Thanks, it started making more sense. But, I am still not sure where 39.78uA current came from, and why at the output of the op-amp, the voltage varies from 0-5?

could yo please explain how does the positive path maintain a constant current?

This makes a lot of sense, thanks. Can we use the regular negative feedback to acmplish the same thing? Something like this? upload.wikimedia.org/wikipedia/commons/thumb/4/41/…

Andy, thanks a lot. This makes perfect sense now. This is exactly what my circuit it diong.

very precise and clear explanation. It makes perfect sense. Thaks a lot for the detalied explanation.

But, could you read my comment above about having independent ground for two seperate units with a differntial signaling? But I thought the answer to this question is that differntial signaling works with no reference needed!

@PhilFrost: Okay even though the noise is only on the path of the positive input, but since it is close to the negative input path, then it will affect it the same way? So, we still call this a balanced line?

@PhilFrost: I posted a question at the comment section in the other post, could you take a look at it please?

I am not sure why driving just one of the inputs will not cause a mis-balance? Why would the noise still see equal impendence on each inputs? How does the noise of V1 (which is on positive input only) affects the negative input?

@PhilFrost: You pointed out to a nice article, which sorts of explains. But now I got even confused on the other article, because I still don't think that "hot" and "cold" need to be flipped of each other. As long as we can assure they are treated equally, and ending up in having the same impedence, I think it is totally fine :)

I see, Thanks. But I am still not sure why we use it in the case of ethernet? Because my impression from this discussion is that we are talking about ground isolation, and if coming out of ethenet phy is already differntial, then it is already isolated !! Then why do we want to isoalte an already isoalted signal?

Thanks. This makes sense. But then if coming out the ethenet chip is already differntial, why can't we just connect it directly to the receive end. Becasue regardless of how the groudning is, since we have a differential signal, we can always get the data, correct? Then what is the main purpose of transformas in this case?