Md Tarek Hassan | PhD Researcher in 6G Wireless, RIS and ISAC

Why Divide a Wideband Signal?

Imagine one fast signal crossing a city. The direct wave and delayed reflections overlap; different parts of the band can fade differently.

Subcarriers are narrow frequency lanes inside the allocated band. Each carries a slower stream of symbols, so the channel often looks approximately flat across one lane.
OFDM transmits many such lanes at once. Their frequencies are chosen to be orthogonal over the useful symbol interval, allowing dense packing without ideal cross-talk when timing and frequency synchronization hold.
FFT/IFFT are efficient ways to separate and combine the lanes. An IFFT creates the time-domain transmit symbol; a receiver FFT recovers its subcarrier values.
Cyclic prefix (CP) copies the end of each OFDM symbol to its beginning. When the important delayed paths fit within the CP and synchronization is adequate, per-subcarrier equalization becomes much simpler.

For orthogonal subcarriers:

\Delta f=\frac{1}{T_u}

$\Delta f$ : spacing between adjacent subcarrier centres, in hertz.
$T_u$ : useful OFDM symbol duration, excluding the cyclic prefix, in seconds.
Trade-off: OFDM simplifies per-subcarrier equalization but adds CP and pilot overhead; high waveform peaks and frequency offset also need attention.

References for this section3

For a simplified 20 MHz sampled band and 1024-point FFT:

Bin spacing: $20\ \mathrm{MHz}/1024\approx19.53$ kHz.
Useful symbol time: about $51.2\ \mu\mathrm{s}$ .
5 $\mu$ s CP overhead: $5/(51.2+5)\approx8.9\%$ of symbol time.
Trade-off: longer CP tolerates more delay but leaves less time for new data. These are teaching values, not an NR numerology.

References for this section3

A pilot is a transmitted symbol whose complex value the receiver knows in advance. It occupies an ordinary time-frequency cell.

What it reveals: comparing a received pilot with its known transmitted value estimates the channel's amplitude and phase at that grid location.
Why several are needed: the channel can differ across subcarriers and can change from one OFDM symbol to the next.
What happens between pilots: interpolation estimates the channel at data cells. Widely spaced pilots can miss a rapidly varying channel.
The cost: pilots occupy time-frequency resources that could otherwise carry data; more pilots are not free.

For one subcarrier in a simplified flat channel,

Y[k]=H[k]X[k]+W[k]

At a known, nonzero pilot, a simple estimate is:

\widehat H[k]=\frac{Y_p[k]}{X_p[k]}

$\widehat H[k]$ : estimated channel on subcarrier $k$ .
$X_p[k]$ : known transmitted pilot; $Y_p[k]$ : received pilot.
Assumption: timing and frequency synchronization are good enough that subcarriers do not strongly leak into each other.

References for this section3

I/Q samples→Sync→FFT→Channel estimate→Equalize→Decode

Synchronize: locate symbol timing and correct carrier-frequency/phase offset. Residual frequency offset $\Delta f$ rotates phase by $2\pi\Delta f T$ over time $T$ and can leak energy between subcarriers.
Transform: remove the CP and use the FFT to inspect each subcarrier.
Estimate: compare pilots against known values, then interpolate across data cells.
Equalize and detect: compensate for the estimated channel and choose likely transmitted symbols.
Decode: use coding redundancy to recover bits; evaluate errors and latency, not just a clean-looking constellation.

References for this section3

Use a noise-free toy channel to follow one symbol:

Transmit a known pilot $X_p=1$ ; receive $Y_p=0.5$ . The estimate is $\widehat H=0.5$ .
Transmit data $X=1+j$ ; receive $Y=0.5+0.5j$ . Divide by $\widehat H$ to recover $1+j$ .
Now let the channel fall to $H=0.05$ . Division multiplies additive noise by 20 in amplitude. An exact inverse is fragile near a deep fade.

MMSE equalizer: uses estimated noise power to limit amplification near a fade, while allowing some residual distortion.
Evaluation: compare post-equalization EVM and decoded BLER, not only a noiseless inverse.

References for this section3

Zero forcing: divides by the channel estimate; simple, but can amplify noise badly near fades.
MMSE: balances channel correction and noise enhancement; depends on a useful noise estimate.
Detection: assigns received points to likely modulation symbols.
Channel decoding: uses redundancy to recover bits. More redundancy improves robustness but reduces the fraction of transmitted bits carrying new information.

Processing stage	Failure you may see	Useful diagnostic
Synchronization	Rotating or smeared constellation	Timing and frequency-offset estimates
Channel estimation	Wrong amplitude/phase correction	Pilot residuals or channel-estimation MSE
Equalization	Noise amplification in faded subcarriers	Post-equalization SINR and EVM
Detection/decoding	Wrong bits or failed blocks	BER and BLER at the chosen modulation/coding

Beamforming: combines antenna-port signals with spatial weights; see phase control.
Evaluation: choose a measurement point and outcome using the performance guide.

References for this section3

A narrow beam can supply strong received power, yet a moving user's channel estimate ages:

Fast motion: pilots, phase tracking, and beam updates may need to occur more often.
Resource trade-off: additional pilots improve tracking but leave fewer cells for data.
Receiver trade-off: more complex equalization may improve BLER but add computation and latency.
Success criterion: judge delivered data and reliability under motion, not pilot MSE alone.
Beyond decoding: pilot observations can carry delay, angle, and Doppler information.
Keep: raw or calibrated complex samples for sensing/localization; bit-only receiver outputs can discard that information.

References for this section3