I am not a native English speaker, but I am wondering, why you call the printings "maps". Aren't they rather high resolution satellite images? Wikipedia defines a map as "a symbolic depiction emphasizing relationships between elements of some space, such as objects, regions, or themes."[1]
And are your images really more detailled than cadastral maps (scale 1:1000) that are nowadays recording the location of a boundary stones with a maximum deviation of ±3 to 5 cm?
Our black and white images (such as the image on the linked article) aren't satellite images at all, they are high resolution elevation models that we've used GIS software to light and create hillshade.
We do have some color images that use satellite imagery and the creation of those images is going to be the subject of my next post. We do things like remove the natural hillshade and apply our own (due to a quirk of human perception where sun coming from below causes terrain to be perceived as inverted) and blending images from different days, etc.
Re: is this really a "map?" That's a comment we get quite a bit, especially in FB comments on our ads. These images are not traditional maps, but a map is a depiction/representation, which these are. Anyway, "map" is certainly shorter than "visual representation of a geographic area" so it's what we go with.
Re: detail. It's all about the size of the area you are printing. Our world map uses 30-meter data, but you'd need to print it on the side of a building to see the limit to the detail. So in our sizes, using 3-5cm data wouldn't improve the maps, you wouldn't be able to see any of that detail. We only make maps if that is true.
> ... they are high resolution elevation models that we've used GIS software to light and create hillshade.
> ... These images are not traditional maps, but a map is a depiction/representation, which these are. Anyway, "map" is certainly shorter than "visual representation of a geographic area" so it's what we go with.
Okay, that makes sense to a certain degree. Seems what you are doing is innovative enough that it deserves a new category of "map".
Have you considered transfering your methods from the macro to the micro world? If I understand it correctly, you could in principle inverse the zoom factors, like using the elevation model of a coin combined with texture data and a lighting model and printing a much larger version of it on a canvas.
These aren't simply photographs but they do seem to fall on the photographic, rather than the map, end of the scale. Raven Maps is a good source of (large) high quality maps that are more traditionally map-like--though some of theirs are pretty photographically oriented as well.
I have a controversial question for you...
let's say I order a map of India from India, would you give me a map with all the sections of the country intact (Kashmir and Arunachal Pradesh) or would you give us a china approved section of India without Kashmir or bits of the east (like how google serves its maps)?
How do you manage to withhold or incorporate sketchy borders with geopolitical issues, or do you refrain from serving such locations altogether to avoid any backlash?
Your contiguous US-states map (and it seems many of your other maps, though not e.g. the North America map or the world map) uses a Mercator map projection: https://ramblemaps.com/continental-us-map
This is a reasonable choice for sailing charts (makes rhumb lines straight) and a reasonable choice for a zoomable web map of the world built around raster tiles (avoids the need for client-side reprojection), but is an uncommon – and in my opinion generally poor – choice for maps of the contiguous US states. What made you settle on that one? Or was it just the existing projection of the data, and you decided not to reproject it?
I think many of your maps would work a bit better (suffer substantially less distortion) if they used a region-specific projection. I have a long-term interest in map projections and would be happy to chat about possibilities if you have questions.
We settled on this choice after originally using alternate projections. The source data is not in mercator, we specifically reproject to mercator.
I don't disagree with you, I personally prefer other projections for many of our maps, but after the hundredth comment saying "you got the shape of [MY STATE] wrong, you idiots!" we realized that people think of their state (and country, in the case of CONUS) in mercator, so we decided it wasn't worth fighting that fight.
For other countries, we tend to use locale specific projections.
One projection question for you, while we're chatting: do you have a favorite Asian projection? I wanted to center the projection on center mass, which would mean changing the poles, but couldn't figure that out!
> hundredth comment saying "you got the shape of [MY STATE] wrong, you idiots!"
Funny enough, when I see either the USA map or the California map in Mercator projection, the first thing that comes to mind is “wow the shape is completely wrong!” :-)
> favorite Asian projection?
Asia is huge, so you are necessarily going to get a lot of distortion of one kind or another. It’s a close-enough-to-circular blobby shape that you might be fine with some azimuthal projection. If you want to preserve local shapes but don’t mind variation in scale, you can use a stereographic projection centered roughly on the circumcenter. If you want to reduce distance errors while allowing a bit of local shape distortion, you could use an azimuthal equidistant projection.
(In either case, you could probably get away with either an assumed spherical earth, or you could properly correct for the ellipsoid. The difference between the two would not be too obvious at a glance.)
There are various other projections you could try. You might find one or another aesthetically better. For example perhaps some oblique conic projection would fit the region slightly better than the azimuthal projection. Or you could try something like https://en.wikipedia.org/wiki/Chamberlin_trimetric_projectio...
Here’s e.g. a nice comparison of some possible projections for Europe:
Thanks for the link. We used Lambert Conformal Conic for Europe and Robinson for Asia, but I'm not thrilled with the Asia projection.
I really want to be able to change the latitude the projection is centered on, but haven't been able to figure out how to move that point yet, so we've got more skew near the pole than we do at the bottom of the map. Feels like it could be better.
One of the more challenging issues I face when walking in the UK, Scotland especially, is knowing the type of surface.
Often something appears passable on Ordinance Survey and OSM maps, but once you actually get there you find it's a bog, marsh or other type of terrain impassable with walking boots. Sometimes it's terrain that would be passable if dry, but when wet becomes too slippery (e.g. lots of very small but steep dirt mounds) or filled with streams.
Your maps are amazing, but the selection is still pretty limited. Are you planning to ever have "DIY" maps, where you could select a chunk of the globe and have that printed? Or is there too much handcrafted custom tweaking required for each map to make that feasible?
If the answer is "no plans/infeasible", please at least consider adding Sydney (maybe in color?), I'm pretty sure it would look incredible.
Thank you! Unfortunately we won't be doing any choose-your-own maps, we really do spend a bunch of time on each one. I don't think we could maintain nearly the level of quality that we like.
I'd love to map Sydney. We're doing a city push right now but starting with US cities, as selling internationally adds some business complexity that we've been trying to avoid so far. As the business grows, we certainly do hope to have more maps outside of the US, and to be able to sell there as well!
All public. At this point if the data doesn't exist, we're out of luck.
I don't anticipate wall art being lucrative enough to commission our own scans. For areas the size of what we tend to map you need an airplane, a drone would take weeks to survey some of them.
I think it could be really amazing for some of your really big maps if you could walk into a room and go "whoa, that's a seriously detailed map" and then get handed the magnifier and go even deeper. Even if you only had super-resolution data for some of the big features stitched in.
A raw piece of imagesetter film behind glass or acrylic and in front of a white background would probably look pretty cool (great contrast), although you'd have some mechanical challenges getting it mounted without bubbles. It may require a fairly thick piece of glass to get enough pressure in the center, and it may require a bezel so you can clamp/glue it to the backing securely.
I plan to write another post on that with some images to show the steps, but the process differs a bit between our black+white and our satellite+hillshade images.
For the black and white, we pull in the relevant elevation models, merging various sources as necessary, compose the map (rotate if it helps), choose hillshade angle to highlight desired terrain, then export to photoshop.
Having been doing this for a while now we spend most of our time in Photoshop, adjusting color curves and healing irregularities (there seem to be more for LiDAR than for the standard 1/3 arc second DEM).
For our satellite + hillshade maps the process is a bit more involved. We actually remove the natural hillshade from the source image and add our own. This helps with perception of terrain (humans tend to perceive terrain as inverted if the sunlight is coming from the bottom of the image) and allows us to really play off the metallic print and get some pop. Even more photoshopping for these "maps."
Really cool looking maps! Most printed large format things I see every day definitely looks worse than e.g 5K iMac screens when you get closer than a couple of feet from it. Would be cool to see something like this IRL
I wonder if it’s possible to use something like photolithography to create incredibly high dpi “3D prints” for elevation maps like this? Maybe some parts would look flat, but you would be able to discern elevation based on touch?
Thanks! Yea, I think we are as close as wall art can get to a high quality screen - we optimize for contrast on the print. And while unlike a screen you need to provide light for a print, the effect of the light on the print is pretty engaging. It changes as you move around it. (It is not, please nobody buy one thinking they are!)
If you're ever looking to print some photos to hang on your wall, it's definitely worth considering metallic paper + face mounted acrylic glass. It has unrivaled "pop" among print options.
I don't know much about photolithography, but a quick search is intriguing. I'll have to put it on my to-learn-about list.
Do you have any plans to create 3d modeled maps that are printed on top of 3d printed plastics or sands? I am in the imagery business and have so many people ask for this, I've pondered what it would take many times.
I have been using GIS data to create 3d carvings out of wood with CNC machines. We've been having tons of fun iterating and learning. We started selling a few here and there and have been trying to optimize our workflow to be able to create a map for any location we can find data for in a reasonable amount of time.
It's a question we often get, but not a product we're looking to pursue. I think it would be prohibitively expensive to do it right and I'm not confident there's a big enough market for it.
The author states that at 300dpi they cannot perceive any loss of detail with 20/15 vision.
But a higher standard would be if at any distance, they cannot perceive a difference between 300 and 400dpi.
Maybe the result would be the same, maybe that was even tested, but when I hear something is the best of the best I am going to look for even a tiny window of improvement!
Some (unsolicited) advice (this is HN after all): there’s a typo in this post (human's should be humans). There are other more subtle grammar issues in the post, too. Some of the language on the front page could be tightened up too.
The product looks awesome though. I’d probably buy one if I lived in the US.
> we print our maps using the highest resolution mediums available. ... There's no sense printing at a higher resolution than humans are able to perceive. For print, 300 dpi is the gold standard.
This is significantly denser than 300 dpi. It's easy to see the difference between a page printed at 300 dpi and one printed at 600 dpi, so I'd say even 300 dpi doesn't reach "higher resolution than humans are able to perceive".
How long have more detailed mediums (or media) been available? In 01949 George Harrison reported improving the control loop of a ruling engine to a precision of, in the medieval units then in use, 0.2 micro-inches (5 nm): https://www.osapublishing.org/josa/abstract.cfm?uri=josa-41-... so that he could cut grooves for a diffraction grating to that precision, which evidently amounts to a precision of 5 million dpi. This seems to have been about a factor of 70 improvement over what Michelson had achieved before 01900. But serious difficulties attended any attempts to use such diffraction ruling engines to cut irregular patterns such as these maps—as well, of course, as limits in the data bandwidth of the necessary control systems.
More detailed media still have been demonstrated; in 01989 hackers at IBM demonstrated the ability to use an STM to position atoms with atomic precision (≈0.1 nm) https://cen.acs.org/analytical-chemistry/imaging/30-years-mo... and in 02013 other hackers at IBM used this technique to make the famous stop-motion animation, "A boy and his atom" out of a few dozen carbon monoxide molecules on a metal surface: https://www.youtube.com/watch?v=oSCX78-8-q0
0.1 nm precision is about 250 million dpi, almost a million times more detailed than Ramble's maps, or a trillion times if you count by detail per unit area. This is almost as high resolution as you're going to get with matter made out of atoms, although you can improve on it by about an order of magnitude by using, say, lithium hydride. But this resolution has been available for something like 30 years now, though you could reasonably argue that xenon atoms adsorbed to cryogenic copper were not an adequately durable medium.
It's an interesting thought to think about a scale model of Earth printed with this resolution. Ramble carefully omitted any quantitative information about the resolution of their elevation models from this post, though in this thread they say their standard DEM data is ⅓", which is 10 meters; you can download free 30-meter-resolution DEM from USGS https://www.usgs.gov/faqs/where-can-i-get-global-elevation-d... and Airbus offers to sell you 12-meter resolution data https://www.intelligence-airbusds.com/imagery/reference-laye.... Much higher-resolution global data almost surely exists but is not available to the public—interferometric microwave SAR from satellites can get down to centimeter resolution https://earthdata.nasa.gov/learn/backgrounders/what-is-sar but is a strategic advantage for change detection (surveillance) and navigation of things like cruise missiles (when GPS is unavailable).
But suppose you have a 1-cm-resolution DEM of Earth, 5.1 exapixels of data (probably about 5 exabytes, 5.1 million terabytes, about US$100 million of disk), as surely the national spying agency of every spacefaring power does. If you were to print a relief map from it at single-atom resolution—0.1 nm—how big would that map be?
Well, the radius of the Earth is 6371 km (the pole-equator distance was supposed to be 10'000 km, which would have made the radius 6366 km, but Humboldt's expedition lamentably made an 0.08% error in their measurements that we must now live with), and scaling that down by the ratio 1cm:0.1nm, or 100 million to 1, we end up with 63.71 mm radius, or 127.4 mm diameter. The scale model of the Grand Canyon would be 19 microns deep and 290 microns wide. The model earth, accurate to the centimeter, would easily fit in your hand, although hopefully it would be equipped with handles so a stray sand grain on your finger wouldn't dig a kilometer-deep trench across Iowa.
You might very reasonably protest that a map that can only be read with an electron microscope, because nearly all its detail is smaller than the wavelength of light, is less than useful. So if we limit the map's resolution to what you can see with visible light—say, 400 nm—its scale is 4000 times larger. Your scale model of Earth would then be 510 meters across, the size of a small town. But you would still need a very fine optical microscope to see most of its detail.
If you printed out sheets of this map on A3 paper, it would take 6.5 million pages, mostly ocean. Each sheet would cover 10.5 km × 7.4 km.
> It's easy to see the difference between a page printed at 300 dpi and one printed at 600 dpi
This depends entirely on viewer and viewing distance. 300 dpi should be roughly at the limit of what someone with 20/20 vision can distinguish from a distance of about a foot.
If you get a teenager with 20/15 vision and put them as close as their eyes can focus (say, 4 or 5 inches), they’ll be able to see a clear difference between these. But if you take an average person and look at the two images from a distance of a few feet, it will be all but impossible to tell the difference.
Subjectively, I think it's easy to see the difference between the output of a "300 dpi" laser printer from a "600 dpi" one. But can we justify that mathematically? I think distinguishing the orientation of a set of bars separated by, in ancient Babylonian units, 1 arc minute, is Snellen's definition of 20/20 vision? 290 microradians? You might have, say, a 1-arc-minute-wide line, followed by a 1-arc-minute-wide space, and then another 1-arc-minute-wide line? At a "foot" an arc minute is 88.7 microns, which works out to 286 "dpi". So, I guess you're right: 300 dpi is fairly precisely the limit of what 20/20 vision can distinguish from one "foot".
In https://dercuano.github.io/notes/bokeh-pointcasting.html I estimated the visual acuity in my good eye experimentally at about 200 μradians, which I guess means I have about 20/15 vision. Too bad I can't focus any closer than a "foot" now that I'm old. I can still see the jaggies on 300-dpi laser prints, though.
But if you have 20/20 vision and can focus at 150 mm (as most of us can before we get old, or as we can with reading glasses, or just if we're a bit nearsighted, or if we're reading in bright sunlight so the bokeh is a bit smaller), then you can distinguish lines separated by a single white pixel at 600 dpi. So in theory you should be able to print out http://canonical.org/~kragen/bible-columns.png on a regular 600-dpi laser printer and then read it with a magnifying glass—the whole KJV Bible on three pages. Two sheets of paper, if you print double-sided. (So far, I haven't managed to get it to print that clearly, but I think that's probably a matter of printer drivers and pixel grid alignment to avoid resampling.)
Now, if you have 20/15 vision, at one foot you should have about 400 dpi of resolution. Or, at the 4 inches you suggest, 1200 dpi! (1145, actually.) You should be able to get the KJV Bible onto one side of one sheet of A4 paper!
Anyway, I think there's a big gap between the ambition described by "The World's Most Detailed Print Maps" and "we print our maps using the highest resolution mediums available" and even "no sense printing at a higher resolution than humans are able to perceive", and the standard you describe, "if you take an average person and look at the two images from a distance of a few feet, it will be all but impossible to tell the difference." I mean there are a lot of things that humans are able to perceive, but not from a distance of a few feet: the flavor of pizza, the scent of most kinds of roses, the difference between good wine and bad wine, the individual scales on a butterfly's wing, bad breath, the subtle impatience in an outwardly tender caress—and, perhaps, even the meaning of most printed text.
And there are many, many things that humans are able to perceive, but average people are not: the meaning of an English sentence (since 83% of people do not speak English), the difference between love and lust, the historical context of colonialism, the particular version of AutoCAD used to design a building, the form of engine malfunction represented by a particular rattling noise, the shoddiness of Dan Brown's writing, the self-serving dishonesty of those who promote the "Law of Attraction," the abusive nature of proprietary-software licenses, global warming, the difference between a Bristol accent and a London accent.
So I think you are redefining the stated goals of this work downward to a really remarkable extent. What an average person can easily see from a distance of a few feet is very far from the limits of human perception.
You are right if they wanted to really push the limits of human perception they could develop custom printers or buy some machines used for physics experiments or integrated circuit fabrication or whatever and try to print at 1200+ pixels/inch.
Realistically, what they are doing is making full resolution prints on commercially available large-format photograph printers, with their effort going into tweaking the shaded relief algorithms and applying photoshop filters to boost local contrast.
I think you're being unjustifiably uncharitable by accusing them of intentional fraud. I think they're just fooling themselves because they're carried away by their enthusiasm for what they've created, which is really pretty cool.
You don't need custom printers to print at 1200+ pixels per "inch". Most off-the-shelf inkjet printers and some laser printers can do it, though at high resolutions there's a serious tradeoff between precise color and precise position. https://news.ycombinator.com/item?id=27965938 points out that standard imagesetters are 2540 "dpi" (and they have been since the 01980s). Photographic processes routinely produce images with a resolution below 10μm (2540 "dpi") and, when desired, below 2μm (12700 "dpi"). I don't mean with some kind of super niche lab equipment; I mean Kodak point-and-shoot cameras with mass-market Kodachrome film. Metal machining routinely hits precisions of 25 μm (1000 "dpi") and has for a century and a half, sometimes using machines made out of junk using nothing but hand tools; when it matters, it gets down to 1 μm (25400 "dpi"), which is roughly the measuring repeatability of any old handheld micrometer. You can buy such a micrometer on eBay for US$30 https://www.ebay.com/itm/363415676100?hash=item549d4324c4:g:... though a quality one probably costs US$60 or more, even used.
What made Harrison's achievements with the diffraction-grating ruling engines 70 years ago remarkable was not that he could position the ruling engine in increments of 25 nanometers or better (a million dpi); that, and even 2-nanometer increments, had been achieved for small gratings about a century earlier. His achievement was maintaining a total error under 5 nanometers across a distance of over 200 mm during the weeks necessary to rule a full diffraction grating. (I think the Richardson Grating Lab still uses the Michelson ruling engine he overhauled in the 01940s, though with further improvements.)
Fraud? Huh? When they say that these maps are the “most detailed” in their marketing materials they clearly mean that they are using a digital elevation model with higher resolution than the ones used to create other similar images available for sale elsewhere. They obviously use relatively standard digital photo printing processes, and are not implying otherwise. If you have your own source file, you can get it printed/mounted using similar technology from your favorite commercial print shop.
One can certainly produce a higher resolution image using various technologies (as you say, a photographic negative potentially has very high resolution). But so what? I really have no idea why you are talking about micrometers, diffraction gratings, and so on.
I came here to tell you, kragen, that I am enjoying your zeros before 4-digit dates, your derision of old units, etc. Oh, that and your informative comments, of course.
> Realistically, what they are doing is making full resolution prints on commercially available large-format photograph printers, with their effort going into tweaking the shaded relief algorithms and applying photoshop filters to boost local contrast.
Yes, exactly this.
> Marketing copy exaggerates a bit, film at 11.
And yes, also this, ha. We sell wall art, so the article is intended to convey that for something you look at on a wall, we're running up against the limits of what a human can see, with normal vision, at any reasonable viewing distance. It is absolutely true that someone with good vision, the right light, and a magnifying glass could probably see some dots, but that's not typically how wall art is viewed, which is what we design for.
You should probably fix that, because lying about your product in order to get people to buy it is fraud. Even if you don't care about the ethical issues, it could have negative repercussions for you in the future. It looks like it's good enough that people would buy it even if you didn't lie about it, so I think you should capitalize on that. (Maybe you disagree? People who notice the lie will think you don't think it's good enough to sell without lying about it, and they'll wonder what else you're lying about.)
Also 300 dpi is coarse enough that, despite my aging vision, I wouldn't need a magnifying glass or particularly good light to see the jaggies from 300 mm, and I think that is a reasonable viewing distance for wall art. 300 dpi is definitely not a "gold standard" for coffee-table books or wall art. Someone with good vision, the right light, and a magnifying glass can see features at 2400 dpi, 64 times the number of pixels you're using. Why do you think Linotype made 2450 dpi the Linotronic resolution 35 years ago? Perhaps you think they didn't have any experience with printing?
I assume you're referring to the azimuth we place the sun for the hillshade? If so, it varies, though it is almost always coming from the top, generally from the upper left.
Exactly where it is coming from depends on the terrain, we'll move it around within a 45 degree window around 337.5 and see what makes the particular landscape look best.
We sell wall art. You likely don't look at a painting by your favorite artist and say "This is overpriced. I bet the canvas and paint only cost $10."
Also, Ramble Maps is a business. We have material costs, shipping costs, advertising costs, rent. We accept returns (and pay for return shipping), spend time making maps, answering emails, and writing posts like this.
And finally, can you let me know where you can get those materials? Those prices are far better than what we pay!
Yeah, no disrespect, I understand you're a business. For example, someone with a little bit of knowledge of GIS and some google skills could create an image of GMTED2010 data and upload it to a printing service for around $260 for a 30"x40". Those with access to printers could do it for about 1/2 that with the right sublimation gear or large format inkjets. I was comparing your 30x45 with 30x40 prints I do of my own works. Granted it's not flyover LiDAR, its space art, but still. I'm not your target customer, but I do wish you the best.
Co-owner and I are long-time HNers. Psyched to be on the front page!