OFDM and Frequency selective channel

If notch exists in the frequency band for OFDM system, some certain subcarriers in the frequency notch will suffer from deep fading and cannot be detected by the receiver, and the offen-used one-tap equalizer will also lose effect because the detected corresponding channel transfer function is too weak.

Except for frequency interleaving and error control coding, is there any other way
to handle this case? such as time domain equalization.

In my opinion, for single carrier system, linear equalization and DFE perform well in such a frequency selective channel because the equalizer can use the signal's energy at the frequency band excluding the notch. How ever, for OFDM, the signal lies only in the faded band, and the time domain equalizer cannot recover it.

Is what I though right, or is there any technique to recover the faded subcarrier?

You are right,

only coding could reduce the effect of deep fades. OFDM is nice because it reduce the channel equalizer calculation from n^2 to n*log. Some try to improve by using a different number of bits for each subcarrier. The problem is that the this information have to send to the receiver before demapping with a more reliable format. So it introduce overhead.

A better way is to use a symbol which covers the hole bandwidth. Then a deep fade affect each symbol only by small amount. In most cases where a fade exist on other frequencies the energy is higher. If you know the channel a matched symbol could be used to improve the link margin. That gives additional 1-2dB. For channel equalisation there exist equalizer operating in the frequency domain having similar n*log advantage.

I cannot quite understand the last sentence of your kind reply. If the OFDM system has been already fixed such as DVB standard, and no changes can be made for the tx side, do you mean that there are still some other equalziation techniques to recover the deep fading channel?

Sorry, there is no way to overcome deep fade losses other than to use a better communication system. For a fixed system there is no implementation headroom to do something better. Only in the phase/frequency/amplitude synchronisation domain is some headroom.

The implementation techniques gives more headroom but only for power. I have heard that factor 5-15 is between optimized full custom and gate level regarding power in the same technology.

You are right. I have already tried linear equalizer and DFE. None of them could help.