|
||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||
|
|
Mark Sauerwald, National Semiconductor Corporation A common problem confronting system design engineers is to get a lot of data from point A to point B as inexpensively as possible. In this problem, there are three parameters, which are all interrelated: the data rate, the distance, and the media that is used. The three most common media are electrical wires, optical fiber and wireless communications. In this paper I will explore the most popular of these media, electrical cables, and look at how the parameters of distance, data rate and cable type are interrelated. Electrical Cables Those of us with a bit more gray and a bit less hair on our heads than we once had, remember when you could go into a coffee shop and order coffee, and a warm caffeine-laden beverage was placed before you. Today you have the option of ordering a tall, skinny, Kenyan cappuccino with hazelnut syrup. All these choices bewilder me, so I usually just ask for a plain coffee. The same tends to happen when I look at cable. There are a plethora of parameters that go into selecting a cable and several of these, that are very important, I am going to ignore here. These include all of the environmental concerns, those parameters that have to do with crosstalk between conductors, etc. What I am not going to ignore are the attenuation characteristics of the cable. Attenuation Characteristics All electrical cable attenuates a signal that is sent through it, and the amount of attenuation is a function of the signal frequency and the length of the cable. The relationship to length is fairly easy -- it is linear -- but the relationship to frequency is somewhat more complicated: it follows a sqrtf characteristic. In the figure below, I have measured data, showing the attenuation of a coax cable (Belden 8281) vs. the frequency, for various lengths of cable. ![]() As can be seen from the plot, the attenuation is linearly related to cable length: the attenuation for 200 m of cable at 10 MHz is about 5 dB, when we double the length to 400 m, the attenuation doubles to about 10dB. The roll-off of attenuation vs. frequency is also readily seen. Cable Type A similar set of curves could have been generated with the variable being different types of cable, all of the same length. In both cases the curves will follow the sqrtf characteristic. All we need to draw this set of curves is one point where we know: the cable length, the frequency that it was measured at, and the measured attenuation. We can then generate the entire set of curves for any particular cable. There is a broad range in the attenuation characteristics (and the costs) of various types of cable. In general, twisted pair cables are lower cost than coaxial cables, but they will give you more attenuation per unit length. If the length that you need to transmit is not very long, or the data rate is not very high, twisted pair may be the best choice for your application. Video is a signal carried primarily on coax, but over short to moderate distances it is possible to use twisted pair. See http://www.belden.com/products/tpvutp85.htm for a more in depth discussion of this. Transmitters and Receivers Cable drivers and receivers will allow various amounts of loss in the system before they begin to have difficulty in recovering a signal. Examining the datasheet for a Serial Digital Video (SDV) cable driver, intended to operate at data rates up to 400 Mb/s ( http://www.national.com/ds/cl/clc007.pdf ), we can see that the drive is 800 mV into the cable, and the minimum input swing is 200 mV, implying that we can have 12 dB of loss in the cable if we use this part for both driver and receiver. Measuring attenuation at twice the data rate of the SDV interface (540 MHz) this means that we can get by with about 50 m of cable (Belden 8281). If we wanted to stretch the link further, we will need to either find a receiver with more sensitivity or a better quality of cable. A table showing Serial Digital Interface transmission distances for various types of cable can be found at http://www.belden.com/products/sdtrans.htm . In this table Belden assumes that the receiver can tolerate up to 30 dB of attenuation. To achieve these higher sensitivities in receivers usually means that there is some cable equalization taking place. An example of a part which does this is the CLC014, automatic cable equalizer ( http://www.national.com/ds/cl/clc014.pdf ). With this part attenuation in excess of 60 dB can be tolerated. It is possible to compensate for the cable characteristics by pre-distorting the signal that is being sent down the cable, but this is much more limited than the receiver design. The basic reason for this is that to provide 40 dB of gain to a small signal in a receiver is not terribly difficult, but to provide 40 dB of gain to a 1 V signal that is being transmitted will require power supplies in excess of 100V! What does the received signal look like? The low-pass characteristic of the cable acts on the edges of the digital data to lower the slew rates. This has the effect of reducing the size of the eye opening in the data. As the cable lengths get longer and longer, the rise time of the edges will exceed the time for a bit cell, and at this point the signal will be attenuated so that its peak-to-peak amplitude gets smaller and smaller. Also, the starting voltage for each bit cell is dependent upon previous history, when all of these patterns are superimposed on the eye diagram, the eye opening gets smaller and smaller, until it vanishes. In the series of oscilloscope photos below the input is a 155-Mb/s signal, and the timebase is 1 ns/div. The upper trace is the output of the cable, and the lower trace is the output of a receiver with an adaptive equalizer (based on the CLC014.) ![]() 0 m Cable: Input And Output Are Clean. ![]() 25 m Cable: Input Edges Are Slowing Down. ![]() 50 m Cable: Rise Time Of Edges Is Almost As Long As The Bit Cell. ![]() 100 m Cable: Eye Is Now Closing, Rise Times Are Greater Than Bit Cell. ![]() 200 m Cable: Input Eyes Are Gone, Signal Is Attenuated, Received Signal Still Quite Clean. ![]() 400 m Cable: Input Eyes Are Completely Gone, Jitter Appearing On Output Edges. When the equalizer is no longer able to compensate for the low-pass characteristic of the cable the internal slew rates begin to degrade, which results in noise on the edges of the filter: jitter. So What Tradeoffs Are Available To Me? Engineering is the art of making compromises for a solution that fits the problem at hand, one that is possible. In the case of data transmission there are several tradeoffs that can be made: Maximum data rate vs. length of cable Length of cable vs. cable type and cost Maximum data rate vs. cable type and cost Receiver cost and complexity vs. cable type and cost, Maximum cable length vs. receiver cost and complexity How Does Cable Equalization Work? Adaptive equalizers, such as the CLC014, try and compensate for the effects of a cable by using a high pass filter. The filter characteristic is matched as closely as possible to the sqrtf characteristic of the cable, then the gain of the filter is adjusted to compensate for cable length. It is not easily possible to generate a filter with a sqrtf characteristic response, so what is done is a filter is generated where the poles and zeroes are placed in such a way as to generate a response which is as close as possible to the sqrtf characteristic. The two most common methods of determining what gain to set are to measure the amplitude of the incoming signal and set the gain to the inverse of this, or, to look at the edge-energy of the incoming signal. When data rates get high the rise time of an edge may be several bit cells long. In this case, the amplitude of the input signal will depend on the digital pattern that is being sent, and the edge-energy method will provide better equalization. The amplitude method has the advantage of being low cost and easy to implement, whereas the edge-energy method will allow equalization to be accurately adjusted regardless of data pattern, but will involve more complex circuitry. Conclusion We have scratched the surface of the kind of things that you need to look at if you want to send high-speed data over long distances. The characteristics of electrical cable lead us to tradeoffs between the primary parameters of maximum data rate, maximum distance, receiver design and type of cable. The best solution will be found when these four parameters are all balance together. Analog Main | Product of the Week | Columns | Editorial | Tech Notes
|
|||||||||||||||||||||||||||||||||||
|
Copyright © 2003 ChipCenter-QuestLink About ChipCenter-Questlink |
||||||||||||||||||||||||||||||||||||