TSW1200 FFT and SFDR analysis problem / question

Hi,

May i ask what version of the TSW1200 software you are running, please?  I cant tell from the saved plot.  The version can be found in the lower left status pane of the GUI right after the GUI initializes, or right after it is re-initialized.  You can re-initialize the GUI by using the menu bar, or by the shortcut control-shift-i.

In general, the SFDR value should be pretty straightforward - it is the largest spur in the FFT plot that is not the bin 0 (DC offset) or the fundamental itself.  In our GUI, to find SFDR we eleiminate bin zero and the fundamental and the chosen harmonics, adn then take the largest item left in the FFT and call that Marker1.  Then for SFDR we take the largest harmonic or the marker1 if it is larger, and subtract that from the fundamental.  In this way, we get the difference between the fundamental and the largets thing in the FFT that is not bin0.

But if the fundamental is 'spread-out' due to leakage, then the marker1 will often land on the slope of the fundamental and may severely limit the SFDR reading.  Our later versions of the GUI allow the user to select Notch Spurs in the menu bar, and choose how many bins to ignore around the fundamental and around DC.  But if the input frequency is not coherent and the leakage effect is huge, then notching out a few bins arround the fundamental won't do it.  For this there are the windowing functions.

When the windowing function is used, the chosen window function is applied the the data samples before the FFT routine, and then the FFT routine is run.  The windowing function tends to 'spread-out' the fundamental, more on some window choices and less so on others.  So the GUI internally chooses to notch a small number of bins around the fundamental, usually about 5 or 6 bins, and a similar number around DC.  And the Marker1 routine is disabled, so there is no Marker1 and the SFDR should just be determined based on the first few harmonics.  (Actually, the code sets Marker1 to (0,0) which puts Marker1 in bin zero, and at a height of 0dB down from the fundamental, so you can see the remnant of the Marker1 in the upper left corner of the plot.  Effectively disabled.)  Our next verison of the GUI should bring back Marker1 calculation while windowing is used.

I don't know why the SFDR you see reported is different than the largest harmonic.  With the version number of the GUI you are using, i can look into the code database and try to see what may be the issue, such as whether during windowing the code is using a different number of harmonics that just HD2. HD3, HD4, HD5 that are displayed or if the number of bins notched out around the fundamental wasn't quite enough for the window function chosen.  That last possibility is one you may test yourself if you wish by choosing to notch maybe 10 bins around the fundamental and see if the SFDR now always matches the worst harmonic.

Regards,

Richard P.

Hi Richard,

I'm using the software version 2p5. 

"But if the fundamental is 'spread-out' due to leakage, then the marker1 will often land on the slope of the fundamental and may severely limit the SFDR reading.  Our later versions of the GUI allow the user to select Notch Spurs in the menu bar, and choose how many bins to ignore around the fundamental and around DC.  But if the input frequency is not coherent and the leakage effect is huge, then notching out a few bins arround the fundamental won't do it.  For this there are the windowing functions."

=> yes, that's exactly what I suspect to see in other FFTs without a window - a "spread-out" fundamental whose slope is 'faking' a high spur.

Thanks for your further tips.

BTW: will the next version be able to run under windows 7? I can't even install the last version under windows 7 - that's a bit annoying because one has to switch computers all the time.

Best Regards,

Mikael