It's just starting to dawn on me that the 2000 never uses sin(x)/x smoothing, unless you have the sample rate cranked all the way to Max (or next to Max with a single channel).
Well, as you surmised later in your post, it likely has to do with the large memory on the DSO (i.e. higher sample rates @ slower time base settings). Using AUTO/14MB, sin(x)/x interpolation is active down to 1ms/div (down to 500us/div with two channels) - and with the 56MB option enabled, down to 2ms/div for one or two channels.
I'm not sure I've ever run into a scope before that enforces that policy (and doesn't let you override it).
Well, as mentioned already, the Agilent X-Series are worse (in terms of what options can't be overridden): it has anti-aliasing and interpolation
always on - and it determines itself whether to use sin(x)/x or linear.
Same here. It wouldn't surprise me, based on the above, if they defaulted sinx to On only at 250 MSa/s rates and up. But let you override that in either direction. (I.e., turn it Off at 500 MSa, or On at 6 kSa, if you wanted.)
Umm... I'd be rather surprised if the DS1000Z gave you MORE control over certain functions than the DS2000. From what I've seen, it's always in the direction of 'less'. Actually, I was wrong. I thought the 'Sin(x)/x' switch replaced the 'Vector' switch on the DS2000, but reading through the manual it appears it has both. So in this regard, the DS1000Z has more functionality than the DS2000, allowing you to manually switch between sin(x)/x interpolation, linear interpolation, or none.
I guess the reason that they HAD to do this was the very reason we mentioned above: forcing linear interpolation at a maximum sample rate of 250MHz was going to cause some bandwidth issues.
Or, as Rigol puts it:
"Enable the dynamic sine interpolation can restore the original waveform more real."