@Dave at 33:33 you talk about the memory. It has 512Mbit. That is enough. It has maximum 24MS that means 24M * 16Bits that is 48MByte. If you use more than one channel you have only 12MS (2 channel), 6MS (3 channel), 6MS (4 channel). Each sample has 16 bits because it need 10 bits in some modes the 1054Z use 12bit ADC.
Edit: it is 12 bit >= 5µs/div, and only 6MS with 3 channel
No, the ADC is only 8 bit. If you're talking about High Res mode, that's just the DSO averaging sequential sample points within the same acquisition to produce added bits of vertical resolution.
The number of added bits is dependent on the time base setting, and the slower the setting, the greater the number of samples that are averaged together for each display point, trading off BW for resolution - since it effectively acts like a low-pass filter which limits the real-time bandwidth. An extra bit of vertical resolution is produced for every factor of 4 averages, so 4, 16, 64, or 256 successive samples are averaged for, respectively, 9, 10, 11, or 12 bits of resolution.
If the DS1000Z is like the DS2000, High Res can be turned on and off for a captured waveform (i.e. when the DSO is stopped). That means that the original samples must not be overwritten or changed since there would be no way to reverse the process and recover them after averaging. So if the DSO *is* doing the averaging in sample memory, it's using different memory locations to store the 9-12 bit sums.
Since the DSO has a maximum of 24M of sample memory - that would mean it would need, at a maximum, an additional 24/4 of memory for storing the sums - or 30M of total sample memory - or the next larger stock DRAM size, which, coincidentally, would be 32M * 16.