ITU - Telecommunication Standardization Sector. G.fast: Far-end crosstalk in twisted pair cabling; measurements and modelling ABSTRACT

Transcription

1 ITU - Telecommunication Standardization Sector STUDY GROUP 15 Temporary Document 11RV-22 Original: English Richmond, VA Nov. 211 Question: 4/15 SOURCE 1 : TNO TITLE: G.ast: Far-end crosstalk in twisted pair cabling; measurements and modelling ABSTRACT This contribution presents measurements on the ar end crosstalk in telephony cables and how they compare with legacy models or FEXT. We show that the measured curves have some similarity with the curves predicted rom legacy models. However, improvement o crosstalk models is recommended. As a irst step, we introduce a irst-order enhancement o that legacy model, which can prevent unrealistic high crosstalk levels at higher requencies, but additional improvements are let or urther study. This contribution is provided or inormation only. 1. Introduction: The estimation o channel capacity or G.ast [4,5] requires adequate models or transmission as well as or crosstalk. Legacy models or FEXT are based on the crosstalk rom identical disturbers in all wire pairs. The validity o these models were never proven or requencies above 3 MHz, and were never meant or use in combination with small numbers o disturbers [6]. This contribution presents measurements on the ar end crosstalk in cables and how they compare with these legacy models or FEXT when applied to a single disturber. It is shown that the measured curves have some similarity with the curves predicted rom legacy models, but one can argue whether these models are good enough or G.ast perormance simulations or not. As a start we will also introduce a irst-order enhancement o that legacy model or preventing unrealistic predictions at higher requencies, and this might be the best solution or the time-being. 2. Modeling EL-FEXT 2.1 Legacy model or FEXT The parametric model or ar-end crosstalk, which is commonly used or al kinds o DSL perormance studies in the past, is shown in the expression below. It evaluates the equivalent FEXT (which is dierent rom an individual wire pair coupling), and a more detailed description o what it represent can be ound in the ETSI Spectral Management standard [6], part 2, chapter 8. H (, L, K ) x = Kx L/ L ext s (, L) Variable identiies the requency. Constant identiies a chosen reerence requency or dimensioning purposes, commonly set to = 1 MHz. T 1 Author: Contact: Rob van den Brink TNO Bas van den Heuvel TNO Tel: rob.vandenbrink@tno.nl Tel: bas.vandenheuvel@tno.nl

2 Variable L identiies the physical length o the loop. Constant L identiies a chosen reerence length or dimensioning purposes, commonly set to L = 1 km. Function s T (, L) represents the requency and length dependent amplitude o the transmission through the actual loop, normalized to a reerence impedance R n. This value equals s T = s 21, where s 21 is the orward scattering parameter o the loop normalized to R n. Constant K x identiies an empirically obtained number that scales the FEXT transer unction H ext (, L, K x ). The value or K x is cable speciic, and is to be speciied or each scenario being studied. A commonly used value (in db) or generic European studies, not dedicated to any particular cable or region, is K x_db = -45 db or = 1 MHz and L = 1 km. 2.2 First order enhancement o legacy model or FEXT One problem o the above legacy model or ar-end crosstalk is that the predicted coupling keeps increasing with the requency and that above a certain requency the predicted FEXT becomes unrealistically higher than the assumed transmission. This was not an issue or VDSL2 studies, but higher requencies are assumed to be used or G.ast. Thereore this models needs to be improved. The legacy model is build-up rom two parts: a coupling part (K x / (L/L )), and a transmission part s T (, L). The coupling part is linear proportional to the requency, like the out-o-band transer o a irst-order high-pass ilter. It scales with (L/L ), and not with (L/L ), since it is a cascade o many short section and the individual coupling values are random in nature. I this coupling is assumed to be purely capacitive and is described by the transer unction o a series capacitance, then we will obtain a coupling that never exceeds the value o 1. K x L/ L R C R jωrc j K x = 1 jωrc 1+ j K ( ) ( ) L / L x L / L + By using this replacement we get a irst order enhancement o the legacy model that can be deined as ollows: ( ) ( ) j K L / L H (, L, K ) x ext x = st (, L) 1 j K x L / L + The magnitude o this transer unction approximates the transer o the legacy model when < /K x (L /L), which holds or instance or requencies below 177MHz when K x = -45dB and = 1 MHz and L = L = 1 km. The phase o this transer unction may be a bit artiicial, although the measurements in the next chapters demonstrate that this phase not a bad estimate o actual values. 3. Measured crosstalk on access wiring (KPN Access cable) We measured ar-end crosstalk between dierent wire pairs in a typical underground access cable being used in the Netherlands (KPN access cable. The cable section was about 1m long (14.1 m), and urther details about its transmission characteristics can be ound in [2]. This chapter concentrates on the results o several FEXT measurements on this cable. 3.1 FEXT Measurements (in-quad, out-o-quad) The FEXT is the transer unction rom one end and wire pair to another end and wire pair. The igures below compares the observed FEXT (between dierent wire pairs) with the transmission (via a single wire pair). The irst two igures show the observed crosstalk between wire pairs situated in the same quads, while the third one does the same or wire pairs situated in dierent quads with dierent twist lengths. As expected, the in-quad crosstalk is higher than the out-o-quad crosstalk. As expected, the curve o the in-quad crosstalk unction is also smoother than or the out-o-quad crosstalk since all in-quad wire pairs ollow the same geometry and have exactly the same twist.

3 2 Wire-pair FEXT crosstalk (di mode, transer) - (KPN, 1m, in-quad #1) di mode FEXT di mode transer (c) TNO req:[mhz] Wire-pair FEXT crosstalk (di mode, transer) - (KPN, 1m, in-quad #2) -2-4 di mode FEXT di mode transer (c) TNO req:[mhz] Wire-pair FEXT crosstalk (di mode, transer) - (KPN, 1m, out o quad) -2-4 di mode FEXT di mode transer (c) TNO req:[mhz] The igures demonstrate that the observed FEXT is so high above about 15MHz that FEXT and transer are in the same order o magnitude. Due to all kinds o resonation eects, the crosstalk above 15 MHz is sometimes a bit higher than the transer and sometimes a bit lower but on average they remain in the same order o magnitude. The observed FEXT is so low below about 2 khz that it becomes hardly measurable with the used measurement setup. As a result we can only draw conclusions about FEXT measurements on this particular cable when the requency is above 2 MHz.

4 3.2 EL-FEXT Measurements (in-quad, out-o-quad) The ratio between FEXT and transmission is indicative or the signal to noise ratio that will be observed at a DSL receiver. Thereore the FEXT itsel is o limited interest, and the EL-FEXT (equal level) is ar more meaningul. EL- FEXT is deined as the ratio between FEXT and transmission, and the igures below show it or several wire pairs. The irst two igures show the observed crosstalk between wire pairs situated in the same quads, while the third one does the same or wire pairs situated in dierent quads. As expected, the in-quad crosstalk is higher than the out-o-quad crosstalk. And the curves o the in-quad crosstalk unctions are also smoother than the one or the out-o-quad crosstalk since all in-quad wire pairs ollow the same geometry and have exactly the same twist. The rippling in crosstalk magnitude between the out-o-quad wire pairs may be caused by the act that the lengths o the involved out-o-quad wire pairs are slightly dierent, while the lengths o the in-quad wire pairs are assumed to be more equal. 2 Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (KPN, 1m, in-quad #1) (c) TNO Len=14.1m EL-FEXT model (CKT1:ETSI, Kx=-45.dB) req:[mhz] Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (KPN, 1m, in-quad #2) Len=14.1m EL-FEXT model (CKT1:ETSI, Kx=-45.dB) (c) TNO req:[mhz] Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (KPN, 1m, out o quad) Len=14.1m EL-FEXT model (CKT1:ETSI, Kx=-45.dB) (c) TNO req:[mhz]

5 The above igures are all decorated with three dierent curves (all in blue) o the enhanced version o the classical EL- FEXT model. The highest line (called CKT1:ETSI) represents a crosstalk level that was commonly used in the past (within ETSI, FSAN, ANSI/ATIS, etc) as a near worst-case crosstalk example or studying ADSL and VDSL perormance ( km, and scaled to a loop length o 14.1m). The curves o our enhanced model bend only away rom what the legacy model would have predicted when it approximates (or exceeds) db. This prevent unrealistic EL-FEXT levels ar above db. This deviation is only visible or curve CKT1:ETSI in the plot or the highest requencies. The other lines are based on the same parametric model but with coupling values that are 12 and 24dB lower. The observed EL-FEXT curves have some similarity with these models above 2 MHz, but one can argue whether these models are good enough or studying the perormance o G.ast. The slope o the measurement is about twice as steep as the legacy model above a certain requency (3 MHz in this example), and the crosstalk coupling gets signiicantly higher than predicted by the (near worst case) assumptions o the past. This can be a signiicant problem since G.ast signals are assumed to operate in this band as well. It may be obvious that the FEXT model needs urther improvement, on top o the enhancement we introduced so ar. 3.3 EL-FEXT Measurements (phase relations) The enhanced model or EL-FEXT does not only predict a magnitude but also a phase. To learn how applicable is phase prediction is, we compared the measured phase o EL-FEXT with the predicted one. The irst two igures below show the observed crosstalk phase between wire pairs situated in the same quads, while the third one does the same or wire pairs situated in dierent quads. Mark that the measurements are only reliable above 2 khz, as explained at the beginning o this chapter Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (KPN, 1m, in-quad #1) [deg] Len=14.1m EL-FEXT model (CKT1:ETSI, Kx=-45.dB) (c) TNO req:[mhz] Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (KPN, 1m, in-quad #2) [deg] Len=14.1m EL-FEXT model (CKT1:ETSI, Kx=-45.dB) (c) TNO req:[mhz]

6 Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (KPN, 1m, out o quad) [deg] Len=14.1m EL-FEXT model (CKT1:ETSI, Kx=-45.dB) (c) TNO req:[mhz] The observed FEXT phase o in-quad crosstalk is roughly at constant distance rom the observed transmission phase over a wide requency interval. Thereore the EL-FEXT is observed to be rather constant over a wide requency band (in the order o +9 degrees, above 2 khz). This may be an advantage when designing crosstalk canceling mechanisms or bonded G.ast systems using all the 4 wires o a quad simultaneously. As such, the phase prediction o the enhanced EL-FEXT model isn t that bad at all, although it suers rom the same limitations as observed beore or magnitude predictions o that model. The observed EL-FEXT phase o out-o-quad crosstalk is not so constant over a wide requency interval, and this may be caused (again) by the act that the lengths o the involved out-o-quad wire pairs are a slightly dierent. 4. Measurements on in-house wiring ( Gamma ) We also measured ar-end crosstalk between the two wire pairs o 5m in-house telephony cabling, which was ound in a consumer shop. The our wires o this cable are more or less organized in a quad, but the geometry o that quad is a bit lexible and can easily change by bending this low quality cable. This causes higher in-quad crosstalk values. The quad is also untwisted, but that has no eect on the in-quad crosstalk. More detail about the characteristics o this cable can be ound in [1] and [2]. The three igures below show the measured FEXT (together with its transmission), the EL-FEXT (together with a model) and the phase o this EL-FEXT (also together with a model). The irst igure below illustrates that above about 6 MHz the FEXT exceeds the transmission. The second igure below illustrates that the measured EL-FEXT is roughly 22 db above the model or EL- FEXT used in many European studies. The shapes o measured and modeled curves have good similarities above 2 MHz, and that our irst order enhancement o the FEXT model becomes essential above about 7 MHz. The third igure below illustrates that the phase predictions o the model isn t that bad above about 2 MHz Wire-pair FEXT crosstalk (di mode, transer) - (Unshielded "Gamma", 5m) di mode FEXT di mode transer (c) TNO req:[mhz]

7 2 Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (Unshielded "Gamma", 5m) Len=51.33m EL-FEXT model (CKT1:ETSI+22dB) EL-FEXT model (CKT1:ETSI, Kx=-45.dB) (c) TNO req:[mhz] [deg] Wire-pair EL-FEXT crosstalk (di mode, CKT1:ETSI model) - (Unshielded "Gamma", 5m) (c) TNO Len=51.33m EL-FEXT model (CKT1:ETSI+22dB) EL-FEXT model (CKT1:ETSI, Kx=-45.dB) req:[mhz] Measurements on other type o wiring We have also measured the ar end crosstalk o the other two cable types that are described in [1] and [2], a Cat5e cable and an in-building multi quad telephony cable. The observed FEXT o the CAT5e cable was so low that it went beyond the measurement capabilities o our setup. Thereore this aspect o these measurements were ignored. The observed FEXT o the multi-quad telephony cable had very dierent characteristics than those presented above. This phenomena is still under study. 7. Summary This paper should be presented under the G.ast agenda item, and addresses issue 5.1 and x The paper shows the results o ar-end crosstalk measurements on real telephony cables and compares the measured curves with predicted curves rom a model. The measured curves have some similarity with the curves predicted rom legacy models, but one can argue whether these models are good enough or G.ast perormance simulations or not. As a irst step, we introduced a irst-order enhancement o that legacy model, which can prevent unrealistic high crosstalk levels at higher requencies, but urther improvements are let or urther study. The inclusion o that model in G.ast is proposed in a separate contribution [3]. Further improvements o the FEXT model are needed as well, to account or slopes that increase above a certain requency. This causes that the actual crosstalk coupling gets signiicantly higher than predicted, even when the near worst case assumptions o the past are used. Ignoring this can be a signiicant problem since G.ast signals are assumed to operate in this band as well. This contribution is provided or inormation only.

8 8. Reerences [1] TNO (Rob van den Brink, Bas van den Heuvel), G.ast: Wideband transer and crosstalk measurements on twisted pair cables, ITU contribution 11BM-21, April 18, 211. [2] TNO (Rob van den Brink, Bas van den Heuvel), G.ast: Wideband modeling o twisted pair cables as twoports, ITU contribution 11GS3-28, Geneva Sept 211. [3] TNO (Rob van den Brink, Bas van den Heuvel), G.ast: Enhhanced model or FEXT, ITU contribution 11RV-23, Richmond, Nov 211. [4] Huawei (Dong Wei, Anni Wu), G.ast: Perormance o G.ast when Coexisting with VDSL2, ITU contribution 11GS3-67, Geneva Sept 211. [5] Ikanos (Massimo Sorbara, Stephanie Pereira), G.ast: Simulations or G.ast at 2m, ITU contribution 11GS3-75, Geneva Sept 211. [6] ETSI TR , Transmission and Multiplexing (TM); Access networks; Spectral management on metallic access networks; Part 2: Technical methods or perormance evaluations. 28.