Cable
Modem Certification
CableLabs’ Data Over Cable Service Interface Specification (DOCSIS) standard paves the way toward vendor-independent interoperability for cable modems.
By Gregory Smith
Why is cable modem certification so important when cable modems operate on
private cable networks owned by the multiple-systems operators (MSOs) such as Time Warner, TCI,
COX, and MediaOne? This article summarizes what certification is all about by explaining why it’s necessary, what the overall goals of the certification process are, and what is required to obtain CableLabs certification.
Data Over Cable Service Interface Specification (DOCSIS) is a series of specifications for Multimedia Cable Network Systems (MCNS). These specifications cover all aspects of a DOCSIS-compatible modem from the physical layer (PHY) through the networking layer of the 7-layer
ISO networking standard. It includes specifications for both a cable modem termination system (CMTS) and the cable modem itself.
The DOCSIS standard evolved from private systems (such as GI and Bay Networks) that added their proprietary systems to a technology pool that could be utilized by any manufacturer wishing to build DOCSIS-compliant modems. To date, over a dozen vendors are trying to manufacture DOCSIS-compatible modems, several companies are building silicon for cable modems, and a host of
organizations are providing support software for debug and service enhancements to the standard.
Retail, here we come
Rather than offer proprietary cable modems that could not interoperate on the same cable network, Cable-Labs wanted to ensure interoperability for any cable modem manufactured by any vendor. The DOCSIS standard was developed to sell services, not hardware, and these services must compete with the price and performance of standard dial-up modems. By allowing an
inexpensive retail channel to develop and capitalize on open architectures (promoting competition), the price of the hardware drops, and the cable modem becomes attractive to the general consumer. For cable operators that can entice 20% of their subscribers to add cable modem service, the numbers are staggering. For a 1-million subscriber network, the annual income can approach $100 million — not a bad additional source of revenue.
Carrying DOCSIS
As shown in
Figure 1
, a typical cable system has many elements in common with a data network. Many cable networks begin with a super-headend connected to a dual redundant SONET ring that links several smaller headends together. These elements of the system (which are monitored and serviced when necessary) are usually remote buildings, housing projects, or pole-mounted equipment. Emanating from each headend are fiber nodes that link multiple cable nodes into a star configuration. The cable branching out
from these nodes is called a trunking cable and typically terminates in multiple distribution cables, which supply the individual customers via drops. Each drop utilizes a small portion of the signal that is sent through a tap to customer-furnished equipment (CFE) such as modems, televisions, and telephony devices. As each drop reduces the signal amplitude on the cable, distribution amplifiers must be placed at frequent intervals along the cable to boost the signal level. A distribution link can have three to
five (or more) distribution amplifiers.
The spectrum on the cable plant allows for downstream, or forward, path signals that occupy the 54- to 860-MHz band, with channels spaced at 6 MHz. NTSC video signals can occupy a significant portion of these channels at lower frequencies, while the higher frequencies may be allocated to DOCSIS signals. The upstream, or return path signals generally occupy from 5 to 42 MHz and can have variable channel spacing, depending on the signal’s type and format.
European DOCSIS is proposing 80 to 860 MHz on the forward path at 8-MHz channel spacing, and 5 to 65 MHz for the reverse path.
RFI, BPI, TRI, and OSSI
This article concentrates on the DOCSIS Radio Frequency Interface (RFI) specification that covers the PHY and data-link layer of the 7-layer ISO network model. The network layer is Internet Protocol (IP) version 4 and allows for migration to version 6. Several network management protocols are also called for, including SNMP, TFTP,
DHCP, and security management protocol as defined in the DOCSIS security system interface specification. The network management protocols are covered by the Operations and System Support Interface (OSSI) specification and cover the management information base (MIB) objects that must be updated and reported through SNMP. The Telephony Return Interface (TRI) and the Baseline Privacy Interface (BPI) specifications outline telephony return modems (no return path) and privacy/access issues. In general, TRI testing
is not as prevalent, as most modems have return-path capability. Baseline privacy testing is generally done, but not all aspects are covered.
The RFI specification defines in detail the PHY transmission formats between the CMTS and the CABLE MODEM. The RFI specification also details the media access control (MAC) sublayer of the data-link layer, but stops short of the logical link control (LLC) sublayer which must be provided in accordance with ISO/IEC 10039.
The first three sections of the
RFI specification cover PHY layer impairments, pertinent reference documents, data forwarding rules, and reference diagrams of a data-over-cable system. Sections 4 and 5 explain the PHY transmission formats, including modulation, symbol rates, channel spacing, filtering, randomization, FEC, and frame structure. Section 6 covers the MAC layer, with an emphasis on MAC framing, time synchronization, upstream MAC message structure, and encryption support. Section 7 defines cable modem/ CMTS interactions, and
sections 8 and 9 cover future enhancements to DOCSIS cable modems. Most of the specifications tested at CableLabs are covered in sections 4 through 7.
Certification “how-to”
In order to verify that the cable modem vendor community is conforming to the DOCSIS specification, a number of testing documents were generated at CableLabs to enhance the certification process. The final result is certification by the Cable- Labs board, which is made up of representatives of the MSOs
that help fund the organization. Three documents cover the testing process: the conformance checklist (PICS Proforma or PICS), the acceptance test plan (ATP), and the ATP development status bulletin. The PICS Proforma is a list of RFI, BPI, TRI, and OSSI subsections correlated with PICS item names, an explanation of the item, and whether the PIC item is mandatory or optional.
The ATP defines a series of tests that are utilized to verify the various PICS items. These tests can cover PHY items, using
appropriate test equipment, or MAC and higher layer items, using browsers that display the MIB values called out in the OSSI specification. The ATP development status bulletin correlates the ATPs with PICS items and the appropriate interface specification.
The risks
To begin the certification process, the over 1,200 testable items in the PICS should be reviewed. Next, the time needed to complete the testing process should be determined. PICS testing can be time consuming and may involve
plenty of personnel, expensive test equipment, specialized software tools, and careful documentation of the results. The testing team will constantly be feeding hardware and software design changes back to the research and development team (see
Figure 2
). These changes will need to be incorporated into the design and a new preproduction run made so that the units can be tested. The typical number of internal cycles for this process is at least three board turns. A
verification of all the PICS items can take six people 4 to 6 months by the time test procedures are written, run, and documented. However, the ATP suite from CableLabs is far from complete, and many cable modem vendors supplement the suite with their own test procedures.
After internal PICS testing, plenty of time should be allowed for the certification process. The actual testing at CableLabs takes about 6 to 8 weeks. Vendors never (as of yet)obtain certification in the first round and must wait 3
months until the next certification interval (known as cert waves, they occur only four times per year). This can make for a total scheduled test time of a year or more, just about the time you’ll be thinking about certifying the next cost-reduced version of the modem. In addition, just getting in the door at CableLabs is $55k for two certification attempts. Typical documentation delivered to support certification can be a stack of binders a foot or more high.
Start at the beginning
A DOCSIS system consists of a CMTS located at the headend and any number of cable modems distributed at various taps on the cable plant. The headends are remote sites, so monitoring capability (covered by the OSSI) is important to ensure that the system is working properly. Unless a CMTS is being built, however, it isn’t necessary to deal with any of the PICS CMTS items. The cable modem is an intense testing effort by itself, with approximately 1,000 PICS items to be tested (see
Table 1
). Because of compatibility issues, testing is always noninvasive, and CableLabs will not use any test ports or open the cable modem to access test points.
DOCSIS utilizes a TDM downstream with DOCSIS MAC frames encapsulated into MPEG frames, called the downstream transmission convergence sublayer. The data rates, symbol rates, FEC encoding, and modulation formats have been around for some time and represent a more mature technology than the return path. Hence, the
testing for forward path signals tends to concentrate on verification of the symbol rates, input level, and bit error rate (BER) performance. The BER performance is measured using MIB values concerned with Reed-Solomon blocks in the FEC. Vendors are encouraged to supplement the CableLabs BER measurements with their own measurements, using a BER tester (BERT) and a suitable interface into the modem.
A majority of the RFI PICS are devoted to testing the return path (or upstream), which utilizes a TDMA
format with timing provided from the CMTS. Bursts can vary in length depending on the type, symbol rate, modulation format, and FEC within the burst. The most important (and difficult) criteria for measurement include spurious, between burst spurious, transient, signal level, and signal distortion. Measuring burst parameters can be difficult and generally involves the use of sophisticated equipment (the most widely used is the HP 89441A transmission analyzer). If the 89441A is the chosen instrument, time
should be allotted to become familiar with its operation.
The remaining RFI items are related to MAC measurements, which include frame formats, MAC management messages, upstream bandwidth allocation, timing reference, encryption, ranging, registration, PHY layer adjustments, fault recovery, configuration, and higher network layer interface issues. A CMTS is needed to perform these and most of the RFI-PHY layer tests. The MAC tests involve tweaking parameters in the CMTS, and verifying them using SNMP to
display the MIB values. Another piece of specialized test equipment that can save time is SmartBits, an Ethernet packet generation device, which can verify end-to-end transmission of packets from the CMTS to the cable modem and back.
BPI, TRI, and OSSI PICS items
Baseline privacy, telephony return and operations, and system support constitute the remainder of the PICS items. I know of no active testing under way to support TRI, except within cable modem vendors that supply this
option. CableLabs test procedures do not cover TRI. BPI tests include unicast, multicast, cryptographic, baseline privacy key management (BPKM), contention packet transfer, and other miscellaneous tests. OSSI tests include MIB walks, SNMP V1 Protocol, CPE-side behavior, MIB verification, network management access, and event reporting. These tests can be tedious, and can depend on an examination of MIB values and determining the accuracy of the values by using actual physical measurements.
Where are
the reinforcements?
DOCSIS testing should not be taken lightly. In fact, the testing aspects for certification can weigh in heavily when it is time for management to cough up research and development funds needed to develop a cable modem. However, help is available and, if used properly, can save considerable amounts of test time.
A silicon provider should be consulted to see if they offer a silicon PICS, which can cover a significant amount of the RFI PICS items. These items may not
need to be verified if CableLabs accepts the silicon PICS. This can save a substantial amount of test time by eliminating the time used to substantiate PICS items that are entirely implemented in silicon. These items include FEC format, filtering, modulation format, and randomization for the RFI PHY, and timing, MAC format, and management messages for the RFI MAC.
It is a good idea to obtain as much information as possible regarding support tools from the silicon vendor. This information may concern
the ability to transmit bursts without a CMTS, measuring BER on the downstream, and providing SNR measurements at the downstream demod. These can be valuable tools when verifying the cable modem in design verification and test (DVT).
It is necessary to understand how the CMTS operates, and training from the CMTS provider should be sought. A good user interface to the CMTS can be valuable for setting various upstream parameters and verifying their operation. Since almost all cable modem
transmission characteristics are controlled by the CMTS, the ability to control the test parameters is important. Additionally, the CMTS vendor should be contacted about diagnostic tools for the upstream burst demodulator. Measurement parameters such as missed packets, errored packets, and corrected packets can be an indication of potential problems in the cable modem.
To obtain more information about the certification process contact CableLabs at www.cablelabs.com. Several of the RFI documents can be
downloaded from their Web site.
The end is near
Getting certification can be a great feeling, particularly if you’ve been busting your hump for 6 months, and are beginning to think that it will never happen (at least while you’re employed at your present occupation). One vendor with whom I consulted handed out T-shirts after getting certified with a timeline of all the major milestones printed on the back. I don’t know about everybody else, but I’m wearing mine to the
Christmas dance.
Gregory Smith received a BSMOfrom the University of Michigan and an MSEEfrom San Diego State University. He can be contacted at
gssmith@incom.net
.
Return to the
Table of Contents
Puzzled by a network processing design issue?
