Communications Technology: Archive

Current Issue

Advertising Information

Meet the Editors

Advisory Board

Annual Awards

Custom Publishing

WebEvents

Show Dailies

Reprints

List Rentals

Archives

Archives

Feature: Upstream Power Measurements: Watts up, Doc?
By and

In this sequel-of-sorts to their September article on downstream power measurements, the authors take a look at upstream digitally modulated carrier power measurements using a procedure known as the zero-span method.

In the September 2000 issue of Communications Technology, we discussed the proper way to use a spectrum analyzer to measure downstream digitally modulated carriers. As we pointed out in that article, the noise-like characteristic of those carriers means care must be taken when performing amplitude measurements. Fortunately, much of today’s available test equipment includes automatic digital channel power measurement capability in the equipment’s operating software, firmware or hardware. Further simplifying downstream digitally modulated carrier amplitude measurement is the fact that the majority of those carriers are on continuously, just like downstream TV channels.

Upstream digitally modulated carriers are another story altogether. Most are bursty in nature, which means they’re on only when a given in-home device actually is transmitting. This lack of continuous carriers makes measuring upstream levels challenging, to say the least.

One method that many use is to take a spectrum analyzer, tune it to the carrier frequency of interest, put the analyzer in max hold, and wait for the bursty upstream carrier to "paint" a display of its apparent amplitude. This is not a good way to measure upstream digitally modulated carriers. More on this later.

Given the so-called long-loop automatic gain control (AGC) operation with regard to most upstream digitally modulated carriers—especially cable modem signals—you might be inclined

to wonder why it’s even necessary to be concerned with their amplitude. In the case of cable modems, the headend’s cable modem termination system (CMTS) manages cable modem upstream levels for every modem in the system. A CMTS that’s compliant with Data Over Cable Service Interface Specification (DOCSIS) will keep all modems to within about 1 dB of each other, unless configured to do otherwise.

Accurately measuring upstream carriers

If you want to do some routine measurements to verify proper operation, you really need to make certain you’re measuring the levels accurately. For instance, it would be nice to know that you’re not overdriving your system’s upstream lasers. Overdriving lasers often causes clipping. The inevitable result of laser clipping is reduced data throughput because when a laser clips, nothing gets through. Why? During the time clipping occurs, a laser has no optical output. Cable modems may tolerate this to a certain extent because missing data may be retransmitted, but services such as voice-over-Internet protocol (VoIP) telephony don’t live well in this environment. One —sult is dropped syl—-les — words, and —-haps even dropped calls! This would not be good if someone were trying to place a 911 call.

So, how do you go about accurately measuring a cable modem’s bursty upstream digitally modulated carrier?

Step one

First, you’ll need a spectrum analyzer that supports sweep settings in the 80 microsecond (µsec) range. Most cable TV spectrum analyzers are capable of this sweep speed. After it has warmed up for a few minutes, tune your spectrum analyzer to the upstream carrier’s known center frequency. Make sure the analyzer’s reference level and attenuation controls are adjusted to keep the expected carrier peaks approximately on-screen. For now, set the sweep to 20 milliseconds (msec). Adjust the span control to 0 Hz, sometimes called zero-span. This essentially converts the analyzer from a frequency domain instrument to a time domain instrument, much like an oscilloscope. Set both the resolution bandwidth (RBW) and video bandwidth (VBW) controls to 3 MHz. Activate and position the trigger line (refer to your analyzer’s instructions if necessary) to about the middle vertical graticule. One or more active cable modems should produce an analyzer display similar to Figure 1.

Step two

Adjust the sweep to 80 µsec. You should see a display similar to Figure 2. Position the trigger line so that it’s about halfway between the highest and lowest parts of the displayed signal.

Now adjust the reference level control so that the top part of digitally modulated carrier is in the top graticule, and move the trigger line up or down as necessary until you get a display like that in Figure 3.

Notice the part of the signal that occupies the first two or three horizontal divisions in the upper left of the figure. Here it appears to have a sine wave component. This is the preamble. Turn on and position a marker about 7/8 of the way into the preamble.

A quick side note here: The preamble you see on your signal may have a different width than what’s shown in Figure 3, depending on the upstream channel’s bandwidth and data rate, modulation format and DOCSIS burst-profile configurations.

In this particular example, the marker amplitude indicates +31.07 dBmV. To validate the accuracy of this procedure, we compared it to an Agilent HP89441A vector signal analyzer and a Boonton burst power meter, and found the zero-span measurement of the preamble’s amplitude to be within 1 dB of the entire packet’s true burst amplitude.

Another side note: If you find the preamble to have a significantly lower amplitude than the rest of the signal, you’re using an RBW setting that’s too narrow for the DOCSIS channel width in use. Incorrect RBW may cause problems if adjacent carriers are present, as well as affect the accuracy of the measurement. The following Web site describes a procedure for measuring upstream carriers under adjacent channel conditions or if the chosen RBW is too wide and is letting too much noise into the channel under measurement: www.cisco.com/univercd/cc/td/doc/product/cable/cab_rout/cr72hig/cr72cnrf.htm.

Other measurements

Besides the amplitude of the upstream digitally modulated carrier, what else can you determine from the zero-span measurement method? Go back to Figure 1. That particular measurement was done in a lab environment, which is why the noise floor is so clean. The first and third spikes in that figure are bandwidth request packets, and the second and fourth spikes are 64-byte ping packet returns. Figure 4 shows 1500-byte ping packet returns, along with bandwidth request packets. Both of these examples are from a 3.2 MHz bandwidth, 16-QAM (quadrature amplitude modulation) digitally modulated carrier.

We stated earlier that a max hold method is not a good way to measure the amplitude of upstream digitally modulated carriers. To see why, look at Figures 5 and 6. They show signals from two cable modems. You can see the modems’ bandwidth request packets (the narrow spikes) and their ping packet returns. However, note that the amplitude of one of the modems is about 3 dB lower than the other.

In this case, the lower modem was transmitting at maximum power, and unable to achieve sufficient input level at the CMTS. A modem operating this way will go offline constantly and have excessive power-adjust messages.

This level difference may not seem like much, but in properly operating systems, the difference between any two modems should be no more than about 1 dB at the input to the CMTS.

Thus, if you measure any one packet, the amplitude of all packets should be within 1 dB of that value. If you were to attempt to measure the upstream digitally modulated carrier using the max hold method, the analyzer would show only the highest modem’s amplitude. If the modem with the highest output level were a malfunctioning one (that is, the modem wasn’t supposed to be transmitting at that high of a level), you might be inclined to think all of your modems were operating too high. Conversely, if you were looking for one particular low level modem, the normal modems’ signals would mask the lower signal level.

Look more closely at Figures 5 and 6. The difference in the spacing between the bandwidth request packets in the two figures is caused by the contention-based nature of multiple remote cable modems operating in the same network. Differences in the spacing between the ping packet returns are a function of packet size and the system’s upstream channel usage. These conditions are normal. In fact, two ping packet returns may be so close together that they might appear on some spectrum analyzers to be one big packet with a slight amplitude change about halfway through the packet. This, too, is normal, and merely indicates the upstream channel is 100 percent occupied during the time period you’re observing.

Real-world conditions

Figure 7 shows conditions in a real two-way plant. Here you can see one bandwidth request packet and one ping packet return. You also can see a lot of noise, including impulsive noise. The difference between the packet and a nearby noise spike is only about 12 dB, clearly below the 25 dB DOCSIS spec. You may count on this particular packet being dropped in real-world operation. In general, if in-channel impulse noise comes within 25 dB of the packet’s peak amplitude, the hybrid fiber/coax (HFC) network technically is not DOCSIS-compliant. This condition may cause lost packets.

If the noise is during the preamble, it generally will force the CMTS to not receive it, while if it is during the data section, it will either cause a correctable or uncorrectable forward error correction (FEC) error. If the error is uncorrectable, the packet is lost entirely.

One critical step to ensure reliable two-way operation is to make sure you’re accurately measuring both downstream and upstream digitally modulated carriers. For the former, use test equipment that has automatic digital channel power measurement capability. The latter is best accomplished with the zero-span method, which also allows you to more readily see the true magnitude of impulse and similar noise that appears in-channel.

Ron Hranac is consulting systems engineer for Cisco Systems, and senior technical editor for Communications Technology. You may reach him at . Mark Millet is consulting systems engineer for Cisco Systems, and is SCTE’s Member of the Year. You may reach him at .

Back to December Issue