Acquisition Bandwidth (Receiver Bandwidth)

MRIninja Knowledge Base | MRI Parameter Deep Dive Version 1.0 — August 2026

1. Introduction and General Purpose

Receiver (acquisition) bandwidth is the range of frequencies sampled during the readout window along the frequency-encoding axis. It is one of the most consequential — and most inconsistently displayed across vendors — parameters on this platform, precisely because it sits at the intersection of three otherwise separate concerns: how much signal-to-noise the final image retains, how many slices can be packed into a single TR, and how severely chemical shift misregistration displaces fat relative to water. This page is the first of two dedicated children of the Chemical Shift — Theoretical Foundations anchor page, developing the "how much" lever announced there in full; its sibling, Fat Shift Direction, developing the "which way" lever, is the next planned addition to this cluster.

Three relationships are deliberately placed at the centre of this page, matching exactly the scope requested for it: how bandwidth is represented across the major vendor platforms (Section 3 — treated in unusual depth, since this is a genuine, practically consequential source of cross-vendor confusion), its relationship with signal (Section 8), its relationship with the number of slices that can be acquired within a TR (Section 5, directly extending the companion Number of Slices page's TR-ceiling framework), and its relationship with chemical shift artefact (Section 9, directly extending the companion Chemical Shift page's pixel-shift equation).

2. Physical Foundations

2.1 What Bandwidth Actually Is

During the frequency-encoding readout, the MR signal is digitised at a fixed sampling rate; bandwidth (BW) is the reciprocal of the interval between consecutive samples (the dwell time, Δt):

BW = 1 / Δt

The total readout duration for one line of k-space is the number of frequency-encoding samples multiplied by the dwell time — equivalently, the number of samples divided by bandwidth:

T_readout = N_freq × Δt = N_freq / BW

This single relationship is the physical root of everything on this page: a wider bandwidth means a shorter dwell time and therefore a shorter readout window — which reduces the range of frequencies each pixel effectively integrates over (improving spatial precision, including for chemical shift, Section 9) at the direct cost of admitting more noise per unit signal (Section 8), and shortens the time each slice occupies within a TR (Section 5).

2.2 A Worked Numeric Example

For a 256-pixel frequency-encoding matrix sampled with a dwell time of roughly 30 μs, total readout time is approximately 256 × 30 μs ≈ 7.7 ms, corresponding to a bandwidth of roughly 1/30 μs ≈ 33 kHz across the full matrix — which, divided across 256 pixels, gives approximately 130 Hz per pixel [1]. This is precisely the kind of calculation that differs in units and framing across vendor consoles (Section 3): the "33 kHz" figure is close to how a GE system would report this bandwidth, while "130 Hz/pixel" is close to how a Siemens or Canon system would report the identical underlying acquisition.

2.3 The Three Relationships This Page Develops

Bandwidth's three central relationships all follow directly from Section 2.1's single equation, but manifest very differently in practice: SNR scales inversely with the square root of bandwidth (Section 8) — a noise-physics relationship, independent of anatomy or sequence family; the number of slices that fit within a TR scales with how much bandwidth shortens each slice's readout contribution to the TR budget (Section 5) — a timing relationship, directly extending the companion Number of Slices page; and chemical shift pixel displacement scales inversely with bandwidth per pixel (Section 9) — a spatial-misregistration relationship, directly extending the companion Chemical Shift page.

3. Units, Terminology and Vendor Nomenclature — In Depth

This is deliberately the most detailed section on this page, because bandwidth is represented in three fundamentally different ways across the major vendors — not merely different labels for the same number, but different physical quantities entirely, a genuine source of cross-vendor protocol-translation error if not understood explicitly.

3.1 The Conversion That Matters: GE Total → Siemens/Canon Per-Pixel

Because GE's rBW is independent of matrix size while Siemens/Canon's Hz/Px is defined per matrix element, the two numbers cannot be compared directly without a conversion — and, critically, the conversion itself depends on the frequency-encoding matrix size, which is easy to overlook mid-translation:

Bandwidth (Hz/Px) = Total receiver bandwidth (Hz) / N_freq

Worked example [2]: a GE protocol specifying rBW = 50 kHz on a 256-pixel frequency-encoding matrix corresponds to 50,000 Hz / 256 ≈ 195 Hz/Px — the number that should actually be entered on a Siemens or Canon system to reproduce the same underlying acquisition. Critically, the same 50 kHz GE value on a different matrix size (say, 320 pixels) would correspond to a different per-pixel figure (≈156 Hz/Px) — the GE number alone, without knowing the matrix size, is not sufficient to translate the protocol correctly.

3.2 The Philips WFS Conversion

Recovering an equivalent Hz/pixel value from a Philips WFS setting requires passing through the chemical-shift equation developed in full on the companion Chemical Shift page (Section 2.2 there): since pixel shift = frequency difference / bandwidth per pixel, bandwidth per pixel = frequency difference / WFS. Using the field-strength-specific water-fat frequency difference (≈220–224 Hz at 1.5T, ≈440–448 Hz at 3T [companion Chemical Shift page; 6]), a Philips WFS setting of, for example, 1.0 pixel at 1.5T corresponds to approximately 220 Hz/pixel — directly translatable to a Siemens/Canon/GE-equivalent value using this relationship.

3.3 "Variable"/"Optimised" Bandwidth — a Fourth Naming Layer

A distinct, more advanced technique — applying a higher bandwidth to early, high-signal echoes and a progressively lower bandwidth to later, lower-signal echoes within a multi-echo readout, to preserve SNR where it is most needed while keeping early echo times short — carries yet another set of vendor-specific names, independently confirmed across a Siemens-published cross-vendor acronym comparison: GE and Fujifilm: "Variable Bandwidth (VB)"; Siemens: "Optimized Bandwidth"; Philips: "Optimized Water/Fat Shift"; Canon: "Matched Bandwidth" [3,7]. This is a genuinely different console feature from the single-value bandwidth setting developed throughout the rest of this page, and is flagged here specifically so it is not confused with it when reading vendor documentation.

4. Typical Value Ranges

5. Parameter Interaction Ecosystem

5.1 Parameter Relationships Matrix

6. Effects on Image Appearance

Narrow bandwidth produces visibly smoother, lower-noise images (the direct visual expression of the SNR relationship in Section 8) but with more conspicuous chemical shift banding and greater susceptibility to motion/flow-related blurring or ghosting. Wide bandwidth produces visibly grainier images at the direct benefit of tighter, less conspicuous chemical shift banding and reduced motion sensitivity. Neither is universally "better" — the correct choice depends entirely on which of these appearance trade-offs matters more for the specific clinical question (Section 13).

7. Effects on Acquisition Time

Bandwidth has no direct effect on the number of phase-encoding steps or excitations, and therefore no direct multiplicative effect on scan time in the way slice count or NEX do. Its acquisition-time relevance is entirely indirect, through two pathways already developed elsewhere on this page and its companion pages: a shorter readout window can permit a shorter minimum TR/TE in some sequence designs, and — more consequentially — a shorter per-slice readout contribution can allow more slices to fit within an unchanged TR (Section 5), avoiding the concatenation/package-splitting time penalty developed in full on the companion Number of Slices page.

8. Effects on Signal and Noise (SNR and CNR) — the Central Cost

Bandwidth's relationship with signal is governed by basic noise physics rather than by anatomy or sequence family: thermal (Johnson) noise power is proportional to bandwidth, but because signal amplitude itself is unaffected by the bandwidth choice, the resulting SNR relationship is:

SNR ∝ 1 / √BW

Worked example: doubling bandwidth from 195 Hz/Px to 390 Hz/Px reduces SNR by a factor of 1/√2 ≈ 0.71 — an approximately 30% SNR reduction for a full doubling of bandwidth, all else equal. This square-root (rather than linear) relationship is the physical reason moderate bandwidth increases (e.g. to manage chemical shift at a joint margin, Section 17 Example 2) are often an acceptable trade, while very large increases carry a correspondingly steeper cumulative SNR cost. CNR, being fundamentally an SNR-difference quantity between two tissues, is affected proportionally in the same direction, though its clinical significance depends on the baseline CNR margin for the specific diagnostic task.

9. Artefacts Associated with Bandwidth — Chemical Shift in Detail

This section develops, with concrete worked numbers, the relationship already established in equation form on the companion Chemical Shift page.

Beyond Type 1 misregistration, narrow bandwidth's longer readout window also modestly increases susceptibility to motion- and flow-related phase error accumulation during each line's acquisition (Section 5.1), a secondary artefact contribution distinct from, but often co-occurring with, chemical shift banding in practice.

10. Behaviour Across Sequence Families

Conventional SE/TSE: routine moderate bandwidth defaults (Section 4) balance SNR against acceptable chemical shift for most clinical applications.

Gradient Echo (GRE/FLASH): bandwidth choice interacts with minimum achievable TE, relevant both for T1-weighting fidelity and for susceptibility-sensitive applications where a very short TE is itself the clinical goal.

EPI (DWI, fMRI, DSC) — the axis-distinction worth restating explicitly: EPI's dominant geometric problem is phase-axis distortion from very low effective bandwidth along that axis (companion Fold-over Direction page), which is a fundamentally different bandwidth than the frequency-axis receiver bandwidth this page addresses. EPI's frequency-axis (readout) bandwidth is typically kept comparatively high specifically to keep each individual readout line brief, which correspondingly keeps frequency-axis chemical shift modest even though the sequence's phase-axis distortion problem remains severe — precisely the contrast already introduced on the companion Chemical Shift page (Section 10 there), restated here from the bandwidth-parameter side.

Multi-echo sequences using Variable/Optimised Bandwidth (Section 3.3): bandwidth is deliberately varied across the echo train — higher for early, high-signal echoes (keeping their TE short), progressively lower for later, lower-signal echoes (preserving SNR where the signal has already decayed) — a more sophisticated application of the same fundamental SNR/readout-duration trade-off developed in Section 8 [3,7].

11. Field Strength Behaviour

Field strength does not change the bandwidth-SNR or bandwidth-slice-count relationships in principle, but it directly changes the chemical-shift stakes of a given bandwidth choice (Section 9): because the water-fat frequency difference doubles from 1.5T to 3T, a bandwidth value that was an acceptable compromise at 1.5T requires deliberate re-evaluation — typically an increase — when a protocol is migrated to 3T, exactly as developed on the companion Chemical Shift page.

12. Vendor-Specific Implementation

See Section 3 for the full terminology and conversion treatment. The single most important vendor-specific practical point: never copy a raw bandwidth number across vendors without converting it — a GE kHz value must be divided by the frequency-encoding matrix size to become a Siemens/Canon-equivalent Hz/Px value (Section 3.1); a Philips WFS value must be passed through the chemical-shift equation to become an Hz/Px-equivalent value (Section 3.2). Treating any of these as directly interchangeable numbers is a documented, avoidable source of unintended protocol drift when translating a protocol between platforms.

13. Practical Optimisation Strategies

13.1 The Central Trade-off Framework

Every bandwidth decision is, at its core, a three-way negotiation between SNR (Section 8), chemical shift/artefact tolerance (Section 9), and slice-packing efficiency (Section 5) for the specific clinical question at hand — not a single "correct" default value to be applied uniformly.

13.2 When to Deliberately Increase Bandwidth

• Fat-water interfaces adjacent to a diagnostically critical structure (orbit/optic nerve, joint cartilage-fat margins) — trading some SNR for reduced misregistration risk (companion Chemical Shift page, Example 2).
• Protocols migrated to a higher field strength, to maintain equivalent chemical shift severity to the original protocol (Section 11, Section 17 Example 5).
• Protocols where an extra slice or two is needed within an unchanged TR — increasing bandwidth to shorten T_slice can avoid triggering a concatenation, at a smaller SNR cost than might be assumed given the square-root relationship (Section 8, Section 17 Example 3).
• Sequences already motion- or flow-artefact-prone, where the secondary readout-duration benefit (Section 9) compounds with the primary chemical-shift benefit.

13.3 When to Deliberately Decrease Bandwidth

• Inherently low-signal sequences or applications where SNR is the binding constraint and chemical shift risk is low (e.g. no critical fat-water interface within the anatomy of interest).
• Protocols with ample slice-count headroom already, where the slice-packing benefit of increased bandwidth would not be exploited anyway.

14. Parameter Extremes

14.1 Very Narrow Bandwidth

Maximises SNR at a substantial cost across every other dimension on this page: pronounced chemical shift, reduced slice-packing efficiency per TR, and increased motion/flow artefact sensitivity from the extended readout window — appropriate only where SNR is unambiguously the binding clinical constraint.

14.2 Very Wide Bandwidth

Minimises chemical shift and readout-duration-related artefact, and maximises slice-packing efficiency, at a substantial, square-root-scaled SNR cost (Section 8) — the operating point EPI-family sequences default toward for their frequency axis specifically because their phase axis already carries the dominant distortion burden (Section 10).

15. Common Optimisation Errors

16. MRI Technologist Pearls

Know your own platform's convention cold — whether you work in Hz/Px, total kHz, or WFS pixels shapes every intuition you have about "high" versus "low" bandwidth, and that intuition does not transfer directly to a different vendor's console without conversion (Section 3).

When a protocol needs "just one more slice" to avoid a concatenation, check bandwidth before reaching for a longer TR — a modest bandwidth increase can shorten T_slice enough to fit the extra slice at a smaller SNR cost than the square-root relationship might suggest at first glance (Section 5, Section 17 Example 3).

For any protocol crossing field strengths, bandwidth deserves the same explicit re-check as TE/TR — not because the physics changes, but because the chemical-shift stakes of an unchanged bandwidth value double (Section 11).

17. Real Clinical Examples

Example 1: Cross-Vendor Protocol Translation

Scenario: a validated GE protocol specifying rBW = 50 kHz on a 256-pixel frequency-encoding matrix needs to be reproduced on a newly installed Siemens system.

Calculation: 50,000 Hz / 256 ≈ 195 Hz/Px — the value to enter on the Siemens console to reproduce the identical underlying acquisition [2].

Lesson: the direct, worked illustration of Section 3.1; skipping this conversion (e.g. entering "50" directly into a Siemens Hz/Px field) would produce a dramatically different, unintended acquisition.

Example 2: Orbit / Optic Nerve — Bandwidth Increase for Chemical Shift

Scenario: axial T1 TSE orbit, orbital fat immediately adjacent to the optic nerve — the same scenario developed on the companion Chemical Shift page, here from the bandwidth-lever side specifically.

Decision: bandwidth increased from a routine ~220 Hz/Px default to ~400 Hz/Px, reducing chemical shift pixel displacement from roughly one pixel to well under half a pixel (Section 9), at an accepted SNR cost given the diagnostic priority of the nerve-fat margin.

Lesson: a deliberate, clinically-justified application of Section 13.2's first bullet.

Example 3: Fitting an Extra Slice Without a Concatenation

Scenario: a TSE protocol at TR = 4000 ms sits exactly at its slice-count ceiling (companion Number of Slices page); one additional slice is needed to complete coverage, which would ordinarily trigger a second concatenation and double scan time.

Decision: bandwidth increased modestly, shortening each slice's readout contribution (T_slice, Section 2.1) enough to raise N_max_per_package by one — avoiding the concatenation entirely, at a bandwidth-proportional SNR cost far smaller than the scan-time cost that would otherwise have been incurred.

Lesson: the direct, worked illustration of Section 5's slice-packing relationship, and a genuinely under-used practical lever compared to the more commonly reached-for options (adding a concatenation, or reducing coverage).

Example 4: Variable Bandwidth in a Multi-Echo Sequence

Scenario: a multi-echo T2-mapping or susceptibility-weighted sequence with several echoes of progressively decaying signal.

Technique: Variable/Optimised Bandwidth (Section 3.3) applies a higher bandwidth to the early, high-signal, short-TE echoes and progressively lowers it for later, lower-signal echoes — improving overall SNR efficiency across the echo train rather than applying a single compromise bandwidth uniformly to every echo [3,7].

Lesson: a more sophisticated, sequence-integrated application of the same fundamental SNR/readout-duration trade-off developed throughout this page.

Example 5: 1.5T-to-3T Protocol Migration

Scenario: the same knee-protocol migration scenario introduced on the companion Chemical Shift page (Example 4 there), developed here from the specific corrective action taken.

Correction: bandwidth deliberately doubled from its original 1.5T value when the protocol is validated on the new 3T system, restoring the chemical shift pixel displacement to its original, clinically-accepted magnitude (Section 9 table) rather than allowing it to silently double.

Lesson: the concrete bandwidth-side resolution of the error most commonly flagged on the companion Chemical Shift page.

18. Visual Educational Material

18.1 The Single Equation Underlying Everything on This Page

BW = 1 / dwell time
T_readout = N_freq / BW

WIDER bandwidth -> SHORTER dwell time -> SHORTER readout
  -> less noise integration time per pixel  -> LOWER SNR
  -> shorter T_slice                        -> MORE slices fit per TR
  -> tighter frequency-per-pixel resolution  -> LESS chemical shift

18.2 Cross-Vendor Conversion Cheat Sheet

GE (total kHz)  ->  Siemens/Canon (Hz/Px):
  Hz/Px = (GE kHz value x 1000) / N_freq (matrix size)
  Example: 50 kHz / 256 matrix = 195 Hz/Px

Philips (WFS, pixels)  ->  Hz/Px:
  Hz/Px = water-fat frequency difference (Hz) / WFS (pixels)
  Example at 1.5T: 220 Hz / 1.0 px = 220 Hz/Px
  Example at 3T:   440 Hz / 1.0 px = 440 Hz/Px

18.3 SNR vs. Bandwidth — the Square-Root Curve

Bandwidth (Hz/Px):     130    195    260    390    520
Relative SNR (approx): 1.00   0.82   0.71   0.58   0.50

Doubling bandwidth does NOT halve SNR -- it reduces SNR
by a factor of 1/sqrt(2) (~0.71), a gentler penalty than
a naive linear assumption would suggest.

19. Evidence Gaps and Ongoing Debate

Formal, quantitative departmental guidance on bandwidth selection by anatomical region and diagnostic priority is not standardised — the worked examples in Section 17 reflect sound clinical/physical reasoning rather than a published, validated threshold table, consistent with the equivalent evidence gaps already documented for related parameters elsewhere in this cluster.

The real-world frequency of avoidable cross-vendor protocol-translation error specifically attributable to bandwidth unit confusion (Section 3, Section 15) has not been systematically audited, despite the clear, mechanistically obvious risk the three incompatible vendor conventions create.

Optimal Variable/Optimised Bandwidth echo-train profiles (Section 3.3, Section 10) are largely vendor-proprietary implementation choices rather than the subject of a single, openly published, cross-vendor-validated optimisation framework.

20. Miscellaneous and Future Directions

This page completes the second of the three-part chemical-shift-management triad announced on the companion Chemical Shift — Theoretical Foundations page. The third and final piece, Fat Shift Direction — developing the "which way" lever in full (worked examples across anatomical regions, vendor implementation, and its interaction with fold-over direction) — is the next planned addition to this parameter cluster.

AI-assisted, anatomy-aware bandwidth recommendation, weighing the specific clinical priority (SNR vs. chemical shift vs. slice-packing need) automatically from the prescribed anatomy and sequence type, is a conceptually natural extension of the decision framework developed in Section 13, though not yet established clinical practice.

End of document — Acquisition Bandwidth (Receiver Bandwidth) — MRIninja v1.0 — August 2026

Parent page: MRI Parameters — Overview and Classification (9501)