first FEC - the Viterbi about nearly one decade depending on the
correction - generates a much steeper slope code rate.
in the diagram BER vs S/N depending on
the Code Rate. This shows the figure 3, As determining the Signal to Noise Ratio
which presents the theoretical values for S/N for a predefined Bit Error Ratio BER =
BER vs S/N for the coded signals 2x 10-4 for a device under test is one of the
referenced to the uncoded 4 QAM signal. most important measurements (the value
BER = 2x 10-4 before Reed Solomon
corresponds to the point where the FEC
Reed Solomon is able to correct errors to
the Quasi Error Free QEF datastream) the
S/N value must be generated in highest
precision. All deviation to the precise S/N
value will cause either a too high Insertion
Loss IL for the system to be tested or also
indicate negative ILs. The accuracy 0.5 dB
is therefore not high enough. To avoid false
interpretations the S/N value corresponding
to a given BER should be at least within the
tolerance of 0.1 dB.
So far the theory.

BROADCASTING DIVISION
The solution of Rohde&Schwarz C/N ratio there is the need to measure this
value. The question is what measuring
The TV Test Transmitter SFQ with the Noise equipment is needed for accurate
Generator option supplies QAM or QPSK determination of the C/N value?
modulated TV signals with selectable C/N Because of the high accuracy of R&S
values dB. The generator furnishes analog equipment, only two instruments are
noise signals and therefore does not required for this purpose: TV Test
produce a spectrum of discrete (although Transmitter SFQ and Spectrum Analyzer
disper-sed) lines as obtained with digital FSE are all that is needed to measure the
noise generators. C/N ratio with highest precision.
Moreover, the superimposed noise referred As a useful signal, a PRBS (pseudo-random
to the symbol rate has to be determined for binary sequence) signal of SFQ with 64
a defined C/N ratio in dB. As the symbol quadrature amplitude modulation (64QAM
rate in Hertz and the signal bandwidth as example) is used. Alternatively, a ”live”
coincide according to the modulation signal can be fed in at one of the TS
formula, the symbol rate is the only (transport stream) inputs - ASI or SPI – of
objective reference for the noise bandwidth SFQ and the output spectrum is displayed
and therefore recommended by Standard at the Spectrum Analyser FSEx.
ETR 290.
SPECTRUM ANALYZER 9kHz ...3.5 GHz FSEA
R ROHDE & SCHWARZ SPECTRUM ANALYZER 9kHz ...3.5 GHz FSEA
S

R ROHDE & SCHWARZ TV MESSENDER TV TEST TRANSMITTER SFQ

S

SFQ

TV TEST TRANSMITTER SFQ

Fig. 4 Block diagram of SFQ and FSE

Fig. 5 PRBS spectrum
Inherent noise of the measuring equipment
has to be taken into account of course. The The spectrum of the PRBS at an RF power
signal spectrum generated by TV Test of Puseful = -33 dBm of SFQ, for example, is
Transmitter SFQ without superimposed displayed on Spectrum Analyzer FSE with
noise has a C/N ratio of > 40 dB, so SFQ the following settings:
makes nearly no contribution to the DETECTOR RND
superimposed noise in the range of interest RANGE 10 dB (1 dB/div)
< 35 dB (see Figure 1). SPAN 20 MHz
With 256QAM signals, a S/N value of 36 dB (for DVB-C with 8 MHz channel)
corresponds to a BER of 1*10-11, which is SPAN 50 MHz
the value for the ”quasi error-free” (QEF) (for DVB-S with 33 MHz
data stream. Here too the effect of SFQ can transponder bandwidth)
be neglected. RES. BANDWIDTH 300 kHz
The equation C/N = S/N - k roll off dB (see VIDEO BANDWIDTH 2 kHz
page 10) defines the the corresponding C/N
and S/N values. As the SFQ defines the
5

BROADCASTING DIVISION
After quadrature amplitude modulation, the The displayed power, with the noise
PRB sequence has an optimally flat spectral generator switched off, can therefore be
distribution in the transmission channel. marked very accurately by means of a
display line (DL).

Fig. 6a PRBS spectrum: level marked with

display line, 10 dB/div

Fig. 6b PRBS sectrum: level marked with

display line, 1 dB/div for highest precision

After switchover to I/Q EXTERN in the

MODULATION menu of the SFQ, the PRBS
signal is switched off. The I/Q inputs should
be terminated with 75 Ω.
75Ω

I/Q external

I/Q Modulator
Attenuator
+ -33+26=7dB
Noise
Generator
C/N=26dB
Part of SFQ Block Diagram

Fig. 7 Useful signal and noise paths in SFQ

BROADCASTING DIVISION
In order to increase the accuracy of this Now, is the noise exactly 26 dB below the
measurement with firmware numbers equal useful signal?
or higher than 1.24, the switching off is no
more done by switching to I/Q EXT, but the This can be verified by changing the setting
RF carrier is switched off via of the internal SFQ attenuator for RF level
MODULATION/NOISE/CARRIER ON/OFF. setting by 26 dB.
75 OHM

The two display lines for the useful signal

and noise should now coincide because
I/Q MODULATOR
Carrier both signals are routed via the internal
ON / OFF Attenuator
+
-33+26=7dB attenuator. If there is no coincidence, the
Noise difference between the useful signal and the
generator
C/N=26dB noise signal can be read from the two lines.
Part of SFQ block diagram

Fig. 8 CODER switched off by firmware and The display lines should be placed on the
noise path in SFQ respective channel spectra as accurately as
possible. While this is a subjective setting, it
Now the noise generator is to be switched to may still be assumed to be correct with an
the SFQ output by MODULATION/NOISE absolute accuracy of < 0.05 dB since it is a
ON. ratio measurement which is performed with
the aid of the display lines.
The power of the SFQ's noise generator is
Pnoise = Puseful - 26 dB for C/N = 26 dB (as
example) referred to the signal bandwidth.
-32.99 dBm

The noise is marked on the display of FSE -33.25 dBm

by means of a line 26 dB below the useful
signal line (see Fig. 9). ∆ = DL1-DL2
= -32.99+33.25 dB
= 0.26 dB

26 dB Fig. 9 Measurement of deviation from

selected C/N value
The absolute overall accuracy of this
measurement is, therefore, determined only
by the accuracy of the SFQ attenuator.
Any overload effects caused by the noise
crest factor or similar factors are excluded
through the use of the high-precision FSE
Fig. 9 PRBS useful signal and noise as a spectrum analyzer.
spectra with 10 dB/div

BROADCASTING DIVISION
 And what about the accuracy of the SFQ
attenuator?

The attenuator error has been shown to be

< 0.01 dB in acceptance testing. This value
is recorded and available together with the
SFQ calibration report.

Level Data sheet tolerance Internal tolerance Error

.... .... .... ....
16 dB ≤0.50 dB ≤0.35 dB -0.08 dB
17 dB ≤0.50 dB ≤0.35 dB 0.00 dB
18 dB ≤0.50 dB ≤0.35 dB 0.01 dB
19 dB ≤0.50 dB ≤0.35 dB -0.01 dB
20 dB ≤0.50 dB ≤0.35 dB -0.05 dB
.... .... .... ....
Table 1 Extract from attenuator test report

If the minimum residual attenuator error plus can be determined with an absolute accuracy of
the previously determined ENDs of the < 0.1 dB by means of the described method.
individual units are taken into account in This accuracy fully meets the requirements for
setting the C/N for a defined BER, the total BER measurements even in the range 1*10-6 to
C/N value 1*10-8 .

BROADCASTING DIVISION
Tip:
S/N = C/N + k roll off dB
Checking the exact C/N value at the output of should not influence this proposal. However, the
SFQ in accordance with the above description accuracy of the S/N value also at different
is in itself a simple procedure. The easy symbolrates should be checked prior to every
calculation to arrive at the S/N value precision BER measurement.

Diagram for QAM

Fig. 10 BER as a function of S/N

BROADCASTING DIVISION
Diagram for QPSK Modulation

0.1

BER

0.01
4QAM
uncoded
0.001

4
10

5
10

1/2
6
10

2/3
7
10

3/4
8
10

5/6
9
10

7/8
10
10

Code Rate
11
10

12
10
2 3 4 5 6 7 8 9 10 11 12

S/N dB

Fig. 11 BER as a function of S/N

10

BROADCASTING DIVISION
2 Conversion of S/N (C/N) to Eb/N0
Often, BER diagrams do not have S/N as abscissa but Eb/N0, which is the energy per useful
information bit Eb referred to the normalized noise power N0. In converting the two quantities
one to the other, some factors have to be taken into account as shown by the following
equations:
C / N =E / N + k FEC + k QPSK /QAM + k P dB or
b 0
E /N = C / N − k FEC − k QPSK /QAM − k P dB or
b 0
E /N = S / N - k roll off - k FEC − k QPSK/QAM − k P dB
b 0
where: C/N = S/N − kroll off dB
To determine S/N dB respectively C/N dB, the logarithmic ratio Eb/N0 is to be corrected by
the following factors (this applies to the determination of Eb/N0 vice versa):

188
k FEC = 10∗ lg ie the factor for FEC to Reed Solomon
204
kFEC = -0.3547 dB

k QPSK / QAM = 10∗ lg(m) ie the factor for the QPSK/QAM modes
Mode m kQPSK/QAM dB
QPSK 2 3.0103
16 QAM 4 6.0206
64 QAM 6 7.7815
256 QAM 8 9.0309

kP = 10∗ lg( P ) ie the factor for the puncturing rate (P=1 for QAM)
Mode P kP dB
QPSK ½ -3.0103
2/3 -1.7609
¾ -1.2494
5/6 -0.7918
7/8 -0.5799
QAM 1 0

α
k roll off = 10∗ lg(1 − ) ie the factor for the cos roll-off filtering in the
4
demodulator/receiver

Mode α kroll off dB

DVB-C 0.15 -0.1660
DVB-S 0.35 (nominal) -0.3977
0.27 (actual in -0.3035
transmitter)

11

BROADCASTING DIVISION
The question of what correction factors are needed depends on whether

• Eb is to be treated as a pure information bit

and on whether measurement is made

• in the transmission channel,

• before or after Viterbi or Reed Solomon correction,
• with QAM or QPSK modulation.

Following are a few examples of conversion equations:

The following applies to in-channel measurements with QAM modulation:

188
E / N = C / N − 10∗ lg − 10∗ lg ( m) dB
b 0 204
The factors for
• cos roll-off filtering and
• the puncturing rate (because only with QPSK necessary)
are not needed.

For measurements in the QAM demodulator, the cos roll-off filtering has to be taken into
account:
 α 188
E / N = S / N − 10∗ lg  1 −  − 10∗ lg − 10∗ lg ( m) dB
b 0  4 204

For measurements in the satellite demodulator with QPSK modulation (for determination of
BER as a function of Eb/N0 after Viterbi FEC), the equation is as follows:
 α
( ) ()
188
E /N = S / N − 10∗ lg  1 −  − 10∗ lg − 10∗ lg m − 10∗ lg P dB
b 0  4 204
All correction factors are included in the equation

If a pure PRBS is used for BER measurements the RS FEC is not inserted and therefore
the equation is as follows:
 α
E / N = S / N − 10∗ lg  1 −  − 10∗ lg ( m) − 10∗ lg ( P ) dB .
b 0  4

12

BROADCASTING DIVISION
Diagram for QAM BER vs Eb/No

0.1
BER

0.01

0.001

256 QAM
4
110

64 QAM
5
110

16 QAM
6
110

4 QAM
7
110

8
110

9
110

10
110

11
110

12
110
4 6 8 10 12 14 16 18 20 22 24 26 28 30

Eb/No dB

Fig. 12 BER as a function of Eb/No

13

BROADCASTING DIVISION
Diagram for QPSK Modulation BER vs Eb/No

0.01

BER
4 QAM (uncoded)
0.001

4
1 10

5
1 10

Code rate
6
1 10

1/2
7
1 10

2/3
8
1 10

3/4
9
1 10

5/6
10
1 10

7/8
11
1 10

12
1 10
2 3 4 5 6 7 8 9

Eb/No dB
Fig. 13 BER as a function of Eb/No

14

BROADCASTING DIVISION
ANNEX 1

Short form discription of SFQ and FSE settings for verifying a C/N of 6.8dB (as
example):

First step:
Set the noise bandwidth.
This bandwidth corresponds always to the symbol rate and not to the bandwidth of the SAW filter of
the STB. So using the normal symbol rate of a satellite system of 27.5 MSymbols/s the noise
bandwidth set at the SFQ is 27.5 MHz.

Second step:
Set the SFQ to:

RF 200 MHz Level x dBm. In our case let’s define x = - 26 dBm.

Modulation QPSK Symbol Rate 27.5 MSymbols/s
Mode PRBS Roll Off 0.25
Noise Menue NOISE OFF
In this frequency range the internal attenuator has the highest precision. If there is some different
frequency response in other settings doesn't matter, because the generated carrier and the noise are
both influenced in the idendical amount. The ratio C/N is therefore constant.

Third step:
Set the spectrum analyzer FSE to:

Center Frequency 200 MHz Span 50 MHz Level Range 10 dB

Mode Low Distortion VBW 30 Hz RBW Coupled
Ref Level -26 dBm Detector RMS Sweep Time 1 s

Fourth step:
The top of the displayed spectrum should be approximately 4 dB above the lower margin of the FSE
display, in order to avoid clipping due to the high crest factor. Set a display line exactly to the
displayed channel power at FSE. The amplitude vs frequency response should of course be flat within
the satellite channel. You should integrate the noise superposed to the displayed trace by your eye.
The value of C (or better RF/IF power) is now marked at the FSE display.

Fifth step:
Set the SFQ to
RF 200 MHz Level -26.0 + 6.8 ⇒ -19.2 dBm.
MODULATION/NOISE/CARRIER OFF This means no RF is output!
(if CARRIER OFF is not available, the firmware update to version higher or equal than 1.24
should be done)
Symbol Rate 27.5 MSymbols/s
Mode PRBS Roll Off 0.35
Noise Menue NOISE ON and select C/N = 6.8 dB @ 27.5 MHz.
BANDWIDTH COUPLING ON this means bandwidth is 27.5MHz

The now displayed NOISE FLOOR at the spectrum analyzer FSE should exactly meet the adjusted
display line. If not so, the difference to the display line shows the deviation between the SFQ
displayed value and the real generated C/N value. In order to calculate the corresponding S/N value
correct C/N: C/N = S/N - kroll off = S/N + 0.3977 dB in case of DVB-S.

15

BROADCASTING DIVISION
Annex 2

Note: Transmitter Output Power

When determining the test transmitter output power, the effect of inherent or superimposed
noise of the test transmitter must be taken into account very accurately because of the high
sensitivity of the bit error ratio BER to even slight changes of C/N or Eb/N0.
The table below shows the deviation of output power as a function of superimposed noise
with the symbol rate in Hz as bandwidth.

Selected C/N Output power Resulting power Deviation

in dB in dBm in dBm in dB
0 -20 -16.990 +3.01
5 -20 -18.807 +1.193
10 -20 -19.586 +0.414
15 -20 -19.865 +0.135
20 -20 -19.957 +0.043
23 -20 -19.978 +0.022
25 -20 -19.986 +0.014
30 -20 -19.996 +0.004
35 -20 -19.999 +0.001
40 -20 -20.000 0

Deviation dB

3,5

2,5

1,5

0,5

0
0 5 10 15 20 25 30 35 40
C/N dB

The above diagram shows that the effect of superimposed noise concerning the output
power is negligible from a C/N value of >23 dB. For values <20 dB, BER measurement is
error-prone already for 64 or 256QAM signals.

16

BROADCASTING DIVISION