cjdelphi gets my point 
@fkeel: you assume that it would take extraordinary high bandwidth to deliver a high fidelity and faithful model of the world. Something like the USS Enterprise holodeck. On the other side I am arguing about the bandwidth that can actually be processed by a human. I am just aiming at the sensors (=eyes, ears, nose, skin). I have very strong doubt that these sensors deliver more than ~10 Mbit per second to the brain.
Let's restrict to the eyes for a moment. There is only so much resolution that your eyes have. Nobody has Gigapixel vision. However the eyes have pan+zoom capabilities. Hence the need for large screens. And hence the need for more bandwidth in the visualization devices. However if we would project pictures right to the retina a much lower resolution would suffice. Still it would be necessary to move the projection according to the movement of the eyes in order to sustain the illusion. But this has nothing to do with bandwidth. It has something to do with us not being able to (a) imperceptibly project to the retina and (b) to adapt for quick movements. (a) is a shortcoming of suitable projectors and (b) is a latency issue.
Now let's have a look at the ears. As cjdelphi noticed mp3 compression reduces bandwidth a lot. However as anyone can easily verify a mp3 song is different from live music. This is because you can walk around and notice how the sound will change according to your position. But consider for a moment we would be able to properly track the head angle and position and deliver suitably adapted sound. Then again mp3 bandwidth would be sufficient. Again we fail to deliver this due to large latency.
In both cases we do not need ultra high bandwidth. Given current technology we can not create such an illusion but this is not due to bandwidth. Acutally even with ultra high bandwidth we can not create such an illusion today. But now consider for a moment we would be able to tap directly into the brains or at least into the nerves to the eyes and ears. Then the bandwidth of the nerves would be relevant. There is nothing outside that must be rendered "just in case". Here I have more than strong doubts that nerves can deliver such high bandwidths. What is the maximum frequency nervous cells can fire? What are the maximum levels they can actually resolve? 8bit? Probably more. 16bit? Probably more. 128 bit? I doubt this (Because this would mean less than 10e-38 of the maximum resolution for the lowest bit). So I think we can assume that no pulse will deliver more than 128 bits. So how many pulses do they send? You argue that the frequency may play a role. Again how luch is this modulation? Again I doubt that you can transmit more 128bit through this additional channel. So far this is an upper bound of 16 Bytes per "fire event". How many events can one cell deliver per second? How many cells are there in parallel? Now where is the hugh bandwidth to be expected from?