In last month’s column
I described the
principles behind
OFDM, how it is generated
and received and
different parameters
that can be adjusted
to optimize the transmission for different
uses. This month I’ll examine how parameters
can be modified for different tradeoffs
between coverage, robustness and
data rate. I’ll use a spreadsheet developed
by Teamcast to show some of the options
available for a 6 MHz U.S. TV channel using
the DVBT2 COFDM standard.
CHOICES, TRADE OFFS
The parameters I will consider are Fast
Fourier Transfer (FFT) size, guard interval
(cyclic prefix) and the pilot pattern.
A channel bandwidth
of 6 MHz
is assumed and
for maximum efficiency
I’ll target
a frame duration
of 250 ms.
The FFT size
is related to the
number of carriers
in the OFDM
signal. All other
parameters being
the same,
doubling the FFT
size will halve the
carrier spacing. As described last
month, the symbol duration is the
reciprocal of the carrier spacing
so halving the carrier spacing will
double the symbol duration.
Since the guard interval is applied
to each symbol, doubling the
symbol duration will reduce the
impact of the guard interval on
the data rate or allow increasing
the guard interval to reject longer
echoes without impacting data
rate. This is an advantage in a single
frequency network as transmitters
can be spaced further apart without
interfering with each other.

Fig. 1: DVBT2 parameters and performance

The downside of the larger FFT
is that as carrier spacing decreases,
Doppler shift becomes more of a
problem. The impact of Doppler
shift is proportional to the carrier
frequency, so VHF will be less impacted
by it than UHF. Narrower channels (6
MHz in the United States versus 8 MHz in
Europe) cram the carriers closer together,
making the impact of Doppler worse.
Mike Simon of Sinclair Broadcast Group
did an analysis of the impact of Doppler on
different FFT sizes in a 6 MHz channel and
found that at 600 MHz the Doppler shift
became a problem over 95 mph with an
8K FFT and dropped to 47 mph with a 16K
FFT. With 8 MHz of channel bandwidth, the
speeds increase to 126 mph and 63 mph,
respectively.
Another disadvantage of larger FFT sizes
is that they are more complicated to demodulate,
requiring more processing power
and, thus, increase power consumption,
a potential problem for handheld devices.
For this month’s examples, I’ll use an 8K
FFT, which will allow mobile reception.
The other parameter to consider is the
guard interval. This will determine the
maximum delay the receiver can tolerate.
Note that, unlike VSB, with COFDM, echo
rejection is set by the transmission parameters,
not the receiver design. This makes it
much easier to design a single frequency
network or distributed transmission system.
Guard interval is usually expressed as
a fraction of the symbol duration. A guard
interval of 1/16 with an FFT of 16K in a
6 MHz bandwidth will have a duration of
149 ms. With an 8K FFT size, a guard interval
of 1/8 will be needed to give the same
duration. I’ll target a guard interval of 75
ms for the examples, which will handle
echoes of the same length.

Fig. 2 
To improve robustness, some data carriers
are replaced with pilot carriers the receiver
uses to perform channel estimation,
equalization, common phase error correction
and synchronization in a manner
somewhat similar to the training signals in
ATSC 8VSB.
The DVBT2 standard specifies eight
different scatter pilot patterns named PP1
through PP8. To find out which patterns
to use in which applications, search online
for EBU Technical Bulletin 3348, “Frequency
and Network Planning Aspects of
DVBT2.”
Moving from PP1 to PP8, overhead decreases, but the carriertonoise (C/N) performance
decreases. The variation in C/N is
not huge, less than 2 dB according to Bulletin
3348, and in practice, welldesigned receivers
are likely to perform better. I’ll use
the highest recommended pilot pattern, up
to PP7, in the examples.
EXAMPLES
For the first example, let’s design a DVBT2
system to match the realistic C/N requirement
for the current ATSC A/53 system.
While 8VSB has an additive white Gaussian
noise (AWGN) of 15 dB, in an environment
with multipath, the required ATSC C/N rises
rapidly.
The Teamcast DVBT2 Multiple PLP Allocator
spreadsheet provides DVBT2 C/N
values for AWGN, fixed reception (Ricean),
portable reception (Rayleigh) and reception
with a zerodB echo.
With an FFT of 8K, a guard interval of
1/16 (75 microseconds), PP5 pilot pattern
and a 249 ms frame length, using 64 QAM
with appropriate coding, we can achieve
a 23.9 Mbps average input data rate (compared
to 19.39 Mbps for ATSC), an AWGN
C/N of 15.1 dB and a worstcase 17.7 dB C/N
in a portable use channel. The symbol duration
is sufficient to allow mobile reception
and multipath should not be an issue with
a 75 microsecond guard band. Fig. 1 shows
the DVBT2 parameters used for these two
examples.
We can trade off some of that data rate
for a more robust signal with lower C/N for
indoor and handheld reception. With the
increased resolution of today’s tablets, we’ll
want HD resolution. The ATSC mobile DTV
standard uses H.264, so let’s use a bit rate of
6 Mbps and see how low we can get the C/N
using DVBT2.
The spreadsheet revealed that for the
same FFT, guard interval and pilot pattern,
with DVBT2 we can transmit 6.3 Mbps using
QPSK modulation and Long or 3/5 rate
coding in a 6 MHz channel. The C/N requirement
is only 2.3 dB AWGN, with a worst case
of 3.6 dB using the portable channel model.
This C/N is better than the theoretical
C/N for a quarterrate A/153 mobile signal,
which would be limited to a maximum data
rate of less than 3 Mbps without using the
“fullchannel” extensions to A/153—not
nearly enough to allow fullresolution HDTV.
In field testing, the required C/N has been
found to be significantly higher, around 7–9
dB, than the predicted C/N for A/153 quarter
rate coding, meaning the DVBT2 data
rate could be increased while maintaining
performance as good or better than ATSC
A/153.
These are just two examples of the performance
improvements available with
COFDM and DVBT2, specifically. Request
the spreadsheet from Teamcast and run
your own studies. Note that with DVBT2’s
physical layer pipes (PLP) you can transmit
multiple streams with different bit rates
and C/N requirements simultaneously. Fig.
2 shows the data rate versus C/N for different
COFDM carrier modulation and coding
parameters in a portable environment (the
most demanding) in relationship to the
Shannon limit (black curve).
There are some tradeoffs—COFDM has
a higher peaktoaverage power requirement,
which means stations may need to
upgrade transmitters or reduce power. What
impact will that have on coverage? I’ll cover
that next month in COFDM Basics Part 3.
Comments and questions are welcome.
Email Doug at dlung@transmitter.com.