You are not logged in.
Hi,
I have some questions regarding to the frequency offset correction.
1. There is a "Coarse Freq Correction" block in ofdm_txrx_mimo. I am just wondering what range of frequency offset can be corrected by this block?
2. In the simulation, if I want to add the frequency offset in the channel I have to put a normalized number(from 0 to 1) to indicate the offset. How do I convert the frequency offset in Hz into the normalized value?
3. In the "Coarse Freq Offset Calc" block in ofdm_txrx_mimo, the phase increment is calculated by taking the product of arctan(y/x) and a scaling factor. The value of the scaling factor is 1/(2*pi*64). Would you please explain how to get this number?
Thank you very much for your help!
Regards,
Lotus
Last edited by Lotus Yi (2007-Dec-12 11:46:13)
Offline
The algorithm is a standard one, presented in the literature. I'd suggest looking through the papers I cited previously.
The basic idea is to compare the phases of two samples which were identical at the transmitter, then compute the frequency offset based on how much the phase changed between the two samples. We use the packet preamble's long symbols for this. The long symbols are each 64 samples long, accounting for the 1/(2*pi*64) normalization factor.
Offline
Hi,
What is the maximum frequency offset ofdm_txrx_mimo can compensate for??
Thank you very much!
Regards,
Lotus
Offline
I believe it's half a subcarrier spacing, or 78kHz in our case (10MHz/64/2).
Offline
Hi,
Does 78kHz include the compensation done by the radio? or it is done by the frequency compensation blocks in ofdm_txrx_mimo??
Thank you very much!
Regards,
Lotus
Offline
There is no CFO compensation in the radio; it's all done digitally.
Offline
Thank you!
Offline
Hi,
I tried to verify the maximum frequency offset the coarse frequency offset correction can compensate for by changing the channel parameter, Freqency Offset (phase inc [0, 1]).
To add the 10kHz frequency offset, I set the parameter to 0.0008.
I calculated it through the following equation: (10KHz * 4) / sampling frequency = (10000 * 4) / (50 * 10^6) = 0.0008. Is this the right way to normalize the frequency offset in Hz to phase increment? If not, how to normalize the frequency offset in Hz to phase increment?
Thank you very much for your help!
Regards,
Lotus
Offline
The normalization factor used in the CFO correction block of reference design is 1/(2*pi*64*72). What does 72 account for?
EDIT: I think I figured it out. The accumulator adds phase changes over 6 short symbols (72 consecutive samples) and hence the normalization, right?
Last edited by Amir (2008-Nov-27 14:33:29)
Offline
Right (sorry for the delay; we were all traveling for the holiday). The coarse CFO is averaged over 72 samples during the 2.5 copies of the 64-sample long training symbol in the preamble.
Offline