# How precise must be the input velocity parameter ?¶

The input velocity (set through the parameter `pos_cov`

) must be given with a good precision. The error should not generally be larger than a spectrum channel (since the FWHM is around 1.5 channels)

Here’s some ways you can determine the size of a channel.

```
[1]:
```

```
from orcs.process import SpectralCube
# load spectral cube
cube = SpectralCube('/home/thomas/M31_SN3.merged.cm1.1.0.hdf5')
```

```
INFO| Data shape : (2048, 2064, 840)
INFO| Cube is in WAVENUMBER (cm-1)
INFO| Cube is CALIBRATED in wavenumber
```

## Compute the channel width (in cm-1)¶

```
[2]:
```

```
# you can simply get it from the axis of the cube (which is in cm-1)
print 'channel width (cm-1)', cube.params.base_axis[1] - cube.params.base_axis[0]
```

```
channel width (cm-1) 2.09831225348
```

```
[3]:
```

```
# you can relate the channel size to the line width as the FWHM of a sinc line is equal to 1.20671 x width
# and the width of a sinc line is equal to 1.25 the channel size by construction (because there are 25% more
# steps made than necessary on the left size of the ZPD to be able to compute the phase)
# the number of steps really contributing to the resolution are the steps made at the right of the ZPD which can be
# esaily calculated
print 'line width (cm-1)', cube.params.line_fwhm / 1.20671
print 'number of steps at the right of ZPD (contributing to the resolution)', cube.params.step_nb - cube.params.zpd_index
print 'ratio of the step number / the number of steps at the right of the ZPD', cube.params.step_nb / float(cube.params.step_nb - cube.params.zpd_index)
print 'channel width (cm-1)', cube.params.line_fwhm / 1.20671 / 1.25
```

```
line width (cm-1) 2.62289031685
number of steps at the right of ZPD (contributing to the resolution) 672
ratio of the step number / the number of steps at the right of the ZPD 1.25
channel width (cm-1) 2.09831225348
```

```
[4]:
```

```
# Finally, you can relate the line width with the resolution at the mean wavenumber of the cube
# and then compute the channel size
print 'Spectral resolution: ', cube.params.resolution
mean_cm1 = (cube.params.base_axis[-1] + cube.params.base_axis[0]) / 2
print 'Mean wavenumber', mean_cm1
line_fwhm = mean_cm1 / cube.params.resolution
print 'Line width (cm-1)', line_fwhm
print 'channel width (cm-1)', line_fwhm / 1.20671 / 1.25
```

```
Spectral resolution: 4733.11731899
Mean wavenumber 14980.9003337
Line width (cm-1) 3.16512338995
channel width (cm-1) 2.09834899186
```

## In terms of velocity¶

The channel width can be related to an uncertainty on the velocity of the lines at a given wavelength

```
[5]:
```

```
import orb.utils.spectrum
channel_width = cube.params.base_axis[1] - cube.params.base_axis[0]
mean_cm1 = (cube.params.base_axis[-1] + cube.params.base_axis[0]) / 2
print 'channel width in km/s at the mean wavenumber of the axis', orb.utils.spectrum.compute_radial_velocity(mean_cm1 + channel_width, mean_cm1, wavenumber=True)
```

```
channel width in km/s at the mean wavenumber of the axis -41.9877390098
```

```
[6]:
```

```
# it could be directly computed with the following formula
c = 3e5 # lightspeed in km/s
print 'channel width in km/s at the mean wavenumber of the axis', c * channel_width / mean_cm1
```

```
channel width in km/s at the mean wavenumber of the axis 42.0197492822
```

```
[ ]:
```

```
```