Wi-Fi as an RF System

01 / 10

A headline rate fails the gateway requirement

The connection display says 77.4 Mbit/s. Can the application rely on 20 Mbit/s while the gateway receives weak node reports?

The first number describes a selected PHY data field. The application needs complete, unique payload at its endpoint. Meanwhile, the node needs a receiver that is both listening and able to distinguish its weak signal. Those requirements share a product, but they do not share a single success metric.

Continue the condition-monitoring case from 07.2. Keep its 32-byte report every 60 s and two-second node-to-gateway deadline. Add P07-M03-HE-BACKHAUL-v1: 20 Mbit/s of gateway-to-peer D0 payload. D0 means application information; D1 means framed/coded bits. A local backhaul test hypothesis uses 1500-byte payload units released every 0.6 ms, with ≥99% complete by 100 ms. These are illustrative requirements, not observed performance.

One gateway, four different quantities
Quantity	Default or requirement	What it does not establish
Nominal data-field rate	77.426470588 Mbit/s	Complete packet airtime or delivered D0 throughput.
D0 service proxy	45.294485294 Mbit/s; offered load 20 Mbit/s	A measured throughput or latency distribution.
Required wall-time service share	0.441554858 under the proxy	Exact RF transmit duty; ACK/listening/backoff have different states.
Weak-node availability	Simultaneous default fails its synthetic RF screen	Being scheduled to listen is not enough when the receiver is desensed.

Think about itWhich number belongs in the backhaul acceptance test?

Answer

Unique D0 bits delivered over a declared interval, together with per-payload latency and omissions. Here use 60 s steady-state windows, recording startup separately. Also measure node delivery while the Wi-Fi load is active. A connection-rate display is context for the test, not its outcome.

Prepare the engineering vocabulary

07.1’s evidence method, 02.5 symbols, 02.6 timing and ISI, 02.7 EVM, 05.4 blockers, 05.5 transmit budgets and 06.4 arrays and diversity own the foundations used here.

The earlier 2.450 GHz QPSK waveform remains a generic teaching baseline. It does not become Wi-Fi, Bluetooth or the narrower synthetic victim merely because they share this story. Start by naming the actual Wi-Fi mode.

02 / 10

Identify the standard, certification, and product

Does a generation label tell you which features this gateway and its peer can use together?

Imagine purchasing a module by its generation and discovering that the host driver exposes fewer simultaneous streams than the radio silicon. There is no contradiction in the physics: you selected a family before establishing the implementation. Separate the IEEE definition, alliance program, product configuration and deployed peer.

Read the identity chain from document to deployed mode
Layer	Evidence to name	Checked example / open question
IEEE base and amendment	Exact edition, amendment and incorporated changes	802.11ax-2021 HE was incorporated into the 802.11-2024 base; the old amendment is listed as superseded.
Generation / certification	Wi-Fi label, program release, role and tested feature list	Wi-Fi CERTIFIED 6 Release 2 and 7 appear as separately listed programs on WFA127760.
Product	Chip/module, board, firmware, host/driver and antenna configuration	G2 is a synthetic two-chain card. No real product certificate is assigned to it.
Peer and active mode	Mutual capabilities plus actual selected resources	AP2 supports the listed two-stream envelope by assumption; real support and adaptation logs are missing.

The IEEE register dated 2026-09-07 lists the 2024 base and 802.11be-2024 as published, while P802.11bn remains in process. A new edition does not erase an older selected mode. These are status observations, not an assertion that every peer implements a later generation. [IEEE-STATUS] [IEEE-HE]

For a concrete document-reading example, certificate WFA127760 identifies an access-point model and firmware, then lists tested features. That is useful evidence at a narrower scope than a generation slogan. The public certificate is readable; the program’s full mandatory/optional requirements were not obtained. [WFA-CERT]

Common misconceptionA Wi-Fi generation means every advertised feature is present.

Separate three questions: is a feature defined, is it required by this exact program/role, and do both products support it in this configuration? An unverified feature stays outside the proposed operating envelope.

Decision: pin HE full-band single-user data as the teaching scope, then check the band and installation before treating any width as a product option.

03 / 10

Band and channel need a regional context

Can a wider channel chosen on a worksheet be enabled in the installed product?

A channel needs a centre and occupied spectrum, not just a width. A 20 MHz allocation fits differently from 80 MHz around a band edge or another user. Device class, permitted installation, interference-avoidance functions and regional requirements constrain which combinations can be used.

Band orientation · questions for the final regional review
Band family	RF/product question	Evidence still required
2.4 GHz	Shared with the node and other users; channel overlap can matter directly.	Local channel/width choices, installed antenna, enabled power and coexistence evidence.
5 GHz	Different channel groups can have different installation and spectrum-sharing conditions.	Exact centre/width, indoor/outdoor use and applicable DFS/TPC requirements.
6 GHz	A label such as 6E does not settle device class or installation.	Market-specific class, spectrum, coordination and final product/antenna conditions.

The Swiss OFCOM orientation points to separate radio interface regulations and distinguishes device classes and installation conditions. The US FCC’s 2026 6 GHz document combines an order and further proposals; reading its title or an earlier summary cannot resolve current applicability. Use these as routes to the responsible review, not a universal legal channel list. [CH-RLAN] [FCC-6G]

Band + centre + width + device class + installation + market → applicability questionThis is a documentary dependency, not a numerical permission function. Conducted power at R1 is not EIRP: the antenna/feed and spatial reference plane must also be established.

The local coupling fixture names mounting A at 25 °C, a 5.180 GHz / 20 MHz Wi-Fi aggressor and a 2.450 GHz synthetic victim, separated by 2.730 GHz. Its 200 kHz noise-equivalent bandwidth is a chosen receiver condition, not a Bluetooth bandwidth. Wider data-resource cases deliberately hold the RF transfer assumption fixed; they do not assert that those centres or transfer values remain valid when hardware changes.

Go deeperWhy even different bands can share a failure path

The receiver sees energy through antennas, package/board paths and finite selectivity. A band label is not infinite isolation. The exact total transfer and the receiver’s behavior at that offset must be supplied; Section 9 uses synthetic values so the accounting can be checked without inventing a measurement.

Keep region and device class as question cards in the experiment. There is no regulatory approval switch. Now count only the resources that actually carry data.

04 / 10

Follow the OFDM resource budget

Why does an FFT with 256 bins not supply 256 payload-bearing subcarriers?

Think of a data symbol as a simultaneous set of frequency-domain assignments. Some assignments carry mapped symbols; pilots support reference tracking; guard and DC bins carry no payload. The transform length sets the representation, while the allocation selects what the receiver can use.

Original accounting diagram. Pilot and null bins are grouped for counting, not drawn at their true spectral positions. The cyclic prefix consumes time but adds no new data symbols. No preamble or complete packet is shown.

Frozen full-channel SU teaching resources · informative source chain
Width	Data N_SD	Pilots	FFT bins	Guard/DC remainder
20 MHz	234	8	256	14
40 MHz	468	16	512	28
80 MHz	980	16	1024	28

These counts agree with the active-tone and pilot accounting in the first-party HE explanation; exact final normative text was not obtained. They are therefore informatively checked teaching resources, not an independently certified reproduction of an IEEE table. MathWorks also separates FFT, active, data and pilot indices. [RS-HE] [MW-OFDM]

$T_{\mathrm{symbol}} = T_{u} + T_{\mathrm{GI}}\qquad \Delta f = \frac{1}{T_{u}} = 78.125 \mathrm{kHz}$ T_u is useful data-symbol time, fixed at 12.8 µs. T_GI is cyclic-prefix duration: .8, 1.6 or 3.2 µs. Δf is subcarrier spacing. SI seconds and hertz are used internally. The prefix repeats existing samples and consumes time without adding new information.

Think about itDouble 20 MHz to 40 MHz, then double again to 80 MHz: should the rate multiply by exactly four?

Answer

The first step doubles data tones, 234 → 468. The final count is 980, so 80 MHz scales the 20 MHz rate by 980/234 = 4.188034188. Guard/pilot occupancy does not stay at one fixed fraction of bandwidth.

Guard time helps preserve the symbol relationship over an appropriately bounded channel. A longer guard is not free throughput: the same information takes longer. The chosen GI/LTF contexts are .8 or 1.6 µs with 2× HE-LTF, and 3.2 µs with 4× HE-LTF; the kernel excludes training and full packet timing. [KEY-GI] [MW-HE]

Common misconceptionAn OFDMA resource unit gets multiplied by the full channel again.

An RU is already an assigned subset of tones. Use its own data count once. This lesson instead allocates the whole selected channel to one user; it does not combine that count with a second RU multiplier or pretend a data symbol is a complete PPDU.

Decision: freeze the allocation first. Then assign mapped bits, coding and streams to those data tones.

05 / 10

MCS and streams change rate and RF demands

What changes when more bits are packed into the same set of data tones?

A mapped 64-QAM symbol selects among 64 constellation points and therefore carries six coded bits per data tone. Coding spends some of that mapped capacity on redundancy. Adding streams adds simultaneous information channels only when the entire implementation supports them.

Selected HE MCS subset · illustrative EVM budgets
HE MCS	Mapping / N_BPSC	Code rate R_c	Synthetic RMS EVM allocation
MCS 0	BPSK / 1 bit per tone	1/2	10%
MCS 6	64-QAM / 6 bit per tone	3/4	3%
MCS 7	64-QAM / 6 bit per tone	5/6	3%

Match the MCS label to the code rate

64-QAM at rate 3/4 is HE MCS 6. HE MCS 5 uses the same constellation at rate 2/3, so it carries less information per symbol. Keep both the mapping and the code rate attached to a mode label when comparing throughput. [MW-HE] [RS-HE]

$R_{\mathrm{PHY}} = N_{\mathrm{SD}} \times N_{\mathrm{BPSC}} \times R_{c} \times \frac{N_{\mathrm{SS}}}{T_{u} + T_{\mathrm{GI}}}$ N_SD is data tones per stream; N_BPSC is mapped coded bits per data tone; R_c is information/coded-bit ratio; N_SS is simultaneous spatial streams. R_PHY is nominal information bit/s in the data field, before preamble, framing, MAC and application overhead.

$\begin{aligned}234\times6\times\frac34\times1&=1053\,\mathrm{bit}\\\frac{1053}{13.6\,\mathrm{\mu s}}&=77.4264705882353\,\mathrm{Mbit/s}\end{aligned}$ Derived default. LDPC is the selected coding family throughout this teaching envelope. No DCM, STBC overhead, puncturing, multi-user division or multi-link summation is included. Supported mathematical resources do not establish complete PPDU validity.

One change at a time · all other default resources fixed
Case	Information bits / data symbol	Nominal rate (Mbit/s)
20 MHz / 0.8 µs GI	1053	77.426470588
40 MHz / 0.8 µs GI	2106	154.852941176
80 MHz / 0.8 µs GI	4410	324.264705882
20 MHz / 3.2 µs GI	1053	65.8125

Closer constellation points make residual errors more consequential, but an EVM percentage needs its reference normalization and processing. Here Q25 assumes 2.5% RMS over all data tones in 1000 steady symbols, normalized to ideal mean symbol energy, after perfect timing/carrier recovery and linear per-tone equalization. No nonlinear correction is assumed. It is an illustrative evaluation, not a captured waveform. Against the chosen 3% MCS 6 budget it leaves 0.5 percentage points; against a normative requirement the result is unknown.

Common misconceptionAn EVM pass proves regulatory compliance.

The 10%/3% allocations are product teaching budgets, not IEEE limits. A budget comparison does not test emissions, power, coexistence or final market requirements. Nor does the AWGN-only EVM/SNR approximation from 02.7 guarantee PER or choose a real MCS.

Go deeperTransmitter headroom needs the actual waveform population

The PA must preserve the selected waveform over its amplitude distribution, operating power, bias, supply, temperature and load. EVM and MCS alone do not determine PAPR or a universal backoff. Use 05.5 to allocate fidelity and DC expenditure, then verify actual modulated output. No assumed crest value is needed for this rate kernel.

Next, ask what physical resources can supply those streams.

06 / 10

Antenna count does not guarantee MIMO rank

Four antennas fit around the enclosure. Why might the backhaul still use one stream?

Antennas are spatial interfaces. Streams are independently recoverable information paths through a channel and receiver. Four interfaces can experience strongly related channels, feed only two simultaneous RF chains, or connect to a peer that supports one stream. The hardware count alone cannot resolve those cases.

Think about itG2 has two RF chains, AP2 supports two streams, and four antennas are fitted. Can you select four streams?

Answer

No. The synthetic capability intersection permits at most two. A four-stream arithmetic result can demonstrate scaling, but it cannot enter this product envelope. With AP1, the intersection falls to one even if the gateway retains its two chains.

N_SS ≤ min(supported TX streams, supported RX streams, usable channel rank)Rank counts independent spatial dimensions under a stated channel model. Having enough rank is necessary, not a guarantee that the weakest supported stream meets the required error rate. The channel, SNR, calibration and adaptation need measurements.

Different antenna mechanisms, different evidence
Mechanism	What changes	What to establish
Spatial multiplexing	Independent information streams share time/frequency resources.	Enough chains at each peer and usable channel dimensions.
Diversity	Alternative observations reduce vulnerability to some fades.	Joint channel statistics and receiver combining behavior.
Beamforming	Coherent weights shape the spatial response.	Channel estimates, phase/gain calibration and the installed antenna response.
Isolation / correlation	Coupling and channel similarity constrain different aspects.	Port transfer does not by itself establish low field/channel correlation.

As a simple mathematical comparison, H = diag(1,1) has two independent columns; a 2×2 matrix whose entries are all one has rank one. Neither is measured gateway data. They explain why the same two-by-two count can expose different spatial dimensions. The detailed array and combining conventions remain in 06.4.

Common misconceptionMIMO equals antenna count.

Specify active chains, supported streams and peer intersection, then request mounted rank/correlation and calibration evidence. G2’s extra antennas might support switching or diversity; they cannot create two additional simultaneous RF chains.

Freeze the supported tuple before budgeting medium access. A valid data-field rate still has to earn wall-clock service time.

07 / 10

Medium access consumes time and energy

Why does the application receive fewer bits than the selected data field could carry?

A transmitter waits for access, sends training and headers, carries framed payload, then handles response and retry behavior. Aggregating payload can amortize some fixed costs. Other stations also need the medium. These costs belong in a traffic model; pretending that every second is payload transmission makes a rate table look like a service promise.

Use a bounded saturated-service proxy. Baseline efficiency η₀=.65 includes protocol payload efficiency, overhead, aggregation, ACK and backoff once. It excludes failed attempts and any explicitly scheduled unavailability. Let p=.10 be an independent failed-attempt probability, including the chosen acknowledgement criterion, with equal service cost for every attempt. This simplified p is not an inferred real PER.

$\begin{aligned}K&=r+1\\A&=\sum_{k=0}^{K-1}p^k\\q&=1-p^K\end{aligned}$ r is additional retry cap; K is maximum attempts; A is expected attempts per offered packet; q is per-packet eventual delivery probability under the finite policy. A and q are dimensionless. Default r=2: A=1.11 and q=.999.

$R_{\mathrm{good}} = \frac{R_{\mathrm{PHY}} \eta _{0} a q}{A}\qquad \frac{q}{A} = 1-p \text{for} p<1$ a is explicitly available Wi-Fi wall-time fraction, independent of η₀. Derived default: 77.4264705882353 × .65 × 1 × .999/1.11 = 45.2944852941176 Mbit/s. At p=1, A=K remains finite and goodput is zero.

Think about itRaise the retry cap without changing p. Does saturated goodput improve?

Answer

Eventual packet delivery improves, but expected work also increases. The two changes cancel in q/A for independent equal-cost attempts. Do not multiply by both q and (1−p) after already dividing by A.

Finite retry policy · p=.1, default rate, η₀=.65 and a=1
Additional retries	Expected attempts A	Packet delivery q	Service proxy (Mbit/s)
0	1	0.9	45.294485294
2	1.11	0.999	45.294485294
10	1.1111111111	1	45.294485294

$u = \frac{D}{R_{\mathrm{good}}} = \frac{20}{45.2944852941176} = 0.4415548575092735$ D is offered D0 bit/s; u is the required wall-time service share under the proxy. u>1 is infeasible; u=1 has no reserve. D=0 has zero demand even if service is zero. Positive load with zero service has no finite share and is reported as infeasible.

Optional energy uses an assumed active-cycle DC power of 2 W, including whatever active access-cycle states η₀ represents. At a=1, it is 44.155486 nJ per successfully delivered payload bit. Reducing a reduces wall-time capacity, not this active-operation energy/bit. Awake/sleep durations, idle power and host consumption remain unknown, so no total product power or battery forecast follows.

Common misconceptionAn average MAC efficiency predicts P95 latency.

The same average can hide very different waiting-time tails. Real ACK timeouts, rate fallback and contention histories also break equal attempt cost. Log release and complete-delivery timestamps; do not manufacture a percentile from η₀ or p.

Use the proxy to screen workload, then explore a finite set of modes rather than declaring one permanent connection rate.

08 / 10

A connection moves through an envelope

Which operating points must remain acceptable when the connection adapts?

Choose low, nominal and high resource cases with fixed names. A real connection can change width, coding, streams and access behavior as conditions change. These three cases are an envelope for investigation, not a timeline or a statistical distribution of selected modes.

Three resource cases · same η₀=.65, p=.1, r=2, a=1, D=20 Mbit/s
Named case	Tuple	R_PHY / R_good (Mbit/s)	Screen / remaining evidence
Low	20 MHz / MCS 0 / 1 stream(s) / 3.1999999999999997 µs	7.3125 / 4.277812	Infeasible · overload; simultaneous default RF failure remains.
Nominal	20 MHz / MCS 6 / 1 stream(s) / 0.7999999999999999 µs	77.426471 / 45.294485	Within the service proxy; simultaneous default RF failure remains.
High	80 MHz / MCS 7 / 2 stream(s) / 0.7999999999999999 µs	720.588235 / 421.544118	Within the service proxy; simultaneous default RF failure remains.

A controlled width increase can improve the proxy, yet a wider real channel can see different contention and interference. Counterexample: the default 20 MHz service is 45.294485 Mbit/s; a 40 MHz case with η₀=.2, p=.5 and a=.8 yields only 12.388235 Mbit/s. These are chosen conditions showing why “wider always wins” is not a valid decision rule.

Go deeperLater-generation features still need a concurrency description

The read WFA127760 AP certificate separately lists EMLSR, load balancing in MLO, STR and static puncturing. A multi-link single-radio mode and simultaneous transmit/receive capability do not mean the same resources are available together. The certificate explains one product’s tested features; it does not define mandatory scope for all Wi-Fi 7 products. The HE kernel here adds no link rates and includes no punctured allocations. [WFA-CERT]

Guided experiment · Class 1 · wifi-mode-envelope/2.0

Wi-Fi Mode Envelope

Predict the resource gain, then test whether the same gateway can still hear its node. The canonical inputs and results below also render without JavaScript.

Before revealing: what should 40 MHz do to the nominal rate, with all other resources fixed?

1 · Resource request and capability intersection

Full-channel width (MHz)Closed choices; default 20. Reset restores the canonical fixture.

HE modulation and codingClosed choices; default MCS 6. Reset restores the canonical fixture.

Spatial streams N_SS (count)Closed choices; default 1. Reset restores the canonical fixture.

Data guard interval T_GI (µs)Closed choices; default 0.8. Reset restores the canonical fixture.

Synthetic gateway capabilityClosed choices; default G2. Reset restores the canonical fixture.

Synthetic peer capabilityClosed choices; default AP2. Reset restores the canonical fixture.

G2 / AP2 are independent synthetic capability cards: all listed widths and MCS/GI contexts, at most two simultaneous chains/streams. Three/four streams can be requested to inspect rejection; four fitted antennas do not expand the cards.

2 · Service, retries and scheduled availability

Baseline efficiency η₀ (fraction)Range 0.2–0.9; arrow step 0.01; default 0.65. Reset restores the canonical fixture.

Failed-attempt probability p (fraction)Range 0–1; arrow step 0.01; default 0.1. Reset restores the canonical fixture.

Retry cap r (additional attempts)Range 0–10; arrow step 1; default 2. Reset restores the canonical fixture.

Offered D0 load D (Mbit/s)Range 0–300; arrow step 1; default 20. Reset restores the canonical fixture.

Concurrency planClosed choices; default Simultaneous. Reset restores the canonical fixture.

Wi-Fi availability a (fraction of wall time)Range 0–1; arrow step 0.01; default 1. Reset restores the canonical fixture.

Protected windows require a ≤ .8. Select the complete protected preset to set both values together. η₀ excludes failures and scheduled unavailability.

3 · RF transfer, quality and unanswered source questions

On-time Wi-Fi power P_TX (dBm at R1-TX)Range -10–30; arrow step 1; default 20. Reset restores the canonical fixture.

Total transfer isolation I (dB, R1-TX → R1-RX)Range 20–100; arrow step 1; default 45. Reset restores the canonical fixture.

Synthetic RMS EVM evaluationClosed choices; default Q25. Reset restores the canonical fixture.

Victim / integrated leakage conditionClosed choices; default V82. Reset restores the canonical fixture.

Region / device-class question cardClosed choices; default CH question. Reset restores the canonical fixture.

Mode-source evidence exerciseClosed choices; default Informative. Reset restores the canonical fixture.

V82: 5.180 GHz Wi-Fi aggressor, 2.450 GHz / 200 kHz synthetic victim, mounting A, 25 °C; no actual device data. V72 changes only integrated leakage to 72 dB as a named stress condition. Neither preset determines lawful transmit power.

Canonical example loaded.

Committed variant: p07-m03-he-envelope-v1. Illustrative model; frozen evaluation 2026-09-08 UTC.

Baseline and committed current · full precision used for decisions
Quantity / unit	Canonical baseline	Committed current
Data-field tuple	20 MHz · MCS 6 · 1 stream(s) · 0.8 µs GI	20 MHz · MCS 6 · 1 stream(s) · 0.8 µs GI
Resources	234 data + 8 pilots; FFT 256; 14 guard/DC	234 data + 8 pilots; FFT 256; 14 guard/DC
Symbol arithmetic	1053 bit / 13.6 µs	1053 bit / 13.6 µs
Nominal information rate R_PHY	77.426470588 Mbit/s	77.426470588 Mbit/s
Expected attempts A / packet delivery q / q÷A	1.11 / 0.999 / 0.9	1.11 / 0.999 / 0.9
Efficiency and explicit availability	η₀=0.65, a=1	η₀=0.65, a=1
Delivered service proxy R_good	45.294485294 Mbit/s	45.294485294 Mbit/s
Offered D0 load / load headroom	20 Mbit/s / 25.294485294 Mbit/s	20 Mbit/s / 25.294485294 Mbit/s
Normalized demand u	0.441554858 wall-time service share	0.441554858 wall-time service share
Assumed 2 W active-cycle DC energy	44.155486 nJ / delivered bit (active only)	44.155486 nJ / delivered bit (active only)

Within the service proxy

Hold the product decision; at least one included screen or evidence gate is unresolved or fails.

Supported by synthetic capability intersection. Neither result grants actual service or market approval.

RF quality and concurrency · separate mechanisms and evidence
Screen / boundary	Committed current result
Synthetic quality budget	2.5% RMS evaluated / 3% allocation · Pass synthetic budget. Normative limit: unknown.
Total blocker at R1-RX	-25 dBm / ceiling −35 dBm; headroom -10 dB · Fail synthetic screen
Noise-like leakage in 200 kHz at R1-RX	-107 dBm with N=−110 dBm; rise 4.764348624 dB / maximum 1 dB · Fail synthetic screen
Required total isolation	Blocker 55 dB; leakage 53.868253244 dB; binding Total-power blocker: 55 dB
Unscheduled simultaneous state	Prohibited · synthetic RF screen fails
Selected window plan	No protected weak-node windows
RF/DC boundary	On-time RF output is independent of the 2 W assumed active-cycle DC power. Scheduling reduces wall-time capacity, not active energy/bit. Whole-product power unknown.

Current simultaneous-state diagnostics at the victim’s component input, referenced to 1 mW (dBm). Solid box: total blocker. Dashed box: separately integrated leakage. V82 is a chosen mounted condition; changes to width/streams hold its transfer fixed as an illustrative sensitivity comparison.

Q25/Q40: assumed RMS error over all selected data tones and 1000 steady data symbols, normalized to mean ideal reference-symbol energy; perfect timing/carrier recovery and per-tone linear equalization assumed, no nonlinear correction. Bandwidth is the selected channel; R1-TX waveform referred to the analyzer’s equalized symbols. These are chosen evaluations, not acquired traces.

Evidence that remains open

Actual D0 latency distribution and P95: unknown.
External aggressors, recovery time, wake/control latency and missed windows: unknown.
CH device class, channel, installation and authorization: unresolved.
Informative mode explanation read; exact normative EVM/conformance limits unknown.
Total product power: unknown without base consumption and awake/sleep durations.

Model scope: full-band SU data field, LDPC, no DCM, STBC, MU allocation, puncturing, MLO or complete PPDU timing. η₀ includes framing/aggregation/ACK/backoff once; p accounts for independent equal-cost failed attempts; a accounts for scheduling once. Service share is not exact RF TX duty. Rates and RF comparisons show up to nine decimal places; retry quantities show up to ten. Decisions retain full precision.

JavaScript is unavailable. Read the complete p07-m03-he-envelope-v1 result above and the static comparison tables throughout the lesson. All prediction and checkpoint answers open with native disclosures.

Decision: retain named tuples and their evidence, not a “best Wi-Fi” score. The next section audits the default failure that remains even when throughput looks comfortable.

09 / 10

Concurrent radios share a real RF product

The gateway’s Wi-Fi channel is far from the node. Why does weak reception still disappear during a burst?

Trace the whole coupled path. R1-TX is the Wi-Fi component RF output; R1-RX is the victim component RF input. R2-TX/R2-RX remain antenna-feed planes and S0 is the spatial/OTA boundary. In p07-m03-coupling-v1, total isolation I=45 dB already includes every transformation between the two R1 planes for mounting A. Subtracting an extra feed loss after it would count that loss twice. R3 remains the receiver decision boundary.

$P_{\mathrm{B}} = P_{\mathrm{TX}} - I = +20 - 45 = -25 \mathrm{dBm}\qquad H_{\mathrm{B}} = B_{\mathrm{lim}} - P_{\mathrm{B}} = -10 \mathrm{dB}$ P_TX is on-time average Wi-Fi RF output, referenced to 1 mW. P_B is total coupled aggressor power at victim R1-RX. B_lim=−35 dBm is an illustrative maximum blocker level for this offset/waveform/victim condition; nonnegative headroom passes this synthetic screen.

Think about itThe throughput target is met. Can this default mode remain simultaneous with weak-node reception?

Answer

No. The coupled blocker exceeds the synthetic ceiling by 10 dB. Throughput and blocking screen different parts of the product. Keep this simultaneous state prohibited even if you later select protected windows.

Now ask a different question: how much noise-like energy is integrated into the victim’s receive bandwidth? Set equivalent input noise N=−110 dBm over a 200 kHz noise-equivalent bandwidth. A separately chosen integrated transfer K_leak=82 dB maps total coupled aggressor power to input-equivalent noise-like leakage. It includes the disclosed spectral/selectivity treatment once; it is neither measured ACLR nor universal rejection.

$P_{\mathrm{leak}} = P_{\mathrm{B}} - K_{\mathrm{leak}} = -107 \mathrm{dBm}\qquad \mathrm{rise} = 10 \log _{10}(1 + 10^{\frac{P_{\mathrm{leak}}-N}{10}})$ The derived rise is 4.764348624365 dB. Convert the two independent powers into a linear ratio before adding them. This white-interference approximation does not predict correlated packet collisions, reciprocal mixing or nonlinear products; those need distinct evidence.

$P_{\mathrm{leak,max}} = N + 10 \log _{10}(10^{\frac{1}{10}}-1) = -115.868253243801 \mathrm{dBm}$ Invert the declared 1 dB maximum rise. Required isolation from leakage is P_TX−K_leak−P_leak,max = 53.868253243801 dB. Blocking requires P_TX−B_lim = 55 dB. The larger requirement, 55 dB, binds this fixture.

One mounted V82 condition · independently checked isolation comparisons
Isolation I (dB)	P_B / H_B (dBm / dB)	Leakage (dBm) / rise (dB)	Synthetic simultaneous state
45	-25 / -10	-107 / 4.764348624	Prohibited; both screens fail
55	-35 / 0	-117 / 0.790097497	Both screens pass; no measured validation
60	-40 / 5	-122 / 0.265723756	Both screens pass; no measured validation

At 55 dB the blocker is exactly on its ceiling: zero reserve, not a robust product margin. At 60 dB, the screen has 5 dB blocker headroom. No layout change is claimed to achieve either value. Obtain measured transfer and receiver behavior over relevant mounting, mode and temperature conditions.

Coordination alternative · original 45 dB isolation retained
State	Wi-Fi service proxy	Weak-node interpretation
Simultaneous · a=1	45.294485294 Mbit/s	Prohibited by V82 internal blocker/leakage screens.
20% protected receive windows · a=.8	36.235588235 Mbit/s	Internal Wi-Fi interference absent only during mutually exclusive windows with adequate settling.
Outside protected windows	Unscheduled simultaneous RF levels remain −25 dBm / −107 dBm	Weak-node reception during Wi-Fi TX remains prohibited. External radios are not removed by this scheduler.

Common misconceptionDifferent channels cannot desensitize the receiver.

The total blocker can stress the receiver even while the integrated in-band leakage follows a different limit. Conversely, a blocker pass need not imply a leakage pass. In the V72 stress preset the leakage requirement becomes 63.868253244 dB and overtakes blocking. Resolve both mechanisms and their conditions.

Protected windows assume ideal no-overlap and adequate settling. Receiver recovery, external aggressors, missed deadlines and wake/control latency are open. A fraction of protected time alone cannot establish the node’s two-second delivery statistic.

10 / 10

Specify allowed modes and a concurrency plan

What should the gateway design record allow, prohibit and ask the lab to establish next?

Keep the 20 Mbit/s common backhaul target. Candidate A is 20 MHz/MCS 6/one stream/.8 µs GI; candidate B is 40 MHz/MCS 6/two streams/.8 µs GI. With protected availability a=.8, their service proxies are 36.235588235 and 144.942352941 Mbit/s respectively. Both satisfy the chosen average load, but neither proves the candidate 100 ms deadline or the inherited node delivery requirement.

Frozen candidate decision · G2 and AP2 capabilities
Candidate	Hardware / service	Concurrency decision
A · retain for bench evaluation	One active chain and antenna path at each endpoint; verify installed behavior.	At 45 dB, protect node windows. Simultaneous reception is prohibited during Wi-Fi TX.
B · retain for bench evaluation	Two simultaneous chains and independent antenna paths at each endpoint; verify rank/correlation and calibration.	Same protected plan; verify actual transfer after width/chain change rather than inheriting 45 dB as measured fact.
40 MHz / MCS 6 / four streams / .8 µs · reject	619.411764706 Mbit/s mathematical data-field rate, unsupported by G2/AP2.	Cannot enter the product recommendation regardless of its calculated rate.

WIFI-NODE-01 · discriminating bench request, not a completed measurement
Test decision	Named conditions / observable
Freeze the DUT	Lab and firmware owners record board revision, firmware hash, host/driver, AP capability, exact Wi-Fi and victim modes, antennas, mounting A, power, supply and actual temperature. Replace V82 with the real receiver waveform and noise bandwidth before a product test.
Freeze planes and wanted level	Calibrate R1-TX on-time output and R1-RX coupled total/leakage with stated uncertainty; keep R2 feed paths named. Sweep a documented weak wanted level around its measured baseline error transition, not an invented universal sensitivity.
Control the comparison	Run A and B with identical D0 offered load 20 Mbit/s; compare quiet node reception, simultaneous Wi-Fi TX and 20% protected reception. Record actual centres, widths, streams, GI, coding, aggregation, retries, interference environment and schedule.
Observe service and recovery	Use synchronized gateway/peer D0 release/delivery logs, node sequence/error counts and TX-enable/receiver-ready timing. Report unique bits per 60 s steady window, lost/late payloads, node errors, recovery delay and observed mode changes. Use a predeclared sample size/confidence method for deployment claims.
What reverses the decision?	Failure of average service rejects that mode/traffic plan; node errors during protected windows challenge isolation/settling/control assumptions. A lower measured blocker ceiling or more leakage raises the isolation requirement. Observe failure tails before claiming either deadline.
Separate owners	Standards owner obtains exact normative mode/test text. Compliance owner resolves region, device class, installation and final configuration. RF/antenna owner measures coupling and rank; firmware owner verifies scheduling and peer behavior.

This complete local snapshot is useful on a first visit. Missing durations, currents and source evidence stay unknown; they are not zero and cannot create a passing result.

Technology evidence card · p07-technology-evidence-card-v1
Field	P07-M03-HE-BACKHAUL-v1 · complete local snapshot
Identity / owner	p07-technology-evidence-card-v1 · P07-M03-HE-BACKHAUL-v1 · RF systems owner
Preserved inherited IDs	P07-M01-LE-EXT-v1 · P07-M02-LE-CONN-v1 · P07-M02-THREAD-SED-v1 · p06-evidence-map-v1 · M01-A · M01-B-LOSS · M01-INSTALL-UNKNOWN
Scenario	Illustrative compact, mains-powered gateway: Wi-Fi backhaul plus weak 2.4 GHz node reception. The generic 2.450 GHz QPSK and earlier LE/Thread variants remain unchanged.
Requirement	Backhaul: 20 Mbit/s unique D0 payload offered continuously, assessed over a declared 60 s steady-state window (startup separate). Local illustrative deadline candidate: ≥99% of 1500-byte backhaul payload units within 100 ms; unverified. Node: retain 32 bytes each 60 s, ≥99% at gateway D0 within 2 s in each site stratum.
Technology family	IEEE 802.11 HE single-user full-channel Wi-Fi backhaul; separate V82 synthetic narrowband victim, not relabelled LE or the generic QPSK baseline.
Standard / profile / mode	Selected IEEE 802.11ax-2021 HE identity, incorporated in 802.11-2024. Data-field kernel only: LDPC, no DCM/STBC/puncturing/MU/MLO. MCS 0/6/7; GI .8/1.6 µs with 2× HE-LTF context, 3.2 µs with 4× context. Exact normative clauses not obtained.
Band / region / class	CH installation candidate; baseline coupling condition uses 5.180 GHz / 20 MHz Wi-Fi and 2.450 GHz / 200 kHz synthetic victim. Centre separation 2.730 GHz. 40/80 MHz are separate resource scenarios with coupling transfer assumed unchanged solely for comparison. Final centres, device class, indoor/outdoor and other markets unresolved.
D0 and latency definition	Backhaul: complete 1500-byte D0 unit released by gateway application to complete receipt at remote peer D0; duplicates count once, loss/late arrivals fail 100 ms candidate deadline. 12000 bits per 0.6 ms = 20 Mbit/s; actual packetization/schedule unknown. Node endpoint remains gateway D0.
PHY/MAC / chains / candidate tuples	Conditional candidates A: 20 MHz/MCS 6/1 stream/.8 µs; B: 40 MHz/MCS 6/2 streams/.8 µs. G2 and AP2 synthetic cards support widths 20/40/80 MHz and at most 2 simultaneous streams, all selected MCS/GI contexts. Four-stream request rejected. Candidate A needs 1 independent chain/path at each peer, B needs 2; channel rank, calibration and installation must be measured.
State assumptions	Wi-Fi active cycle: duration unknown; current unknown · weak-node protected receive window: duration unknown; current unknown · receiver recovery: duration unknown; current unknown · host/base platform: duration unknown; current unknown · sleep: duration unknown; current unknown. Optional active-cycle DC allocation: 2 W; no whole-product power estimate.
Antenna / RF interfaces	D0 application → D1 framed/coded bits → R1-TX component RF port → feed/match → R2-TX → S0 spatial path → R2-RX → receiver → R3 decisions → peer D0. Local R1/R2 suffixes retain portfolio meaning. Total I already includes the full R1-TX to victim R1-RX transfer; no extra feed/filter subtraction. R3 is not an antenna input.
Infrastructure / peer	Powered gateway plus compatible AP/peer, antenna paths, host/network stack, backhaul endpoint and synchronized D0 logs. Product/firmware support and peer deployment evidence unknown.
Qualification lane	Exact Wi-Fi Alliance program release, role, feature list and applicable product/firmware certification record unknown. WFA127760 is a real explanatory example, not this gateway.
Regulatory lane	Compliance owner: CH RIR applicability, final antennas/centres/power/device class/installation; separate US review if added. Conducted +20 dBm is a synthetic input, not permitted EIRP. No market approval.
Claim-level source trace	C-HE-RESOURCE → IEEE-STATUS, IEEE-HE, RS-HE, KEY-GI, MW-HE, MW-OFDM: Informative resources checked; exact normative body unread \| C-SERVICE → chosen fixture / derivation only: Derived from chosen efficiency/attempt/availability assumptions; actual D0 distribution unknown \| C-COUPLING → chosen fixture / derivation only: Illustrative p07-m03-coupling-v1; no measured transfer/blocker/leakage evidence \| C-QUALITY → chosen fixture / derivation only: Illustrative 10%/3% RMS allocations; 2.5% assumed evaluation; normative limit unknown \| C-PROGRAM → WFA-CERT: Public example read; own certificate/feature scope unknown \| C-MARKET → CH-RLAN, FCC-6G: Orientation only; final applicability unknown
Evidence and uncertainty	Actual EVM population, mounted isolation versus frequency/mode, leakage/selectivity, adaptation, state currents/durations, external interference, scheduler/recovery/wake latency and tail distributions unknown. Synthetic evaluation 2026-09-08 UTC.
Conditional decision	Retain A and B as conditional bench candidates. At 45 dB isolation prohibit simultaneous weak-node reception during Wi-Fi TX. Reserve ≥20% mutually exclusive, adequately settled node windows (a≤.8), or demonstrate ≥55 dB total isolation for the V82 condition. Screen equality has zero blocker reserve. Reject four streams on G2/AP2. No guaranteed service or regional approval.
Next evidence / owner	WIFI-NODE-01: lab/firmware owners freeze board, firmware, modes and mounting; measure node errors plus Wi-Fi offered/delivered D0 load in quiet, simultaneous and protected cases; standards owner obtains exact clauses; compliance owner resolves installation/class.
Review date / trigger	2026-12-07 · 2026-12-07 or change in standard/test program, firmware, peer, traffic, band/width, antenna, enclosure, temperature, installation or market, whichever comes first.

Ungraded review

Check your understanding

Answer each question in your own words, then reveal the model answer.

01How many 20 MHz FFT bins carry data, and what changes at 80 MHz?
Model answer
234 data + 8 pilot = 242 active bins, leaving 14 guard/DC bins out of 256. At 80 MHz, 980 data + 16 pilots leaves 28 out of 1024. The data-rate ratio is 980/234 = 4.188034188, not exactly four. These are informative teaching resource counts, not a copied normative table.
02Calculate the default rate and the 3.2 µs GI comparison.
Model answer
MCS 6 is 64-QAM 3/4: 234 × 6 × 3/4 × 1 = 1053 information bits per symbol. Divide by (12.8 + .8) µs to get 77.4264705882353 Mbit/s. With 3.2 µs GI, divide by 16 µs: 65.8125 Mbit/s. The longer prefix adds time, not information. HE MCS 5 would instead use 2/3.
03At p=.1 and two retries, why does adding more retries not raise saturated goodput?
Model answer
A=1+.1+.01=1.11 expected attempts; q=1−.1³=.999. Their ratio is .9. With equal-cost independent attempts, q=(1−p)A, so goodput is R_PHY×η₀×a×.9 regardless of retry cap. At p=1 the finite sum is K and delivery/goodput are zero.
04G2 supports two streams, the peer supports one, and four antennas are fitted. Which stream count can enter the envelope?
Model answer
One. The intersection of G2 and the peer excludes two, three and four streams. Antenna count cannot override RF-chain or peer support, and even a supported tuple needs channel-rank/calibration evidence. Missing capability evidence is unknown, not zero streams or a pass.
05Which of the default blocker and leakage constraints sets required isolation?
Model answer
At I=45 dB, blocker=−25 dBm and headroom=−10 dB. Leakage=−107 dBm; linear noise addition gives 4.764348624365 dB rise. A 1 dB rise allows −115.868253243801 dBm, requiring 53.868253243801 dB isolation. Blocking requires 55 dB, so it binds. At 55 dB, blocker headroom is exactly zero.
06What can protected windows establish, and what experiment comes next?
Model answer
With a=.8 the default Wi-Fi proxy is 36.2355882352941 Mbit/s. Ideal nonoverlap and adequate settling remove the identified internal transmitter only in protected node windows. Simultaneous 45 dB operation remains prohibited; external interference and actual latency remain unknown. Compare synchronized node errors and Wi-Fi offered/delivered D0 logs for named quiet, simultaneous and coordinated modes before accepting the plan.

Handoff: the LPWAN lesson changes the question from abundant local resource capacity to long airtime, coverage and infrastructure tradeoffs. Carry the same D0 endpoints, evidence discipline and concurrency questions into that comparison.

References and further study

Actual access 2026-09-08 UTC; review due 2026-12-07, earlier on document, mode, firmware, antenna, installation or market change. These match the frozen evaluation date; no live freshness service runs in the lesson. Informative sources explain the model; synthetic numbers originate in the named fixtures. IEEE bodies, alliance program requirements and final regional applicability remain unreviewed.

IEEE-STATUS · IEEE 802.11 Working Group
Official project timelines · 2026-09-07 snapshot · 2026-09-07.
Consulted: Published rows: 802.11-2024, 802.11be-2024; in-process P802.11bn row. Official status register; not normative body.
2024 base published 2025-04-28; be amendment published 2025-07-22; bn remains an in-process project. No claim about mandatory features, installed peers or access to clauses.
IEEE-HE · IEEE SA
IEEE 802.11ax-2021 · Enhancements for High-Efficiency WLAN · 802.11ax-2021 · 2021-05-19.
Consulted: Catalogue status, scope, publication history. Catalogue identity only; normative text unavailable.
Selected HE amendment identity; catalogue marks it superseded and the base register lists incorporation into 802.11-2024. Exact mode conformance and normative EVM/test limits remain unknown.
RS-HE · Rohde & Schwarz · Lisa Ward
IEEE 802.11ax technology introduction · 01.00 · PD 3609.9470.52 · 2020-04.
Consulted: §4 Table 2; §4.2.1.1 Figures 4–6; §4.2.1.2 Table 4; §4.2.1.3 Table 5 (printed pp5,9–13). Informative, draft-era first-party explanation.
HE useful time, active-tone counts and pilot positions support the selected accounting; MCS 6 is 64-QAM 3/4. Not final normative authority; no vendor test thresholds or mandatory-feature claims promoted into this lesson.
KEY-GI · Keysight Technologies
Guard Interval (802.11n/ac/ax/be/bn) · 89600 VSA help · 2026 Update 2.0 · 2026; exact publication day unavailable.
Consulted: Guard Interval Time Calculation · 802.11ax rows. Informative instrument documentation.
12.8 µs useful time; cyclic-prefix choices .8, 1.6 and 3.2 µs; time added to useful symbol. Manual analyzer choices are not a declaration that every combination is a valid PPDU.
MW-HE · MathWorks
wlanHESUConfig · Configure HE SU or HE ER SU transmission · Online WLAN Toolbox reference snapshot · Publication day not stated; accessed 2026-09-08.
Consulted: ChannelBandwidth, NumSpaceTimeStreams, STBC, MCS, DCM, ChannelCoding, GuardInterval and HELTFType properties. Informative implementation documentation.
MCS 5 = 64-QAM 2/3, MCS 6 = 3/4; GI/LTF dependencies; LDPC selected for the whole teaching envelope. Software capability is not measured product support or normative conformance; MATLAB was not run.
MW-OFDM · MathWorks
wlanHEOFDMInfo · OFDM information for HE transmission · Online reference; introduced R2019a · Publication day not stated; accessed 2026-09-08.
Consulted: cbw, hegi, ru inputs; info output definitions; FFT-Based Oversampling. Informative implementation documentation.
Distinguishes FFT length, active/data/pilot indices and guard/DC resources. RU is an allocation, not another full-channel multiplier. No calculation was executed with this software; aggregate diagram is original bookkeeping, not a copied tone map.
WFA-CERT · Wi-Fi Alliance
Wi-Fi CERTIFIED certificate WFA127760 · BCM94916R6726W7 · AP · firmware 17.10.373.9 · Certified 2024-01-05; downloaded 2026-09-08.
Consulted: pp1–4: identity, RF architecture and Wi-Fi CERTIFIED 6 Release 2 / 7 feature lists. Actual public certificate; informative product example.
Lists tested features against a named model/role/firmware, including 6E, MLO-related modes and static puncturing; illustrates why the individual record matters. Not the fictional G2 gateway, a recommendation, a universal feature requirement or national permission. Program specification bodies were not obtained.
CH-RLAN · Swiss OFCOM
WLAN / RLAN · Official online orientation snapshot · Publication day not supplied on consulted page.
Consulted: 2.4 GHz, lower/higher 5 GHz, 6 GHz and Legal framework sections. Official informative regulatory orientation.
Device class, installation and applicable RIR must accompany a band question; directs the CH review to RIR1010-01/-04/-05/-11. Underlying RIR bodies and final product applicability not reviewed. No legal channel/power list or authorization supplied.
FCC-6G · US Federal Communications Commission
Unlicensed Use of the 6 GHz Band · FCC 26-1 · Fourth Report and Order / Third Further Notice · Adopted 2026-01-29; released 2026-01-30.
Consulted: Title, contents and introduction ¶¶1–3. Official order/proposal document; limited orientation reading.
US categories and coordination questions require current review; distinguish adopted material from proposed material in the same document. No effective-date analysis, final rules, geofencing implementation or equipment authorization conclusion.

Wi-Fi as anRF System

A headline rate fails the gateway requirement

Identify the standard, certification, and product

Band and channel need a regional context

Follow the OFDM resource budget

MCS and streams change rate and RF demands

Antenna count does not guarantee MIMO rank

Medium access consumes time and energy

A connection moves through an envelope

Concurrent radios share a real RF product

Specify allowed modes and a concurrency plan

Check your understanding

References and further study

Wi-Fi as an
RF System