Scope bandwidth still confuses
In the last month or so (this was written in 2012 [ed.]), several articles have landed on my metaphorical desktop (or is a screen now the real desktop, and that wooden thing the metaphorical one?), all having to do with scope bandwidth/risetime selection. Are they nonsense, or sage advice?
First off the mark is "Scope bandwidth: trickier than you think!," by Tektronix's Dave Pereles. This is a basic intro to the idea that bandwidth (BW) is almost always specified as the -3dB point. This is certainly an important fact that many newbies may not know, or may forget (some oldbies may forget it, too).
What Dave doesn't get into is the fact that not all scopes display the simple single-pole response so nicely drawn on the lovely log-log paper. Two different "500MHz" scopes may have quite different frequency response curves.
It's also important to remember that in some cases, pairing a 500MHz scope with a 500MHz probe may put you as much as 6dB down at 500MHz, but in other cases, a manufacturer will guarantee 500MHz system performance with an appropriately paired probe and scope.
In "Scope bandwidth, rise time, and the 'rule of five'," Dave continues riffing on the risetime/BW theme. In BW terms, his "rule of five" suggests that you select a scope whose BW is five times higher than the highest sine component you'll be viewing in order to maintain amplitude accuracy. That's a lot of headroom! Sure, manufacturers all want to sell high-BW scopes, but I would suggest you look at frequency response graphs or specs before plopping down your hard-earned money. I suspect that you won't usually need quite so much "guardband." As an extreme counterpoint, I've seen recommendations as low as 3.6× the maximum squarewave frequency you want to view! (which might be a reasonable factor for gigabit links for instance)
Dave's rule-of-five is extended to risetime also. Want to see 1ns transitions? Get a scope spec'd at 0.2ns or better. To my mind, this suggestion carries more gravitas than the other "5×" rule, because it's harder to mentally compensate for a scope deficiency when viewing a spectrally complex digital signal, as opposed to a sine wave.
Once again though, if you're willing to do a bit of mental gymnastics, you can relax the rules. A scope spec'd at 0.2ns will only display such an edge when driven by a significantly faster signal. Apply an actual 0.2ns rise input, and the display will show around 0.28ns. As far as I know (feel free to argue), the display behavior can be described by a "root-sum-square" type of response. Which means you can work backwards! The scope shows 0.28ns? Do the math to come up with the actual 0.2ns signal risetime (this is not quite correct with modern digital scopes, but still pretty close).
The third article is "Sharpen rising and falling edges," by Steve Sandler. Steve directly relates BW to risetime, and I'd estimate his "rule" is close to Dave's, if a bit less conservative. Steve also brings sampling rate into the picture, suggesting four samples per edge, or roughly four times the BW.
Steve then talks about probe-scope and probe-circuit interactions, though I find some of the conclusions a bit glib and incomplete. Future blogs will, umm, probe probing in greater depth.