Copyright 1999, Mark Millet
Permission to reproduce, distribute, and display publically is granted until the Jan 1, 2005, provided this Copyright message is included. This content may not be distributed for a fee, nor may any charges be levied beyond the minimal costs associated with media duplication. Additionally, the following text must accompany any distribution of this file. It may be included in the "handouts" or on the same web page, or could be narrated to the "class" during playback.
Spectrum analyzer display for DOCSIS cable modem upstream path. Quicktime (TM) Video format 4 minutes, 10 seconds. 320 x 240 pixels, 8 bit greyscale. 115,225,117 bytes.
By viewing the time index on the spectrum analyzer screen:
The playback may be paused by clicking on the image.
Start time index
Analyzer shows 5-50 Mhz upstream.
3 very active cable modem carriers (3.2 MHz DOCSIS)
A small amount of upstream interference is visable, but the C/anything can be assumed as roughly 25 to 27 dB.
Ingress between 5-18.5 MHz appears
Ingress between 5-18.5 MHz grows substantially stronger.
Lowest frequency channel used for DOCSIS is no longer DOCSIS compliant.
Severe Ingress dissapears.
Common path distortion (CPD) appears. This is from 6 MHz spacing NTSC signals.
Pattern will vary somewhat for 7 MHz or 8 MHz PAL
Approximately 7,500 intermodulation products create this moving interference pattern. Temporarily removing the analog video channels will immediately eliminate this problem, and certify that the cause is indeed CPD. If the downstream DOCSIS signals are removed as well, the modems will stop transmitting, and go offline.
Common path distortion grows in amplitude very severely. Network is effectiely overloaded. Return path laser will clip severely, and almost all data will be lost. -- regardless of center frequency of cable modems, modulation format, or channel width. Once the laser "clips" all data is lost period.
Lowest frequency DOCSIS channel is "shut off" in an attempt to reduce upstream laser loading and maintain traffic on two remaining channels. High packet loss for all customers occurs, modems on all channels may experience difficulties, and the system is certainly NOT DOCSIS compliant.
Should an operator desire continued operation under these circumstances, they should utilize a lower data rate, and narrower RF carrier width, QPSK modulation, and permit frequency hopping. This particular demonstration deliberately precluded frequency hopping or changes in channel width to illustrate the failure case.
Common path distortion principally goes away. Some spikes remain.
Third DOCSIS channel remains offline for the next 30 seconds while the modems attempt to re-register and restart traffic sessions. All customer modems on this channel suffer 100% loss of service during this time.
Return path is "perfectly clean" for a short duration. This is typical during repair and troubleshooting operations as portions of the plant are isolated to find faults. Additionally, intermittant faults on a real system can cause these periods of "perfect performance" for short durations. Only a long term statisticly based test process can truly distinguish the real plant performance.
Lowest frequency DOCSIS channel returns, modems are back online, and customer traffic resumes.
Begin process of adjusting analyzer to "zero span mode."
For further detail, see:
Analyzer is in Zero Span, with 41 MHz center frequency (center of DOCSIS channel) but the resolution and video bandwidth are insufficient for the carrier width.
Resolution bandwidth adjusted to 3 MHz, video bandwidth left at 1 MHz. Resulting amplitude measurement is uncalibrated for 3.2 MHz wide DOCSIS carrier. Should a calibrated amplitude be desired, either the video bandwidth must be 3 MHz, or the center frequency must be 1/2 symbol rate (2560 Ksymbols/sec) offset from the channel center, and 4.25 dB added to the measured value. (300 Khz resolution & video bw).. See chart in upcoming Cisco documentation at the above uRL.
The various pulse "widths" -- the measurement is now in the TIME DOMAIN,
and not the frequency domain, are actually individual docsis packets, sent
from one modem. The actual length in time will vary by number of bits transmitted,
data rate (channel width) and modulation format. The widest packets are
1500 bytes, the smallest are 64 bytes, with MAC control messages even smaller.
Interference returns, and noise floor rises to about 15 dB C/anything. This is approximately the 1% packet loss rate for QPSK modulation format in DOCSIS. (Assuming forward error correction (FEC) is turned off.) Operation at this level during regular performance is both a violation of DOCSIS requirements, and an invitation for customer dissatisfaction.
Interference dissapears, and noise floor drops to nearly 30 dB (in channel). This is DOCSIS compliant, and desired operation.
Sweep time is adjusted, and a trigger line is turned on, but not activated.
Sweep time is not 1.5 msec. Individual DOCSIS packets nearly fill the screen.
Analyzer is brought back to frequency domain mode. Notice how the peeks appear rounded, this is a result of the wide resolution bandwidth necessary to capture the entire packet in zero span mode.
Analyzer returns to frequency mode, with normal settings. Some noise and CPD remain. Analyzer is in 5-45 MHz mode.
Senior System Architect
Direct: 408 526 6275
Cisco Systems Inc. Fax: 650 526 5807
170 West Tasman Dr Pager: 800 365 4578
San Jose, CA 95134 firstname.lastname@example.org
Back to the Cable Home Page
Comments to: Dave Devereaux-Weber
Last Updated: 16-December-1999