| Products > Test Equipment |
| Siglent - 11/20 - New SDS1104X-U, 4 channel 100MHz, 1Gsa/s economy oscilloscope |
| << < (7/32) > >> |
| Mechatrommer:
--- Quote from: nctnico on November 29, 2020, 02:38:01 pm --- --- Quote from: Mechatrommer on November 29, 2020, 02:29:51 pm ---just to be clear, if you want this FFT buzz, just buy a proper Spectrum Analyzer, those 1Mpts DSPed scope will become obsolete quick. --- End quote --- No. Because a spectrum analyser doesn't go to near DC frequencies. Also the frequency resolution of 1Mpts FFT will beat a spectrum analyser hands down at low frequencies. --- End quote --- whats the point if it can have leakage in 500Hz bin? anyway, 1Mpts FFT is not something unsolvable in 1kZ+PC. whats no working solution so far (afaik) is 10Mpts FFT Siglent DSO (even +PC). ps: those who are not aware the deeper details of FFT yet, we cant just use all 17Mpts and expect to get 8.5Mpts FFT, it will take ages to compute O(N^2) of simple "any size" FFT algorithm, we'll need O(N.log(N)) such as DIT or KISS FFT, but that need only of size N=2^y (y must be integer) so your 17Mpts must be downgraded to maybe like 10Mpts data in time domain, hence you'll only get 5Mpts FFT. with 24Mpts data, it will downgraded to 20Mpts, hence we can have 10Mpts FFT (50Hz bin instead of 100Hz bin in 5Mpts FFT). fwiw. |
| 2N3055:
--- Quote from: Fungus on November 29, 2020, 02:52:09 pm --- --- Quote from: nctnico on November 29, 2020, 02:38:01 pm --- --- Quote from: Mechatrommer on November 29, 2020, 02:29:51 pm ---just to be clear, if you want this FFT buzz, just buy a proper Spectrum Analyzer, those 1Mpts DSPed scope will become obsolete quick. --- End quote --- No. Because a spectrum analyser doesn't go to near DC frequencies. Also the frequency resolution of 1Mpts FFT will beat a spectrum analyser hands down at low frequencies. --- End quote --- Where these low-end 'scopes fall down is in their analytical abilities. eg. Here's a scope showing a table of peaks in the signal, with frequency and amplitude. This is way more informative than a picture with wiggly lines in it. (sorry it's a bit out of focus - it's from a youtube video) The manufacturers deliberately don't put this stuff in the low end 'scopes. They leave you to mess around with on-screen cursors, etc. --- End quote --- What, like this? |
| nctnico:
--- Quote from: Fungus on November 29, 2020, 02:52:09 pm --- --- Quote from: nctnico on November 29, 2020, 02:38:01 pm --- --- Quote from: Mechatrommer on November 29, 2020, 02:29:51 pm ---just to be clear, if you want this FFT buzz, just buy a proper Spectrum Analyzer, those 1Mpts DSPed scope will become obsolete quick. --- End quote --- No. Because a spectrum analyser doesn't go to near DC frequencies. Also the frequency resolution of 1Mpts FFT will beat a spectrum analyser hands down at low frequencies. --- End quote --- Where these low-end 'scopes fall down is in their analytical abilities. eg. Here's a scope showing a table of peaks in the signal, with frequency and amplitude. This is way more informative than a picture with wiggly lines in it. --- End quote --- Actually the GWI GDS-1054B is able to show a table with peaks with the search function combined with FFT. I don't know if the MDO mode can also be activated on it though. |
| switchabl:
--- Quote from: Mechatrommer on November 29, 2020, 03:04:22 pm ---ps: those who are not aware the deeper details of FFT yet, we cant just use all 17Mpts and expect to get 8.5Mpts FFT, it will take ages to compute O(N^2) of simple "any size" FFT algorithm, we'll need O(N.log(N)) such as DIT or KISS FFT, but that need only of size N=2^y (y must be integer) so your 17Mpts must be downgraded to maybe like 10Mpts data in time domain, hence you'll only get 5Mpts FFT. with 24Mpts data, it will downgraded to 20Mpts, hence we can have 10Mpts FFT (50Hz bin instead of 100Hz bin in 5Mpts FFT). fwiw. --- End quote --- Fortunately this is not the case. The "normal" Cooley-Tukey FFT algorithm can also be used very efficiently for lengths that can be written as a product of small prime numbers (not just power of 2, but 3, 5, 7 etc as well). Even for large prime lengths, you can use something like Bluestein's algorithm, which is not quite as fast but still O(N log(N)). Often you don't really need to compute a X Mpts FFT anyway. If you only want to see a small frequency span at this kind of resolution, then you can just do DDC (digital down conversion) first and then do a smaller FFT. Now, this is not usually available on low-end scopes, but as we are talking about off-line analysis on a PC, this is not a problem. You don't have to code it yourself either, there are many, many efficient FFT codes for all kind of lengths available. |
| rf-loop:
--- Quote from: Fungus on November 29, 2020, 03:13:48 pm --- --- Quote from: 2N3055 on November 29, 2020, 03:08:16 pm ---What, like this? (Attachment Link) --- End quote --- Is that from the 1000X series? All the 1000X manual has to say about FFT measurement is this: --- End quote --- This is from 1000X-E/X-U manual. Just tiny sample. Of course there is markers and table and so on. But if you read 1000X series manual... its your problem. ETA: and as can see there is err... ( 8 ) |
| Navigation |
| Message Index |
| Next page |
| Previous page |