Pulse Masks Define Signal Limits

Pulse masks (Fig. 1), set the limits for physical parameters such as risetime, falltime, pulse width, amplitude, overshoot, and undershoot. The entire pulse must fall within the mask for it to comply with a standard. Engineers use pulse masks as “sanity checks” for new transmitter and receiver designs. Later, they use the masks in tests to verify that a product complies with industry standards, and they may use the masks in pass-fail tests of production products.

Figure 1. Masks define the amplitude, risetime, falltime, and jitter for pulses in telecom networks.

Some digital scopes include mask options that let you measure the pulses. Figure 2 shows the mask for a 155.52-Mbps STS-3E binary-1 pulse. The mask is typical of that for other electrical signals such as T1, T3, and DS3 that travel over coaxial and twisted-pair wires. Binary 1’s may have two masks with inverse polarities. The polarity depends on the type of bit encoding and on the bit pattern used in the transmission system.

Figure 2. This mask is typical of that for electrical signals that travel over coaxial and twisted-pair wires. (Courtesy of LeCroy.)

If the data-transmission system uses alternate mark inversion (AMI) encoding, then a binary 1 could be either a positive pulse or a negative pulse—no pulse during a clock period indicates a binary 0. With coded mark inversion (CMI) encoding, a binary 0 will have a pulse with a rising edge in the middle of the bit period. Figure 3 shows the mask for a CMI coded binary 1 (the inverse of the mask in Figure 1) and Figure 4 shows a binary 0.^{1, 2, 3}  

Figure 3. A binary 1 of a CMI-coded signal can also be a positive pulse. (Courtesy of Tektronix.)

Figure 4. A CMI-coded binary 0 has one rising and one falling edge. A binary 0 (right) has a rising edge halfway through the bit’s period. (Courtesy of Tektronix.)

You’ll find mask tests difficult to perform on live traffic—which appears as somewhat random bit patterns—because the standards require specific bit patterns. In particular, for testing a binary 1 pulse, you need to “isolate” the pulse so a specified number of 0’s both precede and follow the pulse.

The number of 0’s depends on the standard. Those 0’s guarantee that intersymbol interference from adjacent pulses won’t corrupt the pulse you’re testing. Typically, you measure an AMI-coded binary 1 pulse that’s preceded by two binary 0’s and followed by one binary 0. For measuring a CMI-coded bit stream, you need a binary 1 both preceded by and followed by binary 1’s to get both the rising and falling edges of the pulse. For measuring binary 0’s, you need the measured bit to both follow and precede a binary 0.

A Clean Fit
In Figure 2, the pulse fits cleanly within the mask and the screen shows that the oscilloscope tested 16 pulses with no failures. Even so, the pulse shows a small—although acceptable—amount of ringing and overshoot on the initial falling edge. Overshoot, even while within the masks’ limits, can indicate a line mismatch, meaning that the UUT’s output impedance doesn’t sufficiently match the impedance of the transmission line or of the receiver. Excessive ringing and overshoot can be caused by a line transformer on the output device. In electrical signals, ringing can also indicate that the transmitting device is emitting excessive conducted EMI through the cable. So even though the pulse fits within the mask, the oscilloscope can indicate potential transmission problems.

Masks also indicate if the transmitted signal has sufficient amplitude to arrive at its destination within tolerance. Each standard defines an acceptable range of amplitudes. European standards define amplitude in absolute voltage levels. For example, a 2.048-Mbps pulse has a nominal amplitude of 2.37 V if traveling on a coaxial cable and 3 V if on a twisted pair. In contrast, 34.368-Mbps signals have a nominal amplitude of just 1 V.4 If a transmitted signal’s amplitude is below a standard’s limit, you may have a bad line-driver IC. If the amplitude is too high, the line may not be properly terminated.   

In contrast, North American masks don’t define absolute voltage levels. Instead, they define relative amplitudes between the high and low voltage levels of a pulse.

To minimize overshoot, undershoot, and amplitude problems, you need to match the impedance of your scope to the transmission line. Otherwise, you’ll get incorrect signal measurements and the scope will alter the signal. Scopes can have input impedances of either 50 V or 1 MV, but the transmission lines will have impedances of 50 V,
75 V, 100 V, or 120 V, depending on the standard. If your signal’s impedance is other than 50 V, you need an impedance-matching network, available from your scope manufacturer. Or, you can set the scope to 1 MV and use a feed-through terminator that matches the impedance you need.5 Also, if you want to display differential signals on a scope, you need a differential probe to convert them to single-ended signals.

Because you’ll likely measure many pulses in each test, you need to know how many of those pulses are within tolerance and how the pulses vary relative to an average. Your scope can give you valuable statistics about how your product performs. You can set the scope to count the number of pulses tested through the mask and count errors, but many scopes also do more. For example, if your scope has a persistence mode, then you can see small changes in the pulses. The scope will store each pulse and change color or intensity the more often the pulses attain a given value of timing or amplitude. The persistence-mode display gives you a feel for the pulse’s repeatability.

Another useful tool in mask testing is a histogram. Scopes can produce histograms on the amplitude scale or on the time scale. An amplitude histogram can display variations in pulse amplitude; a time histogram displays a pulse’s jitter. By looking at a time histogram’s width, you can see how many pulses were in or out of tolerance: The wider the histogram, the more jitter in the system.

Bit Errors
Receivers in telecom equipment sample and digitize the incoming signals, then try to interpret them and reconstruct the original bits. If transmitted signals have too much jitter—a jitter histogram or a persistence display exceeds the limits of a mask—than a receiver might misinterpret a bit, causing a bit error. A network’s bit-error rate is proportional to the jitter of an incoming signal. If you find excessive bit errors when testing a transmission line with a bit-error-rate (BER) tester, then use the scope’s mask, color grading, and histogram to find if excessive jitter is causing the problem.

If you find that jitter causes pulses to appear outside of a mask, you can look for clues to find the source of the problem. Interference from other signals may cause jitter. For example, if you look at signals with jitter in the frequency domain and see a strong 20-kHz component, then look for equipment with a switching power supply that switches at that frequency. If you find the switching signal is causing the problem, you may need better grounding or shielding to minimize the jitter.      T&MW

FOOTNOTES
1. Communications Waveform Measurements, Product Note 83475-1, Hewlett-Packard, Santa Clara, CA, September 1995, pp. 11, 12.
2. Freeman, Roger L, ed., Reference Manual for Telecommunications Engineering, 2nd ed., John Wiley and Sons, New York, NY, 1994, p. 448.
3. “Testing 155 Mbps Signals,” Application Note, LeCroy, Chestnut Ridge, NY.
4. Freeman, Roger L, op. cit. pp. 442, 446.
5. Communications Waveform Measurements, op. cit. p. 14

FOR FURTHER READING
ANSI T1.102-1-1993 (R1999), Telecommunications - Digital Hierarchy - Electrical Interfaces, American National Standards Institute, New York, NY, www.ansi.org. 
Recommendation G.703 (10/98)- Physical/ electrical characteristics of hierarchical digital interfaces, International Telecommunication Union, Geneva, Switzerland, www.itu.ch. 
“Testing Telecommunications Tributary Signals,” Application Note, Tektronix, Beaverton, OR, 1998, www.tek.com/Measurement/App_Notes/telcomm_tributary/eng/index.html.

ACKNOWLEDGEMENTS
Thanks to Jay Alexander, R&D engineer at Hewlett-Packard (Colorado Springs, CO), and David Sharpe, senior technical associate at Lucent Technologies (Denver, CO), for their contributions to this article.