Scope Probes With Attitude

For decades, the venerable 10:1 passive scope probe has connected innumerable signals and scopes to each other. And, when neither the probe’s nor the scope’s performance limitations are exceeded, the displayed waveform accurately represents the input. In the 100-MHz world of the 1980s, engineers and technicians used this probe and scope combination with confidence. The results were even better when they remembered to calibrate the probes.

Passive, high-impedance probes use voltage dividers in which each element is an R-C network rather than just a resistor. When the time constants of the two networks are equal, the circuit theoretically provides perfect attenuation independent of frequency. Using a divider with equal time constants is very important because the cable connecting a probe to a scope typically is 3 feet long with a capacitance of 50 pF to 100 pF.

The cable’s capacitance adds to the scope’s input capacitance and shunts the scope’s 1-M? input resistance, resulting in a time constant of approximately 100 µs. Such a severe low-pass filter begins to attenuate signal frequencies greater than a few kilohertz. Nevertheless, because the probe and scope resistances are scaled by 9:1, scaling the shunt capacitances in a similar but inverse ratio achieves equal time constants and constant, wideband attenuation.

Most 10-M? 10:1 passive probes are limited to a maximum bandwidth of about 500 MHz. One way to deal with faster signals is to use a low-impedance passive probe, also known as a transmission-line probe or a low-capacitance probe. These probes are available with attenuation up to 100:1.

Because the lower section of the attenuator behaves like a pure 50-? resistance, there is no need to include a compensating shunt capacitor in the upper section. For best results, your scope should have a 50-? input although these probes also can be used with a 50-? through termination. If the probe has been well constructed, stray capacitance can be as low as 0.25 pF, and a bandwidth of several gigahertz is achievable. For example, the Yokogawa Model 701974 PBL5000 5-GHz Low-Capacitance Probe is a 10:1 500-? probe with 0.25-pF input capacitance.

Of course, neither the high-impedance low-bandwidth nor the low-impedance high-bandwidth probes may help you. What if your circuit produces signal components beyond 1 GHz and has a 1-k? output impedance? A high-impedance probe doesn’t have the needed bandwidth, and a 500-? low-impedance probe loads the circuit too much. Even a 100:1 transmission-line probe has an input resistance of only 5 k?.

A wide range of active probes has been developed to address this problem. They also help at lower frequencies where a passive probe can handle the bandwidth but contributes too much capacitive loading. An active probe can have an input capacitance of less than 1 pF where even the best 10-M? passive probes are closer to 10 pF. LeCroy’s ZS Series High-Impedance Active Probes are good examples, having a 1-M? input impedance, only 0.9-pF input capacitance, and either 1-GHz (Model ZS1000) or 1.5-GHz (Model ZS1500) bandwidth.

As scope bandwidth has continued to increase, probes have kept pace. Today’s high-end bench scopes have 20-GHz or even greater bandwidths and so do the latest active probes. However, this level of scope and probe performance is not easily achieved and has required the development of new technologies.

Comments by Bill Hansen, president of Cal Test Electronics, help to put the latest probe developments into perspective, “The 20-GHz probes represent a very small portion of the market that actually requires such high bandwidths. To operate at very high frequencies, both probes and scopes use expensive, custom ICs and associated components to reduce the capacitive and inductive loading effects.

“As a supplier of replacement probes, we must use off-the-shelf ICs and components in our designs,” he continued. “We address the very large number of 200-MHz, 500-MHz, and 1-GHz bandwidth scopes with our 500-MHz passive probes and the 800-MHz active differential probe with 2-pF input capacitance.”

Bandwidth Relative to What?

Figure 1 shows the variation in input impedance magnitude and phase for a 10-M? 10:1 passive probe, a 500-? transmission-line probe, and a 1-M? 10:1 active probe. The 10:1 passive probe has a 17-pF input capacitance and 200-MHz bandwidth. The 1.5-GHz bandwidth active probe input capacitance is 0.9 pF. And, the 0.25-pF input capacitance of the transmission-line probe limits the bandwidth to 5 GHz. The bandwidths of the three probes are labeled in Figure 1. The phase characteristics are plotted as thin lines in the same color as the corresponding magnitude curves.

Typically, probe bandwidth is measured in a 25-? system because accurate, flat high-frequency signal generators have a 50-? output and are terminated in a 50-? load to make the bandwidth measurement. The input impedance of a 10-M? 10:1 passive probe is almost entirely capacitive near the quoted bandwidth, and for the example given, the specified 200-MHz bandwidth corresponds to the frequency where the capacitive reactance is 50 ?. This point is marked in Figure 1 with crossed lines.

At multigigahertz frequencies, other effects besides input capacitance may limit bandwidth. For the 1-M? active probe and 500-? transmission-line probes shown in Figure 1, the quoted bandwidth also is marked with crossed lines and corresponds to an input impedance of about 120 ?.

The implications of such a large change from a probe’s nominal DC impedance can be great depending on your application. For example, if your circuit generates 1-MHz square waves, a 10:1 10-M? probe loads the 1-MHz fundamental by about 10 k?, but the 10^th harmonic is driving 1 k?. The probe impedance changes by a factor of 10 for each decade in frequency even though you may be operating well within the specified bandwidth.

If you are using a low-impedance passive probe, scope input capacitance will degrade the performance of a 50-? through termination. A scope with a separate 50-? input will present a much better match to this type of probe. The next best approach is to use a scope with a switchable 50-? input.

A passive 10:1 probe probably is the most common type of scope probe, but the compensated attenuator idea is easily extended, and 100:1 probes are readily available. If the probe contains both sections of the attenuator, then the input impedance can be set independent of the attenuation ratio.

For example, when used with a 1-M? scope, the Agilent Technologies Model 10076A Probe has a 100:1 attenuation ratio, but the input impedance is 66.7 M?. Similarly, Agilent’s Model N2771A is a 1,000:1 probe but has a 100-M? input impedance. Both these probes are constructed to handle high voltages: 4 kV for the 10076A and 30 kV for the N2771A. For the 10076A, the bandwidth is listed as 250 MHz, and it’s 50 MHz for the N2771A.

Although high-impedance passive probe design is complex, the input capacitance is the primary parameter that determines bandwidth. Nevertheless, as the signal frequency increases, reflections between the unmatched cable and the scope and between the cable and the probe input resistance become more important. Typically, these probes use cable with a resistive inner conductor to improve damping.

The ground connection is critical to making accurate single-ended measurements. The equivalent circuit model of a single-ended probe (Figure 2a) shows an inductance between the signal ground and the probe ground. If you provide a poor connection such as a long wire instead of a short, low-inductance ground, the equivalent circuit inductance will dominate. Any common-mode transient will affect measurements because the probe input is unbalanced: There’s inductance on the ground input but not on the signal input. For high-frequency signals, single-ended probes must be used with a short, low-inductance ground connection.

An Agilent Technologies application note describes the variation in measured results that can occur when a cable’s external or outside mode is uncontrolled.¹ External mode refers to the transmission line formed between a cable’s outer conductor and its surroundings. If the ground connection is very good, the influence of this mode is minimized. You can move the probe cable around and change the position in which you’re holding the probe with little effect. Conversely, a poor ground allows external-mode effects to compromise measurement accuracy and repeatability.

Active Alternatives

As the name indicates, active probes contain a buffer or amplifier. This means that they can combine high input resistance with very low capacitance because the amplifier isolates the circuit being probed from the capacitance of the probe cable and scope input. On the other hand, active probes are relatively fragile and restricted to smaller signal amplitudes. Their price is higher than for a comparable passive probe, but for bandwidths around 1 GHz, the small cost difference is outweighed by the performance improvement.

For example, LeCroy’s single-ended ZS Series Active Probes handle an input voltage range of ±8 V. The absolute nondestruct input voltage is ±20 V. That’s a long way from ±400 V supported by typical 10:1 passive probes, but this is a DC and low-frequency rating. For frequencies much above 1 MHz, passive probe input voltage derating curves typically fall off quickly, indicating a much lower maximum input at frequencies approaching 1 GHz.

Nevertheless, the only way you can achieve high input impedance at a high frequency is to use an active probe. However, as shown in Figure 1, the input capacitive reactance eventually will follow the same -20-dB/decade slope for both active and passive probes. The important difference is the frequency at which the slope starts.

At low frequencies, single-ended probing can provide accurate results. As the frequency increases, the imbalance from even a very small amount of ground inductance causes variations in the local signal ground voltage that are included in the measurement. The best way to solve this problem is by using a differential probe. As shown in Figure 2b, external mode inductive effects are avoided when a probe has a balanced, symmetrical input.

Multigigahertz Differential Active Probes
Different types of probe tips and adaptors have been developed to help you attach a probe to the desired signal. However, many of these are just too bulky and cause too much signal degradation at very high frequencies.

Instead, probe heads used at these extreme frequencies tend to be directly soldered to the signals where directly means with very short leads. Of course, that isn’t always possible. In these cases, access is provided by adding equal lengths of transmission line to each probe input.

However, these lines effectively are unterminated because they drive an attenuator and the high input impedance of the probe buffer amplifier. A physically and electrically small damping resistor immediately at the signal end of the coax maintains good high-frequency signal fidelity even though the actual probe is a few inches away.

LeCroy’s WaveLink^® Series includes models with bandwidths from 3.5 GHz to 20 GHz, all with minimum AC impedance >175 ? at the listed bandwidths. The Models D310/320 and D610/620 have 3.5-GHz and 6-GHz bandwidths, respectively, and a 100-k? differential input resistance. The WaveLink High Bandwidth Probes have bandwidths of 13, 16, and 20 GHz but a different input architecture. The DC common-mode input resistance is 100 k?, but the differential resistance is 1 k?.

According to Larry Jacobs, senior product development engineer at LeCroy, “A distributed amplifier topology is used for the company’s most recently introduced 13-GHz, 16-GHz, and 20-GHz WaveLink probes. Within a custom-designed IC, a transmission line provides inputs to a succession of eight amplifier stages. A second transmission line, also included in the IC, sums the outputs of the stages. The noise contributions from the stages are uncorrelated, so the signal gain increases without as great an increase in noise as occurs in cascaded amplifier topologies.

“Higher probe gain means that lower scope gain is required, which typically is accompanied by lower noise. In addition,” he explained, “higher bandwidths typically can be achieved with this amplifier architecture. Since a pure distributed amplifier provides insufficient low-frequency response, an additional amplifier is integrated into the IC for this purpose.”

Agilent has two series of InfiniiMax Probes: InfiniiMax I Series 10:1 Probes have 1.5-, 3.5-, 5-, and 7-GHz bandwidths. InfiniiMax II Probes attenuate by a factor of 3.45:1 and have 10-GHz and 12-GHz bandwidths. All provide 50-k? differential or 25-k? single-ended input resistance.

Input capacitance depends on the mode in which the probe is being used. Any differential probe has a small capacitance Cm between the two inputs as well as a larger capacitance Cg from each input to ground. The single-ended input capacitance Cse = Cm + Cg. The differential input capacitance Cdiff = Cm + Cg/2.

For the InfiniiMax II Model 1169A 12-GHz Probe used with the N5381A solder-in differential probe head, Cm = 90 fF, Cg = 260 fF, Cdiff = 210 fF, and Cse = 350 fF. These are measured values, which accounts for the small error between the theoretical Cdiff = 220 fF and the measured Cdiff = 210 fF. You choose between single-ended or differential measurements by attaching the appropriate probe head.

When a differential probe is connected to a differential signal, ideally it only responds to the difference between the complementary inputs. The probe doesn’t measure a signal’s DC offset from ground or a differential signal’s common-mode voltage.

Tektronix has addressed this situation through a series of TriMode™ Probes that buffer the differential input signals as well as a ground connection. As shown in Figure 3, all three signals are combined through various channels on a custom ASIC controlled from the TekConnect probe-scope interface so you can dial-up the measurement you need.

According to Andy Heltborg, product marketing engineer at the company, “Most of the current high-speed serial data standards require measurements to be made on the differential, single-ended, and common-mode characteristics of the signals. A TriMode Probe can make all of these measurements with a single probe tip soldered to D+, D-, and ground. Doing the same amount of testing with fewer test points saves time and frustration when dealing with these high-density, small-feature packages and test points.”

Affordable Signal Access
Agilent first introduced the concept of separate probe heads, and it’s been widely adopted. Because lead length is so important at very high frequencies, the best results are obtained by soldering a probe into the circuit under test. Rather than needing to have several probes, you still can monitor several points with one probe but by plugging into multiple soldered-in heads.

A similar idea was commonly used in the 1980s and 1990s at a few hundred megahertz when solder-in probe-tip connectors gave you the best measurement results with single-ended signals. You first soldered a connector to each signal and adjacent ground that you wanted to measure. Then you could insert the tip of a high-impedance scope probe into the connector. Because the connectors were relatively low cost, you could simultaneously view four scope channels. And, the waveform fidelity was good because the connector contributed virtually zero ground lead inductance.

Several things have changed in the intervening years although the need to view many signals remains:
•?Many standards specify differential signals.
•?Signal frequencies have increased by a factor of 10 to 100.
•?Component size and lead spacing continue to decrease.
•?One 20-GHz differential probe costs more than $10,000.

Differential measurements always give a better representation of the true signal value, but differential probes cost more than single-ended ones, and in many cases, single-ended probes provide good results. Now that several standards specify differential signals, you really need a differential probe to measure them. Further, as signal frequencies become faster, a single-ended probe’s external-mode effects cause significant errors.

The Tektronix P75PMT Probing Module Tip costs about 5% of the price of the P7500 20-GHz Differential Probe. The P75TLRST TriMode Long Reach Solder Tip lists for about 3% of the cost of the probe. So, the economic advantage of separate probe heads is clear even if all the monetary amounts are significantly scaled up from the days of solder-in probe sockets.

High-Frequency Technologies
Solder-in heads are standard accessories for high-frequency probes, but they are not simple assemblies. The transmission lines linking the head to the probe body must be closely matched and the attachment to the signals made with virtually zero lead length. Alternatively, the damping resistor value may be apportioned between a resistor mounted on the probe-head PCB and a series-connected resistor that you solder to the signal point.

The Tektronix P7500 Series TriMode Probes technical reference manual includes a great deal of low-level details about the probe performance when coupled to the various heads. Impedance characteristic curves derived from time-domain reflectometry testing show just how sensitive probe response is to the different tip connection methods.

Figure 4a shows how the tip capacitance dominates the input impedance of the directly connected P75TLRST Probe Head. Figure 4b has a very different shape to the curve because of the relatively large series inductance. The capacitance is about three times smaller when the signal connection is made through separate resistors, but the additional length contributes to the inductance. For both Figures 4a and 4b, the equivalent L and C values are very small and reinforce Tek’s message that your soldering technique affects results at these frequencies.

Summary

A scope probe is the first level of signal conditioning in a measurement system. Obviously, probes are used with scopes but not always. As Rebecca Suemnicht, senior product manager for digitizers at National Instruments (NI), explained, “Automated test environments such as design validation and production test now face the same challenges that the design space encountered many years ago in unobtrusively measuring signals with frequencies beyond 1 GHz.

“Some users of modular digitizers connect to signals via active probes,” Ms. Suemnicht said. “For example, an AC characterization test station for high-performance logic devices was developed using the NI PXI-5154 1-GHz Digitizer with Tektronix P6245 1.5-GHz Active Probes. This solution allowed all channels of the device under test to be measured without the loading that coaxial cables would have.”

The accuracy of the rest of the system only becomes important if the probe output maintains the fidelity of the original waveform. If you really need to deal with 20-GHz signals, you have little choice but to use an appropriate differential probe. However, at lower frequencies, inexpensive, single-ended probes together with good measurement technique often can produce accurate results at low cost.

Reference

1. The Truth About the Fidelity of High-Bandwidth Voltage Probes, Application Note 1404, Agilent Technologies, 2008.

November 2009