Mann.qxd
11/14/03
1:17 PM
Page 403
Digital Wireless Camera Technology Fundamentals By Len Mann
Analog wireless camera systems have been available to the broadcast industry for decades and are in common use at sporting events for sideline coverage of football, pit-lane coverage of car racing, etcetera. Analog systems are essentially line-of-sight systems, meaning that the frequency modulation (FM ) used requires an uninterrupted path between the transmitter and receiver. In situations such as transmission in built-up areas where the transmitted signal can be reflected by a number of routes back to the receiver (multipath interference), analog systems can suffer from picture-quality impairments such as ghosting, noise, chroma flutter, or complete loss. Viewers have become used to these impairments, which can occur when watching in-car cameras or marathon coverage, typically when the transmitter passes under a bridge or behind an obstacle. Traditional analog systems, therefore, have never offered true roaming freedom to cameramen. Analog wireless cameras need laborintensive manual targeting of transmit and receive antennas to ensure a reasonable quality of link.
A contribution received in January 2003. Copyright © 2003 by SMPTE.
SMPTE Motion Imaging, December 2003 • www.smpte.org
S
uppliers of digital wireless camera systems (DWCSs) have made great claims about freeing the cameraman and offering a very rugged link between the transmitter and receiver. This is possible, but only in a correctly engineered system. The purpose of this paper is to highlight some of the important technical factors that separate a truly rugged system from one that is only marginally better than analog.
Modulation The choice of modulation scheme is vital to a rugged wireless camera system, and in this area all manufacturers of DWCSs have agreed on the digital video broadcasting-terrestrial (DVB-T) version of orthogonal frequency division multiplexing (OFDM). OFDM is already accepted as the modulation system for DVB terrestrial broadcasting, digital audio broadcasting (DAB), and IEEE 802.11 wireless local area networks. OFDM is a special case of multicarrier transmission: the datastream is transmitted over a number of lowerrate subcarriers (2,000 in the case of DWCS). One of the main reasons for using OFDM is to increase the robustness against frequency selective fading or narrowband interference. In a single carrier system, a single fade or interferer can cause the entire link to fail; only a small percentage of the available subcarriers will be affected in a multicarrier OFDM system. Error correction coding can then be used to correct for the erroneous subcarriers. All currently available DWCSs use the DVB-T version of OFDM, developed to carry MPEG-2 video and audio data over direct-to-home terrestrial systems. The DVB-T OFDM system has several selectable settings that radically affect the ruggedness of the transmission. DVB-T has a bandwidth of 8 MHz (although 7- and 6-MHz versions are also available). DVB-T allows for each of its subcarriers to be modulated in QPSK, 16QAM, or 64QAM. It also allows for vari403
Mann.qxd
11/14/03
1:17 PM
Page 404
DIGITAL WIRELESS CAMERA TECHNOLOGY FUNDAMENTALS Table 1—Data Rates with a Guard Interval of 1/16 Modulation vs. FEC 1/2 2/3 3/4 5/6 7/8
QPSK 5.85 7.81 8.78 9.76 10.25
16QAM
64QAM
11.71 15.61 17.56 19.52 20.49
17.56 23.42 26.35 29.27 30.74
Table 2—Quasi Error-Free Data Rates FEC
QPSK (CNR in dB)
16QAM (CNR in dB)
64QAM (CNR in dB)
1/2 2/3 3/4 5/6 7/8
5.4 8.4 10.7 13.1 16.3
11.2 14.2 16.7 19.3 22.8
16.0 19.3 21.7 25.3 27.9
able amounts of forward error correction (FEC) to be applied. The choice of modulation and amount of FEC radically affects the total data rate that can be carried over the channel, as can be seen in Table 1. For this table the guard interval was set to 1/16. In DVB-T OFDM systems, QPSK is more rugged than 16QAM, and 16QAM is more rugged than 64QAM. The most rugged signal possible with DVB-T OFDM modulation is achieved by setting the QPSK at 1/2 FEC (maximum error correction). This also gives the lowest data rate. As a result there is a tradeoff between ruggedness and data rate, which is crucial to the performance of the system. QPSK rate 1/2 is massively more rugged and will give far superior results in terms of continuity of the link than 64QAM rate 2/3. Table 2, extracted from the DVB specifications (ETSI EN300-744), illustrates the point. The table shows carrier-to-noise ratios (CNRs) in dBs for various DVB-T settings to achieve bit error ratios of 2 x 10 - 4 in a Rayleigh Channel. Table 2 shows that a link operating at QPSK rate 1/2 requires approximately 14 dBs less CNR than a link operating at 64QAM rate 2/3. It can also be stated that a 64QAM rate 2/3 link would need to transmit approximately 20 times more power to achieve the same ruggedness, translating to 20 times more power for the same performance. 404
Trials have shown that true roving non-line-of-sight performance (ranging over 1 km) can only be achieved at QPSK rate 1/2 at realistic man-portable transmit power levels (100 to 200 mW). Even 16QAM rate 1/2 degrades rapidly outside line-of-sight; 64QAM is practically unusable in non-line-of-sight environments, without transmitting massively more power. To some extent the choice of DVB-T OFDM is a compromise. It was designed mainly for direct-to-home use, and it is expected that over time good systems will start to realize improvements to DVB-T OFDM.1
Video Compression It was previously shown that the choice of OFDM parameters, and therefore the associated data rate, makes a huge difference in the ruggedness of the link for a given power. Use of OFDM at QPSK rate 1/2 is massively more rugged than OFDM at 64QAM rate 2/3 for the same power. However, QPSK rate 1/2 gives a significantly lower data rate. Therefore, the choice of video compression standard is very important to the overall system design. The compression standard selected by different manufacturers varies significantly. Most have selected MPEG2 DVB, some have selected Wavelet, and others DVCPRO. These are fundamental differences and the implications are profound. MPEG-2 DVB is a more favorable choice, for the following reasons: • The use of prediction and motion compensation in MPEG gives significantly better video quality for a given bit rate than either Wavelet or DVCPRO. • It offers a variable rate of compression to suit the data rate of the channel. • MPEG-2 DVB is a widely accepted standard in broadcast; the transport layer is fully defined, allowing a compressed signal to be easily remultiplexed or fed forward in any MPEG-2 system, without the need to decode and re-encode. • MPEG-2 DVB allows for high-quality 4:2:2 video. This format has been rigorously tested and gives excellent results in multiple encode and decode passes with other MPEG systems, and preserves chroma data. The same tests of multiple passes and interaction with MPEG artifacts have not been done in Wavelet or DVCPRO. Each of these points is analyzed in turn to understand the implications on system performance. SMPTE Motion Imaging, December 2003 • www.smpte.org
Mann.qxd
11/14/03
1:17 PM
Page 405
DIGITAL WIRELESS CAMERA TECHNOLOGY FUNDAMENTALS
Better Picture Quality and Variable Bit Rate MPEG-2 DVB is a highly advanced compression system, employing techniques such as DCT, prediction, and motion estimation. This combination allows for a variable data rate system offering rates ranging from 1 to 50 Mbits/sec. Therefore, an MPEG-2 encoder could be set to produce data rates filling any OFDM stream from the most rugged modes of QPSK rate 1/2 (5.85 Mbits/sec) to 64QAM rate 2/3 (23.42 Mbits/sec). By contrast, Wavelet and DVCPRO are simple compression formats that employ none of the advanced techniques used in MPEG-2, such as prediction and motion estimation. Both Wavelet and DVCPRO are intra-only encoders, therefore they only produce quality results at bit rates greater than 20 Mbits/sec. Indeed, DVCPRO only operates at 25 Mbits/sec. These restrictions mean that systems employing DVCPRO or Wavelet will only operate in the higher order modulation modes such as 64QAM. Table 2 shows that fundamentally, DVCPRO and Wavelet systems will never function as true roaming cameras, because the most rugged QPSK modes are not available to them. For a detailed explanation of MPEG DVB, Wavelet, and other forms of compression, see Ref. 2. To summarize the differences, MPEG DVB uses a discrete cosine transform (DCT) technique to convert the spatial picture data into the frequency domain; the resulting frequency data is then run-length coded. MPEG DVB also allows prediction and motion compensation
techniques to exploit temporal redundancy in the picture data. These latter techniques create delay, however they give the most significant compression gains. Wavelet encoders use a Wavelet transform instead of DCT and achieve low delay by not using any prediction or motion compensation, but they do not achieve the same compression ratios. Tests using a Tektronix picture-quality analyzer (PQA) on MPEG-2 encoders versus DVCPRO versus Wavelet encoders have shown that MPEG-2 typically requires 60 to 70% of the bit rate for equivalent picture quality. International work has been done to standardize MPEG-2 video bit rates for acceptable quality transmissions. The EBU (European Broadcast Union) operational settings are a good indicator as to what MPEG-2 bit rates are required. The EBU distributes using two main settings: 19.5 Mbits/sec MPEG (~30 Mbits/sec Wavelet), and 8 Mbits/sec (~15 Mbits/sec Wavelet). Also, most satellite distribution of sporting events from satellite newsgathering (SNG) vehicles is done at 8 Mbits/sec. Using this data and operational experience, the following settings are recommended.
MPEG-2 4:2:2 The launch of the MPEG professional 4:2:2 profiles, which ensure good quality color preservation, has enabled MPEG-2 to reach back into the professional studio domain. The 4:2:2 specification allows material to be repeatedly encoded and decoded without undue degradation. MPEG 4:2:2 profiles are essential on professional contribution wireless cameras.
Table 3—Typical Bit Rates for Given Applications Mode
Typical Usage
Transport Rate
Video Rate
16QAM rate 1/2
High-quality video for sporting events with moderate ruggedness to allow transmission in limited non-line-of-sight situations
11.71 Mbits/sec
11.45 Mbits/sec
QPSK rate 1/2
Moderate quality video for news or basic sporting events with high ruggedness to allow roaming transmission with non-line-of-sight situations
5.85 Mbits/sec
5.6 Mbits/sec
SMPTE Motion Imaging, December 2003 • www.smpte.org
405
Mann.qxd
11/14/03
1:17 PM
Page 406
DIGITAL WIRELESS CAMERA TECHNOLOGY FUNDAMENTALS
Standardized Distribution
“
encoder algorithms makes it possible to develop a high-quality MPEG encoder that operates with delays of about 40 ms (approximately one video frame). This makes for a very practical and usable system, and repeated trials have seen no complaints from directors regarding delay. Compression is not the only source of delay in the system. Usually all the feeds that arrive at the studio or the outside broadcast truck are genlocked to a master clock. The genlock process will introduce a further frame of delay, therefore, having a system that can be genlocked at the source will also save delay.
Another important aspect of The launch of the MPEG using MPEG-2 is its fully standardprofessional 4:2:2 profiles, ized nature. It already dominates which ensure good quality the broadcast chain and is the primary medium for contribution color preservation, has using SNG or electronic newsgathenabled MPEG-2 to reach back ering (ENG) vehicles, distribution into the professional studio over satellite (EBU network) or domain. The 4:2:2 specification asynchronus transfer mode (ATM), and to homes via services allows material to be repeatedsuch as Sky. By choosing an ly encoded and decoded withMPEG-2 DVB camera system, an out undue degradation. MPEG operator has the choice of feeding the stream forward as a com4:2:2 profiles are essential on pressed MPEG-2 stream and professional contribution wireremultiplexing it with other comless cameras. pressed services. This saves the need for extra MPEG encoders, and also preserves picture quality by removing unnecessary encode and decode stages. Receive Solutions and Diversity By contrast, DVCPRO or Wavelet systems can never In many discussions on DWCS there is a high focus be easily remultiplexed or handled in an MPEG world on transmission, with little emphasis on the receive and will never be interoperable; they are closed syssolution. This is unfortunate, because the receiver is tems that will tie a customer to one vendor. A DVCPRO as important as the transmitter in the architecture of a or Wavelet system will typically need to be converted to perfect system. There are three primary differentiators MPEG-2 DVB at the SNG truck or in the studio, which with receive technology: error masking and error would appear to be an unnecessary expense.2 recovery, diversity, and delay.
”
The Importance of Low Delay
Error Masking and Error Recovery
Users of wireless cameras are very aware of the importance of video delay in such systems. Events often use a combination of static and wireless cameras. Static cabled cameras introduce no delay, so wireless cameras must be low delay, otherwise the resulting mixed shots will be disorienting to the viewer. Seeing a shot from one angle in realtime then switching to a time-shifted angle can be very confusing. The conventional wisdom has been that MPEG encoding introduces delays of up to 1/2 sec. Indeed, some DWCS products have been launched using MPEG encoders with up to a 1/2 sec delay. Wavelet and DVCPRO encoders are available with much lower delays (two or three frames in the case of DVCPRO, less than one frame in the case of Wavelet), but both suffer limitations in terms of compression quality at lower bit rates. Careful development of the MPEG
A DWCS will often operate at its maximum range, or in difficult environments in which the received signal is degraded to the point at which it contains non-correctable errors. Therefore, the way a receiver handles the inevitable situation of a repeated stream of errors is very important. Many decoder chips have been developed from the world of satellite set-top boxes; they expect very low error rates and have not been designed to expect regular uncorrected errors. Often, these chips will take significant time to resynchronize to the stream and will not attempt to mask the error, which can mean major loss of video (cuts to black or unnatural blocking effects). Alternatively, a receiver that has been designed to expect errors can relock to the stream quickly and cleanly and also mask errors, making for a far more acceptable result to the viewer.
406
SMPTE Motion Imaging, December 2003 • www.smpte.org
Mann.qxd
11/14/03
1:17 PM
Page 407
DIGITAL WIRELESS CAMERA TECHNOLOGY FUNDAMENTALS
Diversity
Conclusion
Receive diversity is a powerful addition to any DWCS, with the use of multiple receive antennas to improve the reception of a transmitted signal. These antennas eliminate flat fades and extend range and coverage. In enclosed environments, strong reflections can align, causing a total cancellation or null at the receive antenna. This phenomenon is known as a flat fade. At 2.5 GHz, flat fades can be very localized. In practice, when transmitting in enclosed environments such as inside buildings, covered stadiums, or built-up areas, flat fades seriously degrade the performance of the system. The occurrence of flat fades can be eliminated by using two or more receive antennas, spaced a small distance apart, because both antennas will not be nulled simultaneously. This gives a radical improvement to the performance of the link. Extra antennas also allow the range of DWCSs to be extended significantly, because single antennas can be placed to cover a local area from which the main antennas are not receiving signal power. For example, at a recent trial in a major stadium, two antennas were placed at either end of a rugby touchline to cover the action on the pitch; a third antenna was placed close to the players dressing rooms 100 m up the runout tunnel. This allowed the cameraman to achieve a continuous uninterrupted shot, following the players all the way from inside the dressing room as they ran out onto the pitch. Another example is placing antennas on different floors of a building to allow continuous coverage as the cameraman walks from one floor to another. These shots are only possible with receive diversity. There are various ways of achieving diversity, which have been used with an advanced packet switching system that can seamlessly combine the inputs from up to four receive antennas. This gives a great amount of flexibility when placing an antenna.
DVBT COFDM is the accepted standard for modulation by all DWCS manufacturers. To achieve real roaming performance with a DWCS, the transmitter should operate in QPSK rate 1/2, or worst-case 16QAM rate 1/2, because no other modes are sufficiently rugged at usable power levels, when line of sight cannot be guaranteed. Transmit solutions that require the use of 64QAM will typically need 20 times more power than those requiring QPSK, for the same ruggedness. MPEG-2 video compression is essential, because neither of the other standards (DVCPRO or Wavelet) offer sufficient levels of compression to fit a good quality picture into a COFDM QPSK rate 1/2 or 16QAM rate 1/2 stream. MPEG-2 video compression also offers standardization, allowing an open system approach and the ability to downstream remultiplex or modulate onto any MPEG-2 network. MPEG-2 4:2:2 ensures professional picture quality and proven color preservation after multiple encode and decode passes. Low delay is an essential part of any multiangle system, enabling the producer to mix shots from fixed and wireless cameras. Fast error recovery and comprehensive error masking are also important to ensure that the occurrence of errors on the stream is handled gracefully. Receive diversity was shown to be an extremely useful technique for the elimination of flat fades, and for extending coverage range.
Delay Use of conventional MPEG-2 decoder chips in DWCS receivers introduces further unnecessary delay. Tests have shown that careful optimization of the decoder design can further reduce end-to-end system delay.
SMPTE Motion Imaging, December 2003 • www.smpte.org
References 1. Richard Van Nee and Ramjee Prasad, OFDM for Wireless Multimedia Communications, Artech House Publishers: Boston and London, 2000. 2. John Watkinson, Compression in Video and Audio, Focal Press: Oxford, 1995.
THE AUTHOR Len Mann graduated in 1971 with a B.Sc. in electrical engineering and became a chartered engineer (MIEE) in 1977. His career has centered on R&D with companies such as Kontron, Cardiac Recorders, Cabletime, and for the past 11 years with Link Research as managing director.
407
