2

As oscilloscopes get faster and faster, when designing a new scope how do you know that what it's displaying is the same as the actual waveform? For the fastest scopes, you don't have something faster that you can use to verify that the measurement is correct.

For testing a new scope design, is there perhaps some reference circuit that produces a waveform with known characteristics so that you can tweak the probe/amplifier/digitizer until you measure what you "know" the reference circuit is producing?

Do you perhaps measure single frequency sine waves and then assume a certain rise time waveform can be measured based on the bandwidth across which the scope can accurately measure the sine waves? What about intermodulation distortion?

crj11
  • 5,566
  • 1
  • 13
  • 32
  • For RF linearity and rise time common test methods include impulse tests towards 1ps rise time arcs into 50 Ohms using semi-ridged coax and dual RF 3OI measurements . – Tony Stewart EE75 Jul 17 '18 at 13:01
  • Then add network analyzer s parameter tests to calibrate, as VNA’s can go higher than scopes – Tony Stewart EE75 Jul 17 '18 at 13:08
  • Assuming you mean oscilloscopes at the end of the spectrum (100GHz) this is all RF magic that a lot of engineers design that do a lot of different tests with known signal sources and in the end conclude that its probably right. – PlasmaHH Jul 17 '18 at 13:20
  • Probability can now be calibrated using Optical Spectrum Analyzers and deconvolution of PD pulse for return loss and gain with Monte Carlo methods – Tony Stewart EE75 Jul 17 '18 at 13:37
  • 4
    Tony, do you every understand that there are beginners in Electronics? Every time I see your comments or answers you always go into as much technical information as possible, even some that is irrelevant. While most can understand, a lot of the questions are posed by beginners who won't understand what you're saying. Maybe try being more helpful with answers rather than trying to over-complicate things in what I can only assume to be a need to impress people or just being knowingly unhelpful. P.S, this wasn't in response to these comments in particular, just the majority of yours – MCG Jul 17 '18 at 13:46
  • @MCG, the answers and comments shouldn't be stuck in the worriness of OP's level. Answers are for everyone. Let them yield like a spectrum. – Ayhan Jul 17 '18 at 14:06
  • @Ayhan Actually, there's quite a bit of history in Meta about this particular case. – AndrejaKo Jul 17 '18 at 14:17
  • @Ayhan actually, you should take OPs level into consideration. And I was talking about a specific thing, over-complicating it for the sake of it, or for showing off etc as I said. It doesn't help anyone and usually leads to more confusion. Sure, some of us do get it, but the answers are supposed to help everyone, including future users who may have similar issues/questions. – MCG Jul 17 '18 at 15:09

1 Answers1

7

For the fastest scopes, you don't have something faster that you can use to verify that the measurement is correct.

This is not true. We may not have time-domain that can display faster signals, or direct digital synthesis methods capable of generating signals at these frequencies, but we have been able to generate signals in the hundreds of gigahertz for decades.

The LeCroy 100 GHz scope is, to my knowledge, still the only one with .1 THz bandwidth (though I've heard that might change in the next few years). I believe it was first demonstrated in 2014, and then released somewhere in 2015, but don't quote me on that. In any case, that bandwidth of real-time time-domain analysis has only become available within the last decade.

But a quick google will show you people talking about sub-millimeter wavelength systems and physics (generally sub-millimeter is used to refer to signals with frequencies above 300 GHz) since the early 1900s. So, for over a century people have been working with these signals.

Through using physical concepts to generate them, such as cavity resonators etc. we can generate signals that are very high frequency. Using non-linear devices, we can make mixers that operate at 1 THz now. So if we can generate this signal, and know it is a very pure sine we can input this into our new scope and start from there.

When working at these frequencies, we very often don't work with the time-domain (so what an oscilloscope displays) but with the frequency domain (what a spectrum analyzer/network analyzer displays). In fact, I have been working with systems operating significantly above 100 GHz for a few years, but I have not used a scope with a bandwidth over 50 MHz in the last decade.

The front-ends of those scopes tend to operate in a more frequency-domain way than a time domain way - they use mixers and power dividers to cut the input signal into a number of bands (in the case of the LeCroy, I believe it is 3 bands), and then mix each of these down to DC. Then we digitize all of those, and use very complex and smart DSP to stitch them all together. Using careful characterization of the system, we can allow the analog front-end to misbehave to some extent, as we can compensate for it in the DSP (provided it misbehaves in a very predictable and repeatable way).

Joren Vaes
  • 12,502
  • 35
  • 64