A wireless device can report a strong received signal and still struggle to carry useful data. The reason is simple but often hidden: the receiver must distinguish the wanted signal from everything else arriving in the same measurement space. Received power, background noise and signal-to-noise ratio answer different questions. Treating one as a substitute for the others can send an investigation in the wrong direction.
This guide builds the distinction from first principles. It works a complete decibel example, changes the noise while holding signal power constant, follows the result through rate selection and packet delivery, and then places radio evidence inside an end-to-end service test. The numbers are teaching examples rather than promises about a particular access point, client or protocol. The goal is to learn what each observation can support, what it cannot support and which next measurement would reduce uncertainty.
eduKateSG · Wireless Communication
Find your next learning step
Measure wanted power, background and their ratio separately, then connect the radio evidence to the user task.
ROUTE 3 · CHAPTERS 19–27
Change the context
Change conditions and compare what the result can support.
ROUTE 4 · CHAPTERS 28–36
Test the complete service
Add checks, boundaries, misconceptions and practical decisions.
ROUTE 5 · CHAPTERS 37–45
Review and continue
Use transfer tasks, FAQs, evidence limits and authoritative reading.
Full chapter contents · Wireless Communication subject guide
Chapter contents
Define the measurements · Chapters 1–9
1. The short answer: strength is not separation
2. Three quantities, three questions
3. Why dBm and dB must not be confused
4. Worked example: one room, two noise conditions
5. Check the subtraction before interpreting it
6. The linear-ratio meaning behind decibels
Work the link evidence · Chapters 10–18
10. Noise floor is an estimate, not a photograph of all interference
11. RSSI may be a device-specific indicator
12. Time alignment turns numbers into evidence
13. Distributions are better than one favourable snapshot
14. Rate selection responds to more than one displayed number
15. Retries can expose marginal separation
16. Packet loss is not identical to poor radio quality
17. Latency can rise before a link visibly disconnects
18. Interference and ordinary contention are different claims
Change the context · Chapters 19–27
19. Changed context: weaken the signal in a quiet channel
20. Changed context: signal and noise move together
21. Changed context: SNR improves but service remains poor
22. A controlled test changes one factor at a time
23. Antenna orientation can change the received pattern
24. Multipath can create local peaks and nulls
25. Device limits can distort extreme readings
Test the complete service · Chapters 28–36
28. Separate observation, calculation and interpretation
29. Test the arithmetic with invariants
30. Preserve raw values before rounding
31. A decision rule needs an action
32. Communicate uncertainty without becoming vague
33. Common misconception: more bars mean more capacity
34. Common misconception: SNR proves the interferer
35. Common misconception: one endpoint represents the whole link
Review and continue · Chapters 37–45
37. A practical investigation sequence
39. Complete changed-context worksheet
40. Frequently asked questions
42. Compare clients without assuming identical radios
43. Keep scan measurements separate from associated-link measurements
CHAPTER 1 OF 45 · Define the measurements
1. The short answer: strength is not separation
Received signal strength describes the power of the wanted transmission at a receiver under a stated measurement convention. A noise-floor estimate describes unwanted energy the receiver treats as background. Signal-to-noise ratio, or SNR, compares the two. When both power values use dBm, their difference is an SNR in decibels: SNR = received signal level – noise level.
Suppose the signal is -65 dBm and the noise floor is -92 dBm. The SNR is -65 – (-92) = 27 dB. Keep the same -65 dBm signal but raise the noise to -72 dBm, and the SNR falls to 7 dB. The signal-strength reading did not change, yet the receiver now has much less separation between wanted and unwanted energy. That changed context is the central lesson.
Signal strength asks how much wanted radio-frequency power reaches the receiver. Noise floor asks how much unwanted background the measurement process assigns to the relevant channel or bandwidth. SNR asks how large the wanted-to-unwanted ratio is. A link record should keep all three when the equipment exposes them, along with time, channel, bandwidth, device and traffic state.
These quantities are related but not interchangeable. A weak signal in a very quiet setting can have more separation than a stronger signal in a noisy setting. A high SNR can coexist with congestion above the physical layer. A low application throughput can arise while all reported radio measurements look comfortable. The correct question is not “Which single number is the wireless quality?” but “Which part of the path does this number describe?”
dBm is an absolute power level relative to one milliwatt. A value of 0 dBm is one milliwatt; negative values are smaller powers. dB is a ratio. It tells how far apart two power levels are on a logarithmic scale. Writing “the SNR is -65 dBm” mixes an absolute level with a ratio and loses the comparison.
The subtraction shortcut works only because both absolute powers use the same logarithmic reference. Subtracting -92 dBm from -65 dBm cancels the shared reference and leaves 27 dB. If one value came from a different bandwidth, calibration state or convention, the arithmetic could be neat while the comparison was invalid. Units are necessary evidence, but measurement scope matters too.
CHAPTER 4 OF 45 · Define the measurements
4. Worked example: one room, two noise conditions
Imagine a receiver at a fixed desk. During a quiet interval it reports a wanted signal of -65 dBm and a noise floor of -92 dBm. The SNR is 27 dB. Later, another transmitter or electrical source raises the measured noise floor to -72 dBm while the wanted signal remains -65 dBm. The new SNR is 7 dB.
The signal did not become weaker in this simplified record. The environment became less quiet. A diagnosis based only on received signal strength would miss the change. A diagnosis based only on SNR would locate a separation problem but would not identify its source. The next useful evidence might include channel occupancy, a spectrum observation, retries, rate selection and a time-aligned traffic trace.
CHAPTER 5 OF 45 · Define the measurements
5. Check the subtraction before interpreting it
Negative dBm values make sign errors common. Write the operation with parentheses: -65 – (-92) = 27. A quick reasonableness check helps: the distance from -92 to -65 on a number line is 27. For the noisy case, the distance from -72 to -65 is 7. A result of -157 or 157 dB would reveal an arithmetic mistake.
Keep enough precision for the instrument, not for appearance. If a user interface rounds signal and noise to whole decibels, reporting SNR to three decimal places invents resolution. Conversely, a rapidly changing display may need a distribution or time series rather than one rounded snapshot. Good arithmetic cannot rescue an unrepresentative sample.
CHAPTER 6 OF 45 · Define the measurements
6. The linear-ratio meaning behind decibels
For power, an SNR of 10 dB corresponds to a linear ratio of 10:1, 20 dB to 100:1 and 30 dB to 1000:1. The conversion is ratio = 10^(SNR/10). Thus 27 dB is about 501:1, while 7 dB is about 5:1. The 20 dB fall is a hundredfold reduction in the wanted-to-noise power ratio.
This does not mean application throughput must fall by a factor of one hundred. Modulation, coding, channel width, interference timing, device algorithms and protocol overhead mediate the result. The linear conversion clarifies the physical ratio; it is not a direct service forecast. Preserve that boundary whenever translating radio evidence into a user-facing claim.
CHAPTER 7 OF 45 · Define the measurements
7. A stronger signal can still have a worse SNR
Compare two invented links. Link A has -68 dBm signal and -96 dBm noise, giving 28 dB SNR. Link B has a stronger -58 dBm signal but -63 dBm noise, giving only 5 dB SNR. Judged by signal strength alone, B appears better. Judged by separation from background, A is far better.
The comparison also shows why words such as “excellent” or “poor” need a declared rule and context. A device may calculate noise differently, and rapidly varying interference may not appear as a stable floor. Use the raw observations and method first. Add a category only when its thresholds, intended technology and decision purpose are known.
A transmitter does not possess one universal SNR. Each receiver observes a different signal path and local noise environment. The access point may hear a client well while the client hears the access point poorly, or the reverse. Antenna patterns, transmit power, hardware sensitivity and nearby interferers can make the two directions unequal.
Record which endpoint produced each measurement. A dashboard that says “client SNR” may display an access-point estimate of frames received from that client, not a direct measurement of what the client hears. Before comparing two values, read the field definition. Direction and ownership are part of the datum.
CHAPTER 9 OF 45 · Define the measurements
9. Bandwidth belongs beside the noise reading
Noise power generally accumulates across the measurement bandwidth. A wider channel can contain more total background energy than a narrower one, even when the underlying spectral density is similar. Comparing noise floors from different channel widths without declaring bandwidth can therefore mislead.
This is one reason a copied threshold may fail in another configuration. Keep channel width, centre frequency and measurement method with the values. If a change from 20 MHz to 80 MHz coincides with a different noise reading, do not attribute the full difference to a new interferer until the bandwidth effect and device convention have been examined.
CHAPTER 10 OF 45 · Work the link evidence
10. Noise floor is an estimate, not a photograph of all interference
A receiver’s reported noise floor is produced by hardware and firmware under particular conditions. It may reflect thermal noise, receiver electronics and energy classified as background. Short bursts, collisions or transmissions that the device can decode may influence performance without appearing as one stable elevated number.
Therefore, “the noise floor is low” does not prove the channel is uncontested. Pair it with retries, channel utilisation, error counters and time-resolved observations. If the problem occurs only during brief bursts, a long average can hide it. Match the time scale of the evidence to the time scale of the complaint.
CHAPTER 11 OF 45 · Work the link evidence
11. RSSI may be a device-specific indicator
The label RSSI is often used for received signal strength indication, but implementations and scales vary. Some systems expose dBm estimates; others expose an index or a value transformed for a user interface. Two vendors’ numbers can look comparable while representing different calibration and averaging rules.
Do not convert an undocumented index into dBm by guesswork. Preserve the original field, device model, software version and documentation. Within one stable implementation, the indicator may be useful for trends. Across equipment, prefer measurements with defined units and methods, and state any remaining uncertainty.
CHAPTER 12 OF 45 · Work the link evidence
12. Time alignment turns numbers into evidence
A signal reading at 10:00 and a failed transfer at 10:17 do not automatically explain each other. Radio conditions can change as people move, doors close, transmitters start and rate-control algorithms adapt. Build a shared timeline that includes signal, noise, SNR, retries, selected rates, channel use and the user-visible event.
The finest sampling interval is not always best. It must be frequent enough to catch the event and stable enough not to overwhelm interpretation. Keep raw time stamps and note clock differences between devices. A ten-second clock offset can reverse apparent cause and response during a short incident.
CHAPTER 13 OF 45 · Work the link evidence
13. Distributions are better than one favourable snapshot
A single 27 dB observation cannot describe a link that oscillates between 30 dB and 3 dB. Summarise a period with count, median, lower percentiles, minimum where meaningful and the proportion of samples below a declared operational boundary. Show the time series when bursts matter.
Avoid choosing only the sample taken after the connection recovered. Define the observation window from the task: for example, two minutes before to two minutes after each failed upload. A reproducible window reduces confirmation bias and lets later tests use the same comparison.
CHAPTER 14 OF 45 · Work the link evidence
14. Rate selection responds to more than one displayed number
Wireless devices select modulation and coding rates using implementation-specific evidence. Recent successes, losses, retries, channel conditions and probing may all matter. A displayed SNR can help explain why ambitious rates become fragile, but it does not uniquely determine the selected rate or resulting goodput.
When testing, record the rate actually used, not merely the maximum advertised rate. Separate physical-layer rate from payload throughput. A high nominal rate can carry little user data after contention, retransmission and overhead. This connects the present radio question to the existing payload throughput and goodput guide without repeating its accounting lesson.
CHAPTER 15 OF 45 · Work the link evidence
15. Retries can expose marginal separation
If frames arrive near a receiver’s decision boundary, errors and retries may rise. Yet retries also arise from collisions, hidden nodes and other causes. A retry counter supports a symptom claim; it does not by itself identify low SNR as the cause. The strongest record aligns retries with signal, noise and traffic conditions.
Use complete denominators. “Twenty retries” means little without attempted transmissions or frames. Report a defined retry fraction and state whether repeated attempts are counted separately. If the counter resets or wraps, preserve the raw sequence and apply the existing wrapped-counter reasoning before calculating change.
CHAPTER 16 OF 45 · Work the link evidence
16. Packet loss is not identical to poor radio quality
Packets can be lost in queues, software, upstream networks and application paths after a wireless frame succeeds. Conversely, link-layer retries can hide radio errors so the application sees delay rather than loss. A radio diagnosis should therefore identify the observation layer.
The existing deadline delivery and burst-loss guide shows how to keep message deadlines and loss bursts honest. Here the transfer is to add signal, noise and SNR as time-aligned radio evidence, without claiming that either layer alone explains the whole service.
CHAPTER 17 OF 45 · Work the link evidence
17. Latency can rise before a link visibly disconnects
Lower separation can trigger slower rates, retries and backoff while the association remains intact. A user may report pauses rather than a total outage. Measure latency as a distribution and identify where the timestamps are taken. Average latency can remain modest while a few severe delays break an interactive task.
Do not infer low SNR from latency alone. Use the latency-percentile guide to preserve the distribution, then examine whether its tail aligns with radio changes. This layered test turns two partial observations into a stronger diagnosis.
CHAPTER 18 OF 45 · Work the link evidence
18. Interference and ordinary contention are different claims
Several legitimate transmitters sharing a channel create contention; an unrelated or malfunctioning source may create interference. Both can reduce service, but the remedies and evidence differ. A busy channel can show a reasonable noise floor because decodable traffic is not treated as steady background in the same way.
Use careful language: “high channel occupancy coincided with delay” is stronger than “interference caused the delay” when the source is unknown. Spectrum analysis, protocol captures and controlled channel changes can narrow the explanation. Preserve alternative hypotheses until the evidence separates them.
CHAPTER 19 OF 45 · Change the context
19. Changed context: weaken the signal in a quiet channel
Return to the quiet -92 dBm noise floor and weaken the signal from -65 to -79 dBm. The SNR becomes 13 dB. This is worse than the original 27 dB but better than the noisy 7 dB case. Now the problem is loss of wanted power rather than a rise in measured background.
That distinction directs the next test. For a weakening signal, inspect distance, obstructions, antenna orientation, cable loss and transmit configuration. For a rising noise floor, inspect channel conditions and nearby sources. Both paths may coexist, so the record should calculate contributions rather than force every incident into one label.
CHAPTER 20 OF 45 · Change the context
20. Changed context: signal and noise move together
Suppose the signal rises from -70 to -60 dBm while the noise also rises from -95 to -85 dBm. SNR remains 25 dB. A signal-strength graph alone celebrates a ten-decibel improvement, but the separation did not change. The common movement could reflect gain, measurement convention or a broad environmental change.
Do not conclude that performance must be unchanged. Receiver nonlinearities, interference structure and rate choices can still change. The calculation supports a narrow statement: under the reported values, the wanted-to-background power ratio stayed the same. Next inspect outcome measures rather than extending that statement beyond its evidence.
CHAPTER 21 OF 45 · Change the context
21. Changed context: SNR improves but service remains poor
Imagine moving closer to an access point so SNR improves from 12 to 30 dB, yet a file transfer remains slow. The radio separation is now a weaker candidate explanation. Check channel contention, backhaul capacity, server response, transport behaviour and the application itself.
This is a valuable negative result. It prevents repeated antenna adjustments when the bottleneck lies elsewhere. Record that the task was repeated under a materially improved SNR with no corresponding service improvement. A controlled change that fails to move the outcome can be as informative as one that does.
CHAPTER 22 OF 45 · Change the context
22. A controlled test changes one factor at a time
To compare channels, keep client position, orientation, access point, channel width, traffic generator and observation window as stable as practical. Change the channel, allow the link to settle, repeat the same task and capture the same fields. Random movement between trials confounds the comparison.
Perfect control is rare outside a laboratory. State what changed unintentionally and repeat enough times to see whether the result is stable. A classroom exercise can use stipulated data; an operational claim needs observations from the real environment. Never present a one-off controlled teaching example as an instrument prescription.
CHAPTER 23 OF 45 · Change the context
23. Antenna orientation can change the received pattern
Antennas do not radiate or receive equally in every direction. Device orientation, body blockage and nearby materials can change the path. A learner who rotates a device and sees signal shift should not assume the access point changed power.
Record orientation and placement in comparative tests. The antenna engineering guide develops patterns and link-budget reasoning in more depth. This article uses orientation only as an example of why a received-power observation belongs to a complete geometry.
CHAPTER 24 OF 45 · Change the context
24. Multipath can create local peaks and nulls
Reflections allow several copies of a signal to reach a receiver with different delays and phases. Small movements can therefore change their combination. A location with acceptable average coverage can contain a local weak point, and the pattern can change when the environment changes.
Sample a route rather than one convenient point when the user moves. Preserve spatial coordinates or a clear floor-plan reference. The existing phase ambiguity guide treats a different phase-measurement problem; it is useful connected reading, not evidence that every indoor dip has one simple phase explanation.
CHAPTER 25 OF 45 · Change the context
25. Device limits can distort extreme readings
At very high input power a receiver may saturate or behave outside the range in which a display is trustworthy. Near the sensitivity limit, reported values may be quantised, clipped or selectively available only for decoded frames. Missing observations can therefore be informative rather than random.
Inspect documentation and compare repeated readings. If measurements disappear during failures, do not calculate an average from successful frames and call it the incident average. Mark missingness and its likely selection mechanism. A dataset that observes only what the receiver can decode may overstate the quality of the moments that matter most.
A vendor guide may associate ranges of SNR or signal strength with expected service, but requirements vary with modulation, channel width, hardware, mobility and application. Treat such ranges as context-specific planning guidance. Do not turn them into universal physical laws.
Define the purpose first: basic messaging, voice, dense classroom use or a controlled sensor link may tolerate different conditions. Then validate a proposed boundary with the actual devices and task. Report the raw distribution alongside the pass rate so a future system can reinterpret the evidence.
For each trial, record receiver, direction, time window, channel, width, signal statistic, noise statistic, derived SNR, selected rate, retry denominator, payload outcome and user-visible result. Add firmware and configuration versions when they could affect the fields. One row should represent a clearly defined trial, not a mixture of unrelated intervals.
The table supports checks that prose hides. Verify SNR subtraction, inspect missing fields, compare like bandwidths and sort by time. Keep the raw logs beside the summary so outliers can be traced. A clean table is not a substitute for provenance; it is an index into the evidence.
CHAPTER 28 OF 45 · Test the complete service
28. Separate observation, calculation and interpretation
“Signal -65 dBm” is an observation produced by a device. “SNR 27 dB” is a calculation when paired with -92 dBm noise. “The channel was adequate for the tested voice call” is an interpretation tied to an outcome and scope. Label these layers in notes and reports.
The separation makes review easier. Someone can correct the subtraction without disputing the raw log, or question the service claim without denying the measurements. It also prevents automated dashboards from presenting a coloured label as though it were directly observed truth.
CHAPTER 29 OF 45 · Test the complete service
29. Test the arithmetic with invariants
If both signal and noise increase by the same number of decibels, their difference should stay unchanged. If noise rises while signal stays fixed, SNR should fall by the same amount. If signal rises while noise stays fixed, SNR should rise by that amount. These invariants catch sign and column errors.
Add boundary cases to a spreadsheet or script: equal signal and noise gives 0 dB; -60 and -90 gives 30 dB; -90 and -60 gives -30 dB, which indicates the wanted level is below the stated background estimate. Negative SNR is mathematically possible even if a particular receiver cannot decode useful traffic there.
CHAPTER 30 OF 45 · Test the complete service
30. Preserve raw values before rounding
Suppose two samples are -65.4 and -91.6 dBm. Their difference is 26.2 dB. Rounding first gives -65 and -92, then 27 dB. Neither display is inherently dishonest if the device itself supplies limited precision, but the calculation path should be consistent and declared.
Store the most faithful values available, calculate, then round for presentation. Do not manufacture decimals beyond calibration or export precision. Where readings fluctuate more than the rounding difference, focus on distributions and repeatability rather than cosmetic precision.
A boundary is useful only if crossing it changes what the team does. For example, “if the lower tenth percentile of SNR falls below the locally validated boundary during the task, repeat the test on a controlled alternate channel and inspect retry alignment.” This rule links evidence to a next step.
Avoid rules that automatically blame one component. A low value can trigger investigation without proving cause. Record who applies the rule, which data window it uses and how exceptions are handled. Decision discipline is part of measurement quality.
CHAPTER 32 OF 45 · Test the complete service
32. Communicate uncertainty without becoming vague
State what the measurements show, then name the main limit. “The receiver reported a stable -65 dBm signal; its reported noise floor rose from about -92 to -72 dBm during the affected window, reducing derived SNR from about 27 to 7 dB. The device’s noise estimator does not identify the source.” This is both clear and bounded.
Avoid “Wi-Fi was bad” and avoid an exhaustive disclaimer list. Choose the limit that changes interpretation. If no traffic outcome was recorded, say so. If the clocks were not aligned, do not imply second-by-second causation. Precision in scope builds trust.
CHAPTER 33 OF 45 · Test the complete service
33. Common misconception: more bars mean more capacity
Bars usually summarise a device-specific signal indicator. They do not directly display noise, contention, retries, backhaul or application demand. A full-bar icon can coexist with low throughput, and a modest icon can support the user’s task in a quiet channel.
Use bars for the rough purpose the interface intends, not as laboratory evidence. When diagnosing, retrieve defined measurements and run a repeatable task. If only the icon is available, report it as an observation with its limitation instead of translating it into invented dBm.
CHAPTER 34 OF 45 · Test the complete service
34. Common misconception: SNR proves the interferer
A low SNR says the reported wanted-to-background separation is small. It does not name the source, location or technology producing the unwanted energy. It also may not capture short collisions in the way the analyst expects.
Source identification needs additional evidence: spectral shape, time pattern, channel dependence, controlled shutdowns where authorised, or protocol observations. Keep the claim ladder explicit: symptom, association, narrowed cause and verified cause require progressively stronger evidence.
CHAPTER 35 OF 45 · Test the complete service
35. Common misconception: one endpoint represents the whole link
Wireless communication is bidirectional. An access point’s view of uplink frames does not fully describe the client’s downlink environment. Different transmit powers and noise sources can produce asymmetry. A service may fail because acknowledgements or control frames struggle even when bulk data looks strong in one direction.
Collect both directions when the equipment permits. If only one is observable, state that limitation and avoid symmetric language. A careful partial record is more useful than a complete-sounding claim built from one side.
CHAPTER 36 OF 45 · Test the complete service
36. Transfer check: a classroom, warehouse and outdoor link
In a classroom, many legitimate clients may raise contention while the noise floor remains ordinary. In a warehouse, metal surfaces and moving inventory can make the spatial pattern change. In an outdoor point-to-point link, alignment and weather-related path effects may dominate. The same signal-minus-noise arithmetic survives, but the likely mechanisms and tests change.
For each context, write the receiver, bandwidth, time scale, task and alternative causes before interpreting SNR. Transfer is demonstrated when the learner preserves the measurement logic while adapting the evidence plan. Repeating one threshold in every setting is not transfer.
Start with the user-visible task and affected interval. Identify the receiver and direction. Capture signal, noise and derived SNR with channel and width. Add rate, retries, utilisation, latency and payload outcome. Repeat under one controlled change, such as channel or position, and compare aligned distributions.
Then state the narrowest supported conclusion and the next uncertainty-reducing test. Escalate to specialised radio tools only when the decision warrants them and qualified operators are available. This sequence keeps the investigation anchored to the task rather than to whichever dashboard number is easiest to obtain.
Use four fictional intervals from one receiver. In interval A, signal is -67 dBm and noise is -94 dBm, so SNR is 27 dB. In B, signal stays -67 while noise rises to -76, so SNR falls to 9 dB. In C, signal weakens to -81 while noise returns to -94, so SNR is 13 dB. In D, signal improves to -59 but noise rises to -66, so SNR is only 7 dB. Calculate each value before looking at any service outcome.
Now add outcomes: A transfers successfully with few retries; B shows retry bursts; C selects a lower rate but completes; D has a strong displayed signal yet misses a deadline. The table supports different next tests. B points toward time-varying unwanted energy or contention evidence. C points toward path loss, obstruction or orientation. D demonstrates why strong signal alone cannot clear the radio environment. None of the rows identifies a source without additional observations.
Finally, change the reporting task. For coverage planning, compare the spatial distribution of signal and SNR under controlled load. For an incident, align the four fields with retry and application timestamps. For device comparison, repeat the task and avoid treating unlike RSSI implementations as calibrated instruments. The same arithmetic survives, but the claim, sampling and alternative explanations change. A learner has transferred the method when those changes appear in the evidence plan rather than only in the final number.
Which receiver produced the reading? Are signal and noise in the same units, bandwidth and time window? Was the subtraction signed correctly? Are the values raw, averaged or rounded? Did SNR change because signal moved, noise moved or both? Which outcome changed at the same time?
What alternate causes remain? Does the evidence cover both link directions? Would repeating the task on another channel, at another location or with another client distinguish the hypotheses? Which claim is physical, which is protocol-level and which is about the user’s service? These questions turn a dashboard glance into a defensible investigation.
For a reusable wireless record, save the task description, receiver identity, link direction, channel, width, time window, signal and noise fields, derived SNR, selected rate, retry denominator and application outcome. Include tool and software versions because field definitions can change. A screenshot alone rarely preserves enough context to compare another run.
Add one controlled repeat and one changed-context repeat. The first tests stability; the second tests a mechanism by moving signal, noise or load while keeping other factors as steady as practical. Record unexpected differences rather than editing them out. This small package is sufficient for another reviewer to check the arithmetic, question the interpretation and propose the next discriminating test.
Is -65 dBm always good? No. It is a received-power observation whose usefulness depends on noise, technology, task, device and measurement method. Is 20 dB SNR always enough? No universal boundary fits every configuration. Does moving closer always help? It often raises wanted power, but it may not fix contention, interference, backhaul or application limits.
Can SNR be negative? Yes as a mathematical ratio in decibels, though successful decoding depends on the signal and receiver. Should I add dBm values? Not to combine powers directly; logarithmic power addition requires conversion or the appropriate formula. This guide uses subtraction only for a ratio between comparable absolute levels.
CHAPTER 42 OF 45 · Review and continue
42. Compare clients without assuming identical radios
Two clients at the same desk can report different signal and noise values because antennas, calibration, channel width, drivers and averaging differ. A higher reading from one device does not automatically mean the field at its location is stronger. Use each client’s measurements primarily for within-device trends unless cross-device calibration is known.
For a comparative test, run the same task, direction and interval, and retain device metadata. If one client fails while another succeeds, that is useful evidence about the combined client-link system. It does not isolate antenna, firmware or receiver sensitivity without further tests.
CHAPTER 43 OF 45 · Review and continue
43. Keep scan measurements separate from associated-link measurements
A passive or active scan samples channels under different dwell times and conditions from an associated link carrying traffic. Its signal observation can help map coverage, but it may not share the link’s averaging, bandwidth or temporal context. Do not merge scan rows and live-link rows into one distribution without labels.
Use scans to find candidates and associated measurements to test the task. If the tools report different fields, preserve both rather than forcing a common schema. The distinction prevents a quick survey from becoming false proof of sustained service quality.
The examples stipulate clean signal and noise fields so the reasoning can be checked. Real receivers estimate these quantities differently, radio conditions vary quickly and displayed values may omit important interference. This article does not configure transmit power, certify coverage or replace a site survey by a qualified practitioner.
Use the method to structure evidence: declare receiver, units, bandwidth, timing and outcome; preserve raw readings; compute the difference; test a changed context; and stop the conclusion at the supported layer. Safety, regulatory and organisational requirements remain part of any real deployment.
CHAPTER 45 OF 45 · Review and continue
45. Connected reading and authoritative sources
Continue with the Gold Standard of Wireless Communication for the subject map, the Gold Standard of Education for the transfer framework and the How X Works hub for connected mechanisms. Relevant specialist routes include deadline delivery and burst loss, payload length and goodput, latency percentiles and antenna engineering.
For the technical definitions and examples, consult Cisco’s Signal-to-Noise Ratio and Wireless Signal Strength and WLAN site survey guidelines. They provide vendor-specific operational context; use the documentation for the actual equipment when interpreting a deployed system.
Contents · Previous chapter · Return to Wireless Communication