Tech Trends

Posted by JoshDreamland on Oct. 29, 2013, 9:18 p.m.

So, as I assumed, the PhoneBloks idea was not was not novel. Motorola has been developing something like PhoneBloks for about a year now, and has recently partnered with the PhoneBloks guy, according to this article DaSpirit linked me to, earlier, and also according to the project's own homepage.

To be honest, I'm not sure why they partnered with him; if I had to guess, I'd say it was for the PR purposes. The original "proposal" looks as though it was written by an eleven-year-old. It looks like stuff I did back when I was that age, anyway. Only my designs revolved around Game Boys rather than smart phones. But I digress.

What I pointed out in my previous blog as the biggest design flaw in PhoneBloks was the mystical breadboard of infinite interconnectivity. You can just snap a Blok any old place, and it will somehow work. This isn't that far out of the question, and people who say "maybe someday we will have the technology" are just people who are too lazy to look at what we already have and analyze the idea. His original proposal was a glorified USB switch. USB uses four pins; two are for power (one VDD and one ground), and then the other two do I/O, one being reserved for the clock. Our present standing of technology is not the problem. The problem is, how many components of your computer are USB? The smaller the number, the happier you probably are with your computer. With the original PhoneBloks "proposal," everything is essentially USB, including the CPU and RAM! USB 3 is pretty fast. I wouldn't use it to connect a graphics card.

A few days ago I was talking to sirxemic about how I would go about fixing that obvious problem. My solution was simple: Use the USB-like interface to connect the accessories, but let blocks include their own bus types.

Laptop memory chips are connected to a DMA controller by 200 pins, divided into two rows of 100 pins, in groups of 80 and 20. USB's four pins are not exactly equipped for that. A PCIe bus is 164 lines. Again, USB probably won't cut it.

That said, your options are to (1) ensure a bus size of at least 200, or (2) allow individual components to specify their own bus types. I proposed the latter. Basically, we lose the whole god-class idea of having «one big board that holds everything together, hurka durk,» individual blocks would themselves have blocks attached through their own interfaces. For a rough (and very naïve) idea, imagine that we have a logic block which contains the CPU. On its north edge are two 100-pin connectors for RAM blocks. Or for a separate GPU block. A couple hundred pins has proven to be sufficient for either. Thus, to change the RAM, you snap out your logic block, then pull the RAM block out of it, and pop in a new one.

The biggest issue would then be space. As I've indicated previously, the space constraints and additional wall thickness associated with the modularity of this design are going to kill the phone in terms of being cutting edge. If you said to me, "COMPUTAR BLOKS," I would have literally nothing to point out as a problem with the idea. Everything in a desktop computer is PCIe or USB. The issue is, some graphics cards are approaching a foot long. The laptop memory I linked to earlier is probably five of those original blocks long, and three wide. That said, you could probably fit 4GB into a 2x2 block area, as long as heating isn't an issue. Memory of that physical size is now 4GB, and if you broke that chip in half and stacked it, it would fit comfortably into an area of three blocks by two. Moreover, if you could stack that 3×2 area on top of the CPU block, you'd be set.

This is an idealization of what the logic block would look like:

I call it an idealization because it offers roughly the same amount of exhaustive engineering research as the original PhoneBloks proposal. However, I bothered to research dimensionality. The device pictured is 6cm by 4cm by 8mm. It allows for two RAM blocks, 2cm by 2cm by 6mm apiece, which could feasibly be replaced by one big RAM block to avoid having to divide the chips. You might be able to fit 2GB in the whole structure, provided only that my previous suggestion of cutting up notebook memory with scissors and packing it into places works. I know the pins look small, but they are wider than mechanical pencil lead. The RAM blocks will be held into the logic block by four #0 screws, pictured as little gray disks on the front-right face.

Not pictured is the DMA controller, and the bus controller, which are erroneously assumed to fit in the 40×20×2mm volume under the RAM or in the same space as the CPU (totaling 40×40×8mm, less wall size). For reference, the former is just barely wide enough for a single silicon chip. I mean, no one said the blocks had to be 8mm or thinner, but people would probably complain if their phone was more than a centimeter thick, and with the screen and bus board in tow, we're probably at 14mm. But it looked pretty bulky in the original video and no one seemed to mind, so…

Anyway, I don't mean to be a downer, especially considering it's my own doodle. Based on our current phone dimensions, it should be possible to cram decent technology into a block of similar dimensions to the above diagram. Probably nothing cutting-edge, though, as mentioned like eight times over this blog and the last.

In summary,

- The main board is your typical USB switcher, or at best, a PCIe ×1 bus.

- Anything sensitive that needs communicated to the CPU is to be plugged into the CPU block. This means RAM or separate GPUs do not touch the main board; they are designed for any CPU block accepting that interface.

- Peripherals such as wifi or the touch part of the display are -already- USB devices. They should be fine regardless of bus size.

Problems not addressed:

- I haven't specified a main bus layout, or the bus controller scheme

- I haven't specified how the bus system allows for DMA

- I haven't specified whether the actual display picture can be sent over the bus

- I haven't specified how users will choose an appropriate power supply for their CPU

– or how lesser blocks will ensure they are getting the needed voltage

Problems that cannot be addressed:

- A priori, no matter how good our technology gets, a phone will be better whose engineers had the freedom to place components anywhere, rather than in little cubes. Given the same amount of total phone space, you have less working engineering space in a PhoneBloks environment, which is compartmentalized by walls of nonzero thickness. I.e., you still have less space.

Anyway, that's enough of that. Another interesting point is that Vicarious, Inc., the company you've never heard of, has created a CAPTCHA solver which they boast can solve any CAPTCHA. Check the video out. Honestly, this is long overdue, but I'm not sure if they did it "right"—that is, I'm not sure if they found the same type of algorithm that humans employ. It's running time is awful, but that isn't a good metric, because our brains are much more parallel, and there's no saying how much better their algorithm would do if run on a 500-core machine. So I'll have to reserve my comments until such a time as they open-source the algorithm, which probably won't happen.



JoshDreamland 7 years, 2 months ago

I remain skeptical, I just don't want people to mistake me for someone who thinks we don't have the technology. We do have the technology; we've had it for years. My argument is that no matter how good our technology gets, a phone will be better whose engineers had the freedom to place components anywhere, rather than in little cubes.

This blog was meant to give how I'd overcome the design flaws I mentioned in my last blog, but that problem is, by definition, impossible to overcome. A priori, given the same amount of total phone space, you have less working engineering space in a PhoneBloks environment.

I've added that to the blog with a bold heading.

Acid 7 years, 2 months ago

I wasn't skeptical about a modular phone - its absolutely plausible. But the viability of the product, as well as the sustainability, are preposterous. It will be unnecessarily expensive to maintain pertinence until smartphones are obsolete.

Plus, this kind of phone will only be popular among techies and people who have to have the newest thing. Tons of people can barely grasp the capabilities of their current phones. Throw in trying to "manually" trying to keep their phone up to date and they'll just wonder why they didnt get the iPhone 7se.

JoshDreamland 7 years, 2 months ago

A lot of research money will go into this, yes, but people will be interested in the initial idea of reducing electronic waste, and in the idea that repair will probably be cheap. I would actually argue that it is the techies who will be least interested in this phone: they'll want the phones with the latest and greatest performance, and you just won't get that from a compartmentalized version.

What seems to be a big downside of Motorola Ara is that it lacks PhoneBlok's propensity for customization. Originally, PhoneBloks boasted desirability for people who want to be able to choose component sizes, something Ara is blatantly incapable of. Look at the examples from the PhoneBloks idea video:

- If you do all your work on the cloud, you can swap out your storage block for a bigger battery block.

- If you take a lot of pictures, you can upgrade your camera block [for this huge thing].

Ara, as pictured, can't do either of those things, as all blocks have one of two sizes. If you consider the main connector as a big USB array, both of those quickly become possible. Ara seems to have the potential for this, but it has the same flaw not corrected in the original PhoneBloks: the logic chip is one big component including the CPU, RAM, and presumably any GPU.

You could argue that this isn't a problem for Ara, as individual block designers can simply give their blocks those features. But if Ara doesn't specify interface types for RAM, you'll be limited to the same company for all pieces of each block (eg, if you buy a Samsung CPU block which supports separate RAM blocks, you'll need Samsung RAM blocks). Ara should specify standard interface types to help coordinate developers.

Toast 7 years, 2 months ago

The idea of a futureproof phone is attractive I guess, but isn't it just like a "modular" desktop PC? Those are hardly futureproof, and the fact you can add and replace parts appeals only to the nerds, everyone else gets a basic one or a laptop.

JoshDreamland 7 years, 2 months ago

Right, but the ability to have them repaired by a nerd, rather than replacing it entirely, is still pretty appealing. Though Apple's still in business, so I guess it can't be that appealing, yeah?

Toast 7 years, 2 months ago

Well I look forward to seeing it, even though extraneous phone functionality has never been my thing (I have a 3GS which I use to phone, text, check the time, set alarms, timers and stopwatches. That's it).

DaSpirit 7 years, 2 months ago

I see PhoneBloks as a huge ploy. I bet that they will sell the phone for cheap but with low specs. Then they would overprice the tiles (assuming no one else would be creating tiles). I'm sure that other phone companies would attempt to stop this from happening, so they would not contribute in any way and try to advertise their own phones as much as possible. The companies that make the parts will prefer the phone carriers that do not separate their modules also. How else will they sell their cheap shit? This thing will have a lack of support. It's a neat idea, but it would only have worked if it was done like that in the beginning.

svf 7 years, 2 months ago

Did PhoneBloks really promise the i-series of Intel's processors? Dafuq

JoshDreamland 7 years, 2 months ago

@Toast - Yeah, I'm interested in how it turns out, too, especially if Google's involved.

@DaSpirit - I'm assuming that there isn't much room for marking up the blocks before the prices become too great a turnoff. Ideally, the interfaces will be open so that no one company dominates; while I don't necessarily trust Motorola to do this, now that they're owned by Google (assuming this is Mobility they're talking about), I figure there's a much better chance of that. At least, worlds better than if Apple or Microsoft were in charge.

@svf - No, I was exaggerating. I was wondering if anyone would notice that. [:P]

I also didn't want anyone taking the blueprint too seriously.