[Example] is there the function like kspacefirstorder1D in matlab

[YeojunPark]

Hello,

I appreciate to your library.

I can’t find the function like ‘kspacefirstorder1D’ in matlab.

Is there ,in the library, the function analyzing one-dimension like kspacefirstorder1D?

Thanks

[djps]

No, there is not a 1D version, as there is there no one-dimensional c++ or cuda executable to build around.

You could try to use a 2D simulation with one dimension being one I suppose.

[elma16]

Hi,

I tried this suggestion

import matplotlib.pyplot as plt
import numpy as np
from kwave.data import Vector
from kwave.kgrid import kWaveGrid
from kwave.kmedium import kWaveMedium
from kwave.ksensor import kSensor
from kwave.ksource import kSource
from kwave.kspaceFirstOrder2D import kspaceFirstOrder2D
from kwave.kspaceFirstOrder3D import kspaceFirstOrder3D
from kwave.options.simulation_execution_options import SimulationExecutionOptions
from kwave.options.simulation_options import SimulationOptions

# create the computational grid
Nx = 1
N = Vector([Nx, 64])
d = Vector([1e-4, 1e-4])  # grid point spacing [m]
kgrid = kWaveGrid(N, d)

# define the properties of the propagation medium
medium = kWaveMedium(sound_speed=1500)  # [m/s]

p0 = np.zeros(N)
p0[:,32] = 1  # set the initial pressure at the center of the grid

source = kSource()
source.p0 = p0
# define a binary line sensor
sensor = kSensor()
sensor.mask = np.zeros(N)
sensor.mask[..., 0] = 1

# create the time array
kgrid.makeTime(medium.sound_speed)

# set the input arguments: force the PML to be outside the computational grid
simulation_options = SimulationOptions(
    save_to_disk=True,
    pml_inside=True,
    pml_size=0,
    smooth_p0=False,
)
execution_options = SimulationExecutionOptions(is_gpu_simulation=False)

print(f"Shape of p0: {p0.shape}")
print(f"Shape of sensor.mask: {sensor.mask.shape}")
print(f"Type of p0: {p0.dtype}")
print(f"Type of sensor.mask: {sensor.mask.dtype}")

sensor_data = kspaceFirstOrder2D(
    kgrid, source, sensor, medium, simulation_options, execution_options
)

when Nx = 128, this runs as normal. However when Nx = 1 (the 1D setting using a 2D C++ binaries), I get the following error

Shape of p0: (1, 64)
Shape of sensor.mask: (1, 64)
Type of p0: float64
Type of sensor.mask: float64
Traceback (most recent call last):
  File "/Users/elliottmacneil/python/msgb/examples/time-reversal/issue.py", line 49, in <module>
    sensor_data = kspaceFirstOrder2D(
                  ^^^^^^^^^^^^^^^^^^^
  File "/Users/elliottmacneil/python/msgb/.venv/lib/python3.11/site-packages/kwave/kspaceFirstOrder2D.py", line 297, in kspaceFirstOrder2D
    k_sim.input_checking("kspaceFirstOrder2D")
  File "/Users/elliottmacneil/python/msgb/.venv/lib/python3.11/site-packages/kwave/kWaveSimulation.py", line 514, in input_checking
    self.create_sensor_variables()
  File "/Users/elliottmacneil/python/msgb/.venv/lib/python3.11/site-packages/kwave/kWaveSimulation.py", line 1364, in create_sensor_variables
    self.sensor_mask_index = cast_to_type(self.sensor_mask_index, self.index_data_type)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<@beartype(kwave.utils.conversion.cast_to_type) at 0x108dd5300>", line 53, in cast_to_type
  File "/Users/elliottmacneil/python/msgb/.venv/lib/python3.11/site-packages/kwave/utils/conversion.py", line 77, in cast_to_type
    return data.astype(type_map[matlab_type])
                       ~~~~~~~~^^^^^^^^^^^^^
KeyError: 'uint8'

Have I just made a silly mistake somewhere, or does this suggest that there is an issue with using the 2D codes for 1D simulations?

I get a totally different issue when using the 3D codes to solve a 2D problem - that is

import matplotlib.pyplot as plt
import numpy as np
from kwave.data import Vector
from kwave.kgrid import kWaveGrid
from kwave.kmedium import kWaveMedium
from kwave.ksensor import kSensor
from kwave.ksource import kSource
from kwave.kspaceFirstOrder2D import kspaceFirstOrder2D
from kwave.kspaceFirstOrder3D import kspaceFirstOrder3D
from kwave.options.simulation_execution_options import SimulationExecutionOptions
from kwave.options.simulation_options import SimulationOptions

# create the computational grid
N = Vector([1, 64, 64])
d = Vector([1e-4, 1e-4, 1e-4])  # grid point spacing [m]
kgrid = kWaveGrid(N, d)

# define the properties of the propagation medium
medium = kWaveMedium(sound_speed=1500)  # [m/s]

p0 = np.zeros(N)
p0[:,32,32] = 1  # set the initial pressure at the center of the grid

source = kSource()
source.p0 = p0
# define a binary line sensor
sensor = kSensor()
sensor.mask = np.zeros(N)
sensor.mask[..., 0] = 1

# create the time array
kgrid.makeTime(medium.sound_speed)

# set the input arguments: force the PML to be outside the computational grid
simulation_options = SimulationOptions(
    save_to_disk=True,
    pml_inside=True,
    pml_size=0,
    smooth_p0=False,
)
execution_options = SimulationExecutionOptions(is_gpu_simulation=False)

print(f"Shape of p0: {p0.shape}")
print(f"Shape of sensor.mask: {sensor.mask.shape}")
print(f"Type of p0: {p0.dtype}")
print(f"Type of sensor.mask: {sensor.mask.dtype}")

sensor_data = kspaceFirstOrder3D(
    kgrid, source, sensor, medium, simulation_options, execution_options
)

gives me this error

Shape of p0: (1, 64, 64)
Shape of sensor.mask: (1, 64, 64)
Type of p0: float64
Type of sensor.mask: float64
Traceback (most recent call last):
  File "/Users/elliottmacneil/python/msgb/examples/time-reversal/issue.py", line 48, in <module>
    sensor_data = kspaceFirstOrder3D(
                  ^^^^^^^^^^^^^^^^^^^
  File "/Users/elliottmacneil/python/msgb/.venv/lib/python3.11/site-packages/kwave/kspaceFirstOrder3D.py", line 299, in kspaceFirstOrder3D
    k_sim.input_checking("kspaceFirstOrder3D")
  File "/Users/elliottmacneil/python/msgb/.venv/lib/python3.11/site-packages/kwave/kWaveSimulation.py", line 505, in input_checking
    self.check_sensor(k_dim)
  File "/Users/elliottmacneil/python/msgb/.venv/lib/python3.11/site-packages/kwave/kWaveSimulation.py", line 754, in check_sensor
    self.sensor.mask.shape[0] == kgrid_dim and num_dim2(self.sensor.mask) <= 2
AssertionError: Cartesian sensor.mask for a 3D simulation must be given as a 3 by N array.

[waltsims]

I've never personally tried this. @djps will be looking into this to try to help you!

[elma16]

Revisiting this problem for a second, I find that changing the number of gridpoints from 64 to, say, 1024 resolves the issue in the first block. If I had read the error message correctly - as far as I understand it, the issue is an insufficient number of grid points. The same error message can be seen if one constructs an 8 x 8 grid for kspaceFirstOrder2D or 4 x 4 x 4 grid for kspaceFirstOrder3D. I understand that this number of gridpoints for my example is clearly silly, but I think the same issue applies when one wants to use the 2D solver to solve a 1D problem.

[elma16]

Another (probably inconsequential) thing to add is the fact that reconstruction methods in 1D don't work due to the checking in ksource where if p_mask has any singleton dimensions, they are squished.