I'm doing some research on channel eq. for SC-FDMA. Specifically, implementing a combination of CMA, DD and other "classic" techniques. However, I see that most techniques are 'complemented' by linear interpolation using pilot subcarriers or a reference signal.
This process is called frequency domain equalization, as it uses these known values of reference signals to estimate the channel response for that specific symbol.
My question is: would it be possible to do something similar by simply upsampling the transmitted sequence and then using the extra data to estimate the channel response?
I think you posted this around the time dsprelated.com suffered a host outage. It may have gone unnoticed by some.
Remember that the channel here is being estimated in frequency domain with only a few pilot subcarriers, leaving room for getting the intermediate results from interpolation. Try to imagine what happens when you upsample the Tx sequence in time domain. There is no extra data and instead just extra spectral replicas with varying phases in frequency domain are `taken in'. I think you need to understand the channel effect of time domain convolution versus frequency domain multiplication and hence the benefit of the latter over the former.
Thank you for your answer, qasimilyas. Definitely got me a direction to go for :) I'm new to this subject, so some concepts have a bit of a "learning curve"!