Techinque/Idea used in DSL Internet

Hey hi all, i was reading about DSL, i.e., Digital Subscriber Line, that which is used to provide internet facilities. But i'm little confused about it.

One source says:

The multiplexed signal is then carried to the central switching office on significantly fewer wires and for much further distances than a customer's line can practically go. This is likewise also true for digital subscriber lines (DSL).

Click to expand...

and another source says that it doesn't use any multiplexing.

If not multiplexing, then how are 2 signals transferred at same time. What i know is, the 2 signals are added up and mudulated with a carrier, they are sent through the channel and then again split up into 2 signals at the subscriber end.

Another Ques, Why was this idea/technique couldn't be used in dial-up internet?

Ask i know 2 signals can not transferred at the same time but we can send differend time but all data have address so subscriber pich up data that have the same address
sorry if i tell you wrong

In dial-up internet access, the channel is analog (i.e., the telephone 2-wire cable). The same bandwidth with Plain Old Telephone Service (POTS), that is, analog telephony, is used, which is 3.1 kHz (spectrum from 300 Hz up to 3.4 kHz). It turns out from the Shannon-Hartley law,

C=Blog₂(1+S/N)

that for this bandwidth B and the SNR (S/N) (i.e., signal-to-noise power ratio) found in a telephone line, the theoretical capacity C (i.e., maximum data rate) is 35 kbps. In practice, it was about 33.6 kbps due to additional noise not included into the theoretical model and other constraints.

Then came the dial-up modems V.90/V.92. the local loop (that is, the connection between the customer premises equipment (CPE) and the local Central Office (CO) of the telephone company, a.k.a. telco) was still analog, but the line in the telco side was digital (PCM-TDM). Thus, higher data rates could be accomplished. The sampling rate was 8 kHz (2 * 3.1 = 6.2 kHz + 1.8 kHz guardband for protection against alliasing -> oversampling). Each sample was represented by 8 bits. Thus, the bit rate was 8 bits/sample * 8,000 samples/second = 64 kbps. However, only 7 bits were used for data - the remaining bit was used for in-band signaling (in-band: not a separate signal for signaling but signaling carrier inside data. Signaling: control information, no real data). Hence, the data rate was 7 * 8,000 = 56 kbps.

In ISDN, the whole connection was digital, including the local-loop (also called subscriber line). That is, this is actually the first Digital Subscriber Line (DSL) technology! Due to the digital technology, higher bandwidth could be used (120 KHz in Europe and 80 kHz in N. America - the difference is caused due to different line coding), which lead to higher data rate (160 kbps net rate, 64 kbps for data + 64 kbps for digital voice or combined to a 128 kbps signal for data, 16 kbps for signaling, 16 kbps overhead). This is the so-called Basic Rate Access (BRA) ISDN signal.

However, the copper has a bandwidth of about 400 MHz - 600 MHz. So, it would be good to utilize higher portion of the available spectrum. But high frequencies mean high distortion, that is, high data rates mean that the transmitted pulses "spread" in time, and as an effect they interfere at the receiver, which in turn increases the bit error rate (BER). What is the solution?

ADSL! The idea is simple and it is called Discrete Multi-Tone (DMT): Divide the available spectrum in sub-channels. Whereas the channel may be wideband, these sub-channels are narrowband. Modulate data using one sub-carrier for each sub-channel. Therefore, at the end, using these parallel sub-channels you will have high data rate with minimum distortion.

Often adaptive bit loading is used to enhance performance: That is, pilot signals (no data) are used to measure the sub-channels quality. Sub-channels with good quality use higher order modulation schemes (e.g., 64 QAM) whereas not-so-good sub-channels use lower order modulation (e.g., 16 QAM or QPSK). Really bad quality sub-channels are not used at all. That way we ensure the higher possible throughput with the minimum distortion.

Usually the spectrum is divided also in an uplink band (UL: from customer towards telco) and a downlink band (DL: from telco towards customer). Usually the lower frequencies are used for UL because UL is more sensitive in distortion and lower frequencies are better quality (less distortion). Why? You see, when the user signal arrives at the CO, it is multiplexed with other signals. The logic is the same as always in telecommunication networks: start with single-users signals with relative low rate and then multiplex them to signals that carry data from many users and have high rate. The reason? To reduce the number of channels (cables, microwave links, optical fibers, whatever) required to transport the user signals.

As for the better bandwidth utilization that I referred about: ADSL uses bandwidth up to 1.1 MHz (256 sub-channels, sub-channel spacing = 4.3125 MHz). The max. data rates (assuming all the sub-channels are used, and they are used with the higher possible modulation order) are about 8 Mbps for DL and 800 kbps for UL (rule of thumb for data rates in Assymetric DSL (Assymetric = higher data rate at DL): DL/UL = 10). In ADSL2 the same BW is used but due to more efficient encoding we have ~12 Mbps in DL and 1.2 Mbps in UL. In ADSL2+ the spectrum goes up to 2.2 MHz (512 sub-carriers). Thus, the data rates are almost the double of ADSL2, i.e. 24 Mbps for DL and 1.3 - 3.5 Mbps for UL. Note that attenuation increases with distance, so higher data rates are accomplished in relative small distance from the main distribution frame (MDF) which is the last splicing point before the CO. The typical reach for ADSL is from 3 Km up to 5.5 Km from the MDF.

In VDSL the bandwidth goes up to 10 MHz, and in VDSL2 up to 30 MHz. Optical fibers are used until the street cabinet (SC). From there, the max. reach must be from 300 m up to 1.5 Km (the so-called first mile, in contrast with the local-loop which is often called last mile). Multiplexing of the US and DS bands allow high data rates with relative low attenuation and distance penalnty. Several "data rate profiles" are available, as well as symmetric and asymmetric modes. However, two numbers that is good to remember: max. data rate in DL for VDSL is 52 Mbps and for VDSL2 is about 100 Mbps (not 100% sure about that though ... do your search!).

The same idea with DMT is used very often in wireless communication. Under that context, it is called Orthogonal Frequency Division Multiplexing (OFDM).

KostasN said:

In ISDN, the whole connection was digital, including the local-loop (also called subscriber line).

Divide the available spectrum in sub-channels. Whereas the channel may be wideband, these sub-channels are narrowband. Modulate data using one sub-carrier for each sub-channel. Therefore, at the end, using these parallel sub-channels you will have high data rate with minimum distortion.

Click to expand...

Thanks,
So you mean they modulate internet signal with one carrier, and voice/telephone signal with another carrier? How is this different from multiplexing?
ISDN > It became digital many years ago(?), yet there was (only) dial-up internet about 7-8 years ago.

Ok, I will try to explain it differently.

The local loop is the connection between the customer and the telco. Usually it is one or two telephone cables. Each telephone cable consists of two copper wires.

ISDN is a fully digital network, as the name implies (Integrated Services Digital Network). There are two "flavors" of ISDN, namely, Basic Rate ISDN (BRI), also called Basic Rate Access (BRA) ISDN, and Primary Rate ISDN (PRI), also called PRA ISDN. The connection between the Line Termination (LT) unit, located at the end switch in the central office of the telco, and the Network Termination (NT) unit at the customer side is referred to as U-interface. The BRI U-frame consists of the following fields: two bearer channels (B-channels), namely B1 and B2, with bit rate 64 kbps each, one delta channel (D-channel) with bit rate 16 kbps, and overhead for framing and synchronization with a bit rate of 16 kbps, for a total bit rate of 64+64+16+16 = 160 kbps. B1 can be used for digital voice and B2 for data (at the same time), or can be combined ("bonded" as we say) to form a single 128 kbps data channel (in that case, we cannot have voice service). The D channel is used for signaling (i.e., control), but if we wish it can be used also to carry low-rate packet-switched data, such as X.25 packets at a rate of 9.6 kbps. (I remind you that ISDN is a circuit-switched technology, that is, a physical end-to-end path or circuit has to be established prior communication takes place - this is called call setup, and it is one of the reasons that we need the D-channel, i.e. to establish, monitor, and terminate these calls; in other words, the ISDN is in fact a dial-up technology, since we have a call; anyway, what I want to point here is that the B-channels can carry circuit-switched voice or data.) BRI signal is often called 2B+D signal (two B channels and one D channel, plus overhead of course). The spectrum of this signal depends on the line coding used. In N. America the line coding is 2B1Q (2 binary - 1 quartenary). In other words, we have a mapping of pairs of bits (or dibits) to quartenary symbols, that is, we have 4 different symbol states on the line. Thus, the corresponding symbol rate (baud rate, signaling rate) for this bit rate of 160 kbps is 160,000/2 (since we have 2 bits/symbol) = 80 kHz. In Europe is used a line coding technique called 4B3T (4 binary - 3 ternary) which leads to a baud rate of 160 * (3/4) = 120 kHz. The U-interface is a single pair (two wires) connection, but still offers bi-directional communication (i.e., full-duplex). How the different signals are separated? Usign a technique called echo cancellation. It is complex to describe it here, there are many resources on the topic available on the Internet. Simply stated, the system measures the channel and locates the position and duration of echoes, and then removes them. This can be done either using a Digital Signal Processor (DSP) or special software.

The second version of ISDN, PRI, is quite different. In N. America, we have 23 B channels and 1 D channel, but this time the D channel has bit rate of 64 kbps, plus 1 bit for frame synchronization with bit rate of 8 kbps, for a total bit rate of 24 * 64 + 8 = 1.544 Mbps. This is a PCM-TDM frame referred to as PCM24. Each channel is a sample represented by 8 bits. Each sample is a timeslot (TS) - D-channel is placed in TS0. The frame repeats 8,000 times per second, thus it has a duration 1/8,000 = 125 μs. Each TS has a bit rate, as we said, 8 bits/TS * 1 TS / frame * 8,000 frames/s = 64 kbps. We have in total 8 bits/TS * 24 TS + 1 bit = 192+1 = 193 bits. This signal is the Level 1 of the North American Plesiochronous Digital Hierarchy (PDH), that is, the T-carrier system, and it is referred to as T1. In Europe, we have 32 TS - 30 B channels, 1 D channel of 64 kbps (TS16), and 1 TS for frame synchronization (TS0). This is the so-caleld PCM30 frame, with bit rate 32 TS * 64 kbps/TS = 2.048 Mbps. The total number of bits is 32 TS * 8 bits/TS = 256 bits. As you may have already realized, again each TS is a sample represented by 8 bits, and the PCM frame repeats with a frequency of 8 kHz, that is, it has a duration of 125 μs. This is the so-called E1 signal, the Level 1 of the European PDH, a.k.a. E-carrier. As you can see, in PRI ISDN the line is time-shared between the different channels. The line coding used in N. America originally was AMI (Alternate Mark Inversion), now it is B8ZS (Bipolar 8 zeroes Substitution); the line coding used in Europe is HDB3 (High Density Bipolar 3 Zeroes). B8ZS and HDB3 use substitution codes to elliminate the precense of more than 8 or more than 3 consecutive zeroes respectively, which could lead to loss of sycnhronization due to the DC level. Besides that, are very similar in that they are both binary codes, that is, we have two states on the line, a binary 0 and a binary 1. Thus, the baud rate reflects the bit rate, that is, it is about 1.5 MHz in the U.S. and 2 MHz in Europe.

I have to say that in BRI there is also one other interface, called S/T interface, which connects the NT to the Terminal Equipment (TE) - such as the ISDN telephone. This is a 4-wire (2 pair) interface, and it is mainly used in Europe, not in the U.S. Actually, the BRI ISDN was never popular in the U.S., only the PRI was. The S/T frame consits of the 2B+D channel and 48 kbps overhead (for monitoring, testing, etc.), for a total of 192 kbps. The channels are time-multiplexed in a complex way that I will not refer here. The line coding is Inverse AMI, also called Alternate Space Inversion (ASI) or pseudo-ternary code. In AMI, pulses alternate between binary 1 (Mark) adn binary 0 (Space) in order to ensure the absence of a DC level. The same happens in ASI, with one difference: a binary 1 is represented by an absence of pulse (Space) and a binary 0 by a pulse (Mark).

Something else that I forgot to mention is that in PRI , there is a Private Branch Exchange (PBX), also referred to as NT2, before the NT (also referred to as NT1) at the customer side. The PBX is a concentrator/multiplexer, or simply stated, a switch which distributes the calls (B-channels of the PRI U-frame) to the corresponding Terminal Equipments or collects the calls of the TEs and puts them to the corresponding TS of the PCM frame to send them towards the telco.

To sum up: ISDN is digital. ISDN is a dial-up technology (oh yes, it is), but much better than the original dial-up modem technology (also called analog modem, even though digital modulation schemes are used!), due to the digital signaling, the digital TDM technology, the higher bandwidth used, and other stuff that have to do with upper layers (because ISDN is actually a protocol suite, and defines protocols in layers 1, 2, and 3 - not only in the Physical layer of the OSI Reference Model). In U-interface, when we have BRI ISDN, full-duplex communication is achieved via echo cancellation, whereas when we have PRI via TDM (time-sharing).

This is by no means a complete description of ISDN. I have left "blank" many topics on purpose.

In ADSL now, we take blocks of N symbols of the data stream (each symbol is a group of bits, from 1 bit to m bits) and we convert them from serial to parallel. Then, each one of these N parallel symbols modulates one of N sub-carriers, each one with different frequency. Finally, we sum all these signals to take the resulting composite signal. As you can see, we do not have multiplexing here, because all these sub-carriers carry data symbols from the same source. In multiplexing signals from many different sources share a single medium (let's say a cable), either in time (Time-Division Multiplexing, TDM: each source signal is assigned one or more timeslots) or in frequency (Frequency Division Multiplexing, FDM: each source signal is assigned one or more frequency bands), or in code (Code Division Multiplexing, CDM: each source signal is assigned a code; thus, all signals share the whole bandwidth for the whole time but still they are distinguishable at the receiver due to their different code, which serves as an ID), etc. In ADSL we have multi-carrier modulation (MCM), reffered as Discrete Multi-Tone (DMT): data from a single source are converted from serial to parallel (S/P conversion) and modulate multiple parallel carriers (or tones, that is why the name "multi-tone"). The frequency band for US and DS is different, but they can overlap also and separated at the receiver via echo cancellation. The name "discrete" stems from the fact that the signal is designed in frequency as frequency samples and converted in the time domain as time samples via a DSP which implements the Inverse Discrete Fourier Transform (IDFT) operation using the Inverse Fast Fourier Transform (IFFT) algorithm, and then via a Digital-to-Analog Converter (DAC or D/A) this digital signal is converted to an analog signal for transmission over the channel. At the receiver we have the reverse process: A/D, DFT, and P/S conversion.

As you can see, at the local loop we do not have multiplexing. Only at the telco side we have, where multiple DSL signals are multiplexed in the DSLAM (DSL Access Multiplexer) in order to be transported more efficiently to the core network. But this principle stands at every telecommunication network.

Why we use DMT in ADSL? Because we want high data rate R. This implies high bandwidth W (remember Shannon: C = Blog₂(1 + S/N), that is why in ADSL B> 1 MHz (compare with analog telephony, a.k.a. PSTN where the bandwidth is 4 kHz). This implies also low symbol duration T. Thus, typically T<< T_d, the channel impulse response. Thus, we will have inter-symbol interference (ISI), that is, the pulses will spread in time and subsequent pulses will interfere at the receiver (i.e., overlap), causing it possibly to interpret some symbols wrong (simply stated, to think that a 0 is a 1 or vice versa, due to the enrgy spreaded at adjacent symbols) and increasing that way the bit error rate (BER). Instead of using complex and costly channel equalizers (digital filters that remove ISI), in DMT the data stream with rate R is converted to N parallel streams with rate R_N = R/N. That is, the total bandwidth B is divided to N sub-channels with bandwidth B_N = B/N. In other words, the symbol time becomes T_N = NT, which, for N sufficiently large, is greater that T_d. Thus, we have no ISI! Moreover, the data rate is equal to the original of a single carrier modulation (SCM) system: we have N sub-channels with rate R/N, thus the total rate is again N * R/N = R. In fact, we can use rate adaptation, also called adaptive modulation and coding (AMC), to further improve the performance of the system. How that works? We use some pilot tones (unmodulated carriers) to measure the channel in US and DS. In channels with good quality, we use more bits to modulate the carrier, that is, we have more symbols (higher data rate), whereas in channels with not so good quality we have lower data rate (in order to ensure higher tolerance to distortion), and really bad sub-channels are not sued at all.

How ADSL coexists with POTS and ISDN? Simply, the ADSL spectrum starts over the spectrum of these systems. For example, for ADSL over POTS, we have POTS from 0 up to 4 kHz, we have a guard band from 4 up to 25 kHz, and ADSL US starts at 25 kHz. For ADSL over ISDN, ISDN goes up to 120 kHz and ADSL starts at 138 kHz. Moreover, filters are used to separate the spectrums (in case that energy passes from one spectrum to another due to inefficient filtering). There are systems though where ADSL spectrum overlaps with POTS and ISDN spectrum (I think this is called "naked DSL", but I am not sure; search about that ...). Echo cancellation is used once again.

I tried to give you a detailed but still easy to follow answer. I cannot cover everything in a forum, and I am not an expert, I am just a student who is still learning and tries to improve himself. So, if you find anything that I have explain it wrong, please correct me. I hope I helped you a little bit.