Designers of analog and mixed-signal products are frequently faced with the task of selecting and optimizing an interface for transmitting binary data. While this data transmission function may be secondary to the central purpose of the product, most customers in our increasingly interconnected world expect a reliable, efficient means of linking new products together.
Before starting to analyze the general data communications problem, a brief description of some relevant terms is useful.
Signaling rate: The signaling rate of a line is the number of voltage transitions that are made per second expressed in the units bps (bits per second). For example, a square wave with a 1 MHz fundamental frequency (1 us period) has a signaling rate of 2 Mbps.
Figure 1. Signaling Rate
Bit error rate: For each bit of data transmitted, there is a finite possibility of incorrectly receiving the data. Causes may include timing problems (synchronization) between the transmitter and receiver, externally induced electrical noise, and/or signal reflections due to mis-terminated transmission lines. The bit error rate is the number of bits incorrectly received divided by the total number of bits transmitted. For example, if 12 bits are incorrectly received in 1 second over a bus continuously signaling at 1 Mbps, the bit error rate is 12/10^6 = 1.2 x 10^-5 errors/bit.
Simplex, half-duplex, full-duplex: Simplex refers to communications in which data flows in only one direction between nodes. Half-duplex refers to communications in which data may flow in two directions, but not over the same media at the same time. Full-duplex communications allow data flow in both directions at the same time.
Point-to-point, multi-drop, multi-point: Point-to-point systems are comprised of one driver and one receiver, and necessarily operate in simplex mode. Multi-drop systems have one driver and more than one receiver, and also operate in simplex mode. Multipoint systems include more than one driver and more than one receiver, and can operate in either half-duplex or full-duplex mode.
Single-ended and differential signaling: Single-ended signaling schemes use a single path per signal, with data levels interpreted with respect to a common reference. Differential signaling schemes use two channels per signal; the data is detected based on the difference between the values on the two channels. The advantages of single-ended signaling are simplicity and lower cost; the advantages of differential signaling are better data integrity and noise rejection.
Before making any choices towards a solution for high-speed communication, a designer should identify the relevant requirements and constraints for the end application.
Data rate: An initial analysis of the application gives an indication of the data rates required, as well as the advantages and challenges of optimizing for maximum data rate. Thus far we have been loosely using the term "high-speed". What is considered high-speed for one application might be considered sluggish in another application. The required rate of data communication depends upon the number of nodes connected to the network, the data requirements to support each node, the data latency allowable between nodes, and allowances for future upgrades. The following table illustrates these factors.
Table 1. Data rates for sample systems.
Reliability: Reliability encompasses not only preventing component and system failures, but also robustness in the face of harsh environmental influences, low bit error rate, operation with failed network nodes and graceful degradation during local faults.
Cost: The total cost of any data transmission network includes not only the transmitter and receiver components at each node, but also the media itself (wire or fiber), the interconnection hardware, the cost of the installation, software development, and customer support efforts.
Distance: The distance between adjacent connected nodes and the maximum distance between extreme points of the network will constrain the communication method and the possible rate of data transmission.
Connectivity: Different methods of communication facilitate various levels of connectivity, in terms of total number of connected nodes, mode of operation (simplex vs. duplex), topology of interconnection (daisy-chain, star, ring), and data collision and error detection and recovery. The designer must consider these aspects based on the needs of the application.
Communication Methods Comparison
The available techniques for transmitting binary data across a distance can be categorized by several methods. One scheme is to use the type of transmission medium to classify the techniques. This gives solutions based on:
wired mediums, including cabled interconnects and printed wiring backplanes,
optical fiber mediums, and
wireless, including radio frequencies and microwave.
While some technologies may have aspects of more than one of these solutions, e.g. infrared photodiode/detector links, most applications will benefit from proper selection of one of these techniques. Each of these solutions has relative advantages for certain applications. In the next few paragraphs, the strengths of each medium will be briefly discussed.
The advantages of wired data transmission solutions include simplicity of design, security against interception, low power consumption, and a wide selection of components and standards readily available. For low complexity applications involving short distances and moderate data rates, there are a variety of industry standards and component manufacturers to support wired data lines.
Because wired solutions can operate at baseband frequency, the complexity of the transmitter and receiver components is minimized. Using properly-matched copper wire in cables or etch in printed wiring backplanes reduces emission losses, simultaneously controlling the effects of cross-channel interference and delivering a high percentage of the transmitted signal power to the receiver.
The infrastructure for connectors, wiring, and board design also supports the low cost of wired solutions. As witnessed by successive generations of wired modems on telephone wires, for instance, the installed base of wireline infrastructure can support an upgrade path as better performance is available for transmitters and receivers.
Solutions using optical fiber as a medium take advantage of the high data rate capability of optical fiber, low interference between signal channels, and security against data interception. The physical properties of optical fiber can offer lower transmission losses at high frequency than copper media, making possible greater line lengths at high signaling rates. Fiber media is also relatively immune to electromagnetic compatibility issues. The infrastructure for optical fiber has a much shorter history than copper wire, but development in connectors, fiber manufacturing, and component design continue to advance the state of the art.
Wireless applications take advantage of the mobility of the non-attached radio-frequency medium. With the exception of low-speed infrared optical links such as TV remote controls, almost any mobile application will use wireless connectivity. Another consideration of wireless vs. wired solutions is the cost of the cabling and associated infrastructure. As shown in the figure below, the cable cost itself, either optical or copper, can be significant for widely separated connected nodes. Note, however, that for networked nodes distributed along this cable length, the cable cost is shared by each node.
Additional considerations for wireless solutions include the need for a mobile power source (batteries, solar, etc.), interference on the shared frequency spectrum, local regulatory and licensing requirements, and health and safety issues due to the radiated power inherent with wireless applications.
Figure 2. Cable Length vs. Cost
Choosing a Standard
Several organizations are active in setting and maintaining standards for data bus communications. These include the Institute for Electrical and Electronic Engineers (IEEE), the Telecommunications Industry Association/Electronics Industry Association (TIA/EIA), the International Telecommunications Union (ITU), and several trade organizations organized to promote specific bus standards. A multitude of widely-recognized data bus standards exist, covering a wide variety of requirements. Unless overriding reasons force the development of a proprietary solutions, designers will benefit from the advantages of using an existing standard: interchangeability, available components, known characteristics, and ready customer acceptance. Choosing the best solution for any specific application begins with a survey of the most common standards. A brief introduction to a few of these standards is given below, refer to 1 and 2 for a more complete discussion. See also Appendix A.
RS-232: The TIA/EIA-232 standard, originally developed (and still often referred to) as the RS-232 standard, specifies the electrical layer, the connector pin assignments, and the protocol for single-ended data transmission at rates up to 20 kbps. Cable lengths can be up to about 20 meters. For shorter cable lengths this standard can be extended to data rates of over 200 kbps. The 232 standard allows connectivity of 1 driver and 1 receiver; the subsequently derived TIA/EIA-423 standard allows 1 driver and up to 10 receivers, and extends the possible signaling rate and maximum distance to 100 kbps and 1200 meters, respectively.
RS-485 (RS-422, CAN, etc.): The TIA/EIA-485 standard, originally developed (and still often referred to) as the RS-485 standard, specifies only the electrical layer for balanced differential data transmission. The scope of the standard includes (but is not limited to) signaling rates up to 10 Mbps, and wireline lengths up to 1200 meters. Available components and systems designs make it possible to exceed either of these qualifications. RS-422 allows 1 driver and up to 10 receivers; RS-485 specifies up to 32 standard drivers and receivers, operating in half-duplex mode on a single pair of wires.
TIA/EIA-644 (LVDS): Low Voltage Differential Signaling (LVDS) is a wireline solution for applications requiring a higher data rate and/or lower power consumption than is possible with RS-485. LVDS retains the advantages of differential signaling, but with lower signal levels (compared to RS-485), both the switching time and power consumption are reduced. The TIA/EIA-644 standard specifies balanced signaling for rates up to 655 Mbps. Due to attenuation of the signal levels, the transmission distance is effectively limited to about 50 meters.
IEEE 1394 (FireWire): This interface standard defines the transmission method, media and protocol for a both a backplane physical layer and a point-to-point cable-connected virtual bus. The primary application of the cable version is the integration of I/O connectivity at the back panel of personal computers using a low-cost, scalable, high-speed serial interface. The 1394 standard also provides realtime I/O and live connect/disconnect capability for external devices including disk drives, printers and hand-held peripherals such as scanners and cameras. Due to the high speed of 1394, the distance between each node or hop should not exceed 4.5m and the maximum number of hops in a chain is 16, for a total maximum end-to-end distance of 72m. Cable distance between each node is limited primarily by signal attenuation.
Signaling Rate Tradeoffs
Once an appropriate standard has been selected, the bus design can be optimized for reliable transmission of data at high rates. Note that each wireline standard may have different ranges of operation, so that the actual signaling rates considered "high" may vary from standard to standard. The following discussion will in general apply across a wide range of standards, with some tailoring to any specific application.
Error rate: The fundamental limitation on increasing the signaling rate across any bus is the requirement to maintain a low rate of errors in the transmitted data. For wireline data transmission, the major sources of error are:
signal attenuation due to cable effects
signal reflections caused by improper termination
induced voltages coupled from electrical noise sources
synchronization faults due to finite signal transition time and transmission delays
Electrical noise (crosstalk, EMI, etc.): As signaling rate is increased, the effects of electrical noise, especially at high frequencies, become more of a concern. All cables, printed etch, and even component leads tend to act as antennas, both receiving and radiating electrical energy. The dominant wavelength of the antenna is inversely proportional to the fundamental frequency. Therefore, as the fundamental frequency is increased, even relatively short line lengths become efficient antennas for radiating and receiving electrical noise. The noise coupling (both the magnetic field coupling and electrical field coupling) reaches an effective maximum when the wavelength is four times the line length. The table below relates this condition (quarter wave antenna) to the fundamental frequency of the data bus.
In order to reduce the electrical noise generated by the data transmitter, the fundamental frequency should not exceed what is strictly required by the application. This means that the rate of signal transition (slew rate) should be controlled to reduce emitted noise. Similarly, to reduce the received electrical noise on a data bus, the frequency response of the receiver should be limited to only the frequency range of the valid data.
Table 2. Signaling rate and Fundamental Frequency.
Power consumption: The power dissipation of any wireline bus system has both dc and ac components. As the signaling rate increases, the ac power dissipation also increases. The following figure illustrates the increase in supply current at the bus driver for three different bus standards. In all cases, higher data rates incur a cost of supply current; the LVDS curve illustrates the advantage of lower signaling levels and a controlled current-drive circuit.
Figure 3. Current Consumption Issues
For any electrical data bus network, the effects of termination impedance at the nodes and at the line extremes must be considered as the signaling period becomes comparable to the propagation time of a signal across the bus. Signal reflections will be generated at any discontinuity in the bus characteristic impedance. If all reflections (primary and secondary) occur during the transition time of the signaling period, they may be neglected. However, if significant signal changes are caused after the transition time has elapsed, these reflections reduce the margin of the signal compared to a threshold. Along with induced electrical noise and signal attenuation, reflections are a contributor towards data bit errors.
The following figure, from 7, illustrates the effects of non-ideal termination on a wireline data bus. Here the signal is being transmitted from a driver with output impedance RS, over a medium with characteristic impedance Z0, to a receiver with termination impedance RT. In an ideal, matched-impedance case, all the impedance values will be equal. This will result in reflection coefficients at each end of the bus with zero value, since RT=Z0. For cases where the termination resistance is not matched to the characteristic impedance of the cable, a non-zero reflection will be produced at the impedance discontinuity.
Figure 4. Effects of non-ideal termination
Depending upon the degree of impedance mismatch, and also upon the propagation time, tp, along the cable, these reflections will distort the waveform on the data bus at every signal change, and may be interpreted as signal value transitions, causing data bit errors. Therefore, consideration of proper termination is important.
The signal distortion introduced by the electrical characteristics of the wired medium can be compensated for (at least to some extent) at either the transmitter or at the receiver. This can be used to extend the maximum distance over which data may be transmitted, and/or increase the usable signaling rate. The major signal distortion effects introduced by the bus are attenuation due to the line resistance and frequency shaping due to capacitance of the network components.
Wireline Attenuation: As discussed in the previous section, proper line termination is required to reduce signal reflections. For long wirelines, this creates a resistive voltage divider, with signal voltage losses proportional to the length of the wireline. For example, a typical cable for RS-485 applications has a characteristic impedance of 120 Ohms, and a DC resistance of 80 Ohms/km. For a 1 km line with proper termination resistance, the following circuit shows how the signal is attenuated at DC.
Figure 5. Differential termination
For a typical differential output voltage (VOD) of 2 volts, the differential voltage at the input to the receiver (VID) will be reduced to:
To allow for this attenuation effect, the popular standards include voltage margins between the minimum allowable output signal and the maximum required input level. The TIA/EIA-485 standard specifies a differential voltage threshold at the receiver not to exceed 200 mV; therefore, a standard 485 receiver would successfully interpret the attenuated signal in the example above. Of more concern, especially for high signaling rates, is the effect of the line capacitance on the transmitted signal shape.
Wireline Signal Distortion: The capacitance associated with the bus network components has a wave shaping effect, with preferential attenuation of high frequency signals. This imposes a limit on the speed of voltage transitions, effectively constraining the signaling rate. At least two methods exist to compensate for this, pre-shaping (emphasis, boosting) of the signal at the transmitter, and post-shaping (equalization) at the receiver.
Compensation by Pre-Shaping: Pre-shaping applies to the signal a shaping function that amplifies the portion of the waveform that will be attenuated by the wireline medium. The shaping function can be designed in the time domain, boosting the signal voltage for a specified time during each transition, or can be a frequency function, amplifying the frequency components most attenuated by the transmission medium. In either case, the pre-shaping must be selected to match the characteristics of the wireline, which assumes that the designer has knowledge of the length and characteristics of the wired medium to be used for the application. Another considerations when using pre-shaping is that the resulting signal may exceed the allowable signal levels of the applicable standard, limiting interchangeability with other products and applications.
The following figure shows the effect of pre-shaping (boosting the signal at the driver) before transmitting the signal over the data bus. The edges of the original signal are enhanced, compensating for the expected high-frequency losses over the wireline medium. If the exact characteristics of the wireline losses were known, an exact compensation could be accomplished. Here an approximate knowledge of the medium is assumed, and the signal that reaches the receiver is an approximation of the original signal. Note that the amplitude of the boosted signal exceeds the original signal amplitude by over 100%, which may be incompatible with the requirement to conform to the signal levels of the applicable bus standard.
Figure 6. Pre-shaping compensation
Compensation by Post-Shaping: Post-shaping applies a shaping function to the signal at the receiver. This acts to recover the signal as it was originally transmitted, before the effects of the wired medium. As with pre-shaping, the post-shaping function must approximate an inverse of the frequency-dependent attenuation of the wireline. The following figure shows the effect of inexact approximation in the post-shaping process. The original signal is shown, along with the degraded version, with frequency response set by the resistance and capacitance characteristics of the wireline. This degraded signal reaches the receiver, and the post-shaping function is applied to restore the original signal. If the actual characteristics of the channel are known, an exact restoration may be accomplished by applying an inverted version of the degradation function. The two restored signals A and B have approximation errors of plus and minus 10%, respectively, of the wireline RC values. These approximations produce the recovered signals as shown, with much of the original frequency content restored.
As with the pre-shaping method, the success of post-shaping is dependent upon accurate knowledge of the wireline characteristics. One advantage of this method, however, is that the transmitted signal is not altered on the bus; therefore no deviation from accepted standards is necessary. Another advantage is that the restoration function can be tailored at each receiver on a multi-receiver network. For widely separated nodes, this allows optimum signal restoration at each receiver; while pre-shaping must compromise in sending all receivers the same signal. Finally, post-shaping at the receiver enables an adaptive shaping scheme, in which the restoration function is tuned to produce a known signal, based on the applicable bus standard.
Figure 7. Post Shaping
In this section, we will briefly discuss a few example applications, in order to illustrate some of the concepts previously introduced. Although this short overview can be neither rigorous nor comprehensive, it serves to point out selected parts of the thought process involved in the optimization process.
RS-232 Application: Computer Peripheral For low-cost, relatively short-distance applications such as home computer peripherals, the RS-232 standard provides the important features of a well-accepted interface, simplicity, and signaling rates suitable for most non-critical functions. The following figure shows a familiar RS-232 application; connecting two PCs through an RS-232 null modem cable.
Figure 8. A Null Modem connects PCs
RS-485 Application: Digital Motor Control In numerous industrial applications, servomotors must be controlled for quick, accurate motions, often in coordinated multi-axis applications. Digital signal processors (DSPs) are increasingly replacing analog control loops to stabilize and direct these motors. These digital controllers require an interface bus for servo parameter set-up, motion commands, and status feedback.
Signaling rate requirements for such a bus depend on how many motors are being networked, the servo bandwidths involved, and the complexity of the control scheme. Coordinated motion between different axes may require low data latency over the bus, driving signaling rates higher. Typical data rates range from 100 kbps for simple single-axis cases, to 5 Mbps or more for multi-axis coordinated networks with complex trajectory requirements.
The environment for such applications frequently includes high levels of electrical noise from the motor current switching, from high frequency processor electronics, and from associated processes such as welding, hydraulic pumps, etc. Distances between nodes can range from less than a meter for compact motor/controller arrangements, to tens or hundreds of meters for large industrial applications where distributed motors and sensors must be coordinated by a central process controller.
One solution for this application would be RS-485, which provides bi-directional signaling over a balanced, twisted-pair data bus. With signaling rates exceeding 10 Mbps and excellent immunity to electrical noise, RS-485 is well suited to industrial applications such as digital motor control. At data rates up from 100 kbps to 5 Mbps, wireline distances of tens to hundreds of meters are possible with proper termination and shielding. Other similar or derivative standards such as ISO 11898 (CAN), and DeviceNet exhibit these same advantages, and are also widely used. The following figure illustrates a multi-axis DSP-based servomotor network, using the CAN (Controller Area Network) bus standard, which has electrical characteristics very similar to RS-485.
Figure 9. A CAN Network.
LVDS Application: PBX/Central Office Switching As the demand for telecommunication services grows, the need for highly interconnected private business exchange (PBX) and central office functions increases. Typical requirements involve passing digitized data between densely packed modules of switching equipment, with high throughput across several meters. Throughput, connectivity, power requirements and robust operation are paramount concerns for an efficient, reliable network. Typically the data integrity is not subjected to the high degree of electrical noise found in harsh electrical environments, but some amount of electromagnetic interference can be expected due to the close proximity of many high-frequency digital subsystems.
The TIA/EIA-644 standard (LVDS) gives designers many of the advantages of RS-485, such as differential, balanced signaling, with a possible signaling rate to over 600 Mbps. Its lower signal levels allow for fast switching speeds at relatively low power, while maintaining high immunity to electrical noise. LVDS has found increasing use in both point-to-point and multidrop applications, using cable media and backplane solutions. The following figure illustrates such an application that uses LVDS for passing data between circuit boards over a backplane in a rack-mounted chassis, and for passing data between chassis over differential-pair cable media.
Figure 10. Backplane Interconnect
FireWire Application: High Performance PC Peripherals For applications associated with personal computer (PC) peripherals, interchangeability and low cost are important. For peripherals that demand the exchange of large amounts of data, such as digital cameras and scanners, data rates higher than those provided by RS-232 are needed. For example, a simple digital camera providing 8 bits of digitized pixel information for a 640 x 480 pixel sensor must transfer more than 2.5 Mbits of data. At an RS-232 rate of 100 kbps, over 25 seconds would be required to transfer one image.
The IEEE 1394 (FireWire) standard specifies a transmission method, media (either cable or backplane) and the packet-based protocol for these types of applications. The electrical signal specifications enable cable distance of several meters, with signaling rates of up to 400 Mbps. The 1394 protocol uses memory-based addressing, allowing access to each node device with processor to memory transactions.
In summary, the design and optimization of a data bus must follow the same requirements-driven methodology as any other part of the development of the system. Key concerns are the selection of an appropriate industry standard, and identification of the application requirements and constraints on data signaling rate. The number of nodes, data latency, and node bandwidth will drive these requirements. The constraints will include data integrity/allowable bit error rate, power consumption, and electromagnetic interference concerns.
Good engineering practice as far as component selection, medium specification, and proper termination will get the data bus design off to a good start. Advanced techniques, including pre-shaping at the bus driver and/or post-shaping at the bus receivers, can be considered in cases where the signaling rate and wireline length must be extended beyond typical values.
1 "Comparing Bus Solutions", Texas Instruments Applications Report SLLA067, March 2000, http://www.ti.com/sc/docs/psheets/abstract/apps/slla067.htm
2 "Summary of Well Known Interface Standards", National Semiconductor Application Note 216, October 1998, http://www.national.com/apnotes/Analog-Interface.html
3 "Electrical Characteristics of Balanced Voltage Digital Interface Circuits", TIA/EIA-422-B, May 1994, Global Engineering Documents.
4 "Electrical Characteristics of Generators and Receivers for Use in Balanced Digital Multipoint Systems", TIA/EIA-485-A, March 1998, Global Engineering Documents.
5 "Pre-emphasis improves RS-485 communications", T. Salazar and L. Suppan, Engineering Design News (EDN), June 10, 1999, pp. 151-158.
6 Noise Reduction Techniques in Electrical Systems, H. W. Ott, John Wiley & Sons, New York, 1988.
7 "DSP and Data Transmittion", F. Dehmelt, Texas Instruments Digital Signal Processing Application Seminar, Fall 2000.
8 "High Performance Serial Bus", IEEE 1394-1995
9 "Electrical Characteristics of Low Voltage Differential Signaling (LVDS) Interface Circuits", TIA/EIA-644, Global Engineering Documents.
APPENDIX A. BUS INTERFACE SELECTION GUIDE
(from 1 "Comparing Bus Solutions" by F.Alicke, F. Bartholdy, S. Blozis, F. Dehmelt, P. Forstner, N. Holland and J. Huchzermeier, Texas Instruments Applications Report SLLA067, March 2000)
Table 3. A summary of bus standards.