Differential signals are the wave of the future for high-speed, high-volume data transmissions. There's no doubt about it. The prime focus of the convergence industries—television, personal computers, and communications—is not only to transmit more data at faster speeds, but to do it more accurately and more economically. As transmission speeds have increased, digital signal transition levels have correspondingly decreased, making signal integrity and clarity critical concerns. Also, in accordance with Moore's Law, as computer microprocessor speeds continue to double every 18 months, the maximum rate at which ground-referenced signals can reliably transmit data is quickly approaching.
In order to limit the disturbance of signals and reduce power dissipation, more and more designs are turning to differential circuitry. This move to improved differential-signal technology for signal transmission, though, faces challenges—not only at the basic measurement levels, but at the technological-breakthrough level as well. Designers of personal-computer architectures are expanding their use of differential signals to achieve faster speeds. Rambus memory is a prime example, due to its differential clock and data lines. Digital signals aren't just "1s" and "0s" any longer. They behave like analog signals at RF and higher frequencies and require high-frequency analog measurement techniques.
Circuit designers and measurement-instrumentation designers both must invent and improve differential design and measurement techniques to allow technology to jump to the next performance level. The need to understand and use technology-breakthrough measurement tools and proper measurement techniques is of key importance. Many designers still use a pair of probes (single-ended passive or active probes) to measure differential signals. But this technique can be full of hazards, including rapid signal degradation, producing unreliable measurements.
So what's the right way to measure differential signals? We will examine a few key elements that, hopefully, will provide the insight and information required to help define or refine the type of differential measurements required by today's circuitry. We'll also try to identify the pitfalls as well as some of the tools needed to design and verify new products from this differential world.
All voltage measurements are two-point measurements—therefore, they are inherently differential. The measurement is made between two nodes in the circuit. One node is at a potential while the other may be at ground reference or at an elevated voltage level. Single-ended signals are referenced to ground, while differential signals are the difference between two signal lines or test points, neither of which are at ground potential. Many of today's signals fall into the category of true differential or pseudodifferential. These include common telephone lines (balanced nongrounded), battery-powered communication equipment, battery-powered computational devices, disk-drive read-write channel signals, and RF communication ICs.
Many modern RF ICs use differential-signal pairs to provide balanced transmit-and-receive intermediate signals. To ease the matching of these high-frequency signal pairs, the termination impedances sometimes exceed traditional 50-Ω levels. Measuring these impedance levels differentially is easier with a high-impedance differential probe.
While today's signals are becoming more differential in nature, the signal levels themselves are decreasing. Driven by applications that need to draw lower power, such as battery-powered products, this decrease in signal level can result in lower signal-to-noise ratios. And with a lower signal-to-noise ratio, there is a greater need for a measurement technique that can reject the noise or common-mode signals.
An ideal differential amplifier amplifies the difference signal between its two inputs, rejecting signals common to both inputs. Having a high impedance from these two inputs to a common ground helps to eliminate ground loops and their associated problems. The measure of a differential amplifier's ability to eliminate the undesirable common-mode signal is referred to as the common-mode rejection ratio, or CMRR.
CMRR can be degraded by a multitude of factors: amplifier mismatches, poor input connections, long lead lengths, incorrect ground connections, attenuation mismatches, changing source-impedance levels, and increases in signal frequency. The presence of one or more of these factors will degrade the performance of any differential-amplifier design to some extent. It also will determine the amount of common-mode signal that it will be able to reject. Figure 1 shows the effect of unequal source impedance on CMRR, while Figure 2 shows the common-mode error from a differential amplifier with a 10,000:1 CMRR.
So how do you limit the number of errors? Here are a few methods for decreasing the amount of common-mode error introduced.
First, make sure that the measurement tool's bandwidth is sufficient to capture the signal and the noise components, and that it has sufficient CMRR for the signals targeted for capture. Next, keep all signal interconnects to the device under test as short as possible to avoid parasitic inductance and capacitance. If you need to use extended leads, twist them several times to reduce the line pickup. Be careful, though, because this increases the capacitance and inductance to the probe's input. For high-frequency measurements, make ground connections as short as possible (yes, differential probes still have a ground connector). Remember, you may need to connect the differential probes to ground to prevent damage to the device under test. Finally, measure at test points where source-impedance changes are minimal.
Here's a tip for estimating the CMRR error of nonsinusoidal signals: Connect both inputs to the measurement source. The scope will display the common-mode error. Unfortunately, though, this doesn't catch any changes in source impedance at the two measurement points.
Single-ended (ground-referenced) probing requires one test point to be connected to the circuit ground or the circuit common. Using this technique can lead to ground loops, which increase signal degradation. The use of this technique also may increase device-under-test loading to the point of nonoperation.
But, there are situations where single-ended probing solutions may have to be considered. For instance, it may be the only solution available because of bandwidth requirements, or if the measurement solution must be cost-effective. Yet these needs are quickly disappearing with today's improved differential solutions.
Many differential applications, such as disk-drive read-channel signals, Rambus clock signals, and RF amplifiers, are sometimes measured through single-ended techniques. Using a pair of passive or active single-ended probes that have been matched for attenuation, propagation delay, and input/output impedance can produce a certain level of measurement performance.
Sometimes designers don't appreciate or understand the value of differential probes and prefer to integrate single-ended probing solutions into their test regimen. A single-ended probe measurement carries with it a certain signal degradation because it cannot make the appropriate differential-signal adjustments required by a differential measurement. Disturbances that contribute to measurement errors with single-ended probes are: lower common-mode rejection; the insertion of ground-loop paths; lead inductance imbalances; and additional capacitive loads. All of these factors affect the device under test, allowing measurement errors to creep into the design.
The best results that single-ended, quasi-differential techniques can provide are CMRRs of 40 dB (100:1), as compared to a true differential-measurement tool that provides CMRRs of greater than 80 dB (10,000:1) at high bandwidths. Figure 3 shows how single-ended and differential probes can measure differential signals.
Probing Differential Signals
A variety of tools exist for acquiring differential signals. As mentioned above, however, the use of some of these tools may create as many measurement errors as they solve. Some of the tools for acquiring differential signals include:
Quasi-differential solutions: These include single-ended probes that are used with oscilloscopes employing algebraic addition.
Built-in high-performance differential amplifiers: These devices usually are built right into the instrument or into a direct plug-in.
High-gain differential amplifiers: These are generally external accessories and require probes or leads.
Passive differential-voltage pairs: These sets are designed for specific amplifier systems.
High-bandwidth active differential probes: These products provide direct differential input and high CMRR without altering the ground gradients.
Keep in mind that the probe is an extension of the instrument. To reduce the errors inherent to single-ended solutions, the amplifier needs to be placed closer to the DUT. This is accomplished by setting the differential amplifier at the tip of the probe. Placing the amplifier in the probe head as close to the probe's measurement tips as possible greatly improves the measurement accuracy and CMRR components of the measurement solution. Of course, this also minimizes the number of differences introduced into the measurement. At the same time, the probe head is made as small as possible to reduce mechanical stress to the circuit under test.
High-bandwidth active differential probes can be used in place of single-ended probing in most applications, except where the increased cost of this probing solution prohibits its use, or where higher bandwidths provided by single-ended solutions are absolutely required. The main reason the differential probe may be more useful in single-ended measurements is the reduction of the low-impedance ground path. The ground-lead side of a single-ended probing solution, when placed on the circuit under test, creates a series resonant tank circuit. This may lower the measurement system bandwidth. True differential measurements eliminate the low-impedance ground loops created by single-ended probing solutions.
The DUT attachment is a major concern in not only the differential-measurement area, but in all other measurement areas as well. Many of today's differential probing solutions overcome the challenges of DUT attachment by providing a variety of adapters specifically designed for the probe. Many ICs and circuit boards today do not readily allow measurement access to crucial signals. The designer must develop an adapter or include a testing capability right into the circuit design. This is best done in the early stages of the design to ensure that all circuit-design parameters are taken into account.
What seems to be ground may not behave like ground for fast signals where the rise time is less than 2 ns. Ground-distribution problems cannot be isolated with a single-ended probe (even Active FET probes), because the ground, as seen by the IC, isn't always the ground where the probe is attached. What's more, once the probe ground is attached to the circuit, the ground distribution to the IC has been altered.
Many differential-measurement techniques are the same as those mentioned earlier for decreasing the amount of common-mode error. Still, here are a few more procedures to add to your arsenal.
Use interconnects, cables, and adapters designed for the measurement tool you are using. For high-frequency measurements, wind the leads through a ferrite toroid to reduce excessive common mode pickup. Connect the (+) lead to the higher potential level. If measurement errors seem to be apparent, then reverse the leads to check for the level of difference.
When making timing (propagation delay) measurements, ensure that the positive input and negative input of the first test point are in the same direction on the second test point.
Also, remember that measurement-system bandwidth includes scope, probe, and source. Calibrate and characterize instruments, probes, and output devices as a "system." To improve measurement accuracy, system bandwidth should be three to five times the signal to be measured. Knowing your application helps in selecting your probe type. Consider these factors:
Signal type being measured—voltage, current, logic, other
Signal frequency content—dc, Hz, kHz, MHz, GHz
Signal source impedance—resistive, capacitive, inductive
Physical connection considerations —DUT and instrument
Instrument input—50 Ù, 1 MÙ, other
Instrument bandwidth or rise time
Measurement tools have limitations, so there is a need to reduce the amount of known error when making crucial design and engineering measurements. The errors are either electrical (amplitude errors, phase-angle shifts, propagation delays, changing source impedance) or mechanical (physical geometries, attachment degradation, attachment capability). Figure 4 provides an indication of the amount of error involved when the measurement tool isn't fast enough.
Timing measurements are always a concern. As differential and single-ended signals are merged, making these timing measurements will become more and more crucial. For example, when using a differential probe on a processor or bus structure using Rambus, deskew resolution requires timing alignment of the differential signal to the corresponding bus signals. These nanosecond and subnanosecond signals will require additional bandpass from the measurement tool in order to avoid adding errors to the measurement in question.
With any new technology comes new probe-to-DUT interface requirements. The ability to probe the IC will provide the most accurate and reliable signal measurements. Many of today's surface-mount ICs, however, will need to be accessed by probing close proximity pins that require unique probe-tip attachments. Compliance at the probe tips is necessary so probes can be positioned at any angle, and so sufficient force can be applied to assure excellent contact.
Today's computer speeds are limited by the bus structures used by the microprocessor, memory-management chipsets, and peripheral device chipsets. The present architecture used in memory management limits the bus speeds to less than 200 MHz.
Differential-technology applications have improved the speed of processors, displays, and memory buses. Rambus technology, with edge speeds from 600 ps down to 200 ps (600 MHz to 1.6 GHz), represents one of the technology directions for microprocessor memory, displays, and transmission architectures. Rambus technology employs a differential clock as opposed to relying on the present ground-based processor clocking systems. Considering that the clock signal is at the heart of the processor, and that the Rambus clock is differential, it's a fairly safe bet that differential probing technology will be necessary to meet measurement requirements.
IEEE 1394, also known as Firewire, is a rapidly developing plug-and-play technology in the computer peripherals field. It's a pure differential signal. True differential probing will greatly assist this technology's development. The increased dependence on high data-rate signals in video transmission, color printers, and DVD continues to create increased requirements on data-rate transmissions. Signal rise- and fall- speeds reach 260-ps rates with equivalent bandwidths of 1.5 GHz.
Local-area networks with a greater dependence on data transmission have placed increased demands on standard Ethernet connections. Gigabit Ethernet has been developed to alleviate the increased data-transmission demands.
For all of these fast technologies, the low differential-IC's voltage swings (0.8 V to 1.2 V, in some cases) for logic levels increase the susceptibility to noise and signal degradation. Thus, high signal-to-noise ratio is of significant importance. Differential-measurement solutions can provide this level of integrity and performance where a signal-ended system will start to stumble.
Today's signals are measured through all kinds of techniques: single-ended passive probes, single-ended active probes, differential-amplifier systems, battery-powered instruments, and differential probes. The trend is "faster is better." One needs only to look at high-performance technology products and watch how new technology trickles down into consumer products to see this.
As the speeds of devices increase and move into the next generation, the nature of probing will become more complex. Differential probing techniques will continue to evolve and improve. Measured signals will require more direct-connection techniques to provide the reliable and accurate measurement speeds needed to design, verify, and manufacture future products. Today's differential solutions are expanding to meet these and future needs.
Despite all of these voltage-measurement techniques, other measurement styles and approaches may be needed to produce the accurate and reliable measurements required by tomorrow's products. The differential current measurement approach may help improve the speed and accuracy of these measurements, though it's a little harder to use. Higher-frequency current probing solutions are now available for this technique. Electro-optical, another arena that is changing direction quickly, also may lead the way in meeting future measurement needs.
We may not know what the exact structure of future technology will be, but we do know that tomorrow's designs will be faster and smaller. So the measurement tools will have to be faster, more accurate, and easier to use, too.
"ABC's of Probes," Tektronix Inc., Literature number 60W-6053-7, July 1998.
Feign, Eric, "High-frequency probes drive 50-ohm measurements," RF Design, October 1998.
"Measurement Solutions for Disk Drive Design," Tektronix Inc., Literature number 47W-8257-2, April 1992.
Parham, Johnny, "High-Speed Probing," Tektronix, Inc., Literature number 55W-12107-0, June 1998.
Sekel, Steve, "Differential Oscilloscope Measurements—A Primer on Differential Measurements, Types of Amplifiers, Applications, and Avoiding Common Errors," Tektronix Inc., Literature number 51-W-10540-1, July 1997.