How Much Is Too Much?

December 12, 2020 by Elliot Williams 52 Comments

I definitely tend towards minimalism in my personal projects. That often translates into getting stuff done with the smallest number of parts, or the cheapest parts, or the lowest tech. Oddly enough that doesn’t extend to getting the project done in the minimum amount of time, which is a resource no less valuable than money or silicon. The overkill road is often the smoothest road, but I’ll make the case for taking the rocky, muddy path. (At least sometimes.)

There are a bunch of great designs for CNC hot-wire foam cutters out there, and they range from the hacky to the ridiculously over-engineered, with probably most of them falling into the latter pile. Many of the machines you’ll see borrow heavily from their nearest cousins, the CNC mill or the 3D printer, and sport hardened steel rails or ballscrews and are constructed out of thick MDF or even aluminum plates.

All a CNC foam cutter needs to do is hold a little bit of tension on a wire that gets hot, and pass it slowly and accurately through a block of foam, which obligingly melts out of the way. The wire moves slowly, so the frame doesn’t need to handle the acceleration of a 3D printer head, and it faces almost no load so it doesn’t need any of the beefy drives and ways of the CNC mill. But the mechanics of the mill and printer are so well worked out that most makers don’t feel the need to minimize, simply build what they already know, and thereby save time. They build a machine strong enough to carry a small child instead of a 60 cm length of 0.4 mm wire that weighs less than a bird’s feather.

I took the opposite approach, building as light and as minimal as possible from the ground up. (Which is why my machine still isn’t finished yet!) By building too little, too wobbly, or simply too janky, I’ve gotten to see what the advantages of the more robust designs are. Had I started out with an infinite supply of v-slot rail and ballscrews, I wouldn’t have found out that they’re overkill, but if I had started out with a frame that resisted pulling inwards a little bit more, I would be done by now.

Overbuilding is expedient, but it’s also a one-way street. Once you have the gilded version of the machine up and running, there’s little incentive to reduce the cost or complexity of the thing; it’s working and the money is already spent. But when your machine doesn’t quite work well enough yet, it’s easy enough to tell what needs improving, as well as what doesn’t. Overkill is the path of getting it done fast, while iterated failure and improvement is the path of learning along the way. And when it’s done, I’ll have a good story to tell. Or at least that’s what I’m saying to myself as I wait for my third rail-holder block to finish printing.

Sufficiently Advanced Technology And Justice

December 5, 2020 by Elliot Williams 74 Comments

Imagine that you’re serving on a jury, and you’re given an image taken from a surveillance camera. It looks pretty much like the suspect, but the image has been “enhanced” by an AI from the original. Do you convict? How does this weigh out on the scales of reasonable doubt? Should you demand to see the original?

AI-enhanced, upscaled, or otherwise modified images are tremendously realistic. But what they’re showing you isn’t reality. When we wrote about this last week, [Denis Shiryaev], one of the authors of one of the methods we highlighted, weighed in the comments to point out that these modifications aren’t “restorations” of the original. While they might add incredibly fine detail, for instance, they don’t recreate or restore reality. The neural net creates its own reality, out of millions and millions of faces that it’s learned.

And for the purposes of identification, that’s exactly the problem: the facial features of millions of other people have been used to increase the resolution. Can you identify the person in the pixelized image? Can you identify that same person in the resulting up-sampling? If the question put before the jury was “is the defendant a former president of the USA?” you’d answer the question differently depending on which image you were presented. And you’d have a misleading level of confidence in your ability to judge the AI-retouched photo. Clearly, informed skepticism on the part of the jury is required.

Unfortunately, we’ve all seen countless examples of “zoom, enhance” in movies and TV shows being successfully used to nab the perps and nail their convictions. We haven’t seen nearly as much detailed analysis of how adversarial neural networks create faces out of a scant handful of pixels. This, combined with the almost magical resolution of the end product, would certainly sway a jury of normal folks. On the other hand, the popularity of intentionally misleading “deep fakes” might help educate the public to the dangers of believing what they see when AI is involved.

This is just one example, but keeping the public interested in and educated on the deep workings and limitations of the technology that’s running our world is more important than ever before, but some of the material is truly hard. How do we separate the science from the magic?

Why You Need To Finish

November 21, 2020 by Elliot Williams 31 Comments

Mike and I were talking about an interesting smart-glasses hack on the podcast. This was one of those projects where, even if you don’t need a pair of glasses with LEDs on them to help you navigate around, you just couldn’t help but marvel at a lot of the little design choices made throughout.

For instance, I love the way the flex PCB is made to do double duty by wrapping around the battery and forming a battery holder. This struck me as one of those quintessential hacks that only occurs to you because you need it. Necessity is the mother of invention, and all that. There was a problem, how to fit a battery holder in the tiny space, and a set of resources that included a flex PCB substrate. Cleverly mashing that all together ended up with a novel solution. This wouldn’t occur to you if you were just sitting at the beach; you’d have to be designing something electronic, space-constrained, and on a flex PCB to come up with this.

Mike made an offhand comment about how sometimes you just need to finish a project for the good ideas and clever solutions that you’ll come up with along the way, and I think this battery holder example drives that point home. I can’t count the number of my projects that may or may not have been dumb in retrospect, but along the way I came up with a little trick that I’ll end up using in many further projects, outliving the original application.

Finishing up a project on principle is a reasonable goal just on its own. But when the process of seeing something to conclusion is the generator of new and interesting challenges and solutions, it’s even more valuable. So if you’re stuck on a project, and not sure you want to take it all the way, consider if the journey itself could be the destination, and look at it as an opportunity to come up with that next long-lasting trick.

Bad News: Arecibo

If you read the newsletter last week, you heard me wondering aloud if the damage to Arecibo Observatory had crossed the threshold into where it’s no longer economically viable to keep it running, and the sad news has just come in and the battle for Arecibo has been lost. We said we’d shed a tear, and here we are. Sic transit gloria mundi. Here’s hoping something cooler replaces it!

What Is Worth Saving?

November 14, 2020 by Elliot Williams 72 Comments

When it rain, it pours. One of the primary support cables holding up the Arecibo Observatory dish in Puerto Rico has just snapped, leaving its already uncertain fate. It had been badly damaged by Hurricane Maria in 2017, and after a few years of fundraising, the repairs were just about to begin on fixing up that damage, when the cable broke. Because the remaining cables are now holding increased weight, humans aren’t allowed to work on the dome until the risk of catastrophic failure has been ruled out — they’re doing inspection by drone.

Arecibo Observatory has had quite a run. It started out life as part of a Cold War era ICBM-tracking radar, which explains why it can transmit as well as receive. And it was the largest transmitting dish the world had. It was used in SETI, provided the first clues of gravitational waves, and found the first repeating fast radio bursts. Its radar capabilities mean that it could be used in asteroid detection. There are a number of reasons, not the least of which its historic import, to keep it running.

So when we ran this story, many commenters, fearing the worst, wrote in with their condolences. But some wrote in with outrage at the possibility that it might not be repaired. The usual suspects popped up: failure to spend enough on science, or on infrastructure. From the sidelines, however, and probably until further structural studies are done, we have no idea how much a repair of Arecibo will cost. After that, we have to decide if it’s worth it.

Per a 2018 grant, the NSF was splitting the $20 M repair and maintenance bill with a consortium led by the University of Central Florida that will administer the site. With further damage, that might be an underestimate, but we don’t know how much of one yet.

When do you decide to pull the plug on something like this? Although the biggest, Arecibo isn’t the only transmitter out there. The next largest transmitters are part of Deep Space Network, though, and are busy keeping touch with spacecraft all around our solar system. For pure receiving, China’s FAST is bigger and better. And certainly, we’ve learned a lot about radio telescopes since Arecibo was designed.

I’m not saying that we won’t shed a tear if Arecibo doesn’t get repaired, but it’s not the case that the NSF’s budget has been hit dramatically, or that they’re unaware of the comparative value of various big-ticket astronomy projects. Without being in their shoes, and having read through the thousands of competing grant proposals, it’s hard to say that the money spent to prop up a 70 year old telescope wouldn’t be better spent on something else.

Adventures In Overclocking: Which Raspberry Pi 4 Flavor Is Fastest?

November 11, 2020 by Elliot Williams 56 Comments

There are three different versions of the Raspberry Pi 4 out on the market right now: the “normal” Pi 4 Model B, the Compute Module 4, and the just-released Raspberry Pi 400 computer-in-a-keyboard. They’re all riffing on the same tune, but there are enough differences among them that you might be richer for the choice.

The Pi 4B is easiest to integrate into projects, the CM4 is easiest to break out all the system’s features if you’re designing your own PCB, and the Pi 400 is seemingly aimed at the consumer market, but it has a dark secret: it’s an overclocking monster capable of running full-out at 2.15 GHz indefinitely in its stock configuration.

In retrospect, there were hints dropped everywhere. The system-on-a-chip that runs the show on the Model B is a Broadcom 2711ZPKFSB06B0T, while the SOC on the CM4 and Pi 400 is a 2711ZPKFSB06C0T. If you squint just right, you can make out the revision change from “B” to “C”. And in the CM4 datasheet, there’s a throwaway sentence about it running more efficiently than the Model B. And when I looked inside the Pi 400, there was this giant aluminum heat spreader attached to the SOC, presumably to keep it from overheating within the tight keyboard case. But there was one more clue: the Pi 400 comes clocked by default at 1.8 GHz, instead of 1.5 GHz for the other two, which are sold without a heat-sink.

Can the CM4 keep up with the Pi 400 with a little added aluminum? Will the newer siblings leave the Pi 4 Model B in the dust? Time to play a little overclocking!

Continue reading “Adventures In Overclocking: Which Raspberry Pi 4 Flavor Is Fastest?” →

Tired Of The Cat-and-Mouse

November 7, 2020 by Elliot Williams 49 Comments

Facebook just announced their plans for the Oculus Quest 2 VR headset. You probably won’t be surprised, but they want more of your user data, and more control over how you use the hardware. To use the device at all, you’ll need a verified Facebook account. Worse, they’re restricting access to the wide world of community-developed applications by requiring a developer account to be able to “sideload” non-Facebook software onto the device. Guess who decides who gets to be a developer. Hint: it’s not the people developing software.

Our article suggests that this will be the beginning of a race to jailbreak the headset on the community’s part, and to get ahead of the hackers on Facebook’s. Like every new release of iOS gets a jailbreak within a week or two, and then Apple patches it up as fast as they can, are we going to see a continual game of hacker cat-and-mouse with Facebook?

I don’t care. And that’s not because I don’t care about open hardware or indie VR developers. Quite the opposite! But like that romance you used to have with the girl who was absolutely no good for you, the toxic relationship with a company that will not let you run other people’s games on their hardware is one that you’re better off without. Sure, you can try to fix it, or hack it. You can tell yourself that maybe Facebook will come around if you just give them one more chance. It’s going to hurt at first.

But in the end, there is going to be this eternal fight between the user and the company that wants to use them, and that’s just sad. I used to look forward to the odd game of cat and mouse, but nowadays the cats are just too well bankrolled to make it a fair fight. If you’re buying a Quest 2 today with the intent of hacking it, I’d suggest you spend your time with someone else. You’re signing up for a string of heartbreaks. Nip it in the bud. You deserve better. There are too many fish in the sea, right?

What are our options?

New Raspberry Pi 400 Is A Computer In A Keyboard For $70

November 2, 2020 by Elliot Williams 248 Comments

The newest Raspberry Pi 400 almost-all-in-one computer is very, very slick. Fitting in the size of a small portable keyboard, it’s got a Pi 4 processor of the 20% speedier 1.8 GHz variety, 4 GB of RAM, wireless, Ethernet, dual HDMI outputs, and even a 40-pin Raspberry Standard IDE-cable style header on the back. For $70 retail, it’s basically a steal, if it’s the kind of thing you’re looking for because it has $55 dollars worth of Raspberry Pi 4 inside.

In some sense, it’s getting dangerously close to fulfilling the Raspberry Pi Dream. (And it’s got one more trick up it’s sleeve in the form of a huge chunk of aluminum heat-sinked to the CPU that makes us think “overclocking”.)

We remember the founding dream of the Raspberry Pi as if it were just about a decade ago: to build a computer cheap enough that it would be within everyone’s reach, so that every school kid could have one, bringing us into a world of global computer literacy. That’s a damn big goal, and while they succeeded on the first count early on, putting together a $35 single-board computer, the gigantic second part of that master plan is still a work in progress. As ubiquitous as the Raspberry Pi is in our circles, it’s still got a ways to go with the general population.

The Raspberry Pi Model B wasn’t, and isn’t, exactly something that you’d show to my father-in-law without him asking incredulously “That’s a computer?!”. It was a green PCB, and you had to rig up your own beefy 5 V power supply, figure out some kind of enclosure, scrounge up a keyboard and mouse, add in a monitor, and only then did you have a computer. We’ve asked the question a couple of times, can the newest Raspberry Pi 4B be used as a daily-driver desktop, and answered that in the affirmative, certainly in terms of it having adequate performance.

But powerful doesn’t necessarily mean accessible. If you want to build your own cyberdeck, put together an arcade box, screw a computer into the underside of your workbench, or stack together Pi Hats and mount the whole thing on your autonomous vehicle testbed, the Raspberry Pi is just the ticket. But that’s the computer for the Hackaday crowd, not the computer for everybody. It’s just a little bit too involved.

The Raspberry Pi 400, in contrast, is a sleek piece of design. Sure, you still need a power supply, monitor, and mouse, but it’s a lot more of a stand-alone computer than the Pi Model B. It’s made of high-quality plastic, with a decent keyboard. It’s small, it’s light, and frankly, it’s sexy. It’s the kind of thing that would pass the father-in-law test, and we’d suggest that might go a long way toward actually realizing the dream of cheaply available universal (open source) computing. In some sense, it’s the least Hackaday Raspberry Pi. But that’s not saying that you might not want one to slip into your toolbag.