One thing you have to add to your simulation.
The scoop entry is also a low pass filter, in this case a 300 Mhz low pass filter,
This is tricky because the 16pF / 1 MOhm scope input is already part of the 300 MHz spec. But I've created bode plots of the original circuit and with an added first order 300 MHz low pass:
http://imgur.com/a/sMYNHThis set of images also contains a transient simulation of the circuit with the added additional filter. As you can see, the rise time is slightly lower now but it does not really make any difference regarding the qualitative effect of the short transmission line between the termination resistor and the scope input.
thats why signals going to the 300 Mhz always look always like a sine wave, because of the low pass filter.
All the harmonics are gone after 150 Mhz, all is left is a sine wave..
So a scoop of 300 Mhz is usefool to 20 Mhz.
I think you are confusing bandwidth and sampling frequency here. All DS2000 scopes have a sampling frequency of 2 GHz.
Also: The harmonics above the nyquist frequency (1 GHz in this case) are not gone or magically filtered by the sampling. They show up as aliasing frequencies. You have to actively filter those components out using an anti-aliasing filter. If you sample fast enough you already have significant low pass characteristics on your input path and don't need to build a filter, its just implicitly there. That's the 300 MHz in this case.
But those filters never do have an ideal sinc impulse response. So you will never see a signal just morphing into a pure sine wave when approaching the filter edge frequency. (You can build such filters in a DSP of course: Just perform an FFT, mask out the frequencies you do not want, and run an IFFT. But you will never see the equivalent of that in an analog filter.)
There is this rule of thumb that you should have at least a factor 10 between sampling frequency and bandwidth. It is a good rule of thumb, but it is not the ultimate answer. The minimum factor between sampling frequency an signal bandwidth depends on the kind of signal you are interested in, the kind of aliasing filter you are using and the interpolation method you are using. In most RF applications you can get pretty close to the nyquist frequency, because you have extremely band-limited signals, use high order filters and you effectively use a sin(x)/x interpolation (you will never actually look at the signal in the time domain, but the algorithms work with an equivalent representation).
Home
Help
Search
Login
Register
