Addressing industry demands for more efficient bandwidth utilization and building on its experience with 64 QAM transmission over cable, General Instrument has developed a 256 QAM transmission system that provides far more efficient use of cable system bandwidth and expands channel capacity. This expanded channel capacity results in a 44 percent increase in information rate and a 50 percent increase in video content as compared to 64 QAM. With it, broadband network operators will be able to carry two HDTV channels instead of just one in a 6-MHz space. The added capacity enables expanded video, modem, telephony and business data services. 256 QAM transmission also makes it possible to substantially increase the number of cable services on bandwidth-limited networks designed for analog video performance. This capability might allow deferral of costly upgrades/rebuilds.

GI successfully conducted the first extensive field tests of the 256 QAM system in an actual cable environment with Rogers Cablesystems Limited, Canada's largest cable operator. The field testing discussed was performed at 21 locations served by three different Rogers headend sites servicing parts of Toronto, Newmarket, St. Thomas and Woodstock in Ontario, Canada. New and older cable plants were chosen to test the performance of 256 QAM transmission in systems typical of deployment scenarios.


As mentioned above, the 256 QAM system's increased information rate enables a larger number of services to be compressed in a 6 MHz bandwidth. This increased information rate, resulting from 256 QAM's added spectral efficiency, provides the opportunity for carrying additional services such as increased quantities of digitized cable channels, video-on-demand, near-video-on-demand, Internet access and interactivity — without compromising existing features and services — which results in additional revenues for broadband network operators. On average, for equivalent picture quality, nine NTSC signals can be placed in the same bandwidth, as compared with only six signals for 64 QAM. Table 1 provides a comparison of 64 QAM and 256 QAM efficiencies.


These values are based on an average bit stream for each video service. Assuming that film-based services are effectively digitized at a 3 Mbps (Megabits per second) rate, and live video at 4 Mbps, the 256 QAM transmission results in a 50 percent increase in both live video and movies per 6 MHz bandwidth. Also, with the HDTV bit rate specified by ATSC as 19.4 Mbps, 256 QAM is able to transport two HDTV signals in the same bandwidth, while 64 QAM can accommodate only one signal.

The larger constellation size and concomitant reduced Euclidean distance associated with 256 QAM transmission does compromise some of the signal robustness seen with the 64 QAM signal. The recommended carrier-to-noise ratio for operating 256 QAM and 64 QAM through the cable system is 37 dB and 32 dB, respectively. The theoretical BER curve showing carrier level vs. additive white Gaussian noise (AWGN) is shown in Figure 1.


The carrier-to-noise ratio for the theoretical coded 256 QAM signal has a 6 dB shift in noise performance as compared to 64 QAM and is therefore less tolerable to noise. The curve also shows the increase in performance obtained by the use of ITU J.83(B) FEC over the ITU J.83(A) with a 256 QAM constellation at 5.056 MSps (Mega Symbols per second). Parameters such as CNR, CSO and CTB should be well controlled for 256 QAM transmission. It has been observed that peaking in the distortion components is a primary cause of bit errors.

As Figure 2 illustrates, because of the denser 256 QAM constellation, it is less tolerant of these distortions. Therefore, for successful deployment of 256 QAM, cable plants should adhere to FCC technical standards as a minimum.

Test setup

All tests were bit error rate tests and were conducted using Broadcom transmission hardware, ITU J.83(B) Forward Error Correction (FEC) and prototype demodulators. A block diagram of a typical receive site test setup is shown in Figure 3. A pseudorandom data generator and FEC encoder were used to produce the input to the Broadcom 256 QAM modulator. Channel up-conversion was performed using a General Instrument C6M for the 256 QAM signal and was then combined with Rogers headend analog channels for transmission. The QAM signal transmission channels were varied from area to area, with the test channels usually operating at the upper edge of the cable spectrum.


The 256 QAM average signal power level was adjusted at the headend for operation at 10 dB below the adjacent analog video's peak of sync power. The proof of concept receiving equipment which was used consisted of an 860 MHz bandwidth RF tuner and a 64/256 dual QAM demodulator incorporating an ITU J.83(B) FEC at an interleaver depth of 66us. Testing was performed in selected Rogers employee homes and at pedestal taps in residential neighborhoods through 100 feet of coax simulating the drop to other cable subscribers' homes. Extended duration testing was performed in the Rogers employees' homes to both assess longer term error performance as the cable system levels change with temperature and to determine the impact of in-home wiring on 256 QAM modulated signals.

Performance tests at the pedestals consisted of BER measurements and input power level variations of the QAM and analog signals. Two PCs were used for each demodulator/BERT pair during the course of the tests: one for logging errored seconds from the HP3784 BER tester and the other for tuning and controlling the demodulator. Recording of BER data was accomplished via an RS232 link between the BER tester's printer port and a PC. Short-term tests were performed using 15-minute gating periods. Extended duration testing consisted of one-second gating periods for the duration of the test. Each test had an associated error log that recorded the error count and the time duration of the test period. The file was stored in ASCII format for later off-line analysis.

Test results

Initial testing consisted of a lab trial of the 256 QAM signal over an ALS DV6000 (8-bit) digital fiber link. The fiber link consisted of a 1550 nm laser and 20 km of Corning SMF28 fiber optic cable. No problems arose with transmission of the 256 QAM signal through the link. A BER vs. broadband noise response curve was verified for the QAM signal by introducing AWGN into the system after the modulator. Little degradation in BER vs. noise performance was seen on the QAM signal. The link was found to be transparent to the 256 QAM signal and ran error-free. This BER curve is shown in Figure 4.


The IF-RF performance over cable vs. fiber link is virtually identical. System performance, shown in Figure 4, is degraded by approximately 0.6 dB for the following reasons:

  • The 64/256 QAM dual-mode Broadcom demodulator chip, which interfaces directly to the ITU J.83(B) FEC, provides seven soft decision bits rather than the eight required by the FEC in 256 QAM mode. Since the LSB is not used, this results in 0.2 dB of performance loss;
  • In order to transmit 5.356 MSps in a 6-MHz channel, a filter roll-off (a) of 12 percent is required. A filter with an alpha of 12 percent is used in the transmitter, but the Broadcom demodulator chip implements a receive filter with a roll off of 20 percent. This mismatch between the transmitter and the receiver adds 0.4 dB degradation.

The first set of system tests was conducted over a newly-upgraded HFC plant. Two fiber optic links were used and consisted of a 55 km fiber link using the ALS DV6000, and 10 km AM fiber links connecting the headend to several optical hubs, as shown in Figure 5. From the hubs, coaxial distribution was used with the longest runs tested being two equally long active runs. The first consisted of seven trunk amplifiers and two line extenders, and a second consisted of six trunk amplifiers and three line extenders.


The 256 QAM signal was placed on EIA Channel 80. The lower adjacent channel supported cable modem traffic operating at 500 kbps QPSK. The upper adjacent channel was inactive. Five sites were tested under short-term conditions, and all ran error-free. One extended duration test was performed and resulted in 99.97 percent error free seconds (EFS). The test duration was 37 hours, 3 minutes. Table 2 provides a summary of the extended duration tests performed.

Threshold of visibility (TOV) also was performed at this location. TOV is defined by CableLabs as a BER less than or equal to 3E-6, obtained in three consecutive 20-second gating periods. If a BER greater than 3E-6 occurs in one of the three 20-second gating periods, another period is allowed to be tested. The limitation in TOV testing was found to be the signal level at the front end of the tuner. TOV levels were within amplitude variations that are expected to be seen on a typical cable drop over time because of temperature. Digital carrier-to-noise ratios for all sites were found to be between 31 dB and 33 dB. Analog carrier-to-noise ratios up to 45 dB were measured. Carrier-to-noise and distortions did not present a problem at this location.

Subsequent system testing was performed at two different locations on older, non-rebuilt coaxial systems. The first system tested was specified as an "electronics drop-in upgrade" 450 MHz system. This location's longest active run that was tested consisted of a 30-trunk amplifier cascade. The 256-QAM signal was placed on EIA Channel 48. Both lower and upper adjacent channels were present and used sync-suppression for video scrambling. Five short-term tests were run at four locations. The tests ran error-free. One 256-QAM extended duration test was performed and resulted in 99.93 percent EFS. The test duration was 5 hours, 36 minutes. Considerable in-band tilt, (approximately 3 dB), was observed on the 256 QAM signal at this site. The tilt was because of excessive system frequency/amplitude roll-off and exceeded the specification for the demodulator. The tilt is the cause of the degraded BER performance.

TOV testing was performed on one site and found to be consistent with the previous measurement on the recently upgraded HFC system. Digital carrier-to-noise ratios for all sites were found to be between 30.5 and 36.2 dB. Analog carrier-to-noise ratios up to 46.6 dB were measured. The carrier-to-noise ratio did not present a problem at this location. Distortions did not appear to be problematic at this location and met FCC required specifications.

The second fully coaxial system tested was also an older 450 MHz system. The longest active run tested consisted of a 31-trunk amplifier and one-line extender cascade. The 256 QAM signal was placed on EIA Channel 51. The lower adjacent channel was active, and the closest upper channel was EIA Channel 53. Three locations were tested under short-term conditions. All tests ran error-free. Two extended duration tests were performed simultaneously and resulted in 99.996 percent and 99.994 percent EFS. The time duration was 13 hours, 16 minutes. Digital carrier-to-noise ratios for all sites were found to be between 30.8 and 38.4 dB. Analog carrier-to-noise ratios up to 48.4 dB were measured. CNR, CTB, and CSO did not present a problem at this location.


Based on the test results obtained on the Rogers system, 256 QAM is a viable transmission format for properly maintained new and older cable plants and inside wiring. Short-term tests yielded error-free performance, and extended duration test results showed EFS performance of 99.93 percent or better. Test results indicate minimal degradation in performance when operating over a digital fiber link, such as the ALS DV6000. On the headend systems tested, for the most part, RMS distortions measured were below the levels that would induce bit errors.

Distortion levels (rms values) such as CTB and CSO were not the primary cause of errors, but the random peaking of these distortions was a cause for concern. In HFC plants, shorter runs and fewer active components minimize the potential for these effects. The FCC technical standard for cable is still a viable guideline for implementing both 64 and 256 digital transmission. At a minimum, operators should adhere to the FCC specification to ensure successful implementation of digital transmission.


A team of engineers from NEC and California Eastern Laboratories has developed a new miniature double conversion RF receiver for cable modems. Designed to be used as a chipset inserted on a motherboard, or as a miniature, standalone 1.5-inch × 3.0-inch board, the receiver reduces the size and complexity of the RF/analog section of cable modems. More importantly, it will enable OEMs to assemble the RF tuner portion of their high-speed (27 Mbps) digital cable modems for as little as $10, and in turn will enable MSOs to field smaller, lower-cost modems in their HFC networks. CEL and NEC plan to offer the design as a reference standard for all cable modem receivers.

The receiver combines off-the-shelf components with a silicon IC chipset developed at the joint NEC/CEL Product Design Center. This receiver chipset is a product of the partnership. The receiver is designed to process inputs from 250 to 860 MHz, so it can be used in both U.S. and European cable modems. The CEL receiver subsystem was designed to tune and process cable TV or video channels modulated at 64 QAM (a 256 QAM version is planned).

This chipset promises to play a key role in miniaturizing modems. It's designed to replace the "canned" discrete tuner modules now commonly used in cable data modem designs. By packaging the entire frequency conversion function into a small group of ICs, the chipset makes it possible to assemble the entire RF tuner/QAM demod subsystem on a card that measures just 4.5 square inches.

The card combines these functional blocks: High pass filter, designed to reject signals below 200 MHz; Pin attenuator, provides additional signal control ahead of the upconvertor; the upconvertor; an intermediate 915 MHz saw filter; a linear UHF downconvertor; dual low noise synthesizer; 36 or 44 MHz TV saw filter; QAM IF downconvertor; and LO crystal references and voltage regulators.

The first two frequency conversion ICs and the synthesizer make up the tuner section. The third frequency converter further processes the signal to feed the 64 QAM demodulator/forward error correction ICs in the digital section of the modem. The card itself is a four-layer glass epoxy board, with an interface bus for the synthesizer and AGC. The first IC in this line-up is the upconvertor, which combines a 15 dB AGC amplifier, a Gilbert-cell mixer, two stages of local oscillator (LO) buffering, a VCO, and temperature compensation circuitry. Available in a 20-pin SSOP package, the IC provides wide frequency bandwidth and high dynamic range performance, plus 5 dB to 20 dB of conversion gain with an 8 dB noise figure.

The second IC is the downconvertor, which is housed in a compact 8-pin SSOP package and contains a UHF mixer, a VCO and an IF amplifier. It dissipates just 40 mA from a 5V supply and features good linearity with low oscillator phase noise.

The QAM IF downconvertor provides an additional 25 dB of AGC control and performs the final downconversion to 5 MHz. A video amplifier further processes the signal in preparation for digitization and demodulation.