CONTENTS

High Speed Digital

Chris Angove's experience in RF design has proved very useful in the development of circuits and interfaces intended for high speed digital signals as used widely in communications in general and telecommunications in particular. The standards encountered have been synchronous digital hierarchy (SDH) and synchronous optical network (SONET). Often these have been carried in the form of balanced (complimentary) transmission lines, with each line screened, effectively the same as two unbalanced lines but running in parallel. The requirements have been for suitably wideband lines to carry the full bandwidth of the data signal. For example the SDH standard STM-64 (or SONET OC-192) are for a baseband data rate of 9.95328 Gbit/s but the bandwidth extends to nearly 20 GHz.

Return to Chris's Home Page

  1. Overview
  2. Wavelength and Dense Wavelength Division Multiplexing
  3. Synchronous Telecommunications Networks
  4. Synchronous and Asynchronous Data Transmission
  5. Return-to-Zero and Non-Return-to-Zero Data
  6. Optical Components
  7. Eesof/HP ADS RF/Analog Simulation and PCB Layout
  8. Power Feed Equipment
  9. Multi-Layer and Mixed Dielectric PCB Design
  10. High Speed Data Components and Interfacing
  11. Forward Error Correction

  1. Overview
  2. The World Wide Web is responsible for the explosive growth in the volume of worldwide telecommunications in recent years. A few years ago the largest proportion of most telecommunications networks would have been voice traffic and perhaps a small percentage data. Today on many of the high capacity routes, that has reversed. Vast telecommunications revenues are responsible for huge investments in the associated networks and equipment.

    Return to Contents

  3. Wavelength and Dense Wavelength Division Multiplexing
  4. The latest optical fiber cables for long-haul telecommunications are not only extremely low loss, but each separate fiber can carry several optical carriers or wavelengths. This has been enabled by recent optical technology breakthroughs, for example optical couplers, filters (gratings), wavelength multiplexers, wavelength converters and optical isolators. But perhaps the most important was the optical amplifier. This technique of carrying many optical wavelengths over one fiber is called wavelength division multiplexing (WDM), typically up to about 16. The latest 'dense' forms of WDM (DWDM) can carry 40 wavelengths per fiber.

    The total optical fiber capacity depends not only on the number of wavelengths but also the data rate carried per wavelength. Each wavelength was designed to be modulated at the (SDH) STM-64 data rate. It seems like physicists have rather got a headstart over the engineers in optical engineering and they like talk about wavelengths instead of frequencies, so optical components like direct modulators, detectors, phase shifters and optical couplers are defined in terms of operating wavelengths, not frequencies.

    A set of standard optical wavelength channels has been defined by the ITU (ITU G.MCS, Annex A of COM 15-R 67-E). These start at 191.7 THz (terahertz) and cover 0.1 THz steps up to 195.9 THz. That corresponds to a range of optical wavelengths from 1563.86 nm to 1530.33 nm. These assume the light to be in free space, with velocity of 2.99793E8 m/s. For a given frequency, the optical wavelength within a typical optical fiber would be about 0.7 of the freespace wavelength on account of the refractive index of the glass.

    The world's telecommunications capacity or bandwidth is increasing dramatically. Much of this is carried by high capacity trunks over optical fiber cables using synchronous transmission.

    Return to Contents

  5. Synchronous Telecommunications Networks
  6. There are two established systems for the synchronous transmission of data over optical fibers: synchronous digital hierarchy (SDH) and synchronous optical networks (SONET). Generally SONET is used in the United States and Japan and SDH elsewhere. Fortunately the many of the data rates for both systems are compatible. In SONET the basic data rate is 51.84 Mbit/s. Multiples of this are referred to as OC-n, where the actual data rate is 51.84*n Mbit/s. For example OC-3 is 155.52 Mbit/s. SDH has a similar system, but with this the basic data rate is called STM-1 and is equivalent to OC-3, 155.52 Mbit/s. Common faster data rates used on long-haul optical carriers are STM-16 (OC-48) at 2488.32 Mbit/s or STM-64 (OC-192) at 9953.28 Mbit/s. These are often generically referred to as 2.5 Gbit/s and 10 Gbit/s respectively. STM-16 and STM-64 are common data rates used to modulate optical carriers in WDM and DWDM systems.

    Return to Contents

  7. Synchronous and Asynchronous Data Transmission
  8. Two types of data transmission are synchronous and asynchronous.

    In synchronous transmission, the clock is carried in parallel with the data. It may be separate from the data, through a separate cable or even through a separate communication system. Alternative it may be included as part of the data. If it is included with the data when it is transmitted, then it needs to be extracted at the receive end and used in the demodulation process. It does not matter even if the delay (latency) over the clock route differs substantially from that over the data route. The clock waveform is periodic and can easily be synchronised. Most high capacity optical fiber telecommunications links carry synchronous data. SDH and SONET are by definition synchronous.

    Asynchronous data is sent in 'packets' or groups of data. Each group normally has some sort of address of the source and destination and other control data as well as the actual traffic data. It requires recognisable pattern of data to indicate the start or a packet. The average information rate over an asynchronous link is variable, depending on how much data was sent. Asynchronous Transfer Mode (ATM) is an example of an asynchronous transmission.

    Return to Contents

  9. Return-to-Zero and Non-Return-to-Zero Data
  10. Smith provides a good reference for this.

    Non return-to-zero (NRZ) is probably the simplest form of data encoding. For each bit the signal level is held constant at one of either of two voltages for the duration of the bit interval T. If the mark and space voltages are respectively 0 V and E V, the NRZ waveform is unipolar and has a long term average of 0.5 E, assuming approximately equal numbers of 0's and 1's are being carried. In a polar waveform the mark and space voltages will be + V and - V and the long term average voltage will be 0 V under the same conditions. The disadvantage of NRZ is that it is state dependent rather than transition dependent. Long strings of 0's or 1's will result in reduced transitions and poor clock recovery. Unless a separate clock is provided, it is often necessary to pre-encode the data before transmission. NRZ coding is therefore limited to short-haul transmission.

    In return-to-zero (RZ), the signal level representing the mark or space lasts for the first half of the bit period and for the second half it is always at zero. Therefore a 1 is represented by mark-space and a 0 by space-space. For the same data stream, this generates more transitions compared to NRZ thus assisting clock recovery, but the clock rate needs to be twice the net information data rate. A long string of 0's will still result in a shortage of transitions, and some form of pre-coding is usually necessary to avoid these. The long term average of a data stream comprising of approximately an equal numbers of 0's and 1's will be closer to the space voltage than the mark voltage.

    Sometimes, there will be long periods of repeated patterns, without necessarily equal numbers of 1's and 0's. This can create a variation in the mean DC voltage level created by the data waveform known as base-line wonder.

    Return to Contents

  11. Optical Components
  12. Return to Contents

  13. Eesof/HP ADS RF/Analog Simulation and PCB Layout
  14. Return to Contents

  15. Power Feed Equipment

  16. Longer optical fiber cables require regularly spaced repeaters, or amplifiers, to overcome the optical losses along the fibers.

    Repeaters are designed to work from a constant current source. This is normally supplied from one or both terminal stations via copper conductors which run along with the optical fibers.

    Each terminal station may need to apply several thousand volts relative to ground to the end of the cable in order to account for the total voltage drop across the repeaters and cable lengths between them. The return path for the current is through the sea itself, via large sacrificial earth electrodes immersed near the terminal stations.

    I always found this 'earth return' concept a strange thing to grasp. Would not these return currents be huge? Of course not when you think about it. A typical repeater current is around TBA A and of course all repeaters work on the same current, so this is also the current in the sea return path for that particular cable. The sea is not only very large, but also a very good conductor. Even with its finite conductivity, there are a very large number of conductance element in parallel, just like resistors in parallel. Although, by the same reasoning, there are very many in series, and series resistors which add up, the parallel ones dominate. I am sure it is possible to prove this using Maxwells conductivity equation but I have not found a reference for it yet.

    These 'return currents' are in fact small compared to natural currents resulting from the Earth's magnetic field. Sea currents are of course effectively moving conductors. It depends on the locations of the sea currents compared with the cables but apply Faraday's magnetic induction rule and you get typically several hundred volts across the entire cable in the above example. Normally one terminal station is held at a constant voltage and the voltage at the other end adjusted to keep the current constant.

    Return to Contents

  17. Multi-Layer and Mixed Dielectric PCB Design
  18. Return to Contents

  19. High Speed Data Components and Interfacing

  20. Eye Diagrams

    Take the STM-64 (SDH) or OC-192 (SONET) data rate at 9953.28 Mbit/s (see Synchronous Telecommunications Networks). There is no way we could hope to see the individual data directly at this sort of rate, such as on a traditional oscilloscope. However a relatively slow oscilloscope can be used to see the shape of the pulses, and in particular the pulse transitions between high and low. Such displays are know as 'eye diagrams'.

    The HP 83480A Digital Communications Analyzer for example will display such eye diagrams. This comprises a voltage against time display, but the actual connections to the data source under investigation are via a high speed probe and suitable high speed / high frequency connectors. The high speed probe will detect the data but of course the oscilloscope trace will be very slow by comparison, perhaps 1/100 of the speed necessary to see the individual data. Even if it were fast enough, would it be very useful? We would be able to see individual 1's and 0's of an extremely small window.

    What actually happens is that the display, although much slower than the data rate, is synchronised to sample the instantaneous level in a small window of data at regular intervals. For the first sample, a dot is put on the screen corresponding to the voltage detected. The sample point of the next window is arranged to be slightly later, a similar dot is generated. After several samples, a pattern is built up known as an eye diagram because of its similarity to an eye. This shows a build up of such samples over time indicating traces for the low, high and transitions.

    Data Spectrum

    Take the STM-64 (SDH) or OC-192 (SONET) data rate at 9953.28 Mbit/s. If you convert data at this rate to the frequency spectrum, using a Fourier Transform, the spectrum will cover theoretically an infinite bandwidth. In practice however from TBA to something less than 20 GHz.

    Return to Contents

  21. Forward Error Correction

  22. All practical telecommunication links generate errors. Errors are corruption of the intended data. The errors are caused by noise which is always present, the higher the noise level relative to the signal the greater the number of errors. There are many types of noise and many noise sources, both natural and 'man-made'. On any digital communication link, the error performance is described by the bit error rate (BER). The BER is the number of errors received divided by the total number of bits transmitted over the measuring interval. For example, if one error was received out of 1000 successfully transmitted, the BER would be 10-3.

    Forward Error Correction (FEC) is often accomplished by Reed-Solomon coding. A typical system will include a Reed-Solomon encoder at the transmit end of a digital link and a corresponding decoder at the receive end. The encoder takes each block of data and adds extra redundant bits for parity checking. The decoder checks the blocks against the parity bits and applies any necessary corrections.

    The advantage of Reed-Solomon coding is that the probability of an error remaining in the decoded data is lower than the probability of an error if it was not used in the first place. Therefore the power of the signal may be reduced for the same BER performance. This is equivalent to a gain known as the coding gain. On a long haul digital link, for example, this could enable a significant saving in the number of repeaters necessary, or may make them unnecessary altogether on some shorter links. Typically FEC encoder/decoder chipsets would be ASICs designed to work at data rates of a few hundred Mbit/s, which is the present limit of the technology. Several such channels would be processed in parallel from a demultiplexed version of the primary data rate followed by a multiplex stage. To carry the same information, of course the encoded data rate would be faster than the original. Typically this might be 15/14, depending on the encoding algorithm used.

    Return to Contents

    Return to Home Page

    Chris Angove
    Telephone 020 8249 0957

    Copyright ©2015, Chris Angove