Farting around with the idea of flash ADC conversion (utilizing a voltage divider and comparators rather than successive-approximation registers and the like, which flash ADC can get you the conversion instantly without requiring active control. It would take 2^n comparators for an n-bit (precision) sample, but the output would not be in binary, it would be based on the number of comparators outputting "high". You could stack comparators and logic to get a binary output using n strings of comparators (1,2,4...2^(n-1) ), but it also requires external logic. 8 bits would require 255 comparators and a host of AND gates and inverters. 4 bits would require 15 comparators and fewer gates.
If you can pull off a module of 4 bits, you can latch the binary output into the top nibble and recycle the sample for the lower nibble of an 8-bit sample. Back-converting the binary output with the use of a summing amplifier gets a clean signal exactly equal to the binary conversion; a differential amplifier can subtract that from your original signal leaving a residue below the previous sample's LSB and cycling that back through your 4-bit ADC module with a 1/16 reference will complete your 8-bit sample. Technically this process could be repeated but component tolerances and cascading error will have a larger effect with each iteration.