What happens if you normalise it?
Change this one setting from your original set:
REF Level +10 dBm
and then select normalise?
After you normalise then change back to the previous setting.
The key sequence is on this page:
http://www.edn.com/electronics-blogs/emc-emi-rfi-esd/4396464/Using-a-tracking-generator
" So to normalize the sweep, press TG > Normalize > Stor Ref. Now enable normalization by pressing TG > Normalize > Normalize On. You should now have a straight line sweep at the top of the display."
Still same problem..
Its not possible to Normalize it out..
Sorry..
I'm slightly confused here...
I've never used a Rigol 815 but I've used lots of decent lab analysers. Some of them won't allow this combination of settings because it is not a good idea to expect to be able to put +0dBm (let alone +10dBm) into the analyser with 0dB front end attenuation selected and then go on to make valid measurments of a typical spectrum input. So even the combination of 0dBm ref level and 0dB attenuation is deemed illegal in some analysers and it won't allow the user to punch in these settings without overriding the user input in some way.
However, I think the Rigol has more loss in the various switched stages between the input and the first mixer plus it has a permanent fixed attenuator here even with the 0dB setting on the front panel.
So you might be able to put 0dBm into it at 0dBm attenuator setting but it is going to be marginal. At some point the analyser will have an overload warning trip and maybe it is somewhere around 0dBm with 0dB attenuation.
So it may be normal for some analysers to 'just' get away with this and some not. But it's not a typical way to set up an analyser anway.
I'm not sure this behaviour is classed as a fault. I'd class it as unsympathetic settings by the user with unrealistic performance expectations from the analyser (at these settings) rather than a fault.