Motorola created a lot of buzz this week with its announcement of Project Ara, a sort of Lego kit mobile phone that would allow users to pick and choose components. I know this is going to be a disappointment to the folks who build great things from Arduino boards and Raspberry Pi computers, but Project Ara is never going to happen, at least not as a commercial product.
The modular phone is an appealing idea, but there are so many things wrong with it in practice that it’s hard to know where to start. So I’ll just list a few:
Integration. When you are building a desktop PC, you can afford to be sloppy in the efficiency of the design and make of for its shortcomings with pure power, which makes the traditional open design of desktops possible. Laptops offer less leeway, and handhelds provide no tolerance at all. Today’s mobiles are masterpieces of integration. Hardware components are carefully selected and the software painstakingly optimized so that everything works together with maximum efficiency.[pullquote]Today’s mobiles are masterpieces of integration. Hardware components are carefully selected and the software painstakingly optimized so that everything works together with maximum efficiency.[/pullquote]
The iPhone is probably the extreme example of this. When Apple started making its own system-on-chip processors, it eliminated all the components from the ARM reference designs that it would not be using. No SD slot, so no need for an SD slot memory controller. Trimming components translates to a smaller die size and lower power consumption; every milliwatt counts.
Modular design flies in the face of this optimization. The software and SoC have to be designed to accept whatever hardware the user chooses. This means that the operating system cannot be optimized specifically for the hardware in use, a process that all good Android OEMs have to go through. And lack of optimization is going to carry a price in performance, battery life, or both.
Size. Modular systems are necessarily bigger than integrated ones. Modules have to be at least partially self-contained and they must have a way to hook up with each other. Enclosures and connectors add bulk and weight. Just making the battery removable adds a couple of millimeters to the thickness of a handset. A modular phone is going to be much bigger than an integrated phone of the same capabilities.
Physical Integrity. I love Lego. It can be used to built fabulous models. But they are models, not the real things as you find out if you ever drop one. It would be very hard to design any sort of snap-together modular phone that provides the sort of structural integrity and durability we expect for mobile devices. You can improve things by increasing the strength of the “endo” and connectors that hold it all together and, but again you will pay a penalty in size and weight.
I find the idea that there would be much of a market for a modular phone a dubious one. Most people accept the idea that the engineers and designers responsible for handsets have a better idea of how the choose components and but them together than the average user. But even for the makers who really want the freedom to roll their own designs, the modular phone just has too many disadvantages to ever be practical.
This is a prime example of why I love Steve’s writing. He looks at the things that are getting hype and is able to see things for what they are.
I discussed this article with a few of my Android hardware engineer friends, some of whom believe Project Ara will bring about the next iPhone killer.
As Steve points out, if this ever ships, it would target a very niche market, those who value flexibility uber alles. Mind you, the Android camp is a natural haven for Linux hackers and ultra-geeks, who would love to snap in speaker modules, processors and RAM modules. However, most Android users purchased their phones due to pricing concerns, not because they care about the ability to easily hack their phone into submission.
An Ara-esque phone, in production, would not be cheap, would cause immense fragmentation in another dimension (i.e. even on the same phone, app developers wouldn’t be able to count on the platform’s characteristics), would have high QA costs due to the sheer number of permutations, and would be nearly impossible to optimize for power consumption and space efficiency.
Another consideration is how painful FCC certification would be, and whether the carriers would want to ship a phone that would never be a known entity.
Now, it would be interesting as a construction kit for hardware developers, allowing them to prototype new collections of features.
Yet, the average user doesn’t want to think about upgrading the processor in his computer, or laptop, or tablet, much less his phone. It’s a certain type of person who cares about the type of tires on his car; he may support a very lucrative niche industry, but this type of person is an exception to the rule. Most people never think about their car’s tires, or spark plugs. Most don’t know how much RAM is in their phones, and wouldn’t get off the couch to upgrade it.
Thanks.
I didn’t even think about the FCC issues, but they are likely to be a big problem. The regs generally require certification of entire configurations. not individual modules.
Also, it is really hard for a an extreme niche product that is not derivative of a standard product (the Nexus phones, for example, whose hardware is very similar to other, more commercial products) to exist. The economies of scale are very potent in mobile and the diseconomies very punishing.
I appreciate you sharing this blog.Really looking forward to read more. Really Great.