As stated before, bandwidth is limited to 100MHz in 10bit mode.
This is the actual frequency response up to 1GHz:
Given that response, are all the humps past the first one actually aliased?
Huh? I’m not sure whether I understand your question. You’re looking at a spectrum up to 1 GHz that has been calculated out of data sampled at 2 GSa/s. Alias signals within this spectrum could only occur because of input frequencies higher than 1 GHz. Since I’ve created this spectrum using a low distortion signal generator sweeping from 100 kHz up to precisely 1 GHz, there is no such possibility. You just see the typical wobbly stopband response of a simple HiRes or similar FIR filter.
Long story short: The HiRes mode uses all the sample data, there is no decimation hence also no reduction of the effective sample rate. Consequently, the Nyquist frequency does not change and will always be half the sample rate as indicated on the Timebase tab.
Would it make sense to always use the 200M BW limiter in conjunction with 10-bit mode? And if all that is true, perhaps it would be sensible for the scope to automatically switch in the 200M BW when 10-bit is selected? Or am I missing something?
Yes, it does make sense to use the additional 200 MHz bandwidth limit. See the attached screenshots that show the frequency response of the 10 bit mode in full channel mode, i.e. at only 1 GSa/s up to 1 GHz.
First without input bandwidth limit:
SDS2354X Plus_FR_BFull_1GSa_500MHz_10bit
You see the frequency response in the first Nyquist zone up to 500 MHz in orange with only one additional “hump”. Of course, the filter had to be implemented differently for the two different sample rates, in order to provide a similar upper bandwidth limit. At 1 GSa/s it uses just 4 samples, which is enough to calculate a 10 bit result. 8 samples are used at 2 GSa/s, but I suppose the resulting 11 bits are truncated to 10 bits again.
The violet reference trace shows the frequency response from 500 MHz to 1 GHz, from right to left, since this is an aliased signal in the second Nyquist zone. For this I had to deliberately apply a signal above 500 MHz even though the sample rate is only 1 GSa/s.
Now compare this with the second screenshot, where the 200 MHz bandwidth limiter has been activated:
SDS2354X Plus_FR_B200M_1GSa_500MHz_10bit
This is similar to the first screenshot, where the frequency response of the second Nyquist zone is shown in green. Both reference traces for the second Nyquist zone are visible now, so you can easily compare the additional protection against aliasing. At 375 MHz (which is actually 625 MHz!) the aliased signal is about 6 dB lower than without the 200 MHz bandwidth limit.
There is only one reason, why we don’t have (and don’t want) the scope dictating the bandwidth limit – this is the very soft roll-off of a first order gaussian filter. If you compare the two screenshots, then you’ll notice that at the specified bandwidth of 100 MHz the actual attenuation is about 2.34 dB without bandwidth limit, but almost 3.3 dB when the 200 MHz bandwidth limiter is active. Consequently, we are violating the specs with this measure. Of course we still use it anyway, but we always have to be aware of the consequences and we should not complain about nearly 1 dB additional attenuation at the specified bandwidth.