| Products > Test Equipment |
| Functional comparison of R&S RTB2000, Siglent SDS2000X and Keysight DSOX1000 |
| << < (32/39) > >> |
| RBBVNL9:
There were a lot of interesting comments on the last video regarding RBW, Frequency Interval △f, FFT Resolution, and Bin Width on an FFT oscilloscope, and I learned quite a bit myself from that. I now made a shorter video (which benefited from direct input from 2N3055 and Perform01, thanks) that explores these terms in greater detail, in terms of theory but also how we can apply that theory to three oscilloscopes. In the end, the RTB poses a mystery... |
| egonotto:
Hello, Thank you, a little help for the use of the bibliography: https://micronix-jp.com/english/note/application/fundamentals_of_speana_2.html https://www.techplayon.com/spectrum-analyzer/ https://devincody.github.io/Blog/post/an_intuitive_interpretation_of_the_fourier_transform/ http://en.ppt-online.org/269835 https://download.tek.com/document/37W_17249_5_HR_Letter.pdf Best regards egonotto |
| Performa01:
Thank you for this very revealing video! I think you have covered all the important topics quite concisely. I still want to emphasize the importance of knowing the Nyquist-bandwidth of the FFT, which absolutely requires the knowledge of the effective sample rate. I've attached a screenshot to illustrate the problem: SDS2354X Plus_FFT_Aliasing_RT1ns We can see a pulse train with 1 ns rise time, which generates a spectrum of about 950 MHz if we include all the harmonics down to -80 dBm. This is not enough to cause any aliasing at the 2 GSa/s original sample rate as is shown in the Timebase tab. This means that the time domain representation of the signal is free from aliasing atifacts. Actually, even with only 1 GSa/s we wouldn't have much troubles in the time domain; the error caused by aliasing could hardly exceed 1%. Now look at the frequency domain. There the decimated sample rate is just 1 GSa/s and we get an aliased spectrum folding back from 500 MHz down to 50 MHz (actually 950 MHz). This is visible because I chose 12 MHz as the repetition rate of the pulse. If I had chosen 10 MHz, the original and aliased signals would overlap and the aliasing might go unnoticed. In this particular example, the aliasing is easy to detect. But there can be others where it is not so obvious and the aliased signals might accidentally be taken for real. Therefore it is essential to know the usable FFT-bandwidth. In this example it is 500 MHz and we can know in advance that a 950 MHz spectrum will cause false signals. It should be obvious that the risk of a reduced FFT-samplerate gets lower with increased FFT-lengths. Considering this, it is baffling that the instruments with the short FFT lengths of all things don't provide that vital information. |
| RBBVNL9:
--- Quote ---I've attached a screenshot to illustrate the problem: --- End quote --- Excellent demonstration of how aliasing shows up in an FFT analysis! I do not have such a clean pulse source at hand, but if I did, it could be a neat way to investigate the RTB2004 mystery and determine what sample rate it is actually using... |
| Performa01:
--- Quote from: RBBVNL9 on July 17, 2022, 02:36:01 pm ---I quickly checked the ability of the SDS to perform two FFT analyses at the same time, and it looks great! I have the impression that the FFT menu settings are individual for each of these, but settings such as the time base are, of course, common. It never came to my mind to check such simultaneous use, and I don’t think the manual or any other material I have seen refers to it. I think many people are happy to learn about it ;-) --- End quote --- Yes, there might not be too many use cases, but it shouldn't be too much of a surprise: we can have two math channels at the same time and FFT is a math function - so this is to be expected. The SDS6000A can even display four FFTs at the same time... ;) Settings are individual except for the acquisition parameters and the timebase, as is true for all the other math functions likewise. The axis annotations as well as markers are only shown for the FFT in the foreground, that's largely consistent with the behaviour in the time domain view. Now I've demonstrated the simultanous display of the spectra for the sum and difference signals in reply #3546 to the following thread: https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg4312450/#msg4312450 |
| Navigation |
| Message Index |
| Next page |
| Previous page |