Telco Issues


This section discusses some modem issues that are not so easy to identify. Before getting to this, having a lot of information can help to determine the fastest speeds you can get by analyzing what your modem sees after a connection is established. Any form of landline should achieve a V.34 connection, but the two key components of a good V.34 connection are the SNR and the amount of usable voice spectrum. For a connection of 33600 baud, the SNR must be 38 dB or higher and the usable spectrum must range from 244 to 3674 Hz. Otherwise, the speaking signal strength to noise strength ratio has to be higher than 79.44. Anything below these standards degrades service.

NOTE

The decibel (dB) is a unit of measuring the strength of the signal, or of a sound. From "Newton's Telecom Dictionary" (Harry Newton, 1999), "The power in telecommunications is low and normally is measured in milliwatts. However, the milliwatt is not a convenient way to express differences in power level between circuits. Voice frequency circuits are designed around the human ear, which has a logarithmic response to changes in power. Therefore, in telephony, the decibel, which is logarithmic rather than linear measurement, is used as a measure between circuits or transmission level points." As you well know, 1 db = 20 x log(signal/noise).


At least eight factors can reduce the SNR, frequency response, or both:

  • Robbed bit signaling reduces the SNR.

  • Extra analog to digital conversions reduce the SNR.

  • Load coils on the local loop reduce the frequency response.

  • Long local loops reduce frequency response and reduce the SNR.

  • Bridged taps create a loss in signal strength and reduce the SNR. Most of the industry analysts determine bridged taps to be as prominent as on 75 percent of existing copper lines (www.teradyne.com/prods/btd/articles/adsl_test.html).

  • Electrical disturbances including crosstalk, corroded connectors, lines running parallel to fluorescent lights or power cables, and equipment sharing the same power jack as the modem can all reduce the SNR.

  • Voice circuits that pass through any sub-64 k coding reduce the frequency response. Most circuits with sub-64 k coding result in speeds of less than 16,800 baud.

  • If incorrectly configured, Subscriber Loop Concentrators (SLCs), which multiplex multiple loops into a trunk, can limit the voice spectrum and subsequently create a low SNR ratio.

With these and other limitations mentioned later, a V.90 connection rarely obtains its maximum speed. First, only one analog to digital conversion can exist on a circuit, which means that from the moment the call from the client modem hits the first digital switch, the call must stay in digital form over a DS0 all the way until it hits the network access server (NAS). This also indicates that the NAS must interface with only DS0s in the form of a PRI or a channelized T1 (DS1).

The ideal scenario for a V.90 call is a 3200 Hz response range and the local loop should not be longer than 3 miles. At 3.5 miles in the U.S., a load coil is usually applied to the line, which degrades frequency response and in turn eliminates any chance for a V.90 connection. The SNR from the client modem should be 40 dB (40 dB is a SNR of 100), or higher.

Digital Pad

A telephone line connected to a remote user 's minimum point of entry (MPOE) or to a telecommuter's hotel room carries a full-duplex call on a single pair of wires. This means that both send and receive signals are transmitted simultaneously , but on individual wires. When the send and receive signals are broken apart from each other, there is a leakage of audible sound from send to receive and vice versa. This leakage is heard as an echo and becomes more noticeable as the distance increases from source to destination. Because analog lines are primarily used for voice, this echo must be quieted through the use of a pad to remove the echo.

Placement of a pad can cause issues with a 56-bps connection. Both analog and digital pads act as a volume control for both sides of the line. Although a digital pad uses the digital data to adjust the volume, it can affect the speed, but should not prevent a connection. Analog pads have to convert digital data to analog, adjust the volume, and convert back to digital. An analog pad that is located between a residence and the CO can introduce another analog-to-digital conversion into your segment and might prevent a 56-kbps modem connection, although some testing suggests that modems are not adversely affected.

The following are three different call scenarios that demonstrate these issues:

  • In cases where your call originates and terminates within the same CO, there is no pad applied to the line. The distance between source and destination is so small that the echo is not heard and a pad is not required.

  • If your call travels through a long-distance provider to get from source to destination, there is most likely a 6-dB pad on the line. 6 dB (6 dB is a SNR of 2) is only a general rule and can be different depending on your carrier.

  • If your call stays within the same local carrier, but travels from one CO to another, there is a 3-dB pad on the line (signal to noise ratio of approx. 1.4). Again, this is only a general rule and can differ from carrier to carrier. In most cases, this is the type of call that customers make to an Internet service provider (ISP), and an ISP does not generally have equipment at every CO. ISPs do usually have equipment at one CO within a certain calling area or hub, and back haul the traffic from neighboring COs to the location where the equipment resides.

    The reason that the 3-dB pad was mentioned is because it seems to interfere with a lot of 56 k modems in terms of speed and connection quality. There is a way to change your pad to 6 db to see if it makes the connection quality any better. If you dial 1 + area code + number instead of just dialing the number, you are telling the carrier that this is a long-distance call, and that it needs to have a 6-dB pad. If the connection quality is any better, the 3-dB pad is causing the issues.

There is no available data to explain why a 3-dB digital pad adversely affects the connection, nor is there any way to correct this through access server configurations. A carrier can change the padding on a line, but it is uncommon for them to change it for residential customers. In fact, many of the first level support staff that handle trouble calls won't know what this is. A telco technician is the best person to answer these questions and to make this change.

Line Code Errors

On both a PRI and a T1, time slots exist for each of the 64-kbps channels. (See Chapter 2, "Telecommunication Basics," for more information.) As the data is received, a time slot generator places the bits into these time slots. There can never be more than seven 0s in a row, even across time slots, and this is where line coding must be considered . If eight 0s are in a row, the time slot generator converts the eight 0s to a combination of 0s and 1s, depending on the line code. (See Figure 2-10 for a comparison of different coding schemes.)

Alternate mark inversion (AMI) is based on a pseudoternary or three level system signal, which represents binary digits, in which successive marks (logical 1s) are of alternately positive and negative polarity pulses, and the absolute values of their amplitudes are normally equal, and spaces (logical 0s) are of 0 amplitude, or no symbol. Each pulse is approximately 3 volts in amplitude and has a 50 percent duty cycle, which means that it takes up half of the time slot for pulse transmission. Pulses have a tendency to spread out in the time domain as they travel down a line. Restricting the initial transmission to occupy half the time slot helps the repeaters, and the receiving end, find the middle of each time slot and stay synchronized. In AMI, the pattern of bits "1 0 0 0 0 1 1 0" encodes to " + 0 0 0 0 - + 0".

A problem with this coding scheme is the way 0 is interpreted as a no-signal condition. Therefore, an absence of pulses (all 0s) can force repeaters and network equipment (the carriers facility Digital Phase-Locked Loop (DPLL) to lose frame synchronization. To prevent this, all T1s are required to meet the ones density (pulse density) requirement, in which the FCC rule states that no more than 15 consecutive 0s can be transmitted to the line (40 according to some sources). In fact, Cisco and other carriers define the ones density scheme so as to allow a channel service unit/data service unit (CSU/DSU) to recover the data clock reliably, as "no more than eight consecutive 0s." The CSU/DSU derives the data clock from the data that passes through it. To recover the clock, the CSU/DSU hardware must receive at least one 1-bit value for every 8 bits of data that pass through it. The rule is also called pulse density. Often, the carriers use the eighth bit (least significant bit [LSB]) for supervision, which leaves only 7 of the 8 bits for user data, which provide a (7 bits x 8000 frames per second =) 56,000-bps data channel. With AMI, the bits are changed to 0000000P in all time slots when data is being transmitted no matter the bit pattern.

NOTE

In the combination 0000000P, P is a pulse with the opposite polarity as the previous pulse.


The LSB is always 1, which yields a 56-kbps time slot. Only the LSB is altered .

With AMI, throughput suffers when the eighth bit is blindly stuffed with a 1. Maintaining synchronization requires only that enough pulses are sent down the line. Bipolar 8-zero substitution (B8ZS) can transmit an arbitrary bit sequence. When eight consecutive 0 bits are scheduled for transmission, a B8ZS transmitter replaces the eight 0s with a code word that contains intentional bipolar violations (BPVs), as shown in Figure 5-1. This technique guarantees a ones density independent of the data stream.

Figure 5-1. B8ZS Guaranties that Ones Density Is Independent of the Data Stream


The combination of 00000000 is always replaced with the combination 000VP0VP, but the polarity of V and P can differ, based on a polarity of the preceding LSB. In Figure 5-1, the preceding bit in time slot n is positive, and the 0000000 in slot n + 1 is replaced with 000+-0-+. If the polarity of the preceding lsb is negative, the resulting combination will be 000-+0+-.

The most commonly used coding scheme in Europe and Japan is called high density binary 3 (HDB3). This scheme is based again on AMI encoding, but the string of consecutive four 0 is replaced with a sequence that contains one or two pulses. In each case, the fourth bit is replaced with a code violation (v) signal. An additional rule is introduced to ensure effective violation, based on the polarity of the preceding pulse and the number of bipolar pulses since the last substitution:

If the preceding polarity is negative and

  • The number of bipolar pulses since the last substitution is odd, the combination 0000 is represented as 000V (V-negative).

  • The number of bipolar pulses since the last substitution is even, the combination 0000 is represented as V00V (V-positive).

If the preceding polarity is positive and

  • The number of bipolar pulses since the last substitution is odd, the combination 0000 is represented as 000V (V-positive).

  • The number of bipolar pulses since the last substitution is even, the combination 0000 is represented as V00V (V-negative).

Now imagine that an AS5300 router is using B8ZS line coding. The integral CSU/DSU in the router modifies the data prior to sending it across the wire to the telephone company. The other side is expecting AMI coding. It is not difficult to understand that the side using AMI coding will misinterpret the incoming combination in time slot n + 1. While the B8ZS is sending 000VP0VP, the AMI side is expecting 0000000P, because only the LSB is altered in AMI, and as a result, this site passes the data without recognizing that the original combination is eight 0s. The same logic applies if AS5300 is using HDB3. The inaccurate transfer of data doesn't significantly affect voice, but it does extensively interfere with a modem. The challenge of finding this type of problem is that it is well masked. The 108 Bert Test, which is supposed to find any and all possible problems with a T1 or a PRI, never transmits more than seven bits set to zero in a row. This will never force the substitution algorithm and therefore never detect this problem. These types of errors are reported in RFC 1232 (see the BPV errors in Table 7-1 in Chapter 7). However, it is vital in the troubleshooting process to make sure that both sides are using the same coding scheme.




Troubleshooting Remote Access Networks CCIE Professional Development
Troubleshooting Remote Access Networks (CCIE Professional Development)
ISBN: 1587050765
EAN: 2147483647
Year: 2002
Pages: 235

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net