Cable operators are in the process of upgrading their networks and rolling out two-way data services in a systematic plant-by-plant migration to a completely digital infrastructure. The speed at which cable plants are being upgraded is limited by the amount of investment needed to achieve low upstream noise conditions required by today's cable modems. To improve the speed at which plants are migrating to two-way data, next generation modulation technology is required so cable operators can implement a robust, high capacity cable data network at reasonable cost. High capacity modulation in next-generation cable modems promises to improve the economics of the business model so operators can deploy two-way data services on a more widespread, commercial basis.
The technical community, including the Society of Cable Telecommunications Engineers and IEEE 802.14, is in the process of defining a high capacity modulation standard. This standard is being referred to as "HI_PHY" because the technical goal is to deliver HIgh capacity via a new PHYsical layer specification in the ISO protocol stack. This article describes the benefits of a HI_PHY, and it compares several HI_PHY candidates with the current PHY used in MCNS cable modems, namely, QPSK and 16 QAM. Because there is expected to be a negligible cost difference between current PHY and HI_PHY implementations, it can be argued that cable operators can future proof their networks today with a HI_PHY with no cost penalty.
High capacity modulation addresses two major problems for the operators. First, it reduces the amount of upfront investment required by operators to "clean up" their upstream cable plant to support two-way service, thus accelerating service roll-out to more serving areas. Second, it increases the upstream channel efficiency so the operator's investment in data networking equipment can be leveraged across a larger subscriber base. Higher channel efficiency also means more bandwidth available to operators to support enhanced services such as multimedia traffic (telephony, videoconferencing) and virtual private networking services for the business community.Noise immunity
The frequency band used by upstream cable channels is plagued by high-energy ingress, burst and impulse noise. The ingress is caused by public HF and amateur HAM radio stations which operate in the same frequency band used by upstream cable channels. Without careful filtering installed at various locations in the upstream path, these radio signals can penetrate the cable and end up in the headend receiver as ingress energy. Single-carrier QPSK channels can be completely lost in the presence of a single strong narrowband ingress signal.
The current solution to ingress noise is to relocate the subscriber's data traffic into a new upstream channel using a technique called frequency hopping — in effect, hopping away from the ingress. This hopping causes an interruption in the subscriber's traffic flow, which could result in packet re-transmissions and delay.
The upstream cable plant is also exposed to wideband burst noise created by appliances such as garbage disposals, hair dryers and drills. The burst noise energy can be very high, and it is distributed across a wide frequency range that can impact many upstream channels. Burst noise duration is typically a few microseconds, which can destroy several QPSK symbols and severely impact customer transmission. The traditional solution to burst noise is to apply a forward error correction (FEC) code to the data and to interleave the resulting data in the time domain over a long period. This technique can be effective for long data packets, provided the burst duration is only a few microseconds, but it is ineffective for short voice packets or when the burst duration lasts for several microseconds.
A HI_PHY provides high immunity to ingress and burst noise, so frequency hopping is not necessary. It can also automatically adapt to whatever noise is present in the assigned channel, thus eliminating the operator's need to carefully select upstream channels that are free from high energy ingress or burst noise, and eliminating the need to reserve other channels for frequency hopping. Complex frequency management procedures can be replaced by a robust modulation technique that is capable of maintaining a high quality of service in the presence of high energy ingress.Bandwidth efficiency
Among the many factors to be considered when evaluating the overall bandwidth efficiency of an upstream modulation technique, the most obvious one is the channel loading, which is usually described in terms of bits per second per Hertz. Channel loading is a function of the number of bits-per-symbol transmitted (also known as the modulation index), as well as the amount of frequency guardband required to keep one channel from interfering with another.
For example, QPSK has a bit loading of 2 bits-per-symbol, and MCNS requires a frequency guardband of 25 percent. As a result, the bandwidth efficiency of an upstream MCNS QPSK channel is 1.6 bits/sec/Hz. A HI_PHY increases channel loading by increasing the number of bits per symbol transmitted.
Other factors besides channel loading also impact the bandwidth efficiency of the upstream spectrum — such as immunity to noise impairments as well as burst guard time requirements. A HI_PHY opens up regions of the frequency spectrum that are unusable by QPSK because of noise impairments. A HI_PHY allows the entire upstream spectrum to be used because it can continue to operate in channels punctured with ingress and burst noise.
As a result, the entire 5 to 42 MHz (or 5 to 65 MHz in some plants) upstream spectrum can be used for revenue generating traffic. Furthermore, the upstream channels reserved for frequency hopping (when an assigned channel becomes unusable due to noise) can be recovered. High-quality service can be maintained in any assigned channel, so reserved channels are no longer required to be sitting idle waiting for traffic from another channel to be switched into it.
To understand how burst guard time requirements affect channel efficiency, access mechanisms used to insert subscriber data in the upstream channel should be examined. Time Division Multiple Access (TDMA) is the upstream channel access method specified by MCNS and IEEE 802.14. It uses burst transmission in which each subscriber modem gets a grant to transmit for a predefined duration, measured in "mini-slots." When an asynchronous modulation technique is used, a guard time is required between bursts for receiver training and transmitter ramp-up, and no useful subscriber data can be sent or received during this guard time. MCNS calls for a minimum guard time of 5 symbols for QPSK and 16 QAM modulation.
The amount of guard time required between transmissions from different modems is a major factor in determining channel efficiency, especially when the symbol rate of the channel increases, causing the average packet size of the data to decrease. While absolute guard time is typically constant, guard time as a percentage of the total channel capacity increases as packet size decreases. For channels with integrated voice and data traffic, the voice packets are typically small compared to data packets in order to decrease end-to-end delay. As a result, short voice packets will have higher overhead and lower channel efficiency than long data packets.
Many HI_PHY candidates synchronize the cable modems to the same timing reference as the headend. As a result, no guard time is required between successive bursts, which means the channel can support continuous transmission of data from multiple modems. Guard time is an important consideration for HI_PHY candidates because the bit loading per symbol typically increases, and upstream bursts naturally become shorter as fewer symbols are needed to carry the same amount of data. As a result, the guard time overhead increases as a percentage of revenue generating traffic. However, a HI_PHY that requires no guard time makes more of the upstream channel available for revenue generating traffic, as compared to one that requires a guard time.HI_PHY similarities
The HI_PHY modulation techniques that are currently under consideration by the IEEE are listed in Table 1. These candidates can be grouped into two very distinct categories: multi-tone modulation and spread spectrum modulation. Within the multi-tone group, however, there are implementation variations that cause differences in end-to-end performance under various noise conditions.
The most important similarity in all HI_PHY candidates is the use of another dimension to carry transmitted data on the RF carrier. For single carrier QPSK or 16 QAM modulation as specified by MCNS, the only dimension available is time, and as described above, each subscriber is allocated a certain number of time slots in which to transmit data upstream to the headend. In the case of all multi-tone HI_PHYs, the added dimension is frequency, whereas for S-CDMA, the added dimension is codespace. Whether it's frequency or codespace, this extra dimension provided by a HI_PHY increases flexibility to combat the hostile noise environment in upstream channels, which results in higher bandwidth efficiency.
To understand how the extra dimension provided by a HI_PHY increases noise immunity, its behavior in the presence of both ingress and burst noise should be examined. For a multi-tone HI_PHY, the upstream channel is divided into many frequency subchannels, each with its own carrier (or tone) that is individually modulated by the subscriber's data. The carrier in each subchannel is modulated with a QAM constellation that varies from BPSK to 256 QAM in response to the ingress measured in the subchannel at the headend. As the ingress level increases, the subchannel(s) in which the ingress appears is deactivated, and the subscriber's data continues to be carried in subchannels that are ingress-free (see Figure 1, page 49).
As a result, the overall channel will deliver a constant bit error rate (BER) while the channel throughput degrades gracefully. Single-carrier QPSK, on the other hand, delivers degraded BER as the ingress increases, until the channel reaches the point where it can no longer deliver acceptable performance. At this point, the traffic is switched to another channel, which interrupts data transmission.
A spread spectrum HI_PHY uses the codespace dimension to spread the subscriber's data across a wide frequency band. The data is carried via assigned spreading codes on a QAM carrier. At the headend receiver, the data is recovered using a de-spreading process. This de-spreading process causes a narrowband ingress at the receiver input to be spread across a wide frequency band, making the ingress appear as white noise at the output of the de-spreader. As a result, the impact of ingress on data throughput is reduced via the codespace dimension.
While single-carrier QPSK relies exclusively on a time domain solution to burst noise (e.g. FEC and interleaving), a HI_PHY relies on both the time and frequency (or codespace) domains to more effectively combat burst noise. In addition to FEC and data interleaving in the time domain, a HI_PHY can interleave the data across multiple tones (or codes) per frequency channel to increase the immunity to burst noise. A HI_PHY distributes the transmitted data across N frequency subchannels (or N codes), and as a result, the symbol rate per subchannel (or code) is lower and the symbol period is longer than for a single-carrier QPSK channel. This longer symbol period effectively reduces the burst energy in the recovered symbols, thus improving the BER. The amount of improvement in channel performance resulting from the longer symbol periods can be quantified as 10logN, where N is the number of tones (or codes) available in a symbol period.Differences
The differences between HI_PHYs can be described in terms of their implementation details or in terms of their expected performance in real plant environments. Although a discussion of implementation details would reveal some important differences, performance-related issues are focused on here. Table 2 summarizes some of these differences for the most popular approaches.
All HI_PHYs provide improved immunity to ingress as compared to single-carrier QPSK and 16 QAM. However, within the multi-tone group, there are performance differences resulting from the spectral shaping of the transmitted tones. The sharper the frequency rolloff of each tone, the more rejection provided to a narrowband ingress. Because VCMT and DWMT provide very sharp spectral shaping of each tone, they can reject the ingress better than OFDM and DMT.
In the case of CDMA, the ingress immunity is dependent on the number of subscribers that are active in the frequency channel. As contrasted to wireless applications, the total power in the upstream cable plant is limited and has to be shared by all subscribers that are transmitting at the same time. As the number of active subscribers in a channel increases, the power level of each subscriber's transmitter decreases while the noise level in the channel increases, thus decreasing the signal-to-noise ratio at the headend receiver. Therefore, the ability of CDMA to withstand ingress decreases as the number of active subscribers in a channel increases.
Another important difference in performance between approaches is in the area of bandwidth efficiency, which consists of two main factors: channel bit loading and burst guard time. The channel bit loading delivered by a HI_PHY is a function of the number of data bits that can be carried per transmitted symbol. For multi-tone modulation, the bit loading per tone changes from 1 bit/sec/Hz (for BPSK) up to 8 bits/sec/Hz (for 256 QAM) in direct response to the amount of noise measured in each subchannel. Subchannels with low noise carry more bits than subchannels with high noise, with the result being the maximum possible bit loading in the overall channel based on the noise characteristics of that channel.
For S-CDMA, the channel bit loading is dependent on the modulation index of the RF carrier used to send the spreading codes. The higher the modulation index, the higher the SNR required at the receiver to recover the spreading codes. To maximize the number of subscribers that can be transmitting at the same time in a channel, a lower modulation index for the RF carrier (and thus a lower SNR required at the receiver) is prudent. As a result, the channel bit loading for S-CDMA is kept low in order to support the maximum number of subscribers in a given channel.
The second determining factor is burst guard time (discussed earlier). Because the channel overhead increases as a percentage of the revenue-generating traffic when channel bit loading increases, a HI_PHY that requires no guard time makes more of the upstream channel available for revenue generating traffic. Both VCMT and S-CDMA require no guard time between successive bursts, while DWMT requires a 4 symbol guard time. Although S-DMT and OFDM don't require a guard time, a preamble (or cyclic prefix) is used to train the receiver and reduce intersymbol interference between bursts. This preamble is a fixed length and causes the overall channel efficiency to decrease as the burst length decreases (i.e. for higher bit loading or for shorter voice packets).
Finally, differences in sensitivity to ranging offsets between subscribers is important. The collection of cable equipment between the subscriber and the headend typically introduces variations in time delay and phase over the frequency range of a single upstream channel. Delays through block upconvertors and the cable itself vary depending on time of day and outside temperature. These dynamic variations can present particular challenges to some methods that depend on precise phase relationship between adjacent frequency channels and between adjacent symbols. VCMT, S-DMT and OFDM are reasonably tolerant to timing differences between subscribers, while S-CDMA is more sensitive to timing differences between subscribers. DWMT is not only sensitive to timing differences between subscribers, it is also sensitive to RF carrier phase offsets between subscribers.Conclusions
A HI_PHY can more effectively deal with the highly dynamic noise characteristics of upstream cable networks as compared to single-carrier QPSK or 16 QAM. By continuously monitoring the noise on each tone (or code), a HI_PHY receiver spots any noise and then adjusts the mapping of bits onto tones (codes) to minimize the impact of that noise. A HI_PHY solves two important problems for cable operators. First, it reduces the amount of upfront investment and ongoing maintenance costs required to support two-way data service in existing plants, thus accelerating service uptake in more serving areas. Second, a HI_PHY increases the upstream channel efficiency so the operator's investment in networking equipment can support a larger subscriber base.
In addition to opening up regions of the frequency spectrum that are unusable today because of high noise impairments, a HI_PHY increases the efficiency of all upstream channels. Higher channel efficiency means more bandwidth is available to support more revenue-generating traffic. Cable operators can increase the revenue generated by existing upstream spectrum by fourfold.
HI_PHY frequency channels can coexist with already deployed QPSK channels to minimize upgrade costs and to eliminate disruption of existing services while transitioning from first-generation to second-generation network infrastructures. Finally, a HI_PHY can be integrated using readily available semiconductor technology at a cost and power level that is comparable to that of a QPSK upstream solution.
|About the author|
|Rod Gross is VP of marketing with Ultracom Communications.|