Top graph in belt velocity time series has incorrect font.

[skhnat]

I'm not sure why this is happening for both MATLAB (input vs. output and frequency response) plots. It's not in the code itself, but only when it prints to a PDF. If I save a PNG right from MATLAB, this is the plot:

Any suggestions on how to fix this?

[tvdbogert]

That's weird. It's the same in Fig 2 also.

Try to save it as EPS rather than PDF.

I'm not a fan of using the latex (times roman) font for Figures, I think Arial works better especially at small size.

The light blue and purple colors don't look so good to me, was there a reason for picking those?

[moorepants]

Yeh, maybe eps to pdf would work.

I chose the colors from this: http://stanford.edu/~mwaskom/software/seaborn/tutorial/color_palettes.html. This is the issue that the colors came from: https://github.com/csu-hmc/perturbed-data-paper/issues/22. I just want something other than the default matlab colormap, not the black/red that was the original, is from a good color palette, and is consistent among all figures. Any colors that meet that criteria are fine with me.

I like the font in the figures to match the font in the text. I'm not sure what font PeerJ recommends.

[moorepants]

I've followed the "match figure font to paper typeface" for quite a while. But here is something that says sans-serif may be better: http://academia.stackexchange.com/questions/17564/publishing-figures-serif-or-sans-serif-fonts

PeerJ has a figure section: https://peerj.com/about/author-instructions/#figures, only thing it says:

Uniform fonts and font sizes should be used for labels (letter sizing should be readable at ‘actual sized’ reproduction - we suggest a vertical height of 2 mm).

[moorepants]

I'm fine with making these figures just have black and grey lines, i.e. no color, or black lines with different line types. The color really doesn't add anything to these plots.

[tvdbogert]

I think we need the colors, but the purple looks ugly to me. I would try red instead. If we're discussing colors we must be almost done.

[moorepants]

Font is still an issue, so reopening this. We can fix it once we submit the actual article and have information on what the fonts should be from the journal.

[moorepants]

Can you also change the legend in the time series plots to something other than input/output? Commanded/Measured may be more informative.

[tvdbogert]

See if save as EPS will fix the font issue.

[moorepants]

I think that PeerJ will require either PNG or EPS for the final submission anyways, so we'll have to output one of those eventually.

[skhnat]

No, the font is still messed up when I save as EPS. It works if I save as PNG but only when I manually do it in the actual figure. If I try to save it through the script, the axes labels are jumbled.

[skhnat]

It has to do with the settings that I need to use for the Latex interpreter, I think. If I remove those commands and use MATLAB defaults, it prints nicely to PDF, EPS, or PNG.

[moorepants]

Just hold off on this for now, it isn't worth fussing over because we haven't found out what the journal requires. We can submit the pdf as is for peer review, and once accepted we can fiddle with these details. But, from my experience, Matlab gives poor control over this stuff. I've battled it for hours and hours in the past. Their stuff is just buggy.

[moorepants]

If you want an example of how much hacking is needed to get around all of matlab's deficiencies check out this file I wrote years back for creating figures in one of my papers:

https://github.com/moorepants/HumanControl/blob/master/create_ieee_paper_plots.m

Every single figure seems to require a work around to get things to produce correctly. And I still couldn't ever get it to produce the same output simultaneous in the figure window, the pdf, and a png. Bodeplots in matlab were especially difficult to modify. I try to do all my plots these days in either R or matplotlib because the control is just so much easier.

[skhnat]

I just checked and MATLAB only supports a few fonts when printing to PDF or EPS. Otherwise, it defaults to Courier or whatever it interprets is the best font.

[moorepants]

That's too bad. We may have to stick with arial or something.