Fast but forgotten. (sampling oscilloscopes)

Despite the fact that many modern engineers have never seen one, the once popular sampling oscilloscope is still one of the fastest display devices available.

These days about equal numbers of digital and analog oscilloscopes are being sold. One reason is that the technology of analog-to-digital conversion has advanced to the point where it is cheaper to achieve display bandwidth by sampling a signal and storing it in memory than by displaying it directly on a CRT. Besides, once a signal is in numerical form, it can be processed in many useful ways. For example, it can be averaged to remove noise or its frequency spectrum can be computed.

One problem with digital oscilloscopes is that they must be used with care because of aliasing, something that has no counterpart in an analog scope. If the sampling rate is set to a frequency that is close to the repetition rate of the input signal, the display will show the correct waveform. However, if will indicate a time scale that is much slower than the true one and the display may even be time reversed. If the sampling rate is close to a multiple or submultiple of the data rate, the resulting display looks like a combination of several signals and is difficult to interpret.

Can Aliasing Be An Advantage? Aliasing isn't always bad. Back in the middle 1950s a fast transistor, or a good oscilloscope for that matter, had a bandwidth of 5 MHz. (Or 5 Mc/s as it was called then.) Integrated circuits didn't exist, and if you had asked an engineer to design a fast amplifier he would have used vacuum tubes. Despite that, some people were building oscilloscopes that had an effective bandwidth of 300 MHz. They used a few diodes, a few transistors, and a trick or two.

Their secret was a sampling circuit that could measure signal amplitude in a nanosecond or two. If the input was a pulse that was repeated at regular intervals, then successive samples could be taken from different pulses. If each sample was taken from a different place on each pulse, you could build up a picture of its shape without using any really fast electronics.

Because the sampling ran just a bit slower than the repetition rate of the input, it generated an alias of the input signal. The amplitude samples, which were being taken perhaps every 10 ms, were stretched until they could be …