Issues in Migration from DOCSIS 1.0 to 1.1 to 2.0: Issues and Challenges. Victor T. Hou & Burcak Beser

Transcription

1 Issues in Migration from DOCSIS 1.0 to 1.1 to 2.0: Issues and Challenges Introduction Victor T. Hou & Burcak Beser After release of the first-generation specifications at the end of 1996, the DOCSIS 1.0 specifications for interfaces related to cable modem and cable modem termination system technology have essentially captured the market for high-speed data access over broadband cable networks in North America and much of the world. In 1999, a new generation of DOCSIS specifications was issued under the DOCSIS 1.1 umbrella defining enhanced capability in CMs and CMTSs to provide an increased degree of support for Quality of Service (QoS) and security among other enhancements and additions. At the very end of 2001, the DOCSIS 2.0 specifications appeared providing increased capacity and additional channel robustness. Departing somewhat from precedent in 1.0 and 1.1, the specification also mandated support of both advanced Time Division Multiple Access (TDMA) and Synchronous Code Division Multiple Access (S-CDMA) in both CM and CMTSs whereas the traditional pattern was that the CM was required to support all modes or functions while the CMTS could optionally support various modes and functions. Given deployed 1.0 and 1.1 CMs and 2.0-capable CMs to come, issues regarding transitioning from one DOCSIS release are an important discussion. Related to that are issues regarding upgradability of hardware and software in CMs and CMTSs and support of co-existence of CMs of different vintages on one DOCSIS cable modem system. Finally, DOCSIS 2.0 brings potential benefits but also new system considerations that need to be understood for successful deployment when the time is right. DOCSIS 1.0 and 1.1 Despite material that says otherwise, it is not true to say that DOCSIS 1.0 upstream channels can only support a maximum channel rate of 5.12 Mbps. From the very beginning, DOCSIS 1.0 supported a mode to achieve a Mbps channel rate 16- QAM modulation at the symbol rate of 2.56 Msps. DOCSIS 1.1 did not change that. With high performance, low-implementation loss 1.0/1.1 CMTS receiver designs with ingress cancellation and equalization (post-equalization and pre-equalization support) mechanisms, many plants can take advantage of the Mbps channel rate today in DOCSIS 1.0 or 1.1 mode compared with the most prevalent legacy DOCSIS CMTSs. Being able to operate with a channel rate of Mbps is two to four times the channel

2 rate typically used in DOCSIS systems today for DOCSIS 1.0. Thus, it could be said that DOCSIS 1.0 is underutilized today in that respect. In the physical layer, DOCSIS 1.1 made only one improvement. It defined a preequalizer to be implemented in the CM to aid in mitigation of channel distortions. However, before pre-equalization can be turned on for DOCSIS 1.1 CMs on an upstream channel, some 1.0 CMs may require a software update in order to be compatible on an upstream channel with pre-equalization enabled. At the MAC layer, DOCSIS 1.1 included a large number of enhancements related to Quality of Service (QoS) to DOCSIS 1.0. However, it is not completely true to say that DOCSIS 1.0 does not allow for QoS or service tiering to be provided. Some basic QoS features, as well as service tiering, can be provided with DOCSIS 1.0. However, it is helpful to note that the DOCSIS 1.0 specification uses the term of Class of Service (CoS) while DOCSIS 1.1 uses the term QoS. Thus, for dialogue within the DOCSIS community, restricting the term CoS for use in the DOCSIS 1.0 context and the term QoS for use in the DOCSIS 1.1 context is helpful indeed. DOCSIS 1.1 does allow dynamic establishing of service flows as well as dynamic deletion of service flows. On the other hand, for DOCSIS 1.0, CoS associated with a SID could only be established at the time of registration. Another fundamental distinction between DOCSIS 1.0 and 1.1 involves the terms Service Identifier (SID) and service flow. In DOCSIS 1.0, a SID corresponds to a bi-directional flow, both upstream and downstream. The SID is the identifier by which a CM makes requests for upstream bandwidth and by which the CMTS grants upstream transmission opportunities via the MAP messages. Although DOCSIS 1.0 allows for a CM to have more than one SID, in actual practice, not a single DOCSIS 1.0 CM supports more than one SID, and therefore, in practice within a MAC Domain, DOCSIS 1.0 has a one-to-one association between a CM and a SID. SIDs are only unique within a MAC domain, which is simply a collection of downstream channels and upstream channels managed under one DOCSIS MAC entity. In DOCSIS 1.1, the term service flow is introduced and refers to a uni-directional flow. Thus, upstream service flows are distinct from downstream service flows, and each can have different QoS parameters. In DOCSIS 1.1, the term SID is still used, and it applies to an upstream service flow that is admitted or active. At minimum, a CM operating in DOCSIS 1.1 mode will have two service flows, one for the upstream and one for the downstream. Furthermore, in DOCSIS 1.1, CMs will typically have multiple service flows in the upstream and in the downstream in order to support applications such as VoIP that require more than best-effort QoS on top of best-effort or premium data service. With DOCSIS 1.1, a much more complete set of QoS tools is provided in the DOCSIS 1.1 toolkit. QoS mechanisms in the toolkit include fragmentation, payload header suppression, classification, unsolicited grant scheduling and other scheduling services,

3 and so forth. As with DOCSIS 1.0, use of the tools or mechanisms is dependent on service provider choice and CM and CMTS capabilities. Regarding security, DOCSIS 1.1 requires support of Baseline Privacy Plus (BPI+) in CM and CMTSs, as opposed to just Baseline Privacy. BPI+ introduced stronger cryptographic mechanisms as well as CM authentication via X.509 digital certificates. Included in BPI+ is secure software download functionality when CMs need to upgraded with new software images. However, even Baseline Privacy is typically not enabled today in most systems. State of Affairs Today In North America, at the end of 2002, there was virtually no deployment of DOCSIS 1.1 in any cable system. One exception is General Cable Incorporated (GCI), which has reported its experiences with 1.1 and tiered service offerings based upon rate limits and monthly usage limits [3]. Outside of North America, there have reportedly been only a few 1.1 deployments despite availability of DOCSIS 1.1 certified and qualified equipment. A major reason for this is the fact that DOCSIS 1.0 systems operate sufficiently well to produce a substantial revenue stream from users paying a flat fee for a level of service that is typically curbed by upstream and downstream rate limits. However, there are a number of reasons to upgrade to 1.1 and obtain the additional revenue that can be enabled by utilizing DOCSIS 1.1 features. Upgrade to DOCSIS 1.1 Reasons for upgrading to DOCSIS 1.1 involve the additional features and mechanisms that DOCSIS 1.1 makes available. As already mentioned, these are some of the major capabilities introduced by DOCSIS 1.1: Quality of Service Mechanisms Baseline Privacy Plus Standardized and mandatory pre-equalization support This paper does not attempt to discuss the full scope of enhancements in DOCSIS 1.1 over DOCSIS 1.0. For more specific information, consult the CableLabs specifications [5]. In particular, the first item makes possible the support of Voice-over-IP (VoIP), videoconferencing, and other multimedia type services as well as various service tiers including premium type data services with minimum rates in both the upstream and downstream that would be desirable for users with more demanding needs and applications.

4 Some DOCSIS-certified 1.0 CMs are incapable of being upgraded to 1.1 because of hardware limitations, such as lack of sufficient memory or being able to only support one upstream SID. In other cases, no 1.1 capable software version may exist for the particular CM. Some CMs that are upgraded from 1.0 to 1.1 may not be able run BPI+ due to an incorrect or missing X.509 certificate or authorization keys (RSA private key) at manufacture. Some CMTSs and CMTS line cards are recommended not to be upgraded from DOCSIS 1.0 to DOCSIS 1.1 due to real-time processing limitations. However, other CMTSs are software upgradeable from DOCSIS 1.0 to 1.1 or are already running a DOCSIS 1.1- capable software version but configured to operate in 1.0 mode. Before a DOCSIS 1.1 capable CM is provisioned through the configuration file for 1.1, the CMTS must be upgraded and configured to operate in 1.1 mode. DOCSIS 2.0 In contrast to DOCSIS 1.1, which provided enhancements related to the Medium Access Control (MAC), QoS, security, and network management, DOCSIS 2.0 mainly addresses enhancements and new functionality in the upstream physical layer. There was no intent to impact the downstream physical layer. The basic benefits of DOCSIS 2.0 are as follows: Additional capacity Increased robustness There are limits to achieving both objectives. If a plant is not in adequate condition, the robustness mechanisms may not be enough to run at a higher order of modulation. DOCSIS 2.0 is not a cure-all to enable running 64-QAM anywhere and everywhere. In some cases, one may only have a significant benefit in either increased capacity or increased channel robustness but not both. Also, it should be noted that certain CMTSs operating today in DOCSIS 1.0 and 1.1 modes have robustness mechanisms that allow effective operation in certain challenging plants. These would be CMTSs with built-in ingress cancellation and post-equalization, as well as upstream channel monitoring to enable automatic switching between 16-QAM and QPSK or between operator-prescribed symbol rates, or between frequency channels, all with the goal of maximizing total throughput. The first bullet is achieved in multiple ways. Firstly, increased spectral efficiency is possible primarily through more efficient orders of modulation. DOCSIS 2.0 adds 32- QAM, 64-QAM, and 128-QAM corresponding respectively to 5, 6, and 7 bits per modulation symbol. However, it should be mentioned that 128-QAM is only used in conjunction with Trellis Coded Modulation (TCM) for S-CDMA such that the information bit rate is equivalent to uncoded 64-QAM. Also, DOCSIS 2.0 adds 8-QAM so that the selection of modulation schemes provides 2, 3, 4, 5, and effectively 6 bits per

5 symbol. Thus, finer modulation granularity is provided with DOCSIS 2.0. Secondly, capacity is increased through an additional higher symbol rate of 5.12 Msps corresponding to a 6.4 MHz channel bandwidth. The higher symbol rate is equivalent to the other five symbol rates in terms of spectral efficiency, but because it is a wider channel and thus a bigger bandwidth pipe, there is some additional statistical multiplexing gain that increases effective capacity also. As for the second bullet, the mechanisms that increase robustness include: Increased Reed-Solomon error correction Upstream burst interleaving Increased number of taps in the pre-equalizer structure Another major addition to DOCSIS 2.0 is the addition of the multiple access scheme S- CDMA. Various papers have compared CDMA technology with TDMA in contexts other than return path on HFC plants [7] [8], but in addition, there have been previous analyses of TDMA and S-CDMA in DOCSIS 2.0 [2]. Fundamentally, TDMA and CDMA have equivalent theoretical channel capacity, and specifically DOCSIS 2.0 TDMA and S-CDMA have equivalent channel capacity also. Different discussions have focused on particular types of cable plant impairments additive white Gaussian noise (AWGN), impulse noise, narrowband ingress, common path distortion (CPD), amplitude and frequency distortions, and so forth and whether DOCSIS 2.0 TDMA or S-CDMA is more effective. This paper does not attempt to cover that ground. Upgrade from DOCSIS 1.0 to 2.0 or to 1.1 First? A fundamental question being asked by MSOs today is whether to wait for 2.0 or to deploy 1.1. Linked with this question is whether it makes sense to eventually upgrade from DOCSIS 1.0 to 2.0 operation in one swoop or upgrade to 1.1 first. There are several things to consider. There is a learning curve involved in the transition from DOCSIS 1.0 to 1.1. With DOCSIS 1.1, specific line item requirements are increased approximately four fold over DOCSIS 1.0. Although an operator may not configure the use of the full assortment of 1.1 features, there is an educational process involved understanding and working with 1.1. As of the beginning of the year, nearly all CMs deployed in the field today are operating with a 1.0 code base. Many of those CMs are 1.1 upgradeable but will require a software upgrade to a DOCSIS 1.1 code base. A cable plant operating with the full physical layer capabilities of DOCSIS 1.0 (i.e., 16-QAM in the upstream and 256-QAM in the downstream), should be able to support the full implementation of DOCSIS 1.1 since the physical layer remains essentially the same. The purpose for upgrading to DOCSIS 1.1 is not

6 for physical layer benefit although pre-equalization in the CMs does come with DOCSIS 1.1. A number of operational issues are involved in an upgrade from 1.0 to 1.1. Several were discussed in [4]. Some of these issues include: Operating DOCSIS v1.1 CMs in either v1.0 mode or v1.1 mode. There are 4 defined modes of operation for a v1.1 CM. Operators should consider their options and which they plan to support. DOCSIS v1.1 configuration files use Object Identifiers (OIDs) from RFC 2669, the required DOCSIS Cable Device MIB for v1.1, as opposed to OIDs from what is known as the CD-04 MIB that were used with original v1.0 CMs. Secure Software Download (SSD), a new feature for v1.1 CMs. SNMP protocol choices SNMPv1, SNMPv2c, or SNMPv3 Network Management Access. In DOCSIS v1.0, there is the docsdevnmaccess table as defined by RFC2669. DOCSIS v1.1 introduces SNMPv3 and the concept of the View-based Access Control Model (VACM). DOCSIS v1.1 Event Notification, including standardized events and added structure to traps and syslog for ease of parsing. DOCSIS v1.1 includes new Management Information Bases (MIBs). The amount of information available to operators, both for control and reporting, has grown and must be managed properly. Account Management is defined for CMTSs, allowing operators the opportunity to launch scalable services using usage-based billing. The DOCSIS Subscriber Management MIB for protocol filtering at the CMTS. A CPE-side diagnostic address in the CM is defined, including what information should be made available for diagnostics. Likewise, operational issues would be involved in an upgrade from DOCSIS 1.1 to 2.0. There is commercially available management software (i.e., SNMP agents supporting the MIB protocol, etc.), which can help to ease the burden of an upgrade to DOCSIS 1.1, whereas with DOCSIS 2.0, since the changes are mostly at the physical layer, additional technical expertise may be required. There is also a learning process involved in going from DOCSIS 1.1 to 2.0. Because the changes to 2.0 are predominantly with respect to the upstream physical layer, with most of the MAC layer changes related to supporting the additional physical layer functionality, from a technological point of view, 2.0 involves a somewhat different technical area than most of the enhancements in

7 1.1. The major enhancements of 1.1 are related to QoS, enabling PacketCable and tiered service offerings, and security while the major enhancements of 2.0 are related to the upstream physical layer. Depending on the organization, this may involve different subject matter experts. Specifically, there will have to be training regarding the new DOCSIS 2.0 physical layer features. The advanced TDMA features of DOCSIS 2.0 are built upon the DOCSIS 1.x physical layer. For S-CDMA, certain additional concepts need to be inculcated, for example, the familiar haystack spectral content of unspread QAM transmission will typically not be seen on an S-CDMA channel, but instead, spectral content that looks more like white noise. Most DOCSIS 1.0 or 1.1 capable CMs are not software upgradeable to DOCSIS 2.0. Thus, to support DOCSIS 2.0, new CMs are required. Some MSOs have mentioned that they will only purchase DOCSIS 2.0-upgradable CMs after some point in time during However, it is still early in the DOCSIS 2.0 experience. Just as some DOCSIS 1.0 that were claimed to be DOCSIS 1.1 upgradeable via software had certain issues, it is possible in the future that some CMs that are supposedly 2.0 upgradeable may have issue. Although products are now DOCSIS 2.0 certified and qualified, if past experience from 1.0 and 1.1 is a guide, it takes much additional time for interoperability and operational issues to be worked out. First of all, certification testing does not claim to be comprehensive nor exhaustive. As time goes on, more experience is gained in realistic scenarios and more combinations of configurable operational parameters are tried that may reveal further interoperability or operational issues. That does not take away from the fact that tremendous and rapid progress has been made regarding DOCSIS 2.0 such that specifications and test plans have been developed. Since most of the changes in DOCSIS 2.0 involve the physical layer, it is possible that an existing cable plant cannot make use of certain DOCSIS 2.0 capabilities, for example higher orders of modulation, without significant and possibly costly modifications to the plant infrastructure. A case can be made for upgrading from 1.0 to 1.1 then 2.0 or directly from 1.0 to 2.0. Of course, with any debate, a case can be made for either side of an issue. However, given the experience and educational requirements of both 1.1 and 2.0, it appears to be more prudent to subdivide the transition from 1.0 to 2.0 into two steps. In other words, it may be better to implement 1.1 first before fully upgrading to 2.0. In this way, operators can gain experience with 1.1 while offering additional services that result in increased revenues. A subsequent move to 2.0 can then be accomplished as additional upstream capacity is required. Although it may be tempting to convert to 2.0 directly in order to obtain the benefits of the additional robustness mechanisms, the leap could possibly be too large in terms of learning curve requirements. In addition, if higher order modulations are desired to be supported, it may be required to upgrade existing plants.

8 Of course, a CMTS capable of providing a straightforward and cost effective hardware and software upgrade path to 2.0 would allow the MSO to deploy 2.0 when the time is appropriate. Upstream Channels in DOCSIS 2.0 and How 1.0 and 1.1 CMs Can Co-Exist with 2.0 CMs There are three categories of types of upstream channels in DOCSIS 2.0 [6]: Type 1 channel: a channel using DOCSIS 1.x (1.0 or 1.1) burst descriptors only Type 2 channel: a channel using both DOCSIS 2.0 TDMA and DOCSIS 1.x burst descriptors Type 3 channel: a channel using only DOCSIS 2.0 Upon registration, a DOCSIS 2.0 CM categorizes the available upstreams into one of the three categories based on the Upstream Channel Descriptor message. A CM always tries to use the last upstream on which it successfully completed registration. This information would be stored as a channel ID in the non-volatile memory of the CM. If none of the available upstreams advertised in the UCDs matches the stored channel ID, then the CM selects upstream channels in the following order: It first attempts Type 3 channels. It then attempts Type 2 channels. It attempts Type 1 channels last. In this way, a CM attempts to use the most of its capabilities with respect to DOCSIS 2.0. A DOCSIS 1.1 CM that attempts to register will first try to use its last-used upstream channel ID. However, a DOCSIS 1.1 capable CM would only understand a UCD for a Type 1 or Type 2 channel because the UCD for a Type 3 channel uses a new MAC Management Message type value of 29 while a UCD for a Type 1 and Type 2 channel uses type value of 2, the type value specified for a UCD in DOCSIS 1.0 and DOCSIS 1.1. A DOCSIS 1.1 CM does not understand the difference between a Type 1 or Type 2 channel, and thus, cannot preferentially choose one over the other. A DOCSIS 1.0 CM likewise attempts to use its last-used channel ID. If that channel ID is not offered in the UCDs, then it will try a Type 1 or Type 2 channel. It will avoid Type 3 channels because it does not understand the MAC Management type value of 29. Like a DOCSIS 1.1 CM, it does not understand the difference between the two and cannot preferentially choose one over the other. So as just mentioned, a DOCSIS 1.x CM will not attempt to use a DOCSIS 2.0-only channel because it will ignore any UCD message identified with MAC Management Message Type=29. But in addition, a DOCSIS 1.x CM can choose an upstream channel that has both DOCSIS 1.x physical layer parameters and DOCSIS 2.0 TDMA physical layer parameters. In other words, the 1.x CM could select a UCD with MAC

9 Management Message Type=2 and then ignore the DOCSIS 2.0 TLVs within that UCD message. The bottom line is that whether a CM is DOCSIS 1.0 or 1.1, a DOCSIS 2.0 CMTS will be able to be backward compatible with them while being compatible with a CM operating in DOCSIS 2.0 mode. As part of DOCSIS 2.0, a new TLV is defined for the configuration file downloaded from a TFTP server during CM initialization. The Enable 2.0 Mode TLV can be set to Enable or Disable in the configuration file, and thus, the CM can be provisioned via the configuration file specifically to be able to operate in 2.0 mode or not. If upon initialization, the CM s configuration file has the Enable 2.0 Mode TLV set to Disable, and the CM registered as such, the CM will not operate in 2.0 mode until it registers again and downloads a configuration file which does not have this TLV set to Disable. This holds true even if the channel parameters in the UCD used by the CM are changed such that DOCSIS 2.0 parameters are supported. If the channel becomes a Type 3 channel, then the 2.0 CM with 2.0 mode disabled will have to reset and perform registration again. The DOCSIS 1.x CMs on that upstream channel would do the same since a Type 3 channel only supports CMs in 2.0 mode. On the other hand, if a CM registers with 2.0 mode enabled, even if it is only able to register on a channel without 2.0 support, if the channel parameters are later changed to support 2.0 mode, the CM would then operate in 2.0 mode, not in 1.x mode. In other words, the 1.0 CM tries to operate in 2.0 mode when it can but is limited by the Enable 2.0 Mode TLV in the configuration file or the parameters in the UCD on the channel the CM is using. The similar principle holds when a CM is moved to a different upstream channel via Upstream Channel Change and has to deal with different channel parameters, as opposed to merely a change in UCD parameters on the currently used channel. Virtual Upstream Channels Most of the CMTSs that are qualified for DOCSIS 1.0 and 1.1 support only one upstream DOCSIS channel on a single frequency. In this context, a DOCSIS channel is associated with an upstream channel ID, its physical layer parameters are defined in an Upstream Channel Descriptor associated with the channel ID, and its transmission regions are defined in a MAP message associated with the channel ID. The use of one UCD message for all CMs on the upstream channel means that all CMs on that upstream channel have to operate effectively with the upstream physical layer parameters in the UCD. These parameters would have to be set such that the CM or CMs with the most impaired path can work at some reasonable level. However, it is compatible with DOCSIS to have multiple upstream DOCSIS channels within a single upstream frequency channel. The individual DOCSIS channels within the single upstream frequency channel are termed here as Virtual Upstream Channels (VUC). The VUCs that share the same spectrum are kept separate from each other

10 through time division multiplexing that is managed by the CMTS upstream scheduler. A VUC can conceivably support down to one single CM and provide it unique physical layer parameters without affecting the remainder of the channel bandwidth usage. One of the most important aspects of the VUCs is that it is transparent to the Cable Modem that is certified for DOCSIS 1.0 and DOCSIS 1.1. From the viewpoint of the cable modem, the VUC is not any different than if there was only a single DOCSIS upstream channel. The difference is that when the modem is not transmitting, other modems may be transmitting with other physical layer parameters associated with other VUCs. It is possible, for example, to define VUCs in such a way that the only difference is in burst profiles for Short and Long Data while the opportunities for Requests, Request/Data, Initial Maintenance, and Station Maintenance are shared among the VUCs in a frequency channel and thus can be used by any of the CMs on any of the VUCs. An example application of VUCs would be to provide CMs with different orders of modulation or levels of FEC error correction depending on their individual channel paths and impairments. It is understood that a variety of impairments will impact all transmission from nodes combined into a receive port, but on the other hand, there are localized impairments that only impact a few CMs. For example, a malfunctioning or degraded amplifier may impact some CMs or a particular CM may have degraded over time (as has been reportedly observed). In addition, in the coax plant, some CMs may have a transmission plant through very few amplifiers while others may go through larger cascades of amplifiers resulting in significantly different distortion environments for different transmission paths. Thus, VUCs can be used to provide CMs in the same frequency channel with different modulation orders in order to maximize the overall spectral efficiency. In addition, VUCs could be used to have different CMs use different symbol rates on the same channel. Use of Virtual Upstream Channels does not put any restriction on the number of Cable Modems that use a particular upstream channel. Also, VUCs associated with the same frequency channel all use the same upstream receiver and therefore use of the VUC concept does not impact capital cost. Virtual Upstream Channels can be used to segment the cable modems according to the providers in an open access environment such that the total segmentation will be assured. The use of Virtual Upstream Channels also enables a provider to use default upstream settings that are more robust but less efficient for the CMs that are newly installed with unknown transmission performance. Later when the provider gains confidence on the performance of these CMs, they can be migrated to a less robust, but more efficient upstream physical layer settings. Logical channels DOCSIS 2.0 introduces the concept of logical channels in order to handle TDMA and S- CDMA being used on the same upstream frequency channel. Logical channels are a

11 subset of the previously discussed VUC concept. Within one upstream frequency channel, multiple logical channels can exist, each identified with an upstream channel ID within a DOCSIS MAC domain. When there is a transmission opportunity on one logical channel, another logical channel sharing the same spectrum is idle, with the exception of Initial Ranging opportunities that can be shared by multiple logical channels. Per [6], if a CM is using an upstream channel frequency with multiple logical channels, the CM is only operating on one of them. DOCSIS 2.0 defines four types of logical upstream channels. Type 1: DOCSIS 1.x upstream channels that support no DOCSIS 2.0 TDMA features or parameters Type 2: Upstream channels that support DOCSIS 1.x and DOCSIS 2.0 TDMA bursts Type 3A-TDMA: Upstream channels that support DOCSIS 2.0 TDMA mode and do not support DOCSIS 1.x CMs Type 3S-CDMA: Upstream channels that support only CMs operating in DOCSIS 2.0 S-CDMA mode The specification requires the following combination of logical channels to be supported. One Type 1 channel and one Type 3S-CDMA channel with the same modulation rate on both channels One Type 2 channel and one Type 3S-CDMA channel with the same modulation rate on both channels One Type 3A-TDMA channel and one Type 3S-CDMA channel with the same modulation rate on both channels Therefore, in DOCSIS 2.0, logical channels serve only the purpose of enabling TDMA and S-CDMA to co-exist on the same frequency channel. In reality, it is unlikely in real systems that a frequency channel will be configured with both TDMA and S-CDMA logical channels. There have been arguments for running TDMA and S-CDMA on the same plant in different upstream frequency channels based upon understanding that the two different technologies may be more resilient in different types of impairments. The necessity of doing that is left to be determined when more field experience is gained. Although the sole purpose defined for logical channels in DOCSIS 2.0 is being able to run S-CDMA and TDMA mode on the same frequency channel, as seen earlier in the discussion of VUCs, the VUC concept can be useful for a number of other purposes. System Considerations with DOCSIS 2.0 DOCSIS 2.0 requires consideration of various issues related to its successful operation. The various topics to be discussed include:

12 1. System synchronization and timing requirements for DOCSIS Removal of key ambiguities a. Constellation reference levels b. Power level versus power density 3. Further refinement and definition of pre-equalization System Synchronization and Timing Requirements DOCSIS 1.0 and 1.1 achieved system synchronization for upstream TDMA-based transmission through a method involving timestamp messages sent from the CMTS to the CMs [1]. Each CM implements some kind of control loop to receive the timestamp input and replicate the CMTS master timebase such that proper timing of burst transmission from CMs is achieved. The timestamp method was attractive because among other attributes, it was independent of the downstream transmission rate and the downstream modulation scheme. Since DOCSIS 2.0 incorporates S-CDMA technology, which has more stringent timing requirements than TDMA, another method of system synchronization is introduced in DOCSIS 2.0. DOCSIS 2.0 allows three options for the CMTS regarding the relationship between the CMTS master timebase and the downstream symbol clock (referenced to the nominal QAM symbol rate specified for 64-QAM and 256-QAM). Option 1: Not locked Option 2: Downstream symbol clock locked to CMTS master timebase Option 3: CMTS master timebase locked to downstream symbol clock For S-CDMA operation, the CMTS master timebase and the downstream symbol clock must be locked using either the second or third alternative. However, even though the timebase frequency is conveyed via the symbol clock, SYNC messages carrying timestamps are still sent in S-CDMA only operation in order to provide the CMs a reference point on the phase of the CMTS master timebase, in other words the specific value of the CMTS master timebase at a particular instant in time. DOCSIS 1.x did not require either the second or third alternative. For TDMA, the CM must be able to implement a ranging adjustment from the CMTS to an accuracy of ±0.25 microsecond plus ±1/2 symbol. For S-CDMA channels, the CM must implement the ranging adjust to within ±1% of the modulation interval. For example, if the highest modulation rate were used (5.12 MHz), a CM operating in TDMA mode must transmit with an accuracy of ±348 ns while in S-CDMA mode, the burst timing accuracy would have to be ±2 ns. This kind of timing accuracy is required to maintain orthogonality of the codes and prevent performance degradation due to intercode interference. However, it should be noted that this is the maximum allowable timing error allocated to a CM s burst timing based upon receiving the CMTS s ranging adjustment and not the overall system timing error specification, which would be somewhat greater (see [6], Appendix VIII ). Nevertheless, the system timing

13 requirements are much tighter for S-CDMA, and thus, mechanisms are defined in the specification to achieve the stringent system timing. One of the impacts of the tighter timing requirements for S-CDMA is consideration of additional impairments on the system. Timing variation caused by wind and temperature changes may not be a significant issue for TDMA, but they can be for S- CDMA. Wind can apply loading on aerial coax cable such that a change in length occurs and, as a result, impact the propagation delay. (Impact on optical fiber cable due to wind is less of an issue due to the way that a fiber cable typically has the propagation medium, i.e., the fiber, loose and separate from the cable strength members and the sheath). An example calculation given in [6] assumes an 8 km length of coaxial cable with 0.02% length variation due to wind loading: (8 km / (0.87( km/s)) ( ) = s= 6 nanoseconds where the 0.87 factor relates the propagation velocity of a signal through a coax cable to the speed of light in a vacuum, which is km/s. Six nanoseconds is significant in the case of S-CDMA when it is compared to the 2 nanosecond that is allocated to the CMs burst timing error at the highest modulation rate (5.12 MHz). Steep temperature ramps can have impact on optical fiber cable. [6] reports a value of optical fiber propagation delay change due to temperature change as 44 picoseconds per km per degree Celsius. Thus, a product of cable length and temperature change equaling 50 results in approximately a 2 ns change in propagation delay, which is the allocation to CM burst timing error for S-CDMA at the highest modulation rate (5.12 MHz). If the temperature swings are drastic for certain situations during a Periodic Maintenance (PM) interval, it may be required that the CMTS use shorter PM intervals to ensure that S-CDMA CMs continue to transmit successfully during severe temperature ramps or periods of very active wind gusts. As an example, severe temperature ramps could occur when the sun breaks out over the horizon or out of clouds and thermally loads blacksheathed aerial cable when the air temperature is cold. There apparently is no reported data on the effect of temperature and wind upon this aspect of DOCSIS 2.0 S-CDMA, but these are system issues to consider. Again, the CMTS may be able to help by reducing PM intervals. Removal of Key Ambiguities Also, DOCSIS 2.0 clarifies certain aspects of the specification that were not completely clear to everyone in DOCSIS 1.0 and 1.1. In particular, two examples are discussed here: 1. Upon a symbol rate change, is a CM supposed to maintain the same power level or maintain the same power spectral density? 2. In terms of power, what is the relationship between the QPSK constellation and the 16-QAM constellation in the upstream.

14 Regarding the first question, the understanding of the earliest authors of the section on the physical layer in DOCSIS 1.0 was that after a symbol rate change, the power level was to remain the same. However, not all DOCSIS 1.0 or 1.1 CMs for that matter were implemented that way; some attempt to maintain constant power spectral density, resulting in an increase or decrease of transmission level depending on whether the symbol rate was increased or decreased. DOCSIS 2.0 attempts to establish some determinism by placing a parameter in the Upstream Channel Descriptor (UCD) message that specifies whether power spectral density is to be maintained after a modulation (i.e., symbol) rate change or whether power level is to be maintained constant. On the second question, there was some divergence on how the specification was interpreted. Again, the original authors intended that for a given target transmit power level, whether the burst would be sent QPSK or 16-QAM, the power level was scaled such that if all constellation points were used equally, the burst would be equivalent to the target transmit power level. This is shown in the following figure. QPSK constellation 16QAM constellation Relative QPSK/16QAM constellation mapping However, in some actual implementations, some CMs basically scaled 16-QAM transmit power such that the corner points of the 16-QAM constellation overlaid the 4 QPSK constellation points, resulting in QPSK transmissions being approximately 2.7 db higher than 16-QAM. As a result, in the case of burst profiles in which some used QPSK and some used 16-QAM, the CMTS was limited in being able to use bursts other than Maintenance (i.e., ranging) bursts for power adjustment. It also produced confusion in that it seemed that 16-QAM bursts were about 3 db below the target transmit power referenced to QPSK, which is typically used for the Maintenance bursts. DOCSIS 2.0 specifies all constellations for the various upstream modulation formats on a grid that is a uniform reference for all the formats. Given that, the specification gives the

15 straightforwardly computable constellation gain of each constellation relative to the 64- QAM constellation. (The table below shows the various constellation gains.) Therefore, the specification is now clear on the power level to be transmitted by the CM for any constellation, and the CMTS can appropriately adjust around the nominal receive power that it is expecting given the modulation format used for the burst. Given a target transmit power, the specification also clarifies what the CM is to report as its transmit power in the MIB (a transmit power referenced to the 64-QAM constellation). DOCSIS 1.0 and 1.1 never explicitly referenced their constellation points to a grid that was common to both modulation formats that were used. Constellation Constellation Gain Relative to 64-QAM (db) QPSK QAM QAM QAM 0 64-QAM QAM 0.05 For both these issues (constant power spectral density or constant power, relationship of power levels among various constellations), it should be noted the clarification in DOCSIS 2.0 does not fix the existing issue for current DOCSIS 1.0 and 1.1 CMs. Further Refinement and Definition of Pre-equalization for 2.0 DOCSIS 2.0 specifies a 24-tap symbol-spaced pre-equalizer for the CMs. In DOCSIS 1.1, the CM was required to support a pre-equalizer structure with minimum of 8 symbolspaced taps. However, due to varying implementations from different silicon vendors when the 1.1 pre-equalizer specification was being clarified, the DOCSIS 1.1 specification allowed for the CM to support additional taps and other spacings. DOCSIS 2.0 specifies uniform pre-equalization for any CM operating in 2.0 mode. In addition, the 2.0 specification allows for movement of the equalizer main tap location during any maintenance (i.e., ranging) opportunity. On the other hand, with DOCSIS 1.1 the CMTS was restricted to only moving the main tap during Initial Maintenance, limiting the ability of the CMTS to dynamically adjust to varying distortion environments as time went on. Also, DOCSIS 2.0 allows a Ranging Response to specify whether the pre-equalizer coefficients are to be loaded directly or convolved. Loading coefficients directly is helpful when the CMTS knows that the CM sees a new distortion environment such as when it has just moved to a new frequency channel. These refinements involving pre-equalization only apply to DOCSIS 2.0 CMs and do not address pre-equalization implemented by DOCSIS 1.x CMs.

16 Conclusion MSOs face the opportunity and the challenge of going from DOCSIS 1.0 to 1.1 and then eventually to 2.0. The specifications for DOCSIS 1.1 were developed to ensure backward compatibility with DOCSIS 1.0, and the specifications for DOCSIS 2.0 were developed to ensure backward compatibility with DOCSIS 1.0 and 1.1. However, there are several issues to understand before making any migration. Given the learning curve involved with these transitions, it may be more prudent to transition from 1.0 to 1.1 first and then from 1.1 to 2.0 later. To make use of the additional capacity offered by 2.0, improvements may have to be made to cable plants. One technology that can ease migrations is use of Virtual Upstream Channels (VUCs), in which multiple DOCSIS channels exist on the same frequency channel. VUCs have hardware support in certain existing deployed CMTSs today. DOCSIS 2.0 brings a number of new features to possibly increase capacity and spectral efficiency mechanisms that can be configured in the CMTS. Primary among the choices to be made is deciding whether to use TDMA or S-CDMA. In addition, regarding S- CDMA, the tight timing requirements required new synchronization mechanisms. There is also the need to consider additional impairments such as wind and temperature change that may impact how well CMs stay synchronized when using S-CDMA. It should be mentioned that DOCSIS 2.0 also clarifies and provides solutions on certain technical points in DOCSIS 1.0 and 1.1 that resulted in divergent interpretations. Acknowledgements R. Harvey and Fabien Buda of Juniper Networks provided comments and review. References [1] V. T. Hou, R. Mullins, and B. Paratore, TDMA System Synchronization Using Timestamp Messages, Cable Modems: Current Technologies and Applications, ed. Venkata Majeti, IEEE Press, 1999, pp [2] F. Buda, E. Lemois, and H. Sari, An Analysis of the TDMA and S-CDMA Technologies of DOCSIS 2.0, Proceedings of the National Cable Television Association 2002, New Orleans, Louisiana, pp [3] R. Dowling, DOCSIS 1.1 Business Advantage: Worth the Tiers, session at Broadband Plus 2002: The New Western Show, Anaheim, California. [4] P. Siripunkaw and D. Jones, DOCSIS v1.1 Operations Enhancements, Proceedings of 2001 SCTE Cable-Tec Expo, Part One, Orlando, Florida, pp

17 [5] Data-over-Cable Service Interface Specifications RF Interface Specification, SP- RFIv1.1-I , Cable Television Laboratories, August, 2002 Louisville, Colorado. [6] Data-over-Cable Service Interface Specifications RF Interface Specification, SP- RFIv2.0-I , Cable Television Laboratories, December, 2002, Louisville, Colorado. [7] H. Sari, F. Vanhaverbeke, and M. Moeneclaey, Extending the Capacity of Multiple Access Channels, IEEE Communications Magazine, vol. 38, no.1, January 2000, pp [8] M. Moeneclaey, M. Van Bladel, and H. Sari, Sensitivity of Multiple Access Techniques to Narrowband Interference, IEEE Transactions on Communications, vol. 49, no. 3, March 2001, pp