Some days ago I posted a couple of screen captures to show that FFT decimation is just that, i.e. no filtering to prevent alias.
I used the output of my TinySA, since I had it handy, assumed the 2nd harmonic came from that, and had no way to verify.
This morning I used a self built AD9954 based DDS, at 400 MSa/s, with 9th order 35 MHz elliptic low pass, set to 10 MHz and about -3 dBm.
The output spectrum is quite good, here's a TinySA capture (it had been verified also with a R&S costing 3 × my car...unfortunately no longer available to me).
The FFT in the SDS804X (cough, 824) shows quite a pronounced number of harmonics, fiddling with channel 1 vertical setting it is possible to reduce them a bit, but never at the levels shown by the SAs.
More than a bug reporting, this is a question: is this level of harmonic distortion - which I assume coming from the front-end - normal for a scope of this class?
I can live with that, this is just to make sure my scope is OK. With the pitiful FFT of my previous DS1054Z, this kind of check was unthinkable...
EtA: and yes, I see the noise floor is much lower in the FFT. Harmonics are still higher though, and the capture is with the setting for vertical that gave the best results.