A small uncertainty in a waveguide’s cutoff frequency can produce a much larger uncertainty in calculated group delay when the operating frequency is close to cutoff. The same numerical tolerance can matter little far from the boundary and greatly near it. Reporting only the nominal delay hides that change in sensitivity.
The short answer is to preserve the allowed frequency and cutoff values, check that the chosen mode propagates throughout that set, and calculate delay bounds from the complete model. We will do that for an ideal vacuum-filled rectangular conducting guide. Every dimension, interval and frequency used in the worked cases is fictional. This is a mathematical investigation, not a specification, measurement certificate or instruction for operating radio-frequency equipment.
eduKateSG · Waveguides
Find your next learning step
Choose the question closest to your task, or follow the original worked example in order. All teaching chapters remain open.
ROUTE 1 · CHAPTERS 1–6
Define the propagation model
Move from a specified mode to nominal delay and exact cutoff-interval bounds.
ROUTE 2 · CHAPTERS 7–12
Check sensitivity and joint inputs
Verify attainable limits, local sensitivity and the complete propagation domain.
ROUTE 3 · CHAPTERS 13–18
Preserve geometry and signal meaning
Distinguish open boundaries, linked frequencies and simultaneous bandwidth.
ROUTE 4 · CHAPTERS 19–24
Test the wider model and requirements
Separate modes, interfaces, primary derivation, inverse limits and averages.
ROUTE 5 · CHAPTERS 25–29
Repair, transfer and continue
Solve an independent guide example and explain the limits of the answer.
Chapter contents
Define the propagation model · Chapters 1–6
1. Define the delay being investigated
2. State the ideal guide and mode
3. Derive group delay from phase
Check sensitivity and joint inputs · Chapters 7–12
7. Show why the bounds are attainable
8. Compare the exact interval with a local derivative
9. Explain what changes near cutoff
10. Keep length uncertainty in the account
Preserve geometry and signal meaning · Chapters 13–18
13. Distinguish a finite bound from an open-ended one
14. Do not average across a failed domain
15. Convert uncertain geometry into cutoff carefully
16. Keep linked frequency information linked
Test the wider model and requirements · Chapters 19–24
19. Check the single-mode claim separately
20. Separate propagation phase from interface effects
21. Keep the phase-speed result in its proper role
22. Use the primary derivation as a boundary check
Repair, transfer and continue · Chapters 25–29
25. Repair a misleading uncertainty statement
26. Try a changed-context problem independently
27. Check the transfer without losing the domain
CHAPTER 1 OF 29 · Define the propagation model
1. Define the delay being investigated
Group delay in this lesson means the derivative of accumulated propagation phase with respect to angular frequency for a specified uniform guide and mode. For a section of length L, the model gives τ = L/vg, where vg is the mode’s group velocity. The quantity has units of time.
It is not automatically the arrival time of an arbitrary pulse peak or the latency of a complete communication system. Launch transitions, interfaces, detectors, processing and pulse reshaping are outside this initial model. Naming those boundaries lets the calculation remain useful without asking it to explain a system it does not describe.
Our question is narrower than the introductory phase-versus-group-speed distinction: how much can this specified propagation delay vary when cutoff is uncertain? A satisfactory answer must state whether every allowed case supports the same propagating mode. It must also distinguish a delay interval from a probability distribution, because the bounds we begin with carry no probabilities.
Assume an ideal, straight, uniformly vacuum-filled rectangular conducting guide, with wider transverse dimension a and smaller dimension b. We investigate its TE10 mode. In this model the cutoff frequency is fc = c/(2a), where c is the stipulated vacuum wave speed.
We will use c = 3.00 × 10⁸ m/s as an exact teaching constant. Real material dispersion, conductor loss, bends, interfaces and departures from the ideal geometry are not represented. The resulting uncertainty interval therefore concerns the declared model inputs, rather than every possible source of uncertainty in a real installation.
Above cutoff, the longitudinal propagation constant satisfies β = (2π/c)√(f² − fc²). Frequency f and cutoff fc must use the same unit. This is already a domain statement: the freely propagating branch used below requires f greater than fc. A number obtained by applying its delay expression below that boundary is not a slow-propagation answer to the original question.
Write angular frequency as ω = 2πf and cutoff angular frequency as ωc = 2πfc. The propagation constant is β = √(ω² − ωc²)/c. Differentiating with respect to ω gives dβ/dω = ω/[c√(ω² − ωc²)].
Multiplying by length produces τ = (L/c)ω/√(ω² − ωc²), or equivalently τ = (L/c)/√[1 − (fc/f)²]. The factors of 2π cancel when ordinary frequencies are used consistently. They should not be inserted into only one member of the ratio.
The quantity L/c is a useful baseline time. The remaining factor is dimensionless and exceeds one on the above-cutoff branch. This factorisation separates length scaling from the effect of operating relative to cutoff. A student can check units before evaluating a square root, then explain why a result smaller than L/c would contradict this particular model and positive-length assumptions.
Take L = 0.300 m, operating frequency f = 10.0 GHz and nominal cutoff fc = 9.0 GHz. The baseline time L/c is one nanosecond. The frequency ratio is 0.9, so the square-root factor is √0.19, approximately 0.43589.
Nominal group velocity is approximately 0.43589c, and nominal propagation delay is approximately 2.29416 ns. Those numbers are conditional on the nominal model inputs. No interval has yet been calculated, because a single nominal cutoff does not describe its possible variation.
An independent check uses the equivalent expression f/√(f² − fc²). With both frequencies expressed in GHz, this is 10/√19, giving the same dimensionless multiplier. The GHz unit cancels inside that ratio; it would not cancel if one number were left in hertz and the other in gigahertz. Equivalent representations are useful checks when their units and assumptions remain aligned.
CHAPTER 5 OF 29 · Define the propagation model
5. Replace the nominal cutoff with an interval
Now stipulate that the actual cutoff lies anywhere from 8.8 to 9.2 GHz, while L and f remain fixed at the preceding values. These are deterministic permitted values. We have not said that the endpoints are equally likely, that the interval is a standard uncertainty or that every interior value occurs with a particular distribution.
Every allowed cutoff is below the operating frequency of 10 GHz, so this entire interval remains on the chosen propagating branch. That check belongs before the bound calculation. It prevents an interval operation from mixing fundamentally different cases and then presenting one apparently ordinary time range.
The next question is whether delay increases or decreases with cutoff. We can determine that directly from the formula: for fixed positive L and f, increasing fc shrinks the positive denominator and increases τ. Therefore the smallest delay occurs at the smallest cutoff, and the largest occurs at the largest cutoff. This monotonic relationship makes the exact interval calculation simple.
At fc = 8.8 GHz, the square-root factor is √0.2256. The delay is one nanosecond divided by that factor, approximately 2.10538 ns. At fc = 9.2 GHz, the factor is √0.1536 and the delay is approximately 2.55155 ns.
Thus the allowed delay range is approximately 2.10538 to 2.55155 ns under the specified model and interval. The nominal value 2.29416 ns lies within it, but not at the arithmetic midpoint. The nonlinear transformation makes the deviations around nominal unequal.
Subtracting the nominal result gives approximately −0.18878 ns on the lower side and +0.25739 ns on the upper side. Replacing those with an unexplained symmetric error bar would change the information. A symmetric enclosing interval can be constructed if needed, but it should be labelled as an enclosure rather than the exact transformed range. Keep the original input interval available so another reader can reproduce the choice.
| Cutoff in GHz | Group delay in ns | Role in the stipulated cutoff interval |
|---|---|---|
| 8.8 | 2.10538 | Minimum delay |
| 9.0 | 2.29416 | Nominal case |
| 9.2 | 2.55155 | Maximum delay |
CHAPTER 7 OF 29 · Check sensitivity and joint inputs
7. Show why the bounds are attainable
A mathematical enclosure says that all allowed outcomes lie inside it. An exact range additionally says that its claimed extremes are possible within the allowed input set. Here the endpoint cutoffs are explicitly included, so they supply witnesses for both delay extremes.
The delay function is continuous throughout the stipulated interval, because the interval stays strictly below f. Consequently intermediate delay values are also attained as cutoff varies between its endpoints. There is no hidden gap in the output range for this one-parameter model.
That conclusion depends on the permitted set. If the cutoff could take only two discrete calibrated values, the same outer bounds would enclose the outcomes, but most intermediate delays would not be possible. If some combinations of several uncertain inputs were forbidden, independent endpoint choices might not supply valid witnesses. A good uncertainty explanation therefore records both the numerical limits and the structure of the set that makes those limits meaningful.
CHAPTER 8 OF 29 · Check sensitivity and joint inputs
8. Compare the exact interval with a local derivative
Differentiating τ with respect to fc at fixed L and f gives a positive derivative. Its relative form is especially clear: ∂ln τ/∂ln fc = r²/(1 − r²), with r = fc/f. This is a local sensitivity coefficient for fractional changes around a specified above-cutoff state.
At r = 0.9, the coefficient is 0.81/0.19, approximately 4.26316. A sufficiently small one-percent increase in cutoff therefore produces approximately a 4.26-percent increase in delay to first order. The word approximately belongs to the linearisation, not to the exact interval result derived earlier.
The coefficient is not a probability or a measure of how likely the cutoff is to change. It tells us how a local input perturbation affects the output under the model. To use it in an uncertainty calculation, we still need the input information and an appropriate interpretation of that information. A derivative alone does not turn a permitted interval into a statistical distribution.
CHAPTER 9 OF 29 · Check sensitivity and joint inputs
9. Explain what changes near cutoff
The sensitivity coefficient r²/(1 − r²) grows without bound as r approaches one from below. At r = 0.5 it is one third. At r = 0.9 it is about 4.26. At r = 0.99 it is about 49.25.
Those comparisons show why an identical fractional cutoff uncertainty can have very different effects at different operating points. The formula has not become arithmetically incorrect near cutoff; the input-output relationship has become more sensitive. Printing more output digits does not repair uncertain inputs.
This is also a reason to distinguish a mathematically allowed point from a practically adequate model. Loss, imperfect geometry and pulse effects excluded from our ideal calculation may require additional attention in an actual near-cutoff system. We do not quantify those effects here or turn a teaching sensitivity into an operating recommendation. The useful learner result is to notice when a nominal calculation sits close to a boundary that amplifies input variation.
CHAPTER 10 OF 29 · Check sensitivity and joint inputs
10. Keep length uncertainty in the account
Delay is proportional to L when the frequency ratio is fixed. If L is permitted to range from 0.290 to 0.310 m, while cutoff independently ranges from 8.8 to 9.2 GHz and f remains 10 GHz, the minimum uses the shorter length and smaller cutoff. The maximum uses the longer length and larger cutoff.
The bounds are therefore 0.290/c divided by √0.2256, and 0.310/c divided by √0.1536. They are not found by inserting the midpoint length into both extremes. Each endpoint choice must correspond to the objective being bounded.
This calculation assumes the complete rectangular combination of permitted lengths and cutoffs: every pairing is allowed. If manufacturing, measurement or a stated geometry model links the quantities, that assumption needs review. An interval for each input separately does not always establish which joint combinations are possible. Preserve the relationship information before choosing extremes, especially when the same dimension or calibration contributes to more than one reported quantity.
CHAPTER 11 OF 29 · Check sensitivity and joint inputs
11. Allow operating frequency to vary as well
At fixed L and fc, delay decreases as f increases above cutoff. In the equivalent expression f/√(f² − fc²), differentiation with respect to f gives a negative derivative. Raising frequency moves the operating point farther from cutoff in this specific mode model.
Suppose f ranges from 9.9 to 10.1 GHz, cutoff ranges independently from 8.8 to 9.2 GHz, and length is fixed at 0.300 m. The whole set is propagating because the smallest f exceeds the largest fc. The minimum delay uses f = 10.1 and fc = 8.8; the maximum uses f = 9.9 and fc = 9.2.
Write those joint witnesses beside the resulting bounds. Choosing both minimum frequencies together would not give the maximum or minimum automatically: the two inputs act in opposite directions. The monotonicity argument tells us which corners of the allowed input box matter. It also tells us when a numerical search is unnecessary for this simple case.
CHAPTER 12 OF 29 · Check sensitivity and joint inputs
12. Check propagation over the whole input set
For independent intervals, a sufficient and, with included endpoints, necessary condition for every allowed pair to be above cutoff is fMin greater than fcMax. Checking only nominal f greater than nominal fc is weaker and can miss allowed nonpropagating cases.
Change the frequency interval to 9.1 through 10.1 GHz while retaining cutoff 8.8 through 9.2. The nominal values might still suggest an above-cutoff operating point, yet the allowed pair f = 9.1 and fc = 9.2 lies below cutoff. The earlier real propagation-delay formula cannot be applied to every allowed state.
This does not mean that no electromagnetic field can exist in a finite below-cutoff region. It means that the chosen freely propagating branch no longer describes the complete set. Evanescent behaviour and finite interfaces require their own analysis. Report the failed domain guarantee rather than replacing an imaginary square root with zero, its absolute value or another convenient real number.
CHAPTER 13 OF 29 · Preserve geometry and signal meaning
13. Distinguish a finite bound from an open-ended one
Even when every case under consideration is strictly above cutoff, a finite upper delay bound need not exist if the allowed cases can approach cutoff arbitrarily closely. Consider a fixed f and allowed fc values below f but with no positive minimum separation from it.
Each individual allowed case has a finite model delay, but the collection has delays larger than any chosen finite limit. The supremum is unbounded. That is different from a closed interval with fcMax strictly less than f, where the endpoint calculation gives a finite maximum.
A useful example is fc = f(1 − ε), with ε positive and permitted to approach zero. The square-root term becomes √(2ε − ε²), which tends to zero. The model delay therefore increases without bound. This is a mathematical statement about an ideal model and its allowed inputs, not a prediction of infinite latency in a real instrument. Keep that distinction explicit when interpreting an unstable result near a model boundary.
CHAPTER 14 OF 29 · Preserve geometry and signal meaning
14. Do not average across a failed domain
Suppose some allowed cases propagate and others do not. Averaging the input frequencies first and calculating one nominal above-cutoff delay does not establish that the whole uncertain system has that delay. The operation has discarded the distinction between the two kinds of cases.
Nor does assigning zero delay to the nonpropagating cases create a meaningful expected arrival time. A nonarrival or a different transmission regime requires a definition of the outcome being averaged. Conditional delay among transmitted cases, transmission probability and a specified system-level timeout are different quantities.
This lesson uses deterministic bounds, so it does not supply probabilities for those events. If a later analysis introduces a distribution, it must also specify how propagation, observation and nonarrival are represented. Averages become useful only after the outcome and conditioning are clear. The domain check is therefore part of defining the question, not a technical nuisance to be repaired after the calculation.
CHAPTER 15 OF 29 · Preserve geometry and signal meaning
15. Convert uncertain geometry into cutoff carefully
Because fc = c/(2a), a smaller wider dimension produces a larger cutoff. If a lies in an interval of positive widths, the cutoff interval reverses the endpoint order: its minimum uses aMax and its maximum uses aMin.
For a fictional nominal width a0 with a one-percent symmetric tolerance, a lies between 0.99a0 and 1.01a0. The corresponding cutoff factors relative to nominal are 1/1.01 and 1/0.99. They are approximately 0.99010 and 1.01010, not exactly symmetric about one.
A first-order approximation can replace these with roughly one-percent cutoff changes when appropriate, but it should not be confused with the exact reciprocal transformation. Close to cutoff, the next nonlinear delay transformation can amplify the difference. Carrying the original width interval through both steps avoids silently imposing symmetry twice. The student should be able to trace each endpoint back to a permitted physical dimension in the stipulated geometry.
CHAPTER 16 OF 29 · Preserve geometry and signal meaning
16. Keep linked frequency information linked
Now consider a different allowed set: f = 10s GHz and fc = 9s GHz for a common positive scale factor s ranging from 0.98 to 1.02. Length and c remain fixed. Every permitted pair has exactly the same ratio fc/f = 0.9, so every pair gives the same model delay.
The separate marginal intervals are f from 9.8 to 10.2 and fc from 8.82 to 9.18 GHz. Treating those intervals as independently combinable would introduce pairs that the stated shared-scale model forbids. It would produce a wider delay enclosure for a larger input set.
This is not a claim that actual cutoff and source frequency always share one calibration factor. It is an original example showing why dependence information matters. A common frequency-axis scale can behave differently from independent errors or an independently uncertain geometric cutoff. Write the permitted relationships explicitly rather than substituting a vague phrase such as correlated uncertainty for the actual combinations the model allows.
CHAPTER 17 OF 29 · Preserve geometry and signal meaning
17. Compare two different kinds of frequency range
An uncertain operating frequency and a pulse’s spectral bandwidth are different descriptions. The first may mean that one narrowband operating state is not precisely known. The second means that several frequency components are present in the same signal.
Our interval calculation can bound τ(f) across a band that remains above cutoff, but that range is then a variation of group delay with frequency. It is not automatically the uncertainty interval for one unknown carrier. The mathematical function is the same while the interpretation changes.
A complete pulse calculation also needs the input spectrum, phase, amplitude response and observation definition. Two signals sharing the same frequency support can have different temporal profiles. Preserve that missing information instead of calling the difference between maximum and minimum group delays the exact output pulse width. The bound can flag dispersion sensitivity without pretending to reconstruct a waveform from an interval alone.
CHAPTER 18 OF 29 · Preserve geometry and signal meaning
18. Calculate a bandwise delay spread under the model
For a fixed guide and mode with a frequency band entirely above cutoff, τ decreases across the band. The difference between its low-frequency and high-frequency values therefore gives the range of model group delays over that band.
Use the nominal fc = 9 GHz and L = 0.300 m with a band from 9.9 to 10.1 GHz. The two endpoint values follow directly from the same formula. Their difference is positive because the low-frequency component lies closer to cutoff.
This is a group-delay range, not an assertion that two sharply localised packets emerge separated by exactly that amount. To make such a packet statement, the input and propagation approximations must support it. The useful classroom exercise is to compare a narrower and wider band while holding the guide fixed, then explain why the low-frequency side can dominate the variation near cutoff. A numerical result should retain the output quantity named in the question.
CHAPTER 19 OF 29 · Test the wider model and requirements
19. Check the single-mode claim separately
A bound for one mode does not prove that only that mode is present. In an ideal rectangular guide, other modes have their own cutoff conditions. Geometry can allow additional propagating patterns, and excitation determines how the allowed patterns are populated.
For our example, stipulate b = 0.005 m and widths corresponding to the TE10 cutoff interval 8.8 through 9.2 GHz. The TE20 cutoff is twice TE10, so it is at least 17.6 GHz. The TE01 cutoff is c/(2b) = 30 GHz. The other candidate cutoffs of this ideal geometry do not introduce a lower competing propagating mode in the 9.9–10.1 GHz band.
That establishes the relevant ideal single-mode window for the stated dimensions and band; it does not establish the performance of arbitrary transitions or a real manufactured guide. Keep mode availability, launch and loss as separate parts of the explanation. A correct TE10 delay calculation should not silently become a guarantee about a multimode measurement.
CHAPTER 20 OF 29 · Test the wider model and requirements
20. Separate propagation phase from interface effects
The phase accumulated through a uniform ideal section is βL. A measured phase difference between external reference planes can also include transitions, connectors, reflections and instrumentation. Differentiating that complete measured phase may give a different group-delay account from L dβ/dω alone.
This matters when comparing two lengths. If the interfaces are genuinely unchanged and their responses are known or eliminated appropriately, a length-dependent difference can be informative. If changing the specimen also changes the interfaces or produces coherent multiple reflections, a simple subtraction needs further justification.
We do not prescribe a measurement method here. The paper-model lesson is to draw the reference planes and name which phase the derivative concerns. A calculation that is exact for the propagation section can disagree with a complete-system measurement without either number being an arithmetic mistake. The discrepancy becomes useful when the model and observation boundaries are clear enough to identify the missing contribution.
CHAPTER 21 OF 29 · Test the wider model and requirements
21. Keep the phase-speed result in its proper role
For this ideal mode, phase velocity is c divided by √[1 − (fc/f)²], while group velocity is c multiplied by the same square root. Their product is c². The phase pattern can therefore exceed c even though the group velocity in this model is below c.
That relationship does not establish faster-than-light information transfer. A phase crest is not a separately labelled message travelling through the guide, and a narrowband group-delay model does not by itself describe every possible signal front or strongly reshaped pulse.
The Waveguides root already introduces that distinction. Our new calculation uses the group-delay expression to examine uncertainty and domain boundaries. Returning to the definition prevents a large phase speed or a very large near-cutoff group delay from being treated as a paradox. The two quantities answer different mathematical questions, and neither should be substituted for a complete communication latency without the additional system account.
CHAPTER 22 OF 29 · Test the wider model and requirements
22. Use the primary derivation as a boundary check
Richard Fitzpatrick’s University of Texas lecture derives waveguide mode conditions and the above-cutoff phase and group velocities for the ideal vacuum-filled conducting structure. It also distinguishes below-cutoff behaviour and the different TEM situation when an appropriate central conductor is present.
Our intervals, linked-scale example and transfer calculations are original teaching constructions. They use the stated ideal formula; they are not measurements or examples taken from that lecture. The source supplies the physical model context, while the bounds follow from the declared permitted inputs and the monotonicity reasoning developed here.
Read the university waveguide derivation with those assumptions attached. Source route checked on 24 October 2026. A source equation becomes transferable when the learner can identify its geometry, material, mode and domain. The same formula should not be copied into an optical fibre, lossy transition or different guide merely because the general word waveguide appears in both descriptions.
CHAPTER 23 OF 29 · Test the wider model and requirements
23. Turn a delay limit into an input requirement
A specified upper delay limit can be translated back into a permitted cutoff ratio within this model. Let T be the proposed limit and τ0 = L/c. For positive length, a necessary condition is T at least τ0. Above cutoff, the inequality τ0/√(1 − r²) ≤ T gives r ≤ √[1 − (τ0/T)²], with r = fc/f nonnegative.
For our one-nanosecond baseline, a limit of 2.5 ns requires r no larger than √0.84. At fixed f = 10 GHz, the corresponding cutoff ceiling is approximately 9.16515 GHz. The permitted cutoff of 9.2 GHz exceeds it, so the original interval does not guarantee that delay limit.
This is a conditional mathematical requirement, not a recommendation for an actual guide. The guide model, other modes, loss and system reference planes still need their separate accounts. The value of the inverse calculation is that it exposes which input condition the output claim requires. It is more informative than merely saying that the nominal 2.29416 ns happens to be below 2.5. A nominal pass and a guarantee across the allowed set are different statements, and the extremal witness identifies exactly why the latter fails.
CHAPTER 24 OF 29 · Test the wider model and requirements
24. Keep averaging separate from interval transformation
Now add a probability model deliberately: suppose cutoff takes only 8.8 or 9.2 GHz, each with probability one half. This is a new assumption, distinct from the original continuously permitted interval. The mean cutoff is 9 GHz, but the mean delay is the average of the two endpoint delays, approximately 2.32847 ns.
That is larger than the delay of approximately 2.29416 ns evaluated at the mean cutoff. The two operations do not commute because the transformation is nonlinear. In fact, at fixed positive L and f, the delay as a function of cutoff is convex on the nonnegative above-cutoff domain. A mean input can therefore conceal a larger mean transformed output.
Neither result replaces the interval bounds. The two-point probability model still has the same extreme cutoffs, but unlike the continuous interval it has no interior outcome values. A report should say whether it is providing a nominal value, a probabilistic average or a deterministic range. Those summaries can all be useful while answering different questions. Adding probabilities explicitly makes an average meaningful; it does not retrospectively prove that the original interval represented that distribution or that a mean delay determines every case.
CHAPTER 25 OF 29 · Repair, transfer and continue
25. Repair a misleading uncertainty statement
Consider the statement: cutoff is uncertain by about two percent, so the group delay is uncertain by about two percent. That conclusion ignores the nonlinear frequency ratio. At the nominal ratio 0.9, the local relative sensitivity to cutoff is about 4.26, and the exact interval calculation gives asymmetric deviations.
A useful repair begins by recording the actual interval, not only a loose percentage. Check whether all permitted cutoffs remain below the operating frequency. Then apply the monotonic formula or an explicitly justified local approximation and label the result according to its meaning.
If the interval reaches the cutoff boundary, the repair may be a failed propagation guarantee or the absence of a finite upper bound, rather than a larger symmetric error bar. Explain that change in the kind of answer. A strong correction restores the domain and input-set account, so the learner can recognise the same issue in a new numerical setting rather than merely memorise that the original percentage was too small.
CHAPTER 26 OF 29 · Repair, transfer and continue
26. Try a changed-context problem independently
Use a new fictional ideal guide with L = 0.600 m and fixed operating frequency f = 12 GHz. Its TE10 cutoff is permitted to lie from 6 to 9 GHz. Keep the teaching value of c unchanged. Determine the smallest and largest group delays and identify which cutoff gives each extreme.
Then change only the upper cutoff limit to 12 GHz, retaining the included endpoint. Decide whether the complete set still supports the original propagating-delay calculation and whether a finite upper bound remains available as the allowed cutoff approaches the operating frequency from below.
Finally compare a linked family f = 12s and fc = 8s for positive s. State whether the common scale changes the delay when length is fixed. Work from the ratio and domain rather than copying the earlier nanosecond values. The transfer tests three separate skills: scaling length, preserving monotonic endpoint choices and recognising when a changed allowed set changes the kind of conclusion.
CHAPTER 27 OF 29 · Repair, transfer and continue
27. Check the transfer without losing the domain
The new baseline L/c is two nanoseconds. At fc = 6 and f = 12 GHz, the ratio is one half, giving delay 2/√0.75 = 4/√3 ns, approximately 2.30940 ns. At fc = 9, the ratio is three quarters, giving delay 2/√(7/16) = 8/√7 ns, approximately 3.02372 ns.
The smaller cutoff gives the minimum and the larger gives the maximum. Both endpoint states are allowed and above cutoff, so they are witnesses for the exact finite range under the stated model.
With an included upper cutoff of 12 GHz, the full set is not strictly above cutoff. Even restricting attention to the subcases below 12 leaves delays that grow without bound as cutoff approaches 12. The linked family instead keeps fc/f = two thirds, producing a fixed delay of 6/√5 ns for the same length. The separate answers follow from different permitted sets, not from a contradiction in the delay formula.
CHAPTER 28 OF 29 · Repair, transfer and continue
28. Choose a learning route from the unresolved step
For repair, practise converting dimensions and frequencies before evaluating the nominal formula. For stabilisation, reconstruct the endpoint-bound argument and identify its permitted witnesses. For extension, compare independent intervals with a linked family, or separate an unknown carrier frequency from a simultaneous signal bandwidth.
Return to the Waveguides root for modes, cutoff and the guide’s wider system boundary. Follow the Photonics ring-resonator lesson for a different group-related observable: resonance spacing depends on the derivative of accumulated phase, rather than simply a phase index substituted into every formula.
For the separate collection question, the modal-aperture lesson shows why equal scalar modal weights can give different restricted readings. A propagation-delay bound and a spatial collection model describe different observables.
Continue through Learning That Transfers into Life and How X Works. A useful final note names the mode, permitted inputs, domain check, extremal cases and interpretation of the bound. The learner should be able to explain why a nominal answer is insufficient when the uncertainty set approaches a propagation boundary.
CHAPTER 29 OF 29 · Repair, transfer and continue
29. Questions about bounds and propagation
Does a cutoff interval automatically give a symmetric delay interval?
No. The nonlinear transformation can produce unequal deviations around a nominal delay. Calculate the endpoints or state explicitly that a symmetric interval is an approximation or an enclosure.
Does every above-cutoff case guarantee a finite common upper bound?
No. If allowed cases approach cutoff arbitrarily closely, each delay can be finite while the collection has no finite upper bound. A positive separation from the boundary matters.
Are these deterministic bounds standard uncertainties?
No probabilities were assigned to the permitted inputs. Standard deviations, confidence statements and probability coverage require additional definitions and evidence. A bounded allowed set alone does not supply them.
Is the group-delay range the exact pulse width?
No. A waveform calculation needs the input spectrum, phase, amplitude response and observation definition. Bandwise variation of group delay is useful information, but it does not uniquely reconstruct the output pulse.