When is it needed to incorporate frequency offset correction in lteDLPerfe​ctChannelE​stimate?

Hey, I have a few questions regarding the use of time and frequency offset in the function lteDLPerfectChannelEstimate.

My general settings are as follows:

channel.DelayProfile = 'EPA';
channel.DopplerFreq = 0;  
channel.InitTime = 0;
[rxwave, channel_info] = lteFadingChannel(channel,[txwave;zeros(25,1)]);
TimeOffset = lteDLFrameOffset(enb,rxwave);
FreqOffset = lteFrequencyOffset(enb,rxwave);

1) First thing I don't understand is why when I set zero Doppler and zero delay I'm still getting that the frequency offset "FreqOffset = lteFrequencyOffset(enb,rxwave)" and time offset "TimeOffset = lteDLFrameOffset(enb,rxwave)" are not 0.

Isn't frequency offset caused only by doppler effects? does the command lteFadingChannel adds additional delay and frequency offsets?

2) Basically, my main goal is getting the most precise (True) channel response. I don't understand when should I incorporate the timeoffset and frequency offset. I think that if Doppler frequency is set to 0 than the channel impulse response should stay constant. But when I use the command:

[hestIdeal] = lteDLPerfectChannelEstimate(enb,channel,[TimeOffset ,FreqOffset]);

I'm getting that the impulse response changes over the time grid.

However, when I ignore the frequency offset and use:

[hestIdeal2] = lteDLPerfectChannelEstimate(enb,channel,[Timeoffset,0]);

I'm getting that the channel response is indeed constant over time.

What am I missing? I couldn't understand whether should I use the frequency offset? why is it not 0 in the first place? Can anyone please clarify me this issue? Thanks.

Answer 1

[Iain Stirling]

Hello Omer,

The purpose of the frequency offset option on lteDLPerfectChannelEstimate is to match any frequency offset correction applied to the received waveform, so that the perfect channel estimate remains well matched with the received waveform. It plays the same role as the timing offset parameter. So if you have no frequency correction on the received waveform, you do not need to set the frequency offset option on lteDLPerfectChannelEstimate.

Note that even when no frequency offset is present, lteFrequencyOffset will give a non-zero value as noise affects its performance. So if you use that small erroneous offset as an input to lteDLPerfectChannelEstimate, it will actually create a frequency offset that is not present in the channel, hence the variation over time.

In summary, if you are not correcting for a frequency offset on your received waveform, you do not need to set the frequency offset option in lteDLPerfectChannelEstimate.

Regards,

Iain.

[Omer Sholev]

Hey, thanks again for your replay.

I still couldn't figure on why am I getting a frequency offset even though specifying:

channel.InitPhase = 0; 
channel.DopplerFreq = 0;  
% with no added noise at all on the Rx waveform

Nor does the time delay offset, it seems to me that I'm getting a 7 samples time delay offset no matter what I do. i.e., when I specify my own custom delay settings:

channel.DelayProfile = 'Custom';     
channel.DopplerFreq = 0;   
channel.InitTime = 0;
channel.InitPhase = 0;
channel.SamplingRate = info.SamplingRate;
channel.PathDelays =  [0, 500/info.SamplingRate];  % path delays are integer parts of the sampling period 
channel.AveragePathGaindB = [0,-5];

Then

[rxwave, channel_info] = lteFadingChannel(channel,[txwave;zeros(100,1)]);
offset = lteDLFrameOffset(enb,rxwave); % this equals 7 no matter what

offset =7 no matter what I do Do you have any Idea why?

[Iain Stirling]

Hello Omer,

The frequency offset estimation relies on correlating the received waveform with a delayed version of itself, and because the autocorrelation of the waveform is not ideal, some estimation error will result. I would expect that for no Doppler and no noise, the estimation error would be small - what is the estimated frequency offset that you observe? As I mentioned already, I would recommend that you don't perform frequency offset estimation and correction if you know that there is no offset to remove.

Regarding the timing offset, 7 is the implementation delay of the filtering used inside lteFadingChannel, so modelling a fading channel with no delay will still result in a 7 sample delay in the waveform. This minimum delays is returned in the ChannelFilterDelay field of the INFO output of lteFadingChannel.

Regards,

Iain.