Sharpen rising and falling edges
Steve Sandler- August 21, 2012Wherever you find a digital control signal, you find sharp rising and falling edges that need characterization. These edges occur on switching power supplies that drive power MOSFETS and they occur in logic gates. Unless your measurement system—oscilloscope and probe—have sufficient bandwidth and sample rate.
POL (point of load) switching regulators have reached 1 nS edge speed and off-the-shelf high-speed logic gates, such as the NC7SZ04, have rise times of roughly 500 pSec. While many engineers measure edges on these components, they are likely getting incorrect results. The right system bandwidth will result in cleaner, more accurate measurements.
Bandwidth and Sample Rate
The first considerations for measuring high speed edges are the bandwidth and sampling rate required to permit measurement of the edge without limiting the fidelity of the results due to the test equipment. The capabilities of oscilloscopes are varied and often masked; some offer Gaussian response while others offer a flatter response (Ref. 1). Specifications can be misleading. The minimum bandwidth and sample rate required are somewhat, though not strongly, related to this performance and so we can estimate the requirements for a 3% accurate measurement (10%-90%) on a Gaussian scope from Equation 1.
The sample rate should contain no less than two points defining an edge, 4- 5 points is more preferable. The sample rate is calculated in Equation 2.
Using Equation 1 we can estimate that measuring a 1nS edge requires a bandwidth of 950MHz and a sample rate of approximately 4Gs/S while measuring 500pS requires a bandwidth of 1.9GHz and a sample rate of 8Gs/S.
One of the simplest ways to measure a high speed edge is using a 50-Ω matched connection (assuming the signal you want to measure can support the 50 Ohm load). This measurement is the simplest because it doesn’t require a probe, which would further influence the measurement.
Figure 1 shows the measurements of a 500 ps edge signal, generated at the output of a NC7SZ04 logic gate, measured on two similar LeCroy Waverunner oscilloscopes. One oscilloscope is a 1GHz model and the other is a 4GHz version of the same model. Each oscilloscope is set to a sample rate of 20 Gs/s to eliminate sample rate errors. The measurement uses the same signal source and interconnect cables, which eliminates the introduction of additional errors.
The 25% difference in measurements shows that the 1-GHz scope bandwidth is not capable of accurately measuring a 500 ps signal, which is in agreement with Equation 1 and requires a 2-GHz bandwidth to accurately measure a 500-pS signal.
Figure 1. The same 500 ps edge signal from a NC7SZ04 logic gate measured on a 4-GHz oscilloscope (left) and a 1-GHz oscilloscope (right) using the same signal source and interconnect cables. The 1-GHz oscilloscope incorrectly indicates a 25% slower rise and fall time due to its lower bandwidth. Click here to enlarge.
Figure 2 shows the measurement of a 250-ps signal measured on the same two oscilloscopes. Each scope samples at 20 Gs/s to eliminate sample-rate errors. The connection is also made using a 50-Ω coaxial connection to a 50-Ω oscilloscope input so that there is not fidelity limitation from a probe. In this case it is much clearer that 1 GHz is insufficient bandwidth to measure a 250 pS edge.
Figure 2. The same 250 pS edge signal from a fast edge signal is measured on a 4 GHz scope (left) and a 1 GHz scope (right). Both measurements use the same signal source, the same interconnect cable and the same Channel (Channel 1) of the oscilloscope to eliminate other error contribution. The 1 GHz scope does not have sufficient bandwidth even with a 50-Ω connection.
Click here to enlarge.
Page 1 of 2Next >