Possible issue with plotting after defining units

[Christian Buhl Sørensen]

I started using hyperspy to work on hyperspectral camera data, and I'm really impressed.

My experimental setup acquires time-resolved images of a light pulse, and I can then do a FFT to see the images at discrete frequencies. That goes something like this:

s = hs.signals.Signal2D(cube)
s.plot()
\\\\

S = s.transpose().fft().transpose()
S.plot()

That works very well. 

When I want to introduce units, I assume that the scale parameter should be the stepsize. Eg. time step between samples or mm on acquisition plane. I define the units using:

s.axes_manager[0].scale_as_quantity = '{:f} ps'.format(𝝙t*1e12)

s.axes_manager[1].scale_as_quantity = '{:f} mm'.format(30/256)

s.axes_manager[2].scale_as_quantity = '{:f} mm'.format(30/255)

But here is the issue:

When I then plot the temporal data, I get a x-axis that has labels in the range [0; 801] (the first dimensionality of the time data), and not [0-16] as I expect. Furthermore, when I drag the cursor line, it moves only a fraction of the cursor displacement on the screen. I.e. when I drag the red cursor line across the entire plot, it only moves about 1/10 as the cursor. 

I assume I'm using the units wrong, but I have not been able to find any pointers on that. 

Cheers,

Christian

[Josh Taillon]

Hi Christian,

You're right that the 'scale' parameter should be the size of each pixel in the given dimension. I'm not exactly sure of the problem in your code, as I have never used our "scale_as_quantity" option. Perhaps there's an unknown bug there. Can you try setting the scale and units separately to see if it works?

i.e. something like:

s.axes_manager[0].scale = timestep*1e12

s.axes_manager[0].units = 'ps'

s.axes_manager[1].scale = 30/256

s.axes_manager[1].units = 'mm'

s.axes_manager[2].scale = 30/255

s.axes_manager[2].units = 'mm'

- Josh

(also, if you'd like more interactive assistance, you can post in the gitter chat room)

--
You received this message because you are subscribed to the Google Groups "hyperspy-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hyperspy-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Joshua Taillon

www.joshuataillon.com

[Christian Buhl Sørensen]

Hi Josh, 

Thank you for your very swift reply. It helped me significantly further. I'm now able to get it to work, properly, with a little tweak. I will document it here, and start an issue on the github, as there is still a problem that persists.

When I set the parameters as per your suggestion, the time and one of the axes scales correctly. The weird effect with the cursor movement is now negated, and I get the expected time-range. However, the axis

 s.axes_manager[1]

is still not scaled, and the aspect ratio of the otherwise square image is skewed. 

If I define 

s.axes_manager[0].scale = timestep*1e12

s.axes_manager[0].units = 'ps'

s.axes_manager[1].scale = 30/256.1

s.axes_manager[1].units = 'mm'

s.axes_manager[2].scale = 30/255

s.axes_manager[2].units = 'mm'

 *Adding the 0.1 to the /256

It works on both axis. 

I've checked the type(s.axes_manager[2].scale), and that shows float, so flawed typecasting is not the problem, when it fails. 

Cheers,

Christian