Case study of the method to solve the distortion of high-frequency pulse signal measured by oscilloscope

An engineer in Shenzhen Futian Huaqiangbei who specializes in screen development and production needs to use an oscilloscope to measure a series of pulse signals when the Apple tablet computer iPad powers on the screen. After the oscilloscope captures the pulse signal, he can simulate the signal. However, this friend failed to measure it several times, or was not satisfied with the captured signal, so he brought his Macoshine tablet oscilloscope and other related equipment to consult.

First, he demonstrated his measurement method. He needed to measure three signals in total, connected to three channels of the oscilloscope. When channel three was powered on to generate a direct current, channel one and channel two would each generate a pulse with a positive and negative interval and different pulse widths. What he needed to observe was the pulse change pattern of channel one, and use it as a basis for simulation.

The DC generated by channel 3 is around 2.5V, and the pulses of channel 1 and channel 2 are within ±500mV. Therefore, he set the vertical range of channel 1 and channel 2 to 200mV/div, and the vertical range of channel 3 to 1V/div. Then he set the time base of the oscilloscope to 500ms, which means that one screen records a waveform of 500*14ms, that is, a signal of 7 seconds in length.

Then he connected the signal to three channels respectively, and powered on the oscilloscope. The oscilloscope entered the scrolling mode at the 500ms time base, so he could see the signal changes in real time. After capturing a screen of signals, he pressed the pause button, then adjusted the time base to expand the signal and observed the signal at the pulse-intensive part of channel 1. However, the waveform he saw after expansion disappointed him greatly, because the expected square wave turned into a sawtooth wave. Some pulse signals were even lost.

Actually, there is nothing wrong with his operation. The problem is that his operation requires the oscilloscope to have a large storage depth, so that when the time base is increased, the sampling rate will not be reduced too much. The cycle of his pulse signal is actually around 1us, that is, the frequency of 1M. At this time, the bandwidth of the oscilloscope still meets the measurement conditions, but the sampling rate has been reduced too much due to the limitation of the storage depth. The ideal measurement sampling rate should be around 5M/s-20M/s.

Here is a basic knowledge point to share, that is, the real-time sampling rate of the oscilloscope is = oscilloscope storage depth ÷ waveform recording time. From this formula, it can be seen that since the storage depth of the oscilloscope is fixed, the longer the waveform recording time, the lower the real-time sampling rate of the oscilloscope. When we buy an oscilloscope, we always see the oscilloscope marked with a sampling rate of 1G/s or 2G/s, and often ignore the memory depth indicator. In fact, during the measurement process, if the oscilloscope's memory depth is too low, the oscilloscope cannot maintain this marked sampling rate.

Once the problem is found, it is easy to solve. First, we adjust the oscilloscope's memory depth to the maximum of 28Mpts, which is automatic by default. Since the oscilloscope has three channels open, each channel is allocated 7Mpts.

Then, through the overall observation of the previously captured signal, we set the time base to 1ms, set the trigger mode to edge rising trigger, move the trigger level up to 292mV, and then click Single SEQ, intending to use a single trigger mode to capture the signal. After setting up, power on, and then the oscilloscope captures the signal as shown in the figure below.

Then, we stop the signal, adjust the time base and expand the signal, and we can clearly see each pulse of channel 1 and the pulse with a larger pulse width. Users are curious about why there is a more obvious protrusion above the pulse signal, that is, overshoot. In fact, it is because his ground wire is too long. Turning on low-pass filtering can also alleviate this display problem.

Later, we took a prototype that was about to be launched on the market, which had a much larger memory depth. We used his original method to measure it. Because the memory depth was large enough, there was no distortion after the 500ms time base was expanded. Therefore, the importance of memory depth was further verified.

Finally, the user saved the waveform data he needed and returned home with a full load. I believe his problem was finally solved.