Why does my game lag in this room?

Make one move

A button press becomes work for a device. In our game an outline can move using local prediction before the reply arrives. Filling the robot marks a received reply; it does not reproduce a real game’s full server simulation.

Find a turn on the radio

Devices share a radio medium. In the selected access method, a station checks for an idle interval and counts down before transmitting. Recognizable traffic can make it wait; simultaneous uncoordinated interference is a different experiment.

Sending the same bytes can take longer

The selected radio modes carry different numbers of bits per data symbol. A slower transmission uses more time for the same complete frame. That does not mean its electromagnetic wave travels more slowly.

An earlier job can keep a reply waiting

Our access point has a prepared burst of tablet-download frames ahead of a game reply. The laptop’s selected reception can stay the same while the access point needs more time to finish those transmissions.

Receiving data and acknowledging it differ

The reply’s data reaches the laptop before its MAC acknowledgment finishes. A short reply gap and a separate ACK follow. The reply-time metric and complete-exchange time therefore have different endpoints.

The room can change reception

Distance, obstructions, antennas and multiple paths affect the received signal. Our simple room holds antennas and transmit power fixed, and adds declared losses for selected model screens. It does not measure your home.

Two arrivals can combine differently

At one receiver, the direct and reflected fields add. Their relative phase can strengthen or weaken the combined signal. The two-path inset exposes that sum, independently of the simpler room calculation.

A useful measurement has a specific meaning

A material-transmission experiment, a signal indicator and a game-reported reply time answer different questions. Keep the apparatus, reference, units, endpoints and conditions attached to the number.

A precise but bounded radio example

The timing lens uses legacy 5 GHz, 20 MHz, non-HT OFDM: 20 µs of preamble/header, whole 4 µs data symbols, 16 SERVICE bits and six tail bits. T = 20 + 4 ceil[(16 + 8L + 6)/(4R)] µs for complete PSDU length L bytes and selected data-field rate R Mb/s. No aggregation, RTS/CTS or multi-user scheduling is modeled.

A source-checked exchange

For 200 complete frame bytes at 24 Mb/s, the PPDU is 88 µs. A 34 µs DIFS and seven realized 9 µs backoff slots precede it. Data reception occurs at 185 µs; a 16 µs SIFS and a 14-byte ACK at the selected 6 Mb/s basic rate extend completion to 245 µs. Seven is a chosen draw, not the mean of 0–15.

The queue calculation has a declared workload

Each prepared 1500-byte background frame occupies 401 µs of exchange/service time at 54 Mb/s, or 2181 µs at 6 Mb/s. Twenty frames therefore add 8.020 or 43.620 ms of waiting. The quiet reply takes 30.430 ms with the selected 30 ms external term; those two workloads give 38.450 and 74.050 ms.

Press-to-display and model reply are not interchangeable

The displayed reply metric runs from request radio eligibility to response DATA reception. The authored relay forwards the request after its MAC exchange completes. Two selected 8 ms local handling terms give the separately labeled press-to-display result. These are not an observed ICMP RTT or actual game benchmarks.

Free-space propagation and a room assumption

For the stated far-field isotropic comparison, loss is 20 log₁₀(4πdf/c). At 5 m and 5 GHz it is about 60.41 dB. With 20 dBm transmit power and zero antenna gains the received value is −40.41 dBm. Each intersected model screen adds an authored 8 dB, not a published material coefficient.

Logarithmic power needs care

Convert noise and simultaneous interference from dBm to milliwatts before adding: P(mW) = 10^(dBm/10). Divide the desired power by their sum to obtain SINR. A recognized competing Wi-Fi transmission that makes a station defer is not automatically an additive interferer during its own data frame.

A teaching receiver is not a chipset specification

Optional automatic mode selection uses authored thresholds: 8, 18 and 26 dB allow 6, 24 and 54 Mb/s. Below 8 dB it reports no supported mode. Real receivers have error probabilities, frame-length effects, rate adaptation and hardware differences omitted here.

Field amplitude is not power

For normalized direct amplitude 1 and reflected amplitude a, H = 1 + a exp(iφ). Relative power is |H|² = 1 + a² + 2a cosφ. At a = 0.5, in-phase, quarter-turn and opposed cases give 2.25, 1.25 and 0.25. This is a point comparison, not creation or disappearance of total field energy.

The wavelength comparison has a condition

At 5 GHz the vacuum wavelength is 0.059958492 m. A half-wavelength change in path-length difference adds π to propagation phase. Moving a receiver half a wavelength does not necessarily change the path-length difference by that amount. Wideband links and additional paths require a richer channel model.

A published fit is not raw data

The D50H curve reconstructs NISTIR 6055 Table 4.13d coefficients for nominal 13 mm drywall, only from 3 to 8 GHz. At 5 GHz it is about +0.21 dB relative to the experimental free-space reference. A positive reference-normalized receiver value does not imply a passive panel generates energy. It is not substituted for the model-room screens.

Wi-Fi evolved beyond this one exchange

The 1999 IEEE working-group record described 5 GHz OFDM with multiple rates. Current Wi-Fi models also include coordinated multi-user scheduling and OFDMA. OFDM tones are not automatically separate users. Our legacy timing lens remains a selected operating example, rather than a claim about every current device.

A brief burst is not necessarily bufferbloat

The prepared queue here eventually empties. A persistent standing queue can add delay without increasing the useful delivery rate; this motivates queue-management methods such as CoDel. This lesson does not implement CoDel or claim every burst is pathological.

Games and video calls

Look for reception, access time, queues and delays elsewhere. One speed number or full signal bars does not identify the cause of a late reply.

Compare a position change fairly

Use the same device and already available metric at A, B and A again. Returning to A helps reveal changes over time.

Read engineering evidence

Ask what the apparatus measured and which assumptions connect that quantity to your question. A precise formula is not automatically a calibrated home predictor.

Signal bars stay full, but the model reply gets later with a tablet download. What changed?

• The radio wave slowed down
• Time waiting for earlier AP transmissions
• The carrier frequency changed

Moving from 5 m to 10 m adds about 17 ns of direct radio flight. Can that alone explain 20 ms of extra lag?

• Yes
• No
• The signal always stops at 5 m

Two networks have different names but use the same radio channel. Must they be independent?

• Yes; names create separate frequencies
• No; they may share airtime or interfere
• The longer name goes first

The same complete 1500-byte frame changes from 54 to 6 Mb/s in this mode. What happens?

• Its data PPDU lasts longer
• Every nearby device is forced to 6 Mb/s
• Its wave travels nine times slower

If a connection carries ten times more data per second, must every reply arrive ten times sooner?

• Yes
• No; other delays may remain
• Capacity is measured in signal bars

The outline robot moves immediately. Does that prove the reply already arrived?

• Yes
• No; it can be local prediction
• It proves there are no network errors

Five reply times are 30, 31, 30, 120 and 31 ms. Is “31 ms, always smooth” an adequate report?

• Yes, because 31 is the median
• No; the 120 ms observation matters
• Replace 120 with zero

Can the 5 GHz value for NIST’s drywall specimen predict a house’s game ping exactly?

• Yes
• No; its material experiment measures a different quantity
• Old experiments have no value

Channel sensing, idle intervals and backoff

RFC 8325 §6.1 explains DCF and its selected OFDM timing context. Used for the bounded legacy exchange; not represented as all current scheduling behavior.

Legacy OFDM symbol and PPDU timing

ns-3 owner documentation/source describes preamble/header, SERVICE/tail bits and whole-symbol rounding. Formula implemented independently; no ns-3 implementation code copied.

Separate acknowledgment frame

ns-3 3.45 Wi-Fi utilities identify acknowledgment sizing and the selected OFDM timing example. A basic 6 Mb/s ACK is an explicit lesson choice.

Receiver models and modern multi-user scope

Official Wi-Fi design documentation distinguishes PHY reception/error models, spectrum and multi-user scheduling. Our hard thresholds are authored, not extracted from these models.

Free-space attenuation

ITU-R P.525-5, §2. Used under declared distance/frequency/antenna assumptions; it does not provide a furnished-house propagation model.

Actual material experiment, image and fitted coefficients

William C. Stone (1997), NISTIR 6055. Figures 2.3.1–2.3.2, 4.4 and 4.13d/Table 4.13d. D50H, 13 mm drywall, is distinct from the photographed 305 mm concrete specimen C112H. Original report bands are 0.5–2 and 3–8 GHz.

Report scan provenance

Internet Archive scan of the original NIST report, identified by exact source hash in the evidence manifest. Its digitization date is not the experiment date. Complete report pages are converted to WebP without changing plotted values.

NIST republication terms

Technical Series Publications policy supports the selected employee-authored report figures with attribution. Other manufacturer-supplied report material is not bundled; NIST endorsement is not implied.

Multipath and field addition

Tse and Viswanath, author-hosted Fundamentals of Wireless Communication, chapter 2. Original lesson phasor/wave graphics; no textbook illustrations copied.

Standing queues and delay

RFC 8289 explains the motivation for Controlled Delay queue management. The prepared FIFO comparison is not an implementation of CoDel.

Response endpoints and measurement definitions

RFC 2681 defines round-trip metrics; endpoint/processing assumptions are part of the meaning of a result. A game reply differs from a router ICMP response.

Delay variation needs a definition

RFC 3393 distinguishes its formal one-way delay-variation metric. We report a small sample’s median and range, without relabeling it that standards metric.

Local game prediction versus remote state

Yahn Bernier’s original Valve technical paper explains prediction and latency compensation. The lesson’s original robot is a selected illustration, not current universal engine defaults or copied game art.

A browser estimate is not an RF instrument

Network Information API draft values are connection estimates. The lesson deliberately does not turn these into signal power, router ping or a room scan.

Historical 5 GHz standards development

March 1999 IEEE 802.11 working-group report connects OFDM and multiple data rates. Selected contemporary development record; no unsupported single-inventor or first-MIMO claim.

Exact vacuum light-speed constant

c = 299,792,458 m/s. Air is approximated as vacuum for the room-flight comparison.