Much of the discussions regarding and plans for building kites on the internet are contained within decades-old discussion forums and Geocities-era webpages, many of which are disappearing, although the Internet Archive has fortunately preserved many such resources, as has the Kite Plan Base. Rokkaku kites were frequently recommended for kite aerial photography,² and I was able to find several plans on which to base my kite design, using two designs, in particular.³

While the designs call for 3/4 oz ripstop nylon spinnaker cloth, which is sealed to make it low porosity, other fabrics with similar strength and low porosity can also be used. I ended up using a 1.1 oz silicone-coated ripstop polyester as I could get it custom printed, avoiding the need to measure and allowing for a custom design on the kite without the manual effort required for appliqué. Due to the archaic units frequently used with fabric, these two fabrics are actually of similar weight. Sail cloth is measured in “sailmaker’s ounces,” which is the weight of one “sailmaker’s yard” of fabric, which is 36″ by 28.5″; the polyester is measured in more-typical fabric ounces, so the weight of a square yard of fabric. Both quote the weight of the fabric itself prior to sealing the fabric to reduce the porosity. After coating and conversion to standard metric units, one typical variant of 3/4 oz spinnaker cloth, Fibermax 44, is 44 g/m², while the polyester I used is 42 g/m², or more or less the same.

For the spars, I used Easton Expedition-series aluminum 7075 backpacking-tent poles, using stiffer 0.355″ OD (0.025″ wall thickness) tubing for the spine and lighter 0.344″ OD (0.019″ wall thickness) tubing for the spreaders. The spine consists of four 14.5″ sections of tubing, three of which contain an insert tube on one end for connecting the sections together, yielding a 58″ spar. Each of the two spreads consists of two 14.5″ sections, each with an insert tube on one end, and a 16″ section without insert tubes in the center,⁴ yielding two 45″ spars. As the tube sections do not come in these lengths, they were cut down to size with a pipe cutter, and sections with insert tubes were ordered with the tubes preinstalled. The spars are assembled with aluminum dome tip tie offs on the ends and connected with 3/32″ elastic shock cord, which is rated down to −40°C (equal to −40°F).

Using Inkscape, I scaled the design to fit the kite and fabric for a four-yard-long fuzzy tail onto a two-yard section of 60″-wide custom-printed ripstop polyester, including a design based on the True South flag; extra fabric was used to sew an 18″ by 4″ drawstring bag for the kite. The corners where the spars attach, the locations where the bridle passes through the kite to the spreaders, and the locations where the spreaders cross the spine were reinforced with an additional piece of the ripstop polyester sewn on, after which the kite edges were folded over twice and sewn with a zig-zag stitch.

Brass grommets were installed for the bridle pass-throughs, and polyester rectangles were used to create sleeves where the spreaders cross the spine to secure the spine to the kite and the spreaders over it.⁵

To secure the bottom of the spine, a pocket was created using a folded-over section of 1″ polyester webbing and sewn to the kite, with a loop hanging off the end of the kite for attaching a tail.

For the top of the spine, a sleeve was created by sewing the edges of two sections of webbing on top of each other, with one side of hook-and-loop fastener on top, and sewn to the kite; the bottom section extends past the top of the kite and has the other side of the hook-and-loop fastener sewn to it, allowing it to be folded over the end of the spine and pulled tight, allowing the hook-and-loop fastener to be used to tension the kite fabric along the spine, which is installed in the pocket on the bottom end of the kite and slipped through the sleeves where spreaders cross and the sleeve on the top.

Similar sleeves are used for the ends of the spars, except that the hook-and-loop fastener is not installed on the top of the sleeve. Instead, additional sleeves, not sewn to the kite, were created by folding over sections of webbing, and the hook-and-loop fastener was sewn to these sleeves, which also had the end opposite the sleeve folded over and sewn to form a loop. The spreaders are slipped through the webbing sleeves on the ends, as well as the sleeves where they cross the spine, and the standalone sleeves are then slipped onto the ends. Braided line is used to attach the loops on the pairs of standalone sleeves to tension the spreaders, and the hook-and-loop fastener is used to attach the sleeves on the kite to the sleeves on the spreaders, allowing the kite fabric to be tensioned independently of the spreaders. The braided line is attached to one of the standalone sleeves with a figure-eight follow-through knot, with a figure-eight on-a-bight knot on the other end of the line; an alpine butterfly knot is placed along the line—which can be moved to adjust the spreader bow—and the end of the line is looped through the second standalone sleeve and attached to the alpine butterfly with a small quick link.⁶

The bridle is constructed from three sections of braided line. Each spreader has a line with a figure-eight on-a-bight knot on both ends of the line and one in the middle, with Prusik hitches used to attach the end loops to the spreader. The small loops in the middle of each spreader line are attached together with the third, main bridle line using a figure-eight follow-through knot on each end. An additional knot is placed along this section of line for attaching the main kite line using a quick link. The 5″-wide, four-yard-long fuzzy tail was constructed by sewing 3/8″ grosgrain ribbon down the center of the tail fabric using a zig-zag stitch and folding the end of the ribbon over on each end to make a loop. I then cut slits at 1″ intervals to make it fuzzy and attached it to the loop on the bottom of the kite using a small loop of braided line and lark’s head knots. It is approximately 1.0 m from the spreader attachment point to the center of the spreader line, and the line that connects the two spreader lines is approximately 2.5 m long. The completed kite weights approximately 370 g.

For the kite reel, I was again concerned about the performance of the typical mostly-plastic construction in the extreme cold. Thus, I opted to design a reel using a combination of a large 3D-printed ball bearing⁷ and laser-cut aluminum sheet parts. The 3D-printed parts were printed using PLA with a very high infill percentage and were designed to avoid bending moments. They were then tapped with 4-40 thread, and the reel was assembled using 4 mm stainless steel balls and button-head hex-drive screws. A round aluminum standoff and 8-32 socket-head cap screws were used to add a handle, and a ring-grip quick-release pin can be used to lock the reel. I sewed a couple of lengths of webbing into loops such that the lock-off reel can be attached to myself or fixed objects.

For kite line, I opted for braided high-modulus polyethylene line,⁸ sold as fishing line, as this type of line is lighter and thinner than the alternatives. Given the high cost of brand-name line, I opted for the cheaper alternatives available from Chinese vendors on eBay and Amazon, originally purchasing “200 lb” line on eBay. As I did not trust the sellers’ strength ratings, I load tested it using a heavy counterweight, a scale, and a chain hoist, and it failed at approximately half the rated capacity. After getting a refund,⁹ I then ordered “300 lb” line on Amazon, which is the strongest available. This did marginally better and failed at around 120 lb, although that was sufficient and was the strongest line available at a reasonable price. Except for what I used for the main bridle line, I installed the remainder of the 300 m of line on the reel (I used the “200 lb” line for the spreaders and the bridle spreader lines).¹⁰

In December, the day before I left for Antarctica, I took the T out to Wonderland and tested the kite at Revere Beach with 4–5″ of bow in the spreaders. Unfortunately, it was almost dead calm, but I was able to get the kite aloft during a gust and keep it in the air by walking upwind, showing that the kite flew well in extremely-light wind, although the kite line was not taught enough in this wind to attach a camera. Once at the South Pole, it took me several more attempts to get the kite flying reasonably well.¹¹ Unfortunately, the information found online about tuning this type of kite is contradictory.¹² Some folks say that both spreaders should have the same amount of bow, while others say that the bottom spreader should be bowed more; some folks recommend stiffer spreader spars for higher winds, while others recommend more-flexible ones. I eventually settled on 11–12″ of bow in both spreaders, which was considerably more than the 5–6″ I started with, based on what I read online. This was just enough bend to push the aluminum tubes into the plastic deformation regime and permanently bend the tubes (this didn’t happen with 10″ of bow). I found that placing the hitch point on the bridle 18 cm above the center point worked well; moving it up made the kite more unstable, and I couldn’t launch the kite if I moved it down. Originally, I used an alpine butterfly knot for the hitch attachment, although I ended up switching to a simple overhand loop, as it was quicker to tie in the cold; both worked equally well.

With this adjustment, the kite was at least reasonably stable in winds up to around 6 m/s, although it generally wasn’t stable enough for the camera if the wind was much above 5 m/s, particularly when walking the kite upwind; while the kite would fly in less wind, the kite line wasn’t taught enough to attach my camera below around 4 m/s, leaving only a small wind range where it worked well. There were only a couple of days with both appropriate wind and at least mostly-clear skies where I also had sufficient free time to fly the kite and camera.

To pack up the kite, the bow line assemblies are untensioned and removed, the spine spar is removed, and the spreader spars are folded but left attached to the bridle and centered between the grommets. The kite is then folded over the folded spreader spars, the tail is tucked in, the kite is rolled with the folded spine spar (and spreader spars) inside, and the remainder of the bridle is wrapped around the rolled kite. The rolled kite is then placed in its bag, and the bow line assemblies are placed in a separate bag.

The kite had no problems with the cold—although I can’t say the same thing about my fingers when flying it, particularly when I needed to take off my gloves to deal with the camera—so it was a success in that regard. However, in retrospect, a parafoil would have probably been a better option, since there were significantly more days with clear skies and good visibility with wind speeds too fast for this kite than there were days where the wind speed was in this kite’s ideal range. During the two months I was at the Pole, the wind speed was in the 4–5 m/s ideal range only around 20% of the time¹³ and even less time when the skies were also clear. During the same period, the wind was in the 4–7 m/s range, which should work for a parafoil kite, around 50% of the time; while a parafoil kite could work at even higher wind speeds, there is frequently blowing snow obstructing visibility at that point. The spreader spars on this kite appear to be significantly stiffer than what I’ve seen in videos of similar kites, and I suspect that may have hurt its stability in higher winds, since the kite can’t flex to bleed off the extra force from the wind. Furthermore, this kite required a lot more tweaking and adjustment than I would have liked, as my interests lie in aerial photography, not kites; I suspect a commercially-available parafoil kite would require less adjustment.

Details of my camera setup and photos from the kite will follow in a subsequent blog post, as this post is already quite long. I’ve uploaded the design files for this kite, although I know very little about kites, and this kite only worked well in a very narrow wind speed range, so I’m not sure I’d recommend using them. I’ve also uploaded the design files for the kite reel.

1. It was recommended as a replacement for a similar parafoil kite that had also been discontinued, two decades ago.  ↩

2. But only if adjusted correctly, which turned out to be an issue.  ↩

3. There is even a design that was previously used in Antarctica.  ↩

4. Since the joints are the weakest point, it is recommended to avoid having the spreaders cross the spine at a joint on either spar.  ↩

5. The edges of these should have been folded over to prevent tearout, which I later had to fix.  ↩

6. The cutoff lengths of aluminum tube came in handy as handles to pull on the loop on the end of the braided line without the quick link when bending the spreaders; the quick link served as the second handle.  ↩

7. Based on the designs in this video.  ↩

8. Also referred to as UHMWPE, Dyneema, or Spectra.  ↩

9. When I first ordered the line, the seller provided a fraudulent USPS tracking number and instead shipped the item from China, despite what the listing said. During the refund process, the seller tried to get me to contact them outside eBay’s messaging system and then tried to get me to pay for return shipping with the promise of reimbursement, which I assume wouldn’t have happened. When I insisted that they send me a shipping label, they instead processed the refund without having me return the item.  ↩

10. I did not verify the length, so I don’t know if the seller also lied about it.  ↩

11. Including one where I cut my chin open on the laser-cut aluminum edge of the kite reel and needed stitches; I started wearing a neck gaiter while flying the kite after that.  ↩

12. It’s also difficult to look up with the Pole’s extremely-slow and intermittent internet access.  ↩

13. Per SPT’s weather station on DSL.  ↩

To create the custom shapes, I 3D printed molds. To create the molds, I first drew a set of 43 mm by 53 mm designs—the size I measured the commercially-available cookies to be—in Inkscape, making each line at least 2 mm thick, to ensure it would survive the molding and baking process. In the end, I settled on twelve designs: four designs showing E- and B-mode polarization patterns; four designs with Antarctica themes, specifically, an Adélie penguin, the True South flag, an Antarctica outline, and the ceremonial Pole marker; and four designs with BICEP themes, specifically, the Dark Sector Laboratory (DSL) building with BICEP3 ground shield, the BICEP3 cryostat, the BICEP Array tower connected to the Martin A. Pomerantz Observatory (MAPO) building, and a BICEP Array cryostat. After converting the outlines of the designs to paths and taking their union, each design was individually exported in the DXF R14 format and converted into solid models using a custom CadQuery script. The molds were then printed out of PLA using a Prusa MINI+ 3D printer using 0.1 mm layer height. For initial prototypes, the molds were designed to act as both molds and cookie cutters, but the dough would get stuck in the mold. Instead, the design was modified to only cut out the corners, leaving the sides connected to the rest of the rolled out dough; after removing the molds, the sides were cut with a dull knife, freeing the cookies from the dough.

Others have previously baked cookies in this style, so I had an existing reference for a recipe to use. However, I tried to more closely follow Pepperidge Farm’s recipe based on the ingredients list and nutritional information. In the end, I settled on a recipe of 3/4 cup butter, 3/4 cup granulated sugar, 1 3/4 cup all-purpose flour, and 1/2 cup corn starch. Originally, I had included 1 teaspoon of baking powder—as a substitute for the baking soda and cream of tartar in the commercially-available cookies—and less flour and corn starch to better match the fat–sugar–carbohydrate ratio in the nutritional information, but the cookies rose and spread too much during baking, ruining the designs. Thus, I removed the leavening agent and increased the amounts of flour and corn starch to the values listed above. As eggs are listed as being less than 2% of the cookies, I suspect they are only used for the wash, so I left them out of my dough recipe.¹

To make the dough, I partially melted the butter and mixed it with the sugar using a fork. Separately, I mixed the flour and corn starch and then mixed it into the butter–sugar mix, first mixing in half the dry mix and then the rest, until the dough was well mixed. Next, I split the dough in half, wrapped it in plastic wrap, and chilled it. If the dough is too warm, it will stick to the molds, and if it is too cool, it is difficult to roll out, so a happy medium needs to be found. I rolled out the chilled dough between two layers of plastic wrap, to prevent it from sticking, to a thickness of 6 mm, using a rolling pin with disks on the ends that allow setting the thickness. Removing the top layer of plastic wrap, I then dusted flour onto the dough to reduce the risk of it sticking to the molds and used the molds to shape the cookies, pressing down on the molds using much of my body weight.

The mold is then carefully wiggled back and forth to free it from the dough without damaging the design. Originally, I had molded one cookie at a time (as is shown above), but I found that it’s better to press in all of the molds before removing any of them, since this prevents shifting of the dough from later molds from disturbing earlier cookie designs.²

Once the molds are removed, the cookies can be freed from the dough using a knife with a rounded, non-serrated end. Using the same knife (or some sort of small spatula), the cookies are then transferred to parchment paper and placed in a freezer. The remaining dough should be balled back up and placed in a refrigerator; the other half of the dough is then removed from the refrigerator, and the process is repeated. I found that this recipe and mold combination produces around 28 cookies in total.

After freezing the cookies for a least an hour, an egg yolk wash should be applied.³ To do this, I whisked an egg yolk using a fork and then carefully painted it on to just the raised design on the cookies using a foam brush (a suggestion from the recipe I had consulted). The cookies should then be returned to the freezer for at least another 15 minutes, such that they’re frozen when placed in the oven.

Finally, the cookies can be baked. I found this to be a finicky process, since the designs on the cookies would be damaged if the dough rose in the center of the cookies, but the bottom center of the cookies would be under-baked if the cookies were taken out of the oven too soon. In the end, I settled on a procedure using an oven preheated to 425°F, to try to quickly bake in the designs. Start by placing the frozen cookies and parchment paper on a room-temperature baking sheet and baking for 6 minutes; any more than this, and the center of the cookies would begin to rise, damaging the designs. Remove the baking sheet from the oven and allow the sheet and cookies to cool, without removing the cookies. Once the cookies have cooled to near room temperature, return them to the oven for an additional 3 minutes. Again allow them to cool on the baking sheet. Flip the cookies over, return them to the oven for a final 3 minutes, and then allow them to cool. The initial two runs in the oven bake in the designs, while the final oven run finishes baking the bottom of the cookies.

Overall, I was mostly happy with how the cookies came out. The taste and texture were nice, and the designs remained intact, although I would have preferred if the egg yolk wash on the designs made them darker. After dialing in the recipe and baking procedure with two prototype batches, I baked an additional two final batches, producing 60 “good” cookies,⁴ comprised of five of each of the dozen designs.⁵ Between molding the cookies and painting on the wash, it takes over an hour for a batch, and the entire process takes several hours once the cooling time and baking time are included.

My colleagues seem to have enjoyed the cookies.⁶ For anyone wanting to make similar cookies, I’ve uploaded the designs and the script used to create the molds.

1. It would also make it easier to make a vegan version of the recipe. I originally planned to make a vegan and gluten free version for colleagues with dietary restrictions, but I ran out of time.  ↩

2. It also allows the designs to be placed closer together, since the risk of disturbing the adjacent designs is significantly reduced.  ↩

3. If the dough is too warm, applying the wash can damage the designs on the cookies.  ↩

4. I also used some leftover dough from the earlier batches.  ↩

5. Except, I accidentally made one too few of the True South design.  ↩

6. There were several approving comments about how I had gone through the effort of producing both directions of B-mode patterns.  ↩

Starting with prior art, the Planck colormap is a diverging colormap, but it has discontinuities in its perceptual derivative, causing visible banding issues. Here is the output for it from Viscm.

And here is how it performs by the CVD metric I previously developed to evaluate the Turbo colormap (which performs poorly by it). Along with the cusps from the issues with the perceptual derivative, there is a clear imbalance in the metric around the center of the colormap for red and green color-vision deficiencies.

As a first attempt to improve upon this, I fit splines to the color values and applied various levels of smoothing. With small amounts of smoothing, this rounded off the perceptual discontinuities—fixing the worst of the issues—but left much room for improvement; applying more aggressive smoothing further improved the perceptual issues but also more significantly changed the colormap in a manner that my colleagues did not like aesthetically.

For my second attempt, I started by trying to replicate the Planck colormap in Viscm. In the process, I solicited feedback from my colleagues, resulting in tweaks such as moving the mid point from a neutral gray to a slightly warmer color temperature. While this resolved the perceptual uniformity issues, parts of the colormap were less than uniform for individuals with color vision deficiencies. I then iteratively made changes and ran the results through the CVD metric, until the results roughly matched between typical color vision and red and green color-vision deficiencies. Here is the output of Viscm for the new colormap.

And here is how it performs by my previously-developed metric.

While a significant improvement upon the status quo, the new colormap is not perfect, and I think other diverging colormaps such as Peter Kovesi’s CET-D07 colormap (cet_bjy in Python’s Colorcet package), which is both linear and diverging, or some of Fabio Crameri’s colormaps are better choices for most applications. However, the new colormap is probably good enough for CMB maps, where exact values do not need to be read off of the maps by eye, while keeping a similar aesthetic to existing publications.

A Python file that can be imported to use the colormap with Matplotlib or Healpy and the JSCM output from Viscm are available.

1. Y. Li et al., “CLASS Data Pipeline and Maps for 40 GHz Observations through 2022”, arXiv:2305.01045.  ↩

I recently came across a CSS library for emulating L^aT_eX’s default appearance.¹ However, it doesn’t emulate the Knuth–Plass line breaking algorithm, which is one of the things that makes L^aT_eX look good. This got me wondering whether or not it’s possible to emulate this with plain HTML and CSS. A JavaScript library already exists to emulate this, but it adds extra complexity and is a bit slow. It turns out that it is possible to pre-calculate line breaks and hyphenation for specific column widths in a manner that can be encoded in HTML and CSS, as long as web fonts are used to standardize the text appearance across various browsers.

The key is to wrap all the potential line breaks (inserted via ::after pseudo-elements) and hyphens in <span> elements that are hidden by default with display: none;. Media queries are then used to selectively show the line breaks specific to a given column width. Since every line has an explicit line break, justification needs to be enabled using text-align-last: justify;, and word-spacing: -10px; is used to avoid additional automatic line breaks due to slight formatting differences between browsers. However, this presents a problem for the actual last line of each paragraph, since it is now also justified instead of left aligned. This is solved by wrapping each possible last line in a <span> element. Using media queries, the <span> element corresponding to the given column width is set to use display: flex;, which makes the content be left-aligned and take up the minimum space required, thereby undoing the justification; word-spacing: 0; is also set to undo the previous change to it and fix the word spacing. Unfortunately, the nested <span> elements are problematic, because there are no spaces between them; this is fixed by including a space in the HTML markup at the beginning of the <span> and setting white-space: pre; to force the space to appear.

I’ve prepared a demo page demonstrating this technique. It was constructed by calculating line breaks in Firefox 76 using the tex-linebreak bookmarklet and manually inserting the markup corresponding to the line breaks; some fixes were manually made because the library does not properly support em dashes. Line breaks were calculated for column widths between 250 px and 500 px at 50 px increments. The Knuth–Plass line breaks lead to a considerable improvement in the text appearance, particularly for narrower column widths. In addition to the improved line breaks, I also implemented protrusion of hyphens, periods, and commas into the right margin, a microtypography technique, which further improves the appearance. To (hopefully) avoid issues with screen readers, aria-hidden="true" is set on the added markup; user-select: none; is also set, to avoid issues with text copying.

While this technique works fine in Firefox and Chrome, it does not work in Safari, since Safari does not support text-align-last as of Safari 13.² Despite it not working, the corresponding WebKit bug is marked as “resolved fixed”; it seems that support was actually added in 2014, but the support is behind the CSS3_TEXT compile-time flag, which is disabled by default. Thus, I devised an alternative method that used invisible 100% width elements to force line breaks without using explicit line breaks. This again worked in Firefox and Chrome, although it caused minor issues with text selection, but it again had significant issues in Safari. It appears that Safari does not properly handle justified text with negative word spacing; relaxing the word spacing, however, causes extra line breaks due to formatting differences, which breaks the technique. At this point, I gave up on supporting Safari and just set it to use the browser default line breaking by placing the technique’s CSS behind an @supports query for text-align-last: justify.

Automated creation of the markup would be necessary to make this technique more generally useful, but the demo page serves as a proof of concept. Ideally, browsers would implement an improved line breaking algorithm, which would make this technique obsolete.

Update 2024-05: Support for text-align-last was added in Safari 16. Chrome 117 also added support for text-wrap: pretty, but its implementation only looks at the last few lines of a paragraph to avoid orphans and does not improve the typesetting of the rest of the paragraph.

1. Also see corresponding Hacker News discussion.  ↩

2. Even Internet Explorer 6 supports this.  ↩

In particular, I will focus on the labels for canned vegetables. As one would expect, both the old and new label designs feature the name of the vegetable along with a picture of a “serving suggestion.” Since many vegetables are similar in color, it is often easier to find one’s desired vegetable on the shelf by looking for the name, especially when a particular vegetable comes in multiple variants, such as green beans (whole, cut, diagonally cut, and French style). The old design featured a plain sans-serif font in a dark color on a solid white background, resulting in good contrast and readability. The new design, however, is a clear regression; it trades the consistent, easily readable font for a hodgepodge of different heavily-stylized display fonts on a busier background with lower contrast, which results in much worse readability. This loss of readability makes it take longer to locate a particular product on the shelf.

Many of the vegetables come in three variants, regular (“full salt”), low sodium, and no salt added. In the old designs, these were marked using text in a brightly-colored oval. Blue was used for no salt added, and red was used for low sodium; the regular variant did not include an oval. This design allowed one to quickly differentiate between the variants on the shelf. With the new design, these colored ovals were eliminated. The low sodium variant trades the black text on the regular variant for bright blue text and a distinctive blue bar above the text with a clearly readable “low sodium” label. This is an improvement over the old design as it makes the labeling more distinctive and easier to differentiate. Unfortunately, the same is not true for the no salt added variant, which, for some inexplicable reason, is labeled the same as the regular variant except for a small, blandly-colored circular badge in the corner. Instead, it should have been labeled with a distinctive color and a bar with a clearly readable “no salt added” label, similar to the low sodium variant, except using a different color.

The font size was increased for the ingredients list, which is one of the only improvements in the new designs.

1. It’s the closest I’ve seen to what’s suggested by xkcd: Brand Identity.  ↩