Jitter is one of the most common things we need to measure these days. I investigated how far I can get using only a mid-range scope's histogram. Reasonably far, it turns out! Find out how.
We've discussed a lot of jitter here at Scope Junction, ranging from gigabit data to megahertz clocks (see some links below), and in most cases, one expects to need a high-end scope to make meaningful measurements. But with a midrange Tek MDO scope at my disposal, I wanted to see what measurements I could make with it, even though it doesn't explicitly support jitter analysis.
What it (and many other scopes in its performance category) does have is a histogram function. If you don't already know your own scope inside and out, take a look. You may discover histograms!
Now, I needed to figure out a good scope setup for making the measurements. After a bit of fiddling, this is what I came up with:
I concocted a DUT -- a 3.6864MHz crystal oscillator -- and corrupted the power supply with 25kHz modulation. The measurement approach I used was to set the scope's buffer to a good size (5MSa, top of screen), then zoom in to an edge near the end. The trigger level and histogram slicing level should be set to the logic threshold voltage.
The trigger edge and histogrammed edge should both match your circuit's active edge. Ooops. As you can see, I used a rising-edge trigger, but measured the falling edge. If I was making real measurements, I'd redo them all, but not this time.
Now that I had a basic setup that seemed to be working, I delved a bit deeper into the scope to see if I could automate some of the jitter measurements. Aha -- a number of histogram-specific measurement functions are available:
The first line simply says how many acquisitions have occurred. Next is the peak-peak jitter, for which I've also enabled the measurement indicators (the dashed vertical lines). The third line is the standard deviation, or RMS jitter. Stats geeks, help me out here. Does SD always equal RMS, or only for a Gaussian distribution?
The last three lines are perhaps only for you aforementioned geeks. They show what percentage of histogram hits lie within one, two, and three standard deviations.
The image above shows the result of sine modulation to my oscillator power pin. For the next, I've switched to an AC-coupled square wave of slightly higher amplitude. The peak-peak jitter almost doubles, but the RMS more than doubles.
Well, we're off to a good start with our jitter measurements. True, we can't do fancy measurements with this scope, like jitter spectrum or TIE trending, but for many applications, it's enough.
In my next blogs, I'll demonstrate a second method I tried, and discuss some of the finer points and cautions you'll want to consider.