Recalibrating Expectations

The other day I got an e-mail from another engineer who liked everything about our new oscilloscope he was evaluating, but was concerned about the strange display artifacts that our trigger error correction produced. He was seeing something strange and assumed there was something wrong with our product.

The engineer was not prepared for a significant leap in trigger jitter performance. He was seeing characteristics he'd never seen before and he jumped to the wrong conclusion.

Companies compete against each other on the performance, cost, and time-to-market of their products. So it is critical that engineers take advantage of the latest technologies with which to design and build their products. This includes design and measurement tools, as well as electronic components and manufacturing processes.

I was excited when we first introduced our new generation of high-performance oscilloscopes two years ago because we had essentially eliminated trigger jitter in these products. I knew our customers could translate this leap in measurement technology into the improved performance of their new products. The first users who evaluated these oscilloscopes thought so too.

Then I started hearing that some engineers were concerned about trigger jitter-marketing — told to me because I designed the timebase. I was surprised to learn that some users were concerned that our persistence mode display (i.e., many overlaid waveforms) appeared too skinny at the trigger point.

As you know, an oscilloscope's trigger is a mechanism which causes the scope to measure the input waveform when some specified feature (e.g., a rising edge) is detected. Ideally, a scope should display the specified waveform feature at the same position every time the scope measures it. The random error in the displayed waveform's horizontal position is called trigger jitter.

The reason our displayed waveforms appear this way is because our trigger performance is so good they could see waveform characteristics they'd never seen before. It turns out our waveforms are skinny at the trigger point because the trigger jitter is so small. The waveforms get gradually fatter farther from the trigger point because the vertical noise on the displayed waveform gradually becomes uncorrelated to the vertical noise at the trigger point. How quickly the noise at one point on the waveform becomes uncorrelated to another point on the waveform is related to the noise's bandwidth.

The waveforms on the scopes these engineers had been using in the past didn't look like this, and they'd trusted those scopes for years.

In the case of that e-mail I received, it came with a stored waveform file of the engineer's signal. So I analyzed it with our jitter analysis software. It turned out to be a spread-spectrum frequency-modulated clock. It looked just like it was supposed to look, if measured using an ideal trigger. However, the engineer had never seen anything like this before. Instead of gaining new insight into the performance of his circuit he dismissed it as measurement error.

Oscilloscope technology, like that of most tools, improves over time, but occasionally there are leaps in performance which exceed expectations. Just because you couldn't do it or see it yesterday doesn't mean you can't do it or see it today. Be prepared to recalibrate your expectations so you can take advantage of technology leaps when they happen.

Reach Draving at steve_draving@agilent.com.