Editor's Note: Avoidingthese some oscilloscope pitfalls will get you betterdata with minimum effort. Steve Sandlertakes a look at four common scope mistakesand gives details on how to avoid them.This particular mistake focuses onbandwidth and sample rates.
The engineering community usesoscilloscopes more than any other piece ofequipment, and yet many of the publishedresults are questionable at best. Someerrors are very common, and so we caneliminate a great deal of bad data byconsidering a few simple, but key points.This first part of a four-part seriescovers: insufficient bandwidth and/orsample rate .
The majority of engineers only have oneoscilloscope on their bench. This is likelydue to cost. While this may be acceptable,it presumes that the one oscilloscope meetsall of our time-domain measurementrequirements. Let’s consider just a few ofthe scenarios that a typical engineer mightface, and evaluate the bandwidth and samplerate required for the measurement.
Case 1: a simple high-speed CMOSlogic gate
It is rare that any electronics these dayswould not include some type of digital gateor buffer. In this simple case, we are usinga NC7SZ04 Tinylogic™ invertergate, connected to a 10-MHz SMD oscillator.The 10 MHz is not significant. It is just ahandy example. In using such a logic gate,we might consider the rise and fall time ofthe gate. The rise will impact the noisegenerated on the supply voltage rail andwill also provide some guidance on thenarrowest glitch we might need to capture ifwe needed to troubleshoot the circuit.
Most engineers might not guess how fastthese very common gates are. In this firstmeasurement, we connected a 30-Ohm resistorand a 0.01-uF capacitor in series with theoutput of the logic gate. The scope is thenterminated into 50 Ohms, allowing it toperform at its maximum performance level. Inthis measurement, we are using a LeCroy640Zi 4-GHz oscilloscope. The rise and falltime are shown in the oscilloscopemeasurement in Figure 1 with a 40 GS/ssample rate.
Reducing the sample rate to 5 GS/s resultsin a 16.5 percent error in the rise timemeasurement. Likewise, reducing theoscilloscope bandwidth to 2 GHz would alsoresult in a 16.5 percent measurement error.
(466 pSec/559 pS).
Accurately measuring the high speed CMOSgate requires at least 10 GS/s and 2.5 GHzof measurement bandwidth. The measurementbandwidth includes the limitations of boththe probe and the oscilloscope. Since we areusing a 50-Ohm matched connection for thismeasurement, there is no probe limitation.We could use a 2.5-GHz oscilloscope.
Case 2: a linear regulatortransient
A 1-kHz step load applied to a linearregulator results in a narrow voltageexcursion, which exponentially recovers tothe starting voltage. The current step isapproximately from 50 mA to 200 mA, with a15 nS rise time. The voltage transient,recorded at 10 MS/s, 50 MS/s, and 250 MS/sis shown in Figure 3. The transient isapproximately 60 nS wide, as recorded at 250MS/s. The excursion amplitude isapproximately correct at 50 MS/s, though thefidelity of the image is poor. The minimumsample rate for high fidelity should beapproximately 10/ transient width or 166MS/s in this case, while a 20-MHzoscilloscope could easily capture thewaveform.
Figure 3: A linear regulator load stepresponse at 10 MS/s, 50 MS/s, and 250 MS/s.
The results of these two cases might beused as a guide for determining theoscilloscope requirements based on theapplication, though there is a danger indoing so as seen in Figure 4. Figure 4 showsthe output voltage of a linear, low drop out(LDO) voltage regulator, which can be seenoscillating at approximately 140 MHz. If themeasurement was performed using a 20-MHzbandwidth scope, consistent with thetransient response requirements, such anoscillation would easily be missed.
Understand the bandwidth and sample rate youneed for your measurements. You willgenerally require more bandwidth than youthink. For most applications we recommendedat least a 1-GHz bandwidth. If you areworking with high-speed circuits, such asmicroprocessors, FPGA’s, CPU’s, or others,we recommend a minimum of 2.5 GHz and noless than 20 GS/s. FPGA manufacturescurrently require power supply assessmentsto 10 GHz. A 4-GHz scope can often be used,but at least 20 GS/s or preferably 40 GS/sare required.
The measurement bandwidth is limited byboth the scope and the probe. If you areusing or plan to be using a probe you willlikely want to increase the scope bandwidthto account for this.