Last modified: 2025-05-04 12:29:23
< 2025-04-30 2025-05-04 >This evening I sat down to work out how to do the geometry so that the resting surface is tangent to the theoretical resting surface at the corner of the pallet, instead of at the centre of the resting surface.
I wrote this down:
The length of the resting surface (l) is inaccurate, it assumes that the restingsurfaceangleoffset is 0,
but since you always want that surface to have excess length, I think it's good enough.
And I started programming it in Cursor and Cursor was scarily good at suggesting correct auto-complete. And that's just the Cursor tab-complete model! That's not even Claude or whatever. Really impressive.
I think the next useful thing to do would be to give some way to export the actual coordinates of the vertices of the pallets, so that you can reproduce them in CAD.
Sweet, I asked Cursor to add SVG or DXF export of the pallet coordinates, it did both SVG and DXF, and got it almost
completely right on the first
try, just got the <script> imports the wrong way around so it tried to reference something before it was defined.
I've been thinking about sundials again.
If you made a normal horizontal sundial, but made the dial ring able to rotate to compensate for summer time and the equation of time, how bad would the maximum error get?
Obviously on an "equatorial" sundial the error would be 0, but because the angles between the hours are different lengths on a horizontal sundial, rotating the dial ring would introduce an error.
Using https://www.blocklayer.com/sundial I made this template for some random spot in the UK:
By overlaying and rotating in Gimp I found that 1 hour at 6am is about as much angle on the dial as 1h30 at noon.
So if I were to rotate my dial ring by "15 minutes at 6am", that would be equivalent to 22.5 minutes at noon.
The worst case would be a rotation of 75 minutes I think, which combnies the maximum correction for daylight saving with the maximum correction for equation of time. 75 minutes at 6am becomes 112.5 minutes at noon, or an error of 37.5 minutes.
75 minutes at noon would become 50 minutes at 6am, or an error of 25 minutes. If we wanted to minimise the error, we would move it by 75 minutes at some point in between, I think the optimum is an error of 15 minutes at both 6am and noon.
I think 15 minutes is too much to accept.
OK, another idea is we plot summer time on one half of the dial ring and winter time on the other half. Maybe have the inactive half covered up by the housing. To switch between summer and winter time you rotate by 180 degrees. And then to correct for equation of time you are only moving it by 15 mins. Then the maximum error is only about 4 minutes.
< 2025-04-30 2025-05-04 >