It has absolutely nothing to do with the type of RAM technology used. As far as the system is concerned, it can’t tell the difference between SRAM and DRAM. So either type can be used for both normal use, or for storing a hi-res display.
The important point, is that for a true hi-res display, the RAM (either SRAM or DRAM) must be able to respond to normal reads AND respond to the Z80 refresh signal as if it’s a memory read. It’s tied into how the display system works.
The other point, is that some modern SRAM chips are too fast at responding to the control signals, and hence this messes up the timing.
The Sinclair 16K RAM pack (which uses DRAM chips) can be modified to make it hi-res compatible.
I hope this answers the first part your question.
The second part, the ULA basically steals the display data from the Z80, and feeds it a NOP (no operation) code. The Z80 then executes this and moves on to the next byte of display data. Meanwhile, the ULA has grabbed the data that the Z80 got from memory, and now converts it to a serial stream of data, which is the pixel data that goes to the video output, and hence to the screen.
The ZX81 memory map is divided into four.
- The first (lower) 16K part, is the ROM (as it’s only 8K, you get a echo of the first 8K in the second 8K, this is due to partial decoding).
- The second 16K part, is the normal RAM area. If a 16K RAM pack is fitted, it’s simple. If no RAM pack is fitted, there is still 16K of RAM as far as the Z80 is concerned. But the same 1K (for a ZX81) is duplicated/echoed fifteen times.
- The third 16K is more echoes of the ROM.
- The fourth, is more echoes of the RAM. However, it’s important to note that the hardware (the ULA) detects this, and hence ambushes the Z80 so that the display can be generated.
Mark
Standby alert 
