I would expect the eye to diminish to 50%. RS232 is an NRZ formet, and so during a significant portion of the bit interval, you're actually sampling the surrounding data, and not the data you want. As frequencies go up, lines get longer, etc, more or the eye is the wrong data. That's why I am sure this is a bad idea. The more I look at the problem, the more I'm convinced that a small number of samples, or single one at the optimum interval is the best way to go. I would never implement a correction scheme on what the 'likely' outcome might be. Most of the systems I've seen don't even try to correct bits, despite advanced error detection and correction encoding being available. Most use CRC codes and re-transmission requests.
Last edited: