boivinalex / permittivitycalc Goto Github PK

Trace plots appear to "resonate" with some versions of both lmfit and emcee.

No problem with emcee 3.0.2 and lmfit 9.14.

Combinations that have been tested and do not work:

emcee 3.0rc2 with lmfit 9.15, lmfit 9.15rc1, lmfit 1.0.0, lmfit 1.0.1
emcee 3.0rc2 and lmfit 9.14 fails to plot with settings that work for other versions.
emcee 3.0.2 and lmfit 9.15, lmfit 9.15rc1, lmfit 1.0.0, lmfit 1.0.1

Possible problem with the way lmfit handles model parameters after 0.9.14?

Seems to only affect second measurement?

Less weight should be given to frequency points with high uncertainty than those with low uncertainty in the iteration.

Print results when new AirlineData instance is created

When using: airline_name='custom' as input to: run_default() there is no prompt to input a length for the airline and an error occurs. Current fix is to input airline_name='' and then enter 'custom' when prompted, then entering the length at the next prompt.

Include a calculation of resonant frequencies in Airline class.

Currently only nominal values are used for the boundary effect correction. Uncertainties must be manually propagated.

File dialog box hangs on some computers (mac only?).

Maybe this will help:
https://scottwdonald.blogspot.ca/2015/05/python-3-open-file-dialog-window-in.html

Create function that calculates the frequency averaged dielectric constant and loss tangent from the frequency midpoints between resonant frequencies.

Currently confusing as to what is what and where. Needs to be better organized

Better automatic tic locations and grids for log plots

spacing, buffer, xlim, and ylim still need to be implemented. Currently they use default settings for 'lf' and 'lt' plot types.

_dims() function provides inner/outer airline dimensions in units of mm, whereas airline length from get_METAS_data() function is in units of cm.

Since most other lengths are in units of cm, _dims() units should probably be changed to cm.

Use the average measurement of 10 washers in the de-embedding step.

Mu, epsilon, and the predicted s-parameter calculations should all be separate functions. Currently they are repeated several times.

Currently only the nominal values are used for de-embedding. Uncertainties package does not currently have native complex number support and so uncertainties must be propagated manually.

Use the calculated difference from #51 to calculate new Type A uncertainties for #50

Make the value at which the bulk density of corrected to a parameter in AirlineData so it can be easily changed

Include in init

Currently an estimate is used for washer thickness and for adjusted volume calculation for de-embedding. Update using a good average value for washer thickness.

make_plot seems to produce odd looking tick locations when the frequency span is large (kHz - GHz)

Optimized fit currently fails to come to a reasonable fit, especially at low frequency.

Use the average of the forward and reverse permittivity calculations rather than the average s-params