2.6 Impairment Modeling

   


2.6 Impairment Modeling

General impairment modeling is treated in great depth in [1]. Chapters 3 and 7 also study some specifics pertinent to ADSL and VDSL respectively.

2.6.1 Background Noise

Background noise in DSL systems today is often presumed to be “140 dBm/Hz. There is no justification for this level other than early DSL designers agreed that this would represent a noise floor caused by a reasonable ADCs quantization level in 1990. [8] The number stuck. Today, noise floors can be considerably lower as conversion technology has advanced. DSL modems with lower quantization noise will often perform better. The actual thermal limit of line noise at room temperature is “173 dBm/Hz. Usually line noise is slightly higher than this limit and between “160 dBm/Hz and “170 dBm/Hz.

[8] Some claim that exhaustive studies were performed one place or another to get this number, but that is simply not correct. It occurred in a telephone conversation about what was reasonable as no one had studied it in depth just days before the release of a Bellcore Technical Advisory on DSLs in 1990. Subsequent studies sometimes reinforce the number and sometimes show it is too high or too low

2.6.2 Other Noises

The most common and largest noise in DSL is crosstalk, which is studied in detail in [1]. Crosstalk noise in DSLs arises because the individual wires in a cable of twisted pairs radiate electromagnetically. The electric and magnetic fields thus created induce currents in other neighboring twisted pairs, leading to an undesired crosstalk signal on those other pairs. Figure 2.15 illustrates two types of crosstalk commonly encountered in DSLs. Near-end crosstalk ( NEXT ) is the type of crosstalk that occurs from signals traveling in opposite directions on two twisted pairs (or from a transmitter into a "near-end'' receiver). Far-end crosstalk (FEXT) occurs from signals traveling in the same direction on two-twisted pairs (or from a transmitter into a "far-end" receiver).

Figure 2.15. Illustration of crosstalk.

graphics/02fig15.gif

Crosstalk can be the largest noise impairment in a twisted pair and often substantially reduces DSL performance when it cannot be eliminated.

NEXT Modeling

DSL standards and theoretical studies have modeled crosstalk in a 50-pair binder with the coupling function

graphics/02equ51.gif


where K next has been determined by ANSI studies to be

graphics/02equ52.gif


and N is the number of pairs in the binder expected to be carrying similar DSL service. This value is believed to be worse than 99% of the twisted-pair situations.

Thus, for example, to find the crosstalk noise from an ISDN circuit into another twisted pair for a binder containing 24 ISDN circuits, the power spectral density on any line in the binder is modeled by

graphics/02equ53.gif


For more on NEXT models, see [1].

FEXT Modeling

FEXT modeling parallels NEXT modeling with formula

graphics/02equ54.gif


where d is the length in feet, H ( f , d ) 2 is the transfer function from the line input (insertion loss) for the length of transmission line being investigated, S transmit (f) is again the power spectral density input to the line (and not at the source), and finally

graphics/02equ55.gif


FEXT is important in ADSL and VDSL and is considered more in Chapters 3 and 7.

2.6.3 Radio Noise

Radio noise is the remnant of wireless transmission signals on phone lines, particularly AM radio broadcasts and amateur (HAM) operator transmissions.

Radio-frequency signals impinge on twisted-pair phone lines, especially aerial lines. Phone lines, being pieces of copper , make relatively good antennae with electromagnetic waves incident on them leading to an induced charge flux with respect to earth ground. The common-mode voltage for a twisted pair is for either of the two wires with respect to ground ”usually these two voltages are about the same because of the similarity of the two wires in a twisted pair. Well-balanced phone lines thus should see a significant reduction in differential RF signals on the pair with respect to common-mode signals. However, balance decreases with increasing frequency and so at frequencies of DSLs from 560 kHz to 30 MHz, DSL systems can overlap radio bands and will receive some level of RF noise along with the differential DSL signals on the same phone lines. This type of DSL noise is known as RF ingress. Chapter 3 and [1] both have more on RF ingress.

Amateur Radio Interference Ingress

Amateur radio transmissions occur in the bands in Table 2.1. These bands overlap the transmission band of VDSL, but avert the lower transmission bands of other DSLs. Thus, HAM radio interference is largely a problem only for VDSL.

Table 2.1. Amateur Radio Bands

HAM operator bands (MHz)

Lowest Freq.

Highest Freq.

1.81

2.0

3.5

4.0

7.0

7.1

10.1

10.15

14.0

14.35

18.068

18.168

21

21.45

24.89

24.99

28.0

29.7


A HAM operator may use as much as 1.5 KW of power, although such large use of power is rare and may not be in residential neighborhoods or areas with many phone lines. A 400W transmitter at a distance of 10 meters (about 30 feet) leads to an induced common-mode (longitudinal) voltage of approximately 11 Volts on a telephone line. With balance of 33 dB, the corresponding differential (metallic) voltage is about 300 mV, which is 0 dBm of power on a Z = 100 transmission line. HAM operators use a frequency band of 2.5 kHz intermittently with either audio (voice) or digital (Morse code, FSK) signals, leading to a noise PSD of approximately “34 dBm/Hz. More typically, HAM operators transmit at lower levels or may be spaced more than 10 meters when transmitting at higher levels, so that more typical HAM RF ingress levels may be 20 “25 dB less powerful. Nonetheless, this still leads to PSDs for noise in the range from “35 dBm/Hz to “60 dBm/Hz, well above the levels of crosstalk in this section. Furthermore, such high voltage levels may saturate analog front-end electronics.

HAM operators tend to switch carrier frequency every few minutes and the transmitted signal is zero (SSB modulation) when there is no signal. Thus, a receiver may not be able to predict the presence of HAM ingress.

Fortunately, HAM radio signals are narrowband and so transmission methods attempt to notch the relatively few and narrow bands occupied by this noise, essentially avoiding the noise (for some transmission methods, see Chapter 8) rather than try to transmit through it. Some degree of receiver notch filtering is also necessary to eliminate the effect completely (see Chapter 8).

Emissions of signals in radio bands is also of importance. Some studies have found annihilation of radio receivers by VDSL at distances as far as 30 meters from a phone line. This problem is resolved by zeroing tones in the known amateur radio bands in DMT. QAM has no solution to the problem other than to use multiple QAM bands in between the known reserved radio bands. (Notch filters do not solve the problem in QAM, even if decision-feedback equalization is used.)

AM ingress is investigated in Chapter 3 for ADSL.

2.6.4 Combining Different Crosstalk Signals and Noise

Recently, the modeling of several types of different crosstalk signals in a single signal has been investigated by a group of phone companies known as the "full-service access network" group , and has been standardized in [10]. That model is now used in many standards and allows combination of n i crosstalkers of type i by raising each term to the power 1/0.6 before summing. Then, after the summation, the resultant expression is raised to the power 0.6. This can be expressed as:

graphics/02equ56.gif


where X talk is either NEXT or FEXT, n is the total number of crosstalk disturbers, N is the number of types of unlike disturbers, and n j is the number of each type of disturber. Example uses of this equation are given in the following subsections.

Take the case of two sources of NEXT at a given receiver. The NEXT coupling length is the length of line segment where crosstalkers are in the same binder as the affected DSL. In this case there are n 1 disturber systems of spectrum S 1 ( f ) and NEXT coupling length L 1 and n 2 disturber systems of spectrum S 2 ( f ) and NEXT coupling length L 2 . The combined NEXT is expressed as:

graphics/02equ57.gif


The dependency of NEXT on H 1 ( f, L 1 ) and H 2 ( f, L 2 ) is a slight refinement to the model in Section 2.6.2. The 1 “ H 4 factor is usually close to unity, making the two formulae nearly identical in most situations.

For three sources of FEXT at a given receiver, there are n 1 disturber systems of spectrum S 1 ( f ) at range l 1 , a further n 2 disturber systems of spectrum S 2 ( f ) at range l 2 and yet another n 3 disturber systems of spectrum S 3 ( f ) at range l 3 . The expected crosstalk is built in exactly the same way as before, taking the base model for each source, raising it to power 1/0.6, adding these expressions, and raising the sum to power 0.6:

graphics/02equ58.gif


2.6.5 Total Noise Power Spectral Density

The NEXT term and the FEXT term are computed to arrive at separate NEXT and FEXT disturbance power spectra. These power spectra should then be summed with the background noise and radio noise to determine the total disturbance seen by the victim receiver. The total noise seen by the victim receiver is given by:

graphics/02equ59.gif



   
Top


DSL Advances
DSL Advances
ISBN: 0130938106
EAN: 2147483647
Year: 2002
Pages: 154

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