Electronics Explorer oscilloscope bandwidth test (measured at ~88 MHz)

[philipg]

Hi,

I ran a test to measure the bandwidth of the Electronics Explorer's oscilloscope using a fast pulse generator that produces square wave pulses with a rise time of ~2 ns.

The attached image shows the rise time measured by the Electronics Explorer. Note that the oscilloscope was set to oversample and average, as the image also shows. 

Surprisingly, the oscilloscope yielded a rise time of ~4 ns. Using the formula 0.35/(rise time), the calculated bandwidth of the Electronics Explorer's oscilloscope is ~88 MHz. Although oversampling and averaging was required to get this result, this is a surprisingly excellent bandwidth. 

Does this result seem reasonable or am I doing something wrong? 

Thank you.

Phil

[attila]

Hi @philipg

This is a log retired product. In an internal document I see 70MHz bandwidth for the oscilloscope input stage.
https://digilent.com/reference/test-and-measurement/electronics-explorer/start

For better interpolation/oversampling you could set the trigger level to about 3V 

[philipg]

Thank you for responding. Understood that this is a old/retired product, sorry. Here is the oscilloscope result after increasing the trigger level. The resulting rise time is ~21 ns, yielding a calculated bandwidth of ~16 MHz. I'm not sure why increasing the trigger level is better for this type of test. 

[attila]

Hi @philipg

The transitions are usually steeper around the middle and placing the trigger level here makes the triggering more stable, less affected by noise.
The horizontal positioning of the captures for oversampling is done either by FFT phase (if there are multiple periods captured) or by trigger level interpolation (as it is in your experiment). In the second case, a stable trigger is more important.

[philipg]

Since i'm using a Schmitt trigger to produce a fast rising (2 ns) pulse, noise is minimal and the initial voltage rise/slope from zero is very steep. The oscilloscope capture above shows the near vertical rise from 0 volts. I have found that this makes the trigger stable, even when it is set at a very low voltage. I'm wondering if such a low level trigger setting is better for just this specific test -- to estimate the bandwidth using a fast pulse generator. Depending on the trigger setting, the resulting rise times are quite different, yielding very different bandwidth results. The low level trigger yields a bandwidth estimate pretty close to the published 70 MHz value, while the higher trigger is ~5 times lower.

Note that the rise time and bandwidth results would be identical (i.e., ~21 ns and ~16 MHz), if the measurement was taken from 0 volts to the5 volts, rather than from 10% to 90% of the voltage range (as recommended for bandwidth measurements). By setting the trigger level low, the oscilloscope's initial slow response (below 10%) is excluded from the rise time, yielding a much shorter rise time and much higher bandwidth. 

Note that I don't pretend to understand the nuances of this and recognize that you are much more knowledgeable. I'm just commenting on what I'm observing and hoping to learn in the process. Thank you for being so responsive and helpful. 

[attila]

Hi @philipg

The latest version changes the oversample weighting to linear used in combining captures.

Previous versions used exponential weighting to compensate for bandwidth reduction but this could cause artifacts at higher oversampling ratios.

Here C1 is oversampled with EExplorer and R1/2 is captured with a 2GHz 500/20MHz oscilloscope. 

[philipg]

Thank you, Attila. 

The beta s/w certainly changes the oscilloscope results. The rise time now measures ~21 ns, regardless of the trigger level. Also, and the rise time plot is much more linear. The calculated bandwidth is now ~17 MHz. Note that I'm getting similar bandwidth results when I sweep thru frequencies from 10 MHz and 20 MHz. Bandwidth appears to be somewhere in the teens.

Thanks again for your help with this.