Ferkans — Interactive Telecom Tutor

The Wi-Fi Generation Stack: Same BICM, Higher QAM

Wi-Fi 6 (IEEE 802.11ax, 2021) and Wi-Fi 7 (IEEE 802.11be, 2024) are both BICM-LDPC systems in the mould of 5G NR but with three distinguishing engineering decisions:

Higher modulation orders. Wi-Fi 6 adds 1024-QAM ( $Q_m = 10$ ); Wi-Fi 7 adds 4096-QAM ( $Q_m = 12$ ). No cellular standard to date uses 4096-QAM on the data channel.
Same LDPC code family as 802.11n/ac. Wi-Fi's LDPC code, defined in IEEE 802.11-2007, is not the 5G NR code. It uses 12 length- block quasi-cyclic codes at 648/1296/1944 bit codewords and four rates $\{1/2, 2/3, 3/4, 5/6\}$ . The Wi-Fi working group preserved backward compatibility rather than switching to NR-style base graphs.
OFDMA resource units. Wi-Fi 6 introduced the Resource Unit (RU) — a fine-grained spectral allocation replacing the monolithic 20/40/80/160 MHz channel. Each RU carries an independent BICM-LDPC codeword with its own MCS. Wi-Fi 7 adds the 320 MHz channel and Multi-Link Operation (MLO).

The bottom line: the three BICM ingredients are the same as in Chapter 5; only the constellation size and the code-block-length grid differ. In this section we compute the peak rates and the SNR thresholds for the high-QAM MCSs, then explain why Wi-Fi 7's 4096-QAM 5/6 is essentially a benchmarking curiosity rather than a regularly-used mode.

,

Definition:
Wi-Fi MCS Index and Resource Unit

A Wi-Fi MCS is a pair $(Q_m, R)$ indexed by an integer:

802.11ax (Wi-Fi 6): MCS 0-11, corresponding to $Q_m \in \{2, 2, 4, 4, 6, 6, 6, 8, 8, 10, 10\}$ (i.e., BPSK, QPSK at two rates, 16-/64-/256-/1024-QAM) with $R \in \{1/2, 2/3, 3/4, 5/6\}$ .
802.11be (Wi-Fi 7): MCS 0-13, extending 11ax with two more points at $Q_m = 12$ (4096-QAM) and rates $\{3/4, 5/6\}$ .

Unlike 5G NR, the Wi-Fi modulation-rate pairs are coarsely quantised: only 11-13 distinct MCSs per stream vs NR's 28. This reflects Wi-Fi's shorter feedback loop (MPDU-level ACKs rather than sub-slot HARQ) and the unlicensed-spectrum nature of its link adaptation (contention-based rather than scheduled).

A Resource Unit (RU) in 11ax is a contiguous block of 26, 52, 106, 242, 484, or 996 OFDMA subcarriers (with corresponding 2-MHz to 80-MHz bandwidths). Each RU is assigned to one station per TXOP and carries a single MCS. RUs enable multi-user OFDMA, the main throughput innovation in Wi-Fi 6.

Wi-Fi's MCS system is both simpler and more exposed to users than NR's: users routinely see MCS numbers via iw or a router GUI, whereas NR MCS indices are buried in DCI payloads. The operational advice for Wi-Fi 6/7 engineering is simple: if MCS $> 9$ happens rarely, your PHY SNR is below about 30 dB.

,

Wi-Fi MCS Spectral Efficiency vs SNR Threshold

Wi-Fi 6 and Wi-Fi 7 MCS points plotted against the minimum SNR needed to achieve PER $\le 10\%$ on AWGN. The operational "useful" range is typically 5-30 dB for residential networks. Note the Shannon bound (dashed grey) — the $\sim 1$ dB gap between 4096-QAM 5/6 and Shannon illustrates how close BICM with strong LDPC gets to the fundamental limit. Also note that MCSs above 9 (Wi-Fi 6) or 11 (Wi-Fi 7) require SNRs above 30 dB, achievable only in close-range, line-of-sight, low-interference conditions.

Parameters

Standard

Example: Peak PHY Rate: 4096-QAM 5/6 at 320 MHz, 2 Streams

Compute the peak PHY rate of a Wi-Fi 7 device using MCS 13 (4096-QAM, rate 5/6), on a 320 MHz channel, with 2 spatial streams (2SS). Use the 11be numerology: 4096 subcarriers per 320 MHz channel with 78.125 kHz subcarrier spacing and a 14.4 $\mu$ s OFDM symbol (0.8 $\mu$ s guard interval).

Solution

Useable subcarriers

A 320 MHz channel in 11be has 4096 total subcarriers, of which 3920 are data subcarriers (the rest are pilots, DC null, and guard). This is with the 78.125 kHz subcarrier spacing inherited from 11ax. Other overheads (LTFs, STFs, preamble) do not count toward the payload rate.

Bits per symbol per stream

$Q_m R = 12 \cdot 5/6 = 10$ information bits per subcarrier per stream. Over 3920 data subcarriers and 2 streams, one OFDM payload symbol carries $3920 \cdot 10 \cdot 2 = 78{,}400$ info bits.

Symbol duration and rate

Symbol duration: $T_{\rm OFDM} = 12.8 \;\mu s + 0.8 \;\mu s = 13.6 \;\mu s$ (with short GI) or $14.4\;\mu s$ (with long GI). Use the 11be "short GI" value $13.6\;\mu s$ for peak: $R_{\rm PHY} = \frac{78400 \text{ bits}}{13.6 \times 10^{-6} \text{ s}} \approx 5.76 \text{ Gbps}.$ The IEEE table lists 5764.7 Mbps for MCS 13, 2SS, 320 MHz, 0.8 $\mu$ s GI — matching our calculation.

Operational commentary

This rate is per spatial stream pair. Real Wi-Fi 7 devices with 4 spatial streams and MU-MIMO can hit $\sim 23$ Gbps aggregate, but only with the transmitter and receiver within a few metres, line-of-sight, and interference-free. In a typical residential environment with a single wall between devices the achievable MCS is around 9 (1024-QAM 5/6), giving $\sim 1.2$ Gbps per stream. 4096-QAM is realistically a benchmarking mode. $\blacksquare$

⚠️Engineering Note

Wi-Fi 7 4096-QAM: Why You Almost Never See It

The 802.11be specification requires a receiver to be able to decode 4096-QAM 5/6 at an input SNR of $\ge 32.5$ dB on AWGN with PER $\le 10\%$ . (Some vendor datasheets quote 34-35 dB to leave margin.) This is a high bar:

Typical residential Wi-Fi SNR at 5 m, one wall, is 25-30 dB.
At 10 m or through two walls, SNR drops to 15-20 dB, well below the 4096-QAM operating point.
Implementation losses (phase noise, IQ imbalance, EVM floor of the TX PA) eat another 2-4 dB.

Practically, 4096-QAM works in three scenarios: (i) very short-range "two-room" wireless-backhaul links, (ii) benchmarking testbeds in RF-isolated chambers, and (iii) 6 GHz-band operation where clean spectrum pushes SNRs higher. For the vast majority of consumer Wi-Fi 7 traffic, the active MCS is in the range 7-11 (256-QAM or 1024-QAM).

The gap from 4096-QAM 5/6 threshold (32.5 dB) to the Shannon limit at 10 b/s/Hz (29 dB) is less than 4 dB — an impressive BICM design, but reaching the required SNR is the real engineering challenge, not the coding gain.

Practical Constraints

•
4096-QAM 5/6 minimum SNR: 32.5 dB per 11be spec
•
Typical residential SNR ceiling: 25-30 dB
•
TX EVM limit (3.5% at 4096-QAM per spec): corresponds to -29 dB EVM
•
PHY rate 5.76 Gbps/stream achievable only at short range with clean spectrum

📋 Ref: IEEE Std 802.11be-2024

🔧Engineering Note

Wi-Fi LDPC: Legacy Codes, Simple Padding

Wi-Fi 6/7 LDPC is the same code family introduced in 802.11n (2009):

Code-block lengths: 648, 1296, 1944 bits.
Rates: 1/2, 2/3, 3/4, 5/6.
Structure: quasi-cyclic with $Z \in \{27, 54, 81\}$ lifting, depending on code-block length.

No rate matching, no base-graph switching, no puncturing tables. If the number of payload bits does not match one of the 12 $(N, R)$ codewords exactly, the transmitter uses shortening (pad information bits with zeros) and puncturing (drop parity bits) in simple ratios fixed by the spec.

Compared to 5G NR this is rigid: Wi-Fi cannot use an arbitrary rate like $R = 0.73$ . But the reduced flexibility is compensated by very simple encoder/decoder hardware and by the fact that Wi-Fi re-selects the MCS on every frame (no HARQ), so non-optimal codeword selection costs only one frame's worth of efficiency.

Theorem: Ergodic Throughput Bound for Wi-Fi BICM-LDPC

For a Wi-Fi 6/7 BICM-LDPC link operating on a block-fading channel with fade distribution $p_\Gamma(\gamma)$ (where $\Gamma$ is the SNR on each OFDM symbol), let $\eta^\star(\gamma)$ denote the maximum achievable spectral efficiency over the discrete Wi-Fi MCS set at SNR $\gamma$ (with BLER $\le 10\%$ ). Then the ergodic throughput is $\bar\eta_{\rm WiFi} = \mathbb{E}_\Gamma[\eta^\star(\Gamma)] \le \mathbb{E}_\Gamma[\log_2(1 + \Gamma)].$ The gap between the two is bounded by $\mathbb{E}_\Gamma[\log_2(1 + \Gamma)] - \bar\eta_{\rm WiFi} \le \sup_\gamma \bigl[\log_2(1 + \gamma) - \eta^\star(\gamma)\bigr],$ which for Wi-Fi 7 with Gray-labelled QAM is approximately 1.5 bit/s/Hz at high SNR (the Gray-BICM gap plus the MCS-quantisation staircase loss).

This is the pointwise-bound version of the AMC throughput theorem of s04. It says Wi-Fi throughput is upper-bounded by ergodic capacity, and the gap decomposes into (i) the 0.25 bit/s/Hz Gray-BICM gap from Chapter 5, and (ii) a staircase loss from having only 14 discrete MCSs rather than a continuum.

Show Hint

The ergodic throughput is the expected max-over-MCS rate at the instantaneous SNR.

Apply Jensen-like bound: maximum over discrete $\eta^\star(\gamma)$ is at most the Shannon curve.

Subtract and bound by the pointwise gap.

Proof

Step 1: Pointwise throughput

At instantaneous SNR $\gamma$ , the station selects the highest MCS $i^\star(\gamma)$ that gives BLER $\le 10\%$ . Its rate is $\eta^\star(\gamma) = Q_{m,i^\star} R_{i^\star}$ . Since BICM capacity $\le$ Shannon, $\eta^\star(\gamma) \le \log_2(1 + \gamma)$ .

Step 2: Ergodic expectation

Expected throughput: $\bar\eta = \mathbb{E}_\Gamma[\eta^\star(\Gamma)] \le \mathbb{E}_\Gamma[\log_2(1 + \Gamma)]$ by monotonicity of the expectation.

Step 3: Pointwise gap upper bound

$\mathbb{E}[\log_2(1 + \Gamma)] - \bar\eta = \mathbb{E}[\log_2(1 + \Gamma) - \eta^\star(\Gamma)] \le \sup_\gamma [\log_2(1 + \gamma) - \eta^\star(\gamma)]$ , and this sup is finite (about 1.5 b/s/Hz for Wi-Fi 7 Gray-QAM). $\blacksquare$

Common Mistake: Wi-Fi "SNR" Is Not Shannon SNR

Mistake:

Comparing the SNR threshold table in 11be (e.g., "MCS 13 needs 32.5 dB SNR") with the Shannon bound from Chapter 5 is tempting but often misleading.

Correction:

The 11be "SNR" is an equivalent-AWGN input SNR, measured at the receiver ADC after the LNA, and assuming an ideal receiver EVM. It includes the receiver noise figure (typically 5-8 dB) but does not include transmitter EVM (typically -29 dB at 4096-QAM, corresponding to 3 dB of implementation loss). To convert to Shannon SNR for capacity comparisons, add the TX EVM and any channel- estimation penalty (1-2 dB at high QAM). The actual "Shannon margin" of 4096-QAM 5/6 is about 2-3 dB, not 4.

Wi-Fi 6 (11ax) vs Wi-Fi 7 (11be): BICM Ingredients

Parameter	Wi-Fi 6 (11ax)	Wi-Fi 7 (11be)
Max modulation	1024-QAM ( $Q_m = 10$ )	4096-QAM ( $Q_m = 12$ )
Max code rate	5/6	5/6
LDPC code family	IEEE 802.11n (2009)	Same as 11ax (unchanged)
Max bandwidth	160 MHz	320 MHz
Max spatial streams (HE-SU)	8	8 (+ MLO across bands)
Peak PHY rate (1 SS, max BW)	1.2 Gbps	2.88 Gbps
Peak PHY rate (all-stream)	9.6 Gbps	~46 Gbps (with MLO)
MCS set size	12 (MCS 0-11)	14 (MCS 0-13)
Min SNR for max MCS	$\sim 30$ dB	$\sim 33$ dB
HARQ	None	Optional (not widely deployed)

Historical Note: The QAM Arms Race: From 64-QAM to 4096-QAM in 20 Years

2004-2024

Wi-Fi's modulation-order ladder tracks the evolution of consumer wireless:

802.11a/g (1999/2003): 64-QAM, peak 54 Mbps.
802.11n (2009): 64-QAM, up to 4 streams, peak 600 Mbps. LDPC introduced as an optional code.
802.11ac (2013): 256-QAM, up to 8 streams, peak 6.9 Gbps. LDPC becomes the recommended code.
802.11ax/Wi-Fi 6 (2021): 1024-QAM, OFDMA, peak 9.6 Gbps.
802.11be/Wi-Fi 7 (2024): 4096-QAM, 320 MHz, Multi-Link Operation, peak 46 Gbps.

Each generation doubles the QAM alphabet size (4-6 new information bits per symbol) and demands 3-6 dB more SNR. The driving force is not user demand (which has grown an order of magnitude slower) but benchmarking competition among silicon vendors and the desire to advertise higher peak rates than LTE/5G cellular. Whether 4096-QAM is actually used in real traffic is less clear: most internet-bound packets suffice with MCS 7-9 (256-QAM or 1024-QAM) even on excellent links, because the bottleneck moves to the upper-layer TCP/IP stack.

,

Quick Check

A Wi-Fi 7 device reports MCS 13 on a 320 MHz channel with 2 spatial streams. Which of the following is the most likely deployment scenario?

A residential living-room connection 5 m from the access point, through a thin wall

An office hallway 15 m from the access point, behind a metal door

A 6 GHz dedicated backhaul link between two APs in the same room, line-of-sight

Public cafe Wi-Fi at a busy hour with 20 associated clients

Correction:

A 6 GHz dedicated backhaul link between two APs in the same room, line-of-sight

Clean 6 GHz spectrum plus short-range line-of-sight easily gives 35-40 dB SNR with low interference. This is precisely the niche for 4096-QAM.

Wi-Fi MCS (Modulation and Coding Scheme)

A small integer (MCS 0-11 in Wi-Fi 6, MCS 0-13 in Wi-Fi 7) selecting a modulation order $Q_m$ and code rate $R$ from a table in the IEEE standard. Coarser than the 5G NR MCS set (which has 28 primary entries) because Wi-Fi adapts at the frame level rather than sub-slot.

Related: Wifi Ldpc, Ofdma Ru, MCS Index

Resource Unit (RU)

A contiguous block of 26, 52, 106, 242, 484, or 996 OFDMA subcarriers in 802.11ax/be, allocated to a single station for one transmission opportunity. Enables multi-user OFDMA in Wi-Fi 6/7.

Related: Ofdma, Wifi Mcs

Key Takeaway

Wi-Fi 6/7 is BICM-LDPC with high-order QAM. The LDPC code family is inherited from 11n (2009); the big changes per generation are constellation size (up to 4096-QAM in 11be), bandwidth (320 MHz), and multi-user features (OFDMA, MLO). The 4096-QAM operating point sits at $\sim 33$ dB SNR — impressive capacity per Hz but realistically a benchmarking mode, not a residential workhorse.

Wi-Fi 6/7: LDPC + 1024/4096-QAM