EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: Free_WiFi on December 04, 2019, 07:49:19 pm
-
https://hackaday.io/project/167292-8-ghz-sampling-oscilloscope (https://hackaday.io/project/167292-8-ghz-sampling-oscilloscope)
https://www.youtube.com/watch?v=99u53V7uDFY (https://www.youtube.com/watch?v=99u53V7uDFY)
-
While interesting topic, there are other two forum threads about it.
https://www.eevblog.com/forum/testgear/ted-yapo-towards-an-open-source-multi-ghz-sampling-oscilloscope/ (https://www.eevblog.com/forum/testgear/ted-yapo-towards-an-open-source-multi-ghz-sampling-oscilloscope/)
https://www.eevblog.com/forum/chat/ted-yapo-towards-an-open-source-multi-ghz-sampling-oscilloscope-218878/ (https://www.eevblog.com/forum/chat/ted-yapo-towards-an-open-source-multi-ghz-sampling-oscilloscope-218878/)
-
:-//
-
Calling these comparator based devices samplers is inaccurate IMO. They have much slower acquisition than true samplers and are kind of boring.
-
SyntheSys Research used a similar technique in their BertScopes.
A set of fast comparators with DACs driving the set points and accurate gating delays for the latching.
You can then sweep the contours of the eye diagram as well as quickly establishing a bit error rate.
SyntheSys Research got bought by Tektronix, who extended the product line, with the fastest BertScope covering 28.6 Gbps.
Nowadays you can get real time scopes that are fast enough to do the same thing.
-
Calling these comparator based devices samplers is inaccurate IMO. They have much slower acquisition than true samplers and are kind of boring.
But are there any readily available, not too expensive, and reasonably easy to implement 'real' samplers around?
-
Calling these comparator based devices samplers is inaccurate IMO. They have much slower acquisition than true samplers and are kind of boring.
No, the term 'sampler' is correct. You don't need to sample a repetitive signal with a samplerate which is higher than the Nyquist frequency.
-
They aren't sampling though, they are comparing. They use a gated comparator, not a sampling gate.
-
Calling these comparator based devices samplers is inaccurate IMO. They have much slower acquisition than true samplers and are kind of boring.
No, the term 'sampler' is correct. You don't need to sample a repetitive signal with a samplerate which is higher than the Nyquist frequency.
A sampler convolves the sampling gate with the input signal which in the ideal case produces a non-linear sin(x)/x frequency response which is why the 0.35 rule does not apply. This explains the discrepancies he noted in his presentation. The sin(x)/x frequency response also means that you can completely characterize the non-linear frequency response of a sampler by measuring its first null.
In contrast a comparator input stage is linear which produces some other limitations that a true sampler does not have. Comparators for instance have limited input full power bandwidth.
Still considering the cost of the comparator method, he gets very good results. But having seen the sampling breadboard he made, I am not surprised he only got poor performance with it. He was right about the contrast between the complex engineering design yet simple construction; you really do need to know what is going on and see the physical circuit in a non-straightforward way.
I suspect part of his problem with his analog sampler experiment was confusing sequential and random sampling. He described random sampling but built something suitable only for sequential sampling or at least it looked that way. The two have very different circuit requirements.
-
With a proper sampling gate you can also get away with having an inaccurate sampling time base. Just sample a low passed clock signal from a crystal, both during triggering and at the same time as sampling ... the samples from the clock will give you an accurate time, even if the timing of the sample pulse was not very accurate. This should allow you to get away with simply discharging a capacitor to determine the timing of the sample pulse, which would usually be too sensitive to temperature for long timespans.
-
That is how an inphase and quadrature phase sinewave sampling time delay counter (http://www.ko4bb.com/~bruce/TDC.html) works. They are commonly used where high sample rates are required. Some very fast frequency counters use that method and it would also be suitable for a digital storage oscilloscope that supports a high waveform acquisition rate.
Older digital sampling oscilloscopes did not bother because the avalanche pulse generator which produces the sampling gate could not operate fast enough to require it.
-
You don't need to sample a repetitive signal with a samplerate which is higher than the Nyquist frequency.
Or even a non-repetitive signal. :popcorn:
The devil is in the phrase "Nyquist frequency". A single rising edge can easily break your Nyquist frequency.
-
Calling these comparator based devices samplers is inaccurate IMO. They have much slower acquisition than true samplers and are kind of boring.
No, the term 'sampler' is correct. You don't need to sample a repetitive signal with a samplerate which is higher than the Nyquist frequency.
The term sample is indeed correct.
Nyquist's frequency is always valid, but you need to understand the definition of "signal bandwidth". If you don't understand that then Tayloe mixers (commonly found in cheap SDR dongles) will seem to be impossible magic.