Digital-ATV – Understanding Symbol-Rates, FEC and RF Bandwidth
(TechTalk #76)
In the May 2009 newsletter, TechTalk presented an introduction to D-ATV. Then in
the June newsletter, Robbie-KB6CJZ and I teamed-up to present a top-down approach
for planning a DATV Station that resulted in selecting the DVB-S standard. This
month, TechTalk will explain a few Digital-ATV concepts that are typically not understood
by hams and even analog ATVers. Using the DVB-S standard to transmit a digital ATV
signal involves:
- QPSK (Quadrature Phase Shift Keying) modulation
- FEC (Forward Error Correction) algorithms
- MPEG-2 compression data rates for video
- Video bit-rate needed
- Net Data bit-rate available
- Symbol-Rates
- RF Bandwidth
This article will now walk through these various DATV factors and arrive at determining
the resulting RF bandwidth for DVB-S.
For DATV, the analog camera output is first digitized by the MPEG-2 Encoder board
shown in Fig 1, and then compressed by the MPEG-2 algorithm. The reason the
compressed video data rate varies in Table 1 is that the low value means
little motion in the video scene and the higher value means a lot of motion. Notice
in Table 1 that the uncompressed NTSC camera video stream is 168 Mbits/sec,
while the uncompressed PAL camera video stream is 216 Mbits/sec. The NTSC video
stream data-rate is a 22% reduction from PAL.
| Video Data Stream | Data-Rate | Notes |
| Analog NTSC Camera | 168 Mbits/sec | A/D digitized, uncompressed |
| NTSC MPEG-2 | 2-3 Mbits/sec | compressed |
| VHS MPEG-2 | 1-2 Mbits/sec | compressed |
| Analog PAL camera | 216 Mbits/sec | A/D digitized, Uncompressed |
| PAL MPEG-2 | 2.5-6 Mbits/sec | compressed |
| HDTV Camera | 1-1.5 Gbits/sec | uncompressed |
| HDTV MPEG-2 | 12-20 Mbits/sec | compressed |
Table 1 – Camera Video Data Streams and MPEG-2 Data Streams
Stefan-DG8FAC of SR-Systems (located in Germany...see links at the end) has explained
to me that in Europe many hams set the MPEG-2 output data-rate to be 2.5 Mbits/sec
for PAL. Stephan further suggests that the MPEG-2 output data-rate for NTSC would
be about the same. I suspect that there should be about a 22% reduction in MPEG-2
output data-rate from PAL, to about 2.0 Mbits/sec. I will plan for a 2.5 Mbits/sec
video stream, but when I finally put together my station DVB-S transmitter, I will
measure the NTSC MPEG-2 output to see if the data-rate can be reduced to a 2.0 Mbits/sec
video stream.
FEC Inflation of Video Stream Data-Rate
Forward Error Correction (FEC) is a technology that not only can detect an error
on the received signal, but adds enough redundancy of the data so that it can correct
the wrong bit. It can correct two wrong bits. Since redundancy increases the data-rate
of the video stream, there is a trade-off between more redundancy and the required
video data-rate becoming too large. As we will see a little later in this article,
the larger the video stream data rate, the higher the required RF bandwidth. So
at some point the FEC algorithm will not have enough redundancy to correct too many
errors, and the DATV screen will go blank.
Figure 1 – DATV Block Diagram Showing Various Data-Rates and Symbol-Rates for DVB-S
QPSK (for 2.25 Msymbols-per-sec, the Bandwidth is 3 MHz)
DVB-S commercial television standard uses two different Forward-Error-Correction
(FEC) algorithms together in order to provide protection against noise errors and
multipath errors. The first FEC algorithm is called Viterbi. The second FEC algorithm
is called Reed-Solomon.
The Viterbi FEC algorithm can be configured for different levels of error correction.
Theses different Viterbi configuration/redundancy settings are usually called: 1/2,
2/3, 3/4, 5/6 and 7/8. The first number (“1” in the case of configuration 1/2) is
the number of input bits. The second number (“2” in the case of configuration 1/2)
is the number of output bits from the FECviterbi algorithm. So the MPEG-2 output
data stream is “inflated” 100% by this FEC algorithm configured for 1/2. That is...for
every bit going into the FEC engine, two bits come out. A FECviterbi algorithm configured
for 3/4, for example, would inflate the MPEG-2 output data stream by 33%. So FEC
levels can really inflate the data-bit-rate going to the RF modulator; the MPEG-2
algorithm compresses the video stream, but the FEC algorithms start to expand the
required data-bit-rates again. The Reed-Solomon FEC algorithm has a fixed configuration.
Its data stream “inflation rate” is 188/204. So for every 188 bits going into the
FEC Reed-Solomon algorithm, 204 bits come out...an additional FEC inflation of 8.5%.
Digital Modulation Symbols and Symbol-Rates Digital modulation technology
like BPSK (for example PSK-31), QPSK (Quad Phase Shift Keying – like DVB-S) and
QAM256 (Quadrature Amplitude Modulation with 256 “constellation points”) have the
ability to put more information into a narrow frequency spectrum than analog modulation.
The complexity of the digital modulation scheme, allows us to pack more “data bits”
into each SYMBOL. Table 2 lists out how many data bits can be packed into a symbol
for several well known digital modulation technologies.
| Modulation Scheme | Data Bits per Symbol (Me) |
| BPSK | 1 |
| QPSK | 2 |
| 8-VSB | 3 |
| QAM16 | 4 |
| QAM256 | 8 |
Table 2 – Symbol Bit-Packing for Various Digital Modulation Technologies
Table 2 means that QPSK will pack two data bits into each symbol being modulated.
If we know the final output data-bitrate (I will call this inflated data rate the
“Gross Data-Bit- Rate”) we need for the television signal, then the “symbol-rate”
we need is exactly one-half of that data-bit-rate. For example:
Gross Data-Bit-Rate = 4.5 Mbits/sec Symbol-Rate Needed = 2.25 Msymbols/sec
The formula to calculate the Symbol-Rate setting that I need for my DVB-S transmitter
is:
Symbol-Rate Needed = NDBR / (Me x CRv x CRrs)
Where:
- NDBR = Net Data Bit Rate (aka the information rate) Same as MPEG-2
output data rate in Table 1
- Me = Modulation Efficiency (2 for QPSK in Table 2)
- CRv = Correction Rate setting for Viterbi (1/2, 3/4, etc)
- CRrs = Correction Rate value for Reed-Solomon is 188/204
I will now calculate an example for QPSK where the output of MPEG-2 is 2.4 Mbits/sec
and FEC viterbi is set to 1/2.
Table 3 - Net Data Bit-Rates for DVB-S at a given RF Bandwidth
(NOTE-1: NTSC Analog Camera produces about 2.4 to 2.5 Mbits-per-sec MPEG-2 output
for Ham Radio type broadcasts)
(NOTE-2: The Net Data Bit-Rate values inside the Table need to be at 2.4 Mbps or
larger to support the expected camera data rate coming from MPEG-2 encoder)
Symbol-Rate Needed = [2.4 Mbit/sec 2 bits]/[symb * (1/2) * (188/204)]
Symbol-Rate Needed = [2.4 Mbit/sec 0.921 bits]/[symbol]
Symbol-Rate Needed = 2.65 Msymbol/sec
If I change the FECviterbi setting to 3/4, then the CRv value becomes 3/4 and the
results are:
Symbol-Rate Needed = 1.73 Msymbol/sec
The Symbol-Rate that is needed was reduced because the “inflated data-rate” caused
by a lot of FEC redundancy was reduced. If you look at Table 3 on the preceding
page, it shows the Net Data Bit Rate that can be supported by a particular Symbol-Rate
using several FEC settings. The FEC setting needs to result in a number of Net Data
Bit Rate that is at least 2.4 Mbits/sec. The red values in the table show FEC settings
or Symbol-Rates that result in a Net Data Rate of less than 2.4 Mbits/sec that I
set as my goal for MPEG-2 video stream output.
RF BandWidth for DVB-S DATV
It turns out, one of the advantages of digital-ATV is it can be more bandwidth-efficient
than analog ATV. With QSPK modulation you actually have the ability to easily make
the DATV RF bandwidth as narrow as 2 MHz or 3 MHz without giving up any noticeable
quality. This is because the commercial DTV standards planned to transmit several
Television streams inside one normal (old) RF TV bandwidth. The final formula is
for DATV Bandwidth (BW). For QPSK modulation, the formula for RF BW is:
RF BW = 1.33 x Symbol-Rate
This Bandwidth is the spacing that can be used for placing adjacent DATV station
center-frequencies. This value of Bandwidth is where the signal is down about -15
dB or more. The expression "occupied bandwidth" is sometimes used to refer to a
bandwidth that is 1.19 times the symbol rate, where the signal is down by approximately
-10 dB. If the Symbol-Rate used is 2.25 Msymbols-per-sec, then:
RF BW = 1.33 x 2.25 Msymbols/sec = 3.0 MHz
If we can use a Symbol-Rate of only 1.5 Msymbols/sec, then the bandwidth reduces
to:
RF BW = 1.33 x 1.5 Msymbols/sec = 2.0 MHz
Again, Table 3 on the preceding page provides an overview of what RF Bandwidth you
can choose and what the resulting Net Data Bit Rate will be for various FEC selections.
Conclusion
In reviewing the results in Table 3, I have concluded that I will use an
RF Bandwidth of 2.5 MHz to support an NTSC MPEG-2 output of 2.4 Mbits/sec by selecting
FEC to be 3/4. I plan to put together a DATV station soon. When I do, I will measure
the NTSC MPEG-2 video stream that is really required. If my suspicions that I will
see a NTSC MPEG-2 video stream at around 2Mbits/sec are confirmed, then I probably
will change to a 3 MHz RF BW by using the FEC setting of 1/2. This FEC setting will
produce high DATV signal correction capability in one-half of the normal 6 MHz analog
ATV bandwidth.
Useful DATV Links
Click
here
if you'd like to download a PDF copy of TechTalk #75 and TechTalk #76.