comparator design cmos
If we go to the definition of an comparator, it should be able to give a digital data for an error signal (e). Now to do this it need some time. An op-amp in open loop can do this, but at much higher cost of circuit/power and also takes more time. Latch, if we look its operation, is again an amplifier basically. The difference is that, it works in re-generative mode. Thus its feeds its own output to itself and re-generates unless it saturates at the digital output voltage. Again, this is done at lower transistor count and power consumption. This is the reason why re-generayive amplifiers (the latch) are popular in comparator that simple linear amplifiers. The other reson (to store data) mentioned above is also true once the data is in digital domain. Thus here, latch serves both the purpose.
For reference, you can go through a review paper by Razavi (in JSSC) , also availble in this forum.
sankudey