vendor.bruker

vendor.bruker

Functions¶

build_fid¶

vendor.bruker.build_fid(data, dims, pv_params)

Construct a strict xmris FID DataArray from an N-dimensional numpy array.

Expected Bruker Parameters in pv_params:¶

• PVM_SpecSWH (float): Spectral width in Hz. Used to calculate dwell time.

• PVM_RepetitionTime (float): TR in ms. Used to calculate the repetition coordinate.

• PVM_FrqRef (float): Reference Larmor frequency in MHz. (Required for to_ppm)

• PVM_FrqWorkPpm (float): Carrier chemical shift in ppm. (Required for to_ppm)

• groupDelay (float): Bruker specific FID delay. In ACQ_RxFilterInfo.

Parameters¶

Returns¶

estimate_group_delay¶

vendor.bruker.estimate_group_delay(
    da,
    dim=DIMS.time,
    *,
    search_range=None,
    header_hint=None,
    window=16.0,
    metric='acme',
    refine=True,
    return_profile=False,
)

Measure the true digital-filter group delay by minimizing residual phase.

The vendor header value (Bruker ACQ_RxFilterInfo/GRPDLY) can under-count the real receiver digital-filter group delay for some ParaVision/probe combinations, leaving a residual first-order (frequency-dependent) phase error after :func:remove_digital_filter — negligible near the carrier but large for peaks far from it. This estimator finds the delay that removes that residual.

An incorrect delay is mathematically a residual linear phase across the spectrum, so the correct delay is the one that, after removal, makes the spectrum maximally absorptive under a single global zero-order phase (φ0). The discriminating power comes from forbidding first-order phase (φ1): delay and φ1 are the same linear-phase degree of freedom, so any φ1 freedom would make every delay look equally good. This is the peak-agnostic generalization of a tied-phase model residual. argmax(|FID|) is deliberately not used — it lands on the filter’s ringing, not the true delay.

Parameters¶

Returns¶

Notes¶

Emits a UserWarning when the measured delay deviates strongly from the header (a likely under-counting header), or when the cost profile is ambiguous (several delays give a near-minimal residual — linear-phase aliasing).

remove_digital_filter¶

vendor.bruker.remove_digital_filter(
    da,
    group_delay='header',
    dim=DIMS.time,
    keep_length=True,
)

Remove the hardware digital filter group delay from Bruker FID data.

Bruker consoles use a cascade of digital FIR filters during analog-to-digital conversion. Because these filters calculate a moving average, they require time to “wake up”, introducing a causality delay at the start of the Free Induction Decay (FID). This manifests as a time-shift, effectively prepending the actual signal with a specific number of filter transient points (often appearing as a flatline or wavy noise).

If left uncorrected, this time shift causes a massive, rolling first-order phase error in the frequency-domain spectrum.

This function realigns the signal to t=0 by:

1. Truncating the integer portion of the delay.

2. Applying a first-order phase correction to exactly compensate for the remaining fractional sub-point delay.

Parameters¶

Returns¶

reshape_bruker_raw¶

vendor.bruker.reshape_bruker_raw(raw_data_1d, pv_params)

Reshape a flat Bruker rawdata.job0 array into a squeezed N-dimensional numpy array.

Bruker stores multi-dimensional data sequentially. This function parses the method parameters to determine the shape, filters out empty dimensions, and reshapes the data to match xmris conventions.

Expected Bruker Parameters in pv_params:¶

• PVM_SpecMatrix (int): Number of points in the FID.

• PVM_EncNReceivers (int): Number of receive channels (Default: 1).

• PVM_NAverages (int): Number of averages (Default: 1).

• PVM_NRepetitions (int): Number of repetitions (Default: 1).

Spatial encoding is currently unsupported, we assume:¶

n_slices = 1 # number of slices / slabs n_ph1 = 1 # phase encoding direction A n_ph2 = 1 # phase encoding direction B

Parameters¶

Returns¶