By Jack Webb, Sunrise Telecom
Broadband Gear Report
DOCSIS is effectively a transparent Ethernet bridge over an HFC network. There are two functional components in a DOCSIS network - the cable modem (CM) on the subscriber side, and the cable modem termination system (CMTS) in the headend or hub site. The CMTS communicates with the CMs on a 6-MHz wide (8 MHz in EuroDOCSIS deployments), 64- or 256-QAM digitally encoded RF signal on the downstream path of an HFC network between 88 and 860 MHz. The CMs communicate with the CMTS using a quadrature phase shift keying (QPSK), 8-, 16-, 32- or 64-QAM digitally encoded RF signal transmitted on the upstream of the HFC plant at a frequency between 5 to 42 MHz (5-65 MHz Euro-DOCSIS).
The digital data includes media access control (MAC) information, which enables the CMs to coexist with other CMs by using a time division multiple access (TDMA) scheme. (DOCSIS 2.0 also supports synchronous "spread spectrum" code division multiple access - S-CDMA.) In essence, the CMTS is the system scheduler that coordinates the power level, frequency, transmit time and pre-equalization of all CM signals on the DOCSIS network.
Because the CMs and the CMTS are able to convey digital data between each other over the HFC network for "command-and-control" processes, they also are able to transmit packets containing other non-DOCSIS MAC related data. This fundamentally facilitates the ability to send Ethernet traffic bi-directionally over an HFC network. The CMTS-CM DOCSIS network transports IP-based traffic in the same method that is used to communicate MAC protocol between the devices. Now that the IP traffic can traverse the HFC network, end users are able to utilize this network to transmit/receive content destined for/from the multitude of available data network services such as email, Web browsing, IP video and VoIP.
In summary, each user is assigned a unique cable modem with a unique MAC address. The CMTS works as a system controller/scheduler enabling many cable modems to reside on the same RF network. TDMA and/or S-CDMA are employed to allocate a finite time for each cable modem to transmit and receive IP data. To the user the network appears to be a simple Ethernet bridge.
DOCSIS-Compliant Networks
Table 1 is from the DOCSIS 2.0 Radio Frequency Interface (RFI) specification and details RF specifications for a DOCSIS-compliant downstream HFC plant.
Table 1
| Parameter |
Value |
| Frequency Range |
Cable system normal downstream operating range is from 50 MHz to as
high as 860 MHz. However, the values in this table apply only at
frequencies >= 88 MHz. |
| RF channel spacing (design bandwidth) |
6 MHz |
| Transit delay from headend to most distant customer |
<= 0.800 msec (typically much less) |
| Carrier-to-noise ratio in a 6-MHz band |
Not less than 35 dB (see note 2,3) |
| Carrier-to-composite triple beat distortion ratio |
Not less than 41 dB (see note 2,3) |
| Carrier-to-composite second order distortion ratio |
Not less than 41 dB (see note 2,3) |
| Carrier-to-cross-modulation ratio |
Not less than 41 dB (see note 2,3) |
| Carrier-to-any other discrete interference (ingress) |
Not less than 41 dB (see note 2,3) |
| Amplitude ripple |
3 dB within the design bandwidth (see note 2) |
| Group delay ripple in the spectrum occupied by the CMTS |
75 ns within the design bandwidth (see note 2) |
| Microreflections bound for dominant echo |
-20 dBc @ <= 1.5 µsec, -30 dBc @ > 1.5 µsec -10 dBc @ <=
0.5 µsec, -15 dBc @ <= 1.0 µsec (see note 2) |
| Carrier hum modulation |
Not greater than -26 dBc (5%) (see note 2) |
| Burst noise |
Not longer than 25 µsec at a 10 Hz average rate (see note 2) |
| Maximum analog video carrier level at the CM input |
17 dBmV |
| Maximum number of analog carriers |
121 |
Notes:
1. Transmission is from the headend combiner to the CM input at the customer
location.
2. Measurement methods defined in [NCTA] or [CableLabs1].
3. Measured relative to a QAM signal that is equal to the nominal video level in
the plant.
Although carrier-to-noise ratio (CNR), composite second order (CSO) and composite triple beat (CTB) are generally considered to be "analog" impairments, they also impact RF modulated digital signals. CSO and CTB from intermodulating analog signals will fall under digital signals. These beats are not coherent, but will degrade the modulation error ratio (MER).
Microreflections impact the transmissions of a cable modem by reflecting the transmitted signal back toward the signal source. The incident and reflected signals interact to produce amplitude and group delay ripple, causing degraded MER and intersymbol interference (ISI). This may cause unexpected results including modem registration failures, intermittent data loss and poor voice quality in VoIP networks.
The DOCSIS standard requires a post forward error correction (FEC) bit error rate (BER) of 10-8 in a production network. This translates to no more than one error in every 100 million bits of data transmitted. To achieve this post-FEC BER, Table 2 shows the respective MER required for both 64- and 256-QAM downstream DOCSIS channels.
The carrier-to-interference specification of >25 dB is effectively CNR plus any ingress that may exist in the return path. The upstream DOCSIS carrier is especially susceptible to impairments such as impulse noise, which causes errors in the transmitted data. Higher order modulations, such as 64-QAM, may require more than 25 dB CNR to operate without errors.
Amplitude ripple and group delay can be grouped into a class together; if there are problems with one, it is likely there will be problems with the other. Amplitude vs. frequency or flatness, should ideally be flat. As ripples in amplitude increase, so will group delay. Both will impact a CMTS's ability to recover the signals transmitted by cable modems. Signals impaired by frequency perturbations and/or excessive group delay will suffer packet loss at the CMTS. While group delay has typically been considered an impairment at the return path band edges due to the roll-off of diplex filters, it can be present throughout the return path at any spectrum location suffering from frequency response impairments.
Microreflections are caused by impedance mismatches (poor return loss). Detecting microreflections is challenging, but it can be done with a time domain reflectometer (TDR), upstream sweep system, upstream characterization toolkits or, though ill-advised without proper testing, by replacing suspect cables and components.
While seemingly simple, the actual implementation of DOCSIS networks has many complex pitfalls that cause impairments and network failures. Impairments at the physical RF transport layer can result in poor or lost communications at the IP layer. Interoperability issues between various DOCSIS devices (CMTS and CM devices from multiple vendors) in addition to over-utilization of the DOCSIS network can result in packet loss, delay and jitter. Finally, all of the standard impairments that exist in Ethernet networks also are present in Ethernet networks over DOCSIS, such as collisions, delay, buffer overflows and routing errors, which result in packet loss, delay and jitter.
Upstream RF Impairments
The upstream path in an HFC network can be considered the "Achilles heel" of a VoIP system since it usually contains the greatest source of impairments. A short list and their DOCSIS specification follow:
Linear impairments:
| Microreflections | -10 dBc @ <= 0.5 µsec |
| -20 dBc @ <= 1.0 µsec |
| -30 dBc @ > 1.0 µsec |
| Amplitude ripple | (<0.5 dB/MHz) |
| Group delay | (<200 ns/MHz) |
Non-linear impairments:
| Common path distortion (CPD) | (> 25 dB) |
| Return laser clipping | (> 25 dB) |
The wide variety of upstream impairments can cause data from cable modems and embedded multimedia terminal adapters (EMTAs) to become corrupted before reaching the CMTS. If the CMTS is unable to properly demodulate a corrupted signal, it discards the frame. In normal data traffic, the data will be retransmitted by a higher level application, but for VoIP, there is no retransmission since VoIP is a real-time protocol (RTP). Lost frames are gone for good.
Carrier-to-Noise
To detect noise in the upstream, set a spectrum analyzer on MAX-HOLD (maximum hold) at the headend or hub on the return path that goes to the CMTS upstream port. For a basic CNR measurement, place a marker at the top of the DOCSIS haystack and a delta marker on the peak of the noise floor adjacent to the DOCSIS haystack. The delta marker reading will yield the CNR. DOCSIS requires 25 dB CNR or greater.
 |
| Figure 1: >45 dB CNR |
|
 |
| Figure 2: ~25 dB CNR |
|
Group Delay
Group delay is an RF impairment that will manifest a number of symptoms on a DOCSIS network including modems failing to register, frequent de-registration, slow data rates, inability to support higher modulation orders, poor VoIP voice quality on calls, calls failing to connect and more. One troublesome function of group delay is that it often appears to be an IP-related impairment because it is virtually invisible in the RF domain. Group delay occurs when phase vs. frequency is not linear (amplitude ripple/tilt occurs when amplitude vs. frequency is not linear). Group delay usually occurs at the roll-off points of the diplex filters and its impact increases with more filters in the cascade. Group delay also can occur due to amplitude changes throughout the transmitted spectrum.
The best method to observe group delay is to use a special QAM generator and a spectrum analyzer with QAM demodulator specifically designed to characterize the upstream as shown in Figure 3. The top trace is the adaptive equalizer graph, showing a ~2.5 µs microreflection at about -23 dBc. The microreflection caused 1.6 dB peak-to-valley in-channel amplitude ripple (second trace) with the ripples spaced 400 kHz apart, and ~270 ns peak-to-peak in-channel group delay (bottom trace). Note that the group delay ripple is off-scale in the third trace, but shown in full scale in the second screen shot (Figure 4).
 |
| Figure 3: Group Delay |
|
 |
| Figure 4: Group Delay of 270 ns/MHz |
|
The DOCSIS specification for upstream group delay is <200 nsec/MHz, but this channel has 270 ns/MHz far exceeding the specification - an indication of a severe impedance mismatch.
One can use the formula D = 492 x Vp/F to calculate the approximate distance to an impedance mismatch. D is the distance in feet to the fault from the test point; Vp is the cable's velocity of propagation (typically ~0.87 for hardline cable); and F is the frequency delta in MHz between successive standing wave peaks on the sweep trace. The 400 kHz-spaced amplitude ripple suggests an impedance mismatch about 1,070 feet from the test point. In order to see the 400 kHz-spaced ripples on a conventional reverse sweep, it would be necessary to have sweep points at least every 200 kHz.
Laser Clipping
Although the return path in a DOCSIS network is defined as 5-42 MHz, the return path lasers typically transport 0-200 MHz. Therefore, much more information is available if the spectrum analyzer is set for 0 to 200 MHz. This provides an ideal method for identifying lasers operating in a non-linear mode of operation, commonly referred to as "clipping."
Figure 5 shows a spectrum analyzer set with a span of 200 MHz. The upstream DOCSIS channel is the highest signal. There also is a third harmonic and a complete image of the 5-42 MHz pass-band and DOCSIS carrier above 140 MHz. This is a clear indication that the laser is in compression and is likely causing dropped VoIP packets due to laser clipping.
Note: Always view below 5 MHz for AM broadcast radio ingress (0.5 to 1.7 MHz) and/or ham radio ingress near 1.8 and 3.5 MHz.
 |
| Figure 5: Return Path Spectrum with 200 MHz Span and Laser in Compression |
Summary
Troubleshooting a DOCSIS network extends beyond brute force, poke-and-hope methods. Today's subscribers using high-end services demand quick resolutions to any impairment that may impact their lifestyle. In order to be effective solution providers, cable operators and their technicians must first understand the requirements to sustain a DOCSIS network and the impairments that impact it. In addition, they must be versed on the techniques to quickly conquer and divide the network to rapidly identify the root cause of any problem.
Jack Webb is senior product manager at Sunrise Telecom. Reach him at jwebb@sunrisetelecom.com or (408) 363-8000.