Making sense of scope specifications: why sampling rate value can be so misleading
Having a good sampling rate (also known as sample rate) is important for capturing precise, detailed waveforms. For example, for high-speed CAN bus communicating at 500 kbit/s, sampling at 5MS/s will place at least 10 points for each pulse, providing a picture with clear transitions. So, should we just take specifications for various oscilloscopes, select the ones with sampling rates higher than 5MS/s, and skip the rest? Well, the whole purpose of this post is to show that this is, in general, not a good idea.
I'm going to consider a specific example, Snap-on M4 Scope Multimeter Module, found in products such as Verus and Zeus. The maximum sampling rate for it is 6MS/s, so we should be good, right? For a definitive answer we need to know the details, and Snap-on is kind enough to provide them in a document called "Zeus bid specification" (see the link section). Here is a table from it (I labelled the rows in red for easy reference):
In row 1 we see that sample rate of 6MS/s is achieved only for 50 microsecond sweep and 1 channel only. But we need two channels to monitor CAN-High and CAN-Low, so this does not work for us. We have to move to the second row, immediately sinking the sample rate down to 3MS/s. This is below our self-imposed 5MS/s value, but, oh well, not that bad.
With 100us sweep, many CAN bus packets will not fit the whole screen, so we might want (and Snap-on recommends that in one of the articles, see the links section) to switch to a sweep of 200us. That is row 3, and the sample rate is 1.5MS/s now, with transitions having a somewhat trapezoid shape:
Things get more complicated if the issue with the CAN bus is intermittent. Suppose we are watching the scope and notice some change in the waveform, or, perhaps, the car illuminates the MIL. What are the chances we catch the issue when we stop the scope? Take a look at the Total time column for the 200us sweep -- it is just 87 milliseconds. If the issue is truly intermittent, by the time we stop the scope, it is gone. We can try to increase the total time to half-a-second or one second (rows 6 and 7), but then the sampling rate drops even further, to 250kS/s and 100kS/s, respectively*.
Such sneaky intermittents are relatively rare, but this example shows that, compared to Verus or Zeus, we have a higher chance of catching an intermittent with, say, a scope that has a maximum sampling rate 500kS/s for each channel, but can run at that rate for 5-10 seconds. Ironically, that scope would have been disqualified if we used the "sampling rate must be at least 5MS/s" rule for scope selection. Of course, ideally, we would like to use a scope that can run at 5MS/s for 5-10 seconds, but that is a whole different story.
To summarize, the maximum sampling rate listed in the specifications is just a number until it is put in the context of the tests that are to be carried out. This allows to determine the effective sampling rate and see if it is adequate for the task.
Well, this is it for now, thank you for sticking with me to end of the post. I have a lot more examples illustrating various scope specifications. If you found this discussion helpful and would like me to go over other values commonly found in oscilloscope specifications, possibly via a Zoom talk or YouTube Live, please let me know.
*EDIT: Some issues may still be visible at longer sweeps if the scope is switched to the Peak Detect mode.
Links: snapon.com/display/1068/B….pdf snapon.com/diagnostics/uk…
So what you are saying is sample rate is nearly useless without a decent buffer size, at least when looking for intermittents?
Pretty much so for scopes that use the buffer in the standard way. However, there are other types of scopes, see Olle's reply.
Thank you, for the valuable information. Go to automotivetestsolutions.com/all-ats-articl… read article “common questions asked about scope speed” by Bernie Thompson it a very interesting article shout out to “The Bernie master”.
Very interesting Dmitriy...Thank you for sharing this.
There are some scopes that may not have the greatest sample rate, but they continuously record data to an external buffer, such as an hard drive. With USB v3 transfer rates are more than adequate. That way, you have a practically limitless buffer.
Are you using the Autoscope4? I personally think that scope is one of the best scopes I own, not very convenient for roadtest situations, but anything in the bay it's my go-to scope. And I have a Pico 4425, Verus Edge scope, Autel and the uScope to choose from.
Yes, I do have an Autoscope4. Other than some sensitivity to strong electrical fields, I really do like it.
Great post Dmitriy, although those Snap-On specs may not be to impressive, someone can still certainly get the job done. For many years that was my only scope and I'm sure for many others reading this post they might be in that situation. I just don't want to see good Techs get discouraged if they are limited to certain equipment.
Can't argue with that, David -- the M4 Scope Multimeter module works just fine for many many tests.
If you acquire a 500k can signal , using a scope with a max sample rate of 500k/s, how do you know you are not aliasing? Nyquist theorem states to accurately recreate a digital signal it must be recorded at twice the frequency interval. 500 kbps, bits per second is just that 500,000 bits in 1 second. Since the bits have the ability to be high or low, a zero or a 1 , the baud rate is actually…
Hi Jose, the waveform at 500kS/s, indeed, will not look pretty, but I have never claimed it would. I said that the chance of noticing an intermittent will be higher for a continuous 500kS/s capture than for a scope with a short total collection time. This is because the intermittent will not even be stored in the buffer when the latter scope is stopped, and that was the point of the comparison.