Dear All,
I have an OFDM TX block connected to an OFDM RX block through a 'Frequency Selective Fading Model' block, implemented in GRC. The parameters used in the model are:
Num sinusoids (for sum of sinusoids): 8
Max Doppler: 0
LOS Model: NLOS
PDP Delays: (0, 1.6, 2.8)
PDP Magnitudes: (0.4, 0.4, 0.2)
Num Taps: 100
When a packet is received by the channel payload equalizer block, I simply print out the "Initial Taps", which are the initial channel estimated, and plot the channel as a function of subcarrier index. I observe that at subcarrier 0 (the midpoint, since subcarrier indices are centered around 0), there is a large phase shift. In my experiment, the fft_len is 64, used as follows:
Subcarrier (-32 through -27): Unused
Subcarrier (-26 through -1) : Carry data/pilots
Subcarrier 0 : Unused
Subcarrier (+1 through +26) : Carry data/pilots
Subcarrier (+27 through +31): Unused
What is the cause of this large discontinuity between subcarriers (-1 and +1)? Obviously, ignore subcarrier 0 since it is unused. I've pasted the 'Initial Taps' below. The columns are: subcarrier index, real part, imag part, magnitude of channel, respectively.
| -32 |
0 |
0 |
0 |
| -31 |
0 |
0 |
0 |
| -30 |
0 |
0 |
0 |
| -29 |
0 |
0 |
0 |
| -28 |
0 |
0 |
0 |
| -27 |
0 |
0 |
0 |
| -26 |
-0.241 |
-0.245 |
0.343 |
| -25 |
-0.164 |
-0.203 |
0.261 |
| -24 |
-0.08 |
-0.183 |
0.199 |
| -23 |
0.006 |
-0.185 |
0.185 |
| -22 |
0.087 |
-0.21 |
0.228 |
| -21 |
0.157 |
-0.255 |
0.3 |
| -20 |
0.212 |
-0.318 |
0.382 |
| -19 |
0.247 |
-0.391 |
0.463 |
| -18 |
0.262 |
-0.47 |
0.538 |
| -17 |
0.254 |
-0.549 |
0.605 |
| -16 |
0.225 |
-0.622 |
0.662 |
| -15 |
0.178 |
-0.683 |
0.706 |
| -14 |
0.117 |
-0.728 |
0.737 |
| -13 |
0.046 |
-0.754 |
0.755 |
| -12 |
-0.028 |
-0.758 |
0.758 |
| -11 |
-0.099 |
-0.741 |
0.748 |
| -10 |
-0.163 |
-0.706 |
0.724 |
| -9 |
-0.211 |
-0.653 |
0.686 |
| -8 |
-0.244 |
-0.589 |
0.638 |
| -7 |
-0.255 |
-0.519 |
0.578 |
| -6 |
-0.246 |
-0.448 |
0.511 |
| -5 |
-0.213 |
-0.381 |
0.437 |
| -4 |
-0.165 |
-0.327 |
0.366 |
| -3 |
-0.098 |
-0.289 |
0.306 |
| -2 |
-0.016 |
-0.265 |
0.266 |
| -1 |
0.079 |
-0.264 |
0.275 |
| 0 |
0 |
0 |
0 |
| 1 |
0.232 |
-0.506 |
0.557 |
| 2 |
0.191 |
-0.587 |
0.617 |
| 3 |
0.143 |
-0.654 |
0.67 |
| 4 |
0.079 |
-0.7 |
0.704 |
| 5 |
0.01 |
-0.727 |
0.727 |
| 6 |
-0.065 |
-0.729 |
0.732 |
| 7 |
-0.135 |
-0.712 |
0.724 |
| 8 |
-0.199 |
-0.675 |
0.703 |
| 9 |
-0.25 |
-0.62 |
0.669 |
| 10 |
-0.282 |
-0.552 |
0.62 |
| 11 |
-0.295 |
-0.478 |
0.561 |
| 12 |
-0.285 |
-0.401 |
0.492 |
| 13 |
-0.254 |
-0.329 |
0.416 |
| 14 |
-0.204 |
-0.268 |
0.337 |
| 15 |
-0.138 |
-0.221 |
0.26 |
| 16 |
-0.06 |
-0.195 |
0.204 |
| 17 |
0.022 |
-0.189 |
0.191 |
| 18 |
0.105 |
-0.207 |
0.232 |
| 19 |
0.181 |
-0.246 |
0.306 |
| 20 |
0.245 |
-0.306 |
0.392 |
| 21 |
0.29 |
-0.38 |
0.478 |
| 22 |
0.316 |
-0.465 |
0.562 |
| 23 |
0.318 |
-0.554 |
0.639 |
| 24 |
0.296 |
-0.641 |
0.706 |
| 25 |
0.253 |
-0.72 |
0.763 |
| 26 |
0.189 |
-0.786 |
0.808 |
| 27 |
0 |
0 |
0 |
| 28 |
0 |
0 |
0 |
| 29 |
0 |
0 |
0 |
| 30 |
0 |
0 |
0 |
| 31 |
0 |
0 |
0 |
Why is there such a large phase shift taking place around 0? Any pointers would be greatly appreciated.
This effect does not show up with a standard frequency xlating fir filter used instead of this frequency selective fading model block. Nor does the problem show up when the experiment is run over USRP hardware instead of the channel model block.
Thanks in advance!
Best regards,
Aditya