Triple-Play Service Testing Explore the key performance indicators that operators should monitor in order to successfully deploy and maintain data and voice services over a two-way network Carlene Gardner Strategic Marketing Manager JDSU Cable Networks Division For CCTA Training, San Juan PR, August 26, 2009

Problems in reverse path Outline Problems in reverse path What to look for to check carrier health What causes performance to degrade DOCSIS testing Recap of process to get online Tips for testing to check where the problem is VoIP testing Specific parameters for voice service Why it is different Future services

Upstream Test Parameters

Metric: Carrier to Noise, Signal to Noise

Metric: Carrier to Noise, Signal to Noise C/N and SNR, versus the noise floor, are somewhat predictive of BER and MER

Metric: MER Modulation Error Ratio (MER) on downstream QAM carrier is similar to S/N or C/N MER on upstream QAMs is similar to SNR read at CMTS MER determines how much margin the system has before failure

Average symbol magnitude Metric: MER MER math Indicates how close symbols are to ideal point MER(dB) = 10 x log RMS error magnitude average symbol magnitude Upstream channel fails if MER is not better than QPSK: 18 dB MER 16 QAM: 24 dB MER 64 QAM: 27 dB MER RMS error magnitude Average symbol magnitude

Metric: MER of constellation

Metric: MER degradation due to noise Symbols cannot reach ideal point due to noise Ideal points

What Causes MER and BER to degrade? Noise Ingress in Upstream Cause: Often originating from subscriber’s homes, loose RF connectors, or faulty coax cable Upstream noise is worse at headend due to “funneling” A little bit of noise from many locations becomes a lot of noise at the headend Effect When noise ingress gets too high, data symbols start falling outside their constellation boundaries The CMTS can no longer correctly determine good data from bad FEC is no longer able to correct corrupted data packets The CMTS discards packets with too much data corruption Web & email traffic will re-transmit and may eventually get through VoIP traffic is lost forever!

What Causes MER and BER to degrade? Noise Note that the higher noise floor causes the MER measurement To degrade

What Causes MER and BER to degrade? Coherent Ingress Cause Coherent Ingress are carrier waves or other constant carrier signals that exist in the HFC network Ingress: Often originating from subscriber’s homes, loose RF connectors, or faulty coax cable Internally-generated impairment: Common Path Distortion (CPD) Effect Within a DOCSIS upstream channel, can cause intermittent, complete data loss or cause a complete DOCSIS outage Subscribers may complain about a problem that comes and goes 24x7 monitoring of ingress is the only way to confidently identify intermittent ingress problems

What Causes MER and BER to degrade? Compression – Laser / RF Amplifier Cause Excessive input levels into an active device causing the laser or amplifier to “clip” or not be able to transmit the highest amplitude voltage signals Not rebalancing upstream devices after a change such as adding another DOCSIS channel Effect Laser or amplifier clipping causes data loss because the highest amplitude symbols (bits) are pushed into other symbol decision boundaries Data loss can be continuous or sporadic depending upon system and device conditions Web traffic can be re-transmitted, but VoIP traffic is permanently lost

What Causes MER and BER to degrade? Group Delay Cause Linear changes in phase of a signal Inherent difficulty of maintaining even speed of propagation through multiple devices, long amplifier cascades Group delay can also occur due to amplitude changes throughout the upstream band Effect Frequencies propagate at different speeds through the HFC plant Group delay affects cable modem signal quality and thus MER If group delay is bad enough, the CMTS will be unable to recover the transmitted signal and data will be lost (<200 nsec/MHz is spec) Data and VoIP communications will be impacted or lost completely

DOCSIS Testing Return path is ½ the communication.

DOCSIS Test – Initialization of a Modem “Ranging” Scan and synchronize to downstream Obtain upstream parameters Range “Registration” Send device class Establish IP connectivity Establish time of day Get operational parameters

Qualify the Drop Verify receive level at cable modem MER shows that downstream is clean and clear with margin BER shows that downstream is clean and clear of impulse noise Upstream is properly aligned and CMTS has “ideal” receive level Throughput Packet Loss

Registration Full list of registration addresses, Service Identifier (SID) and config file assigned Error messages detail where a failed test halted, can help locate source of problem

IP or RF?

IP Impairments Traffic Congestion CMTS Over-utilization Switch / Router Over-utilization Viruses, Worms or just General Killer Apps Routing Errors Cable Modem routes MTA routes IP Gateways Provisioning Issues Subnet Rules Configuration Files such as TFTP Files SNMP BPI & PacketCable Certificates

VoIP Testing

VoIP – Bullet Train Analogy Ideal World: Packets like train Cars through a station – 1 at a time, evenly spaced, and Fast

VoIP – Train Analogy Real World VoIP Packets don’t always do what you want…

Metric: MOS Score, R-Factor Test the HFC Performance VoIP Quality MOS R-Value Processing Packet Loss Jitter Delay

Metrics: Delay, Packet Loss and Jitter Time it takes a packet to ‘transverse’ the network Too much delay affects the quality of a call Over-talk and Echo Usually an architecture (traffic/capacity) issue Generally not a HFC issue with equipment such as amplifiers X-Time X-Time Network D D C C B B A A Point A Point B Aim for 100ms or better

Metrics: Delay, Packet Loss and Jitter Packet did not arrive (Point B) or out of sequence Worse if it is ‘bursty’, many lost in a row – “lossy” Can be architecture or physical layer Ingress (especially upstream) Routers over capacity (too full to hold any more) ∞ -Time X-Time Network D D C C B B A A Point A Point B Aim for 1% or better

Metrics: Delay, Packet Loss and Jitter Packets not arriving with the same timing (different from X-Time) – time between packets is different You never notice with Data, doesn’t matter how the information arrives, just care that it shows up but VoIP is Real-Time Key Causes are IP packet routing, IP based equipment X-Time Network D D C C B B A A C B D A Point A Point B Slower than X Slower than X Faster than X Aim for 10ms or better

Use Metrics to Segment HFC and IP layer Segment HFC and IP impairments Identify if issues are occurring in HFC Plant or in the IP network Check MOS of VoIP over DOCSIS channel Check VoIP packet statistics Noise and Ingress on plant are major causes of Packet Loss

Testing the Home for Ingress Contribution 7 dB TAP Return Equalizer House Drop Cable OLDER TV SET WIRELESS LAPTOP DIGITAL SET-TOP The subscriber drop remains the weakest link in the cable network Seven out of ten service calls are generated by problems at the drop Ingress caused in the home wreaks havoc on the reverse path Must be found in the home before connecting to network when possible Must be monitored continuously and eliminated quickly Home wiring may be made with inferior components and craftsmanship Micro-reflections Frequency response variation & excess losses Leakage / Ingress Replacing all home wiring is economically unacceptable, testing is required to find faults and bring the home wiring up to standards necessary for new services. Kinked or damaged cable (including cracked cable, which causes a reflection and ingress). Use of staples that perforate or compress coaxial cable Cable-ready TVs and VCRs connected directly to the drop. (Return loss on most cable-ready devices is poor) Older splitters may not be rated for 750MHz, 860MHz or 1GHz Some traps and filters have been found to have poor return loss in the upstream, especially those used for data-only service. Most common source of leakage is within the home wiring (approximately 75%) and drop cable (approximately 20%). There’s a lot of homes that still have the original wiring from 20-30 years ago! Inferior quality coaxial cable, passives, connectors Poor installation of splices and connectors - water and weather can result in pulled out, loose or corroded connectors Illegal connections to cable Some of the older TV sets with poor tuner shielding can produce leakage and ingress problems If you are in the home for a video install, take the time to make sure it is ready for any triple play or added value service. Much of the growth is the result of the rapid penetration of the telephony market. More than half of new clients were recruited in that market. The number of new Internet accounts for homes and businesses grew 12.5% increase. Subscription revenues from non-traditional services accounted for 39.4% of all subscription revenues for the industry in 2007, as compared with 24.3% in 2003 and 3.8% in 1999. COMPUTER VoIP GROUND BLOCK 2-Way Amplifier High Pass Filter ETHERNET 3-Way Splitter ONLINE GAMING eMTA-CABLE MODEM INGRESS SPECTRUM MEASUREMENTS

Adjust Goals Per Location Chart for goal direction, for example keep delay less than 100 ms, jitter less than 10 ms, packet loss less than 1, and keep R-Value above 70. Will have issues if delay starts getting above 150 ms one way, above 15 ms for jitter, above 2% packet loss, etc.

Future Return path is ½ the communication.

DOCSIS® Versions at a Glance DOCSIS 1.0 (High Speed Internet Access) 23 million products shipped worldwide as of YE2002 228 CM Certified, 29 CMTS Qualified DOCSIS 1.1 (Voice, Gaming, Streaming) Interoperable and backwards-compatible with DOCSIS 1.0 “Quality of Service” and dynamic services, a MUST for PacketCable™ In the field NOW - 64 CM Certified, 22 CMTS Qualified DOCSIS 2.0 (Capacity for Symmetric Services) Interoperable and backwards compatible with DOCSIS 1.x More upstream capacity than DOCSIS 1.0 (x6) & DOCSIS 1.1 (x3) Improved robustness against interference (A-TDMA and S-CDMA) DOCSIS 3.0 (Channel Bonding) Interoperable and backwards compatible with DOCSIS 1.x & 2.0 Specification released for about 2 years More deployments under way

DOCSIS® 3.0 – Channel Bonding Individual 256 QAM DOCSIS channel Versions 1.0/1.1/2.0 used only one channel for upstream and one channel for downstream communications 256QAM = ~40Mbps (38.8 Mbps) 4 x 256QAM 4 x ~40Mbps = ~160 Mbps BONDED

Expanding digital environment Trends in the upstream Expanding digital environment Upstream becoming increasingly crowded More carriers More challenging than before Wider carriers offer a wider target for interference Higher modulation has more sensitivity to ingress Here is the expanding digital environment. Even though the return path is shrinking in terms of homes passed per node, it is expanding in terms of activity and performance. There are several reasons for the expansion of activity and expectations on the upstream, all driven by the need for more bandwidth. First of all we are seeing more carriers in the upstream. Before, it was typical to have just a STB carrier and one modem carrier, but now we see multiple DOCSIS carriers used to get more bandwidth out of the return, creating a crowded upstream. Second, we see changes in the kind of carriers operators are using. QPSK was the modulation of choice for a long time because of its reliability, but it doesn’t provide the bandwidth operators need today to satisfy the demands on usage. Instead of adding more QPSK, operators now tent to take the modulation of the DOCSIS carrier up to 16QAM or even 64 QAM. The higher orders of modulation carry more data but are less robust, so we need to make sure that the upstream is cleaner that it was for QPSK, in particular making sure the noise floor is low and that the channel or channels we are using can offer a good carrier to noise ratio. The typical strategy has been to use the upper end of the return path for DOCSIS carriers since the noise floor is better. Finding the space in the return path for multiple DOCSIS carriers that are 16 or 64 QAM modulation is a challenge because most operators consider the band below 10 MHz to be unusable. Some operators will use 10-18 MHz for robust set-top box carriers or telemetry signals but not for DOCSIS. So that leaves the band from about 18 to 42 MHz for DOCSIS carriers, or about 60% of the upstream that we can use. With wider carriers, there is the additional challenge of actually finding completely clear blocks of 3.2 or 6.4 MHz that can be used as DOCSIS channels. These wider channels are great for adding bandwidth but they also great targets for noise, meaning that there is now a greater probability that impulse noise will fall in the band that is being used. 34

Use constellations to view impairments Microreflections etc. Microreflection pattern over time Cloud pattern over time Constellations are of particular benefit to view microreflections since they cannot be seen in a spectrum analyzer. The interferer in this case is the QAM carrier itself, but at a lower amplitude and with a delay, like an echo. When the two signals mix, the reflection distorts the true QAM constellation and creates a square pattern within each cell of the constellation. When the points are allowed to build up over several hundred packets, the characteristic diamond pattern of microreflections is clear. The diamond shape can be at any angle, depending on the phase of the reflected signal, so the square shape may be set at more or less of an angle than is shown in this example. Micro-reflections are one of the impairments that vary from modem to modem, so by tracking the best and worst unequalized MER, you can tell if the problem is generalized or if it is likely to be isolated to one area. If the overall average is good, but there are some outliers with very poor MER, then the problem is likely to be in a particular home, possibly due to a missing load or loose connector or some faulty in-home wiring. If the MER is consistently poor on the overall modem traffic, then there is likely to be a problem that affects all modems such as an impedance mismatch, CW interference, noise or a microreflection whose origin is in a shared or common part of the transmission path, perhaps a faulty piece of line equipment that is close to the fiber node. Remember that microreflections are only one of the kinds of impairments that contribute to poor MER.. 35

Back to the Basics Most problems are still physical layer issues Most of the test strategy remains the same Divide and conquer Check forward and return RF levels analog and digital Check for leakage Sweep the forward / reverse to detect issues Replace questionable connectors / passives Tighten F-connectors … but not too tight Robust plant will be ready for the next great thing

Future Services

JDSU Cable Networks Division Thank you! Carlene Gardner JDSU Cable Networks Division 38