if you estimate blindly the channel coefficients, you cannot also estimate the offset of carrier frequency...

Dear Jiajun,

Eliminating the frequency offset means that , other way expressed, you must synchronize the received carrier at the receiver site with the that locally generated at the receiver for synchronous demodulation of the received modulated signal.This in normally accomplished by A carrier regeneration circuit from the received signal and then using such copy to synchronize the locally generated carrier using an appropriate phase locked loop.The appropriate phase locked loop must have a frequency detector for aided acquisition. The frequency detector works at large frequency deviation and the phase detector works in at small frequency deviations.Normally one uses a Costas PLLore for such systems as the carrier is usually BPSKed. This is the conventional way for synchronization and eliminating the frequency off sets whether small or large.

wish you uccess.

If I had to implement an actual system, here's how I might start off doing it.

If you are close enough to the expected carrier frequency, it is possible to aquire an estimate of the PN sequence phase using non-coherent acquisition.  Once this is done, it is possible to estimate the carrier frequency using a fourier transform or a chirp frequency sweep/correlation step.  Shift the receiver local oscillator frequency to that value then activate your carrier tracking loop to make final phase/freq offset corrections and perform carrier tracking for coherent demodulation.

Every cancellation before channel estimation should be kind of self-cancellation. as we've saw carrier offset cancellation in OFDM systems, there are such algorithms to estimate offset without any need to channel information and only by using your own sequences...

In DS-CDMA uplink in reception there are many users at same time. i suppose that the frequency offset impairment compensation must be done in the baseband sampled signal because each user have a diferent offset and the hardware can only compensate for one user. There must some kind individual compensation (each for each user) but I think it must be more complicated than single user one because of the interference.

I think that the estimation of the frequency offset must be done with the training information (TDD-Midamble). In that way there is not the interference penalty.

Hi everyone,

Thanks for your comments on my question.

As for DS-CDMA with training sequence, I think that cancellation of CFO before channel estimation should follow the type of communication system or communication standard. Here, I hold an example (IEEE 802.11).

The tranining symbol repeatly transmits in preamble. Hence, the ISI of every training symbol is the same. In this case, previous symbol and current symbol have the same values. Thus, we dont need to consider ISI when we estimate CFO. the method of CFO estimation can be cross-correlation between current symbol and preivous symbol, because the spreading sequence of IEEE 802.11 is Barker sequence (11 chips) which has a good auto-correlation. Therefore, the cross-correlation will maximize SNR and accurately obtain the CFO value.

CFO = ( 1/(2*pi*Ts) ) * angle( cs*ps' );

where cs and ps are current symbol and previous symbol respectively. cs and ps are the matrix of 1*11. Ts is the duration of symbol.

Kind regards

Jiajun

What about using pll as i proposed in my comment. In this case you donot need a training sequence and symbol timing. Normally the signal chips modulate the carrier by BPSK ORPQK.

wish you uccess

Hi Abdelhalim Zekry,

Of course, I think that PLL is a another good way to eliminate CFO and to improve bandwidth efficiency without training sequence in packet.

Personally, high accurate PLL techniques may bring high complexity of circuits. In addition, phase shift keying (PSK) is very sensetive to phase rotation. Therefore, sometimes we need use some digital algorithms to compensate the inaccurate PLL techniques for reducing the complexity. For example, Differential PSK is a good way to reduce the complexity. 

Overall, I think that the researches for both PLL technique and baseband algorithm are important for wireless communications. But, now I am focusing on algorithm study.

Thanks again for your idea.

Kind regards 

First I must define the term "carrier frequency offset." In synchronous communication system theory, the receiver must produce a replica of the expected carrier frequency. Due to equipment imperfections and doppler shift, It is impossible to have perfect carrier frequency knowledge. To overcome this in practical systems, we do carrier frequency and phase estimation by means of frequency error estimation and a feedback tracking loop. Decades ago, this was done with a discriminator circuit in FM radio broadcast receivers. In simple comm systems, we can get the offset estimation using a simple phase-locked loop (PPL). In more exotic receiver designs, we use a Costas Demodulator feeback loop. In software  radion systems, we digitize the signal and use software implimentaitons of PLLs and Costas. We can also calculate an FFT spectrum and control a feedback loop.

In practical receiver designs, the feeback control signal controls a clock or frequency oscillator that synchronizes receiver frequency and phase estimates to the received signal.

Good luck.

Sorry I forgot to answer the question about channel estimation. The short answer is that it is the most practical to remove the dopper shift before processing the received signal.

As you are aware, a well-designed testing signal (training signal) must be sent over the channel and its signal properly recieved and analyzed at the receiver. The receiver must have an exact replica of the initial transmitted signal. It is equivalent to measuring the impulse respone of the channel although we often use the Fourier Transform equivalent and do the analysis in the frequency domain, i. e., use the transfer function. Doppler shift of the transmitted signal affects all parts of the signal spectrum but if we send a signal that has a carrier component, we can correct for doppler shift by estimating the received carrier frequency as compared to the local receiver clock.

Hi Steve F. Russell,

Thanks your comments.

But I am confused by one part of your comments.

As I know, CFO is caused by dismatched oscillators between Tx and Rx. In addition, Doppler shift exists in a moving MS. Normally, it determines the speed of time varing Rayleigh fading and Rician fading.

Why the issue of carrier frequency offset should consider Doppler shift ?

Many thanks

Jiajun