Are smartphones any good? (was Re: Oneplus One support)

Dr. H. Nikolaus Schaller hns at goldelico.com
Sat Oct 4 13:59:05 UTC 2014


Hi,

Am 04.10.2014 um 15:03 schrieb Michael von Glasow:

> On 04/10/14 02:29, Jonathan Petruska wrote:
>> I don't think a fully free cellphone is too unthinkable. Has anyone thought of porting to mini arm PCs like the Pi, or has this already been done
> A proof-of-concept phone has already been built, albeit not with Android:
> http://www.davidhunt.ie/piphone-a-raspberry-pi-based-smartphone/
> 
> As for Android on the Pi, some attempts have been made, but a lot of things are still missing:
> http://www.intorobotics.com/raspberry-pi-android-guides-resources/
> 
> Of course, there's also the GTA04, for which a port is under active development.
> 
> Such projects, however currently have a few hurdles to overcome:
> 
> - Hardware sourcing: The typical smartphone nowadays is more than just a Raspberry Pi with a touchscreen, modem and battery. Hardware typically includes one or two cameras, GPS, WiFi and Bluetooth modules, as well as a bunch of environment sensors (accelerometer, magnetometer, gyroscope, proximity, ambient light, temperature, humidity). The challenge is to find freedom-friendly components, for which free drivers either exist already or are doable (i.e. documentation of interface and protocols is freely available).

It has even become worse in the last 3 years. With only two major platforms (Apple and Android) everything has been streamlined so that newes silicon is directly going to big factories (in China). So these latest and greatest components are no longer available through the component distribution channels. Try to get an Apple A8 processor... Try to buy the latest MediaTek SoC. Try to get a display comparable to the latest Samsung Galaxy. Try to get a PoP RAM with 2 GByte. It is easier to by a device where it is already installed :)

> 
> - Momentum: A certain number of people who own and use such a device, and a certain percentage of them who contribute to development are necessary for any such project to take off. Without this critical mass, development won't be able to keep up with the pace of the industry and the product will be outdated before it is even finished. Openmoko and the GTA02 aka Freerunner was an example of where this worked quite well, but IMHO the Freerunner community was just about big enough to get off the ground.

Fortunately the pace of the industry has reduced. It was a head-to-head race between Apple and Samsung in 2012. Since then there have only been marginal improvemens. Some more pixels here, some MHz and mAh there. This will give more time until some fundamental improvement becomes necessary and will make our work a little relaxed.

> 
> - Cost: With the above constraints on hardware, a relatively low number of users (i.e. small quantities) and most likely some quality demands, such a device is likely to be costly. Take the GTA04, which has a 500 to 600 euro price tag just for the mainboard. With fully functional phones selling for half this price, one does need a lot of enthusiasm to go down that road.
> 
> - Form factor: The GTA04, just like its GTA02 predecessor, is quite a big brick. That probably has to do with hardware sourcing, production quantities and also cost.

Well, any Raspberry Pi based design is even worse on that aspect :)

The Neo900 is a step to bring the GTA04 forwards - while reusing much of the GTA04 design (incl. Replicant).

And yes, it has something to do with quantities. The barrier is around 5000 units. Then, the cost of a professional industrial designer + mould design becomes neglectible (let's say 30 EUR per unit - but it is 300 EUR for 500 units and 3000 EUR for 50).

> 
> So how can we tackle this?

> - Start with the easier tasks: A ten-inch tablet has fewer issues with form factor than a phone – its dimensions are largely determined by the screen size, which takes up the bulk of the space. The next biggest part is the battery – but the other components are the same size as in a phone, thus less is gained or lost from shrinking the mainboard or not doing so. Also a tablet does not need to include all the hardware components of a phone: a basic tablet can work on WiFi only, thus the cell modem is not a concern at that time. Once the tablet takes on shape and has evolved to a certain degree of maturity, cell connectivity can be tackled.
> 
> - Plan for a lifecycle: Over time, hardware requirements evolve as faster processors, more performant graphics and systems with more memory become the norm. Just having finished one platform doesn't mean work stops there – rather, this is where work on the next generation begins.
> 
> - Keep specs stable: We have just established that hardware evolves – and new hardware requires adaptations in software. However, such improvements should be as gentle as possible in order to limit the changes in software needed. Where possible, newly introduced hardware should be compatible with existing drivers, or at the most require only smaller modifications to existing drivers, rather than requiring completely new drivers.

That is (one of the) idea(s) behind the Neo900 evolving from the GTA04.

So far, I agree with everything you wrote up to here.

> 
> - Get a company on board: When it comes to industrial manufacturing processes and building devices in professional quality, a different set of expertise than that of a software developer is needed, and the financial limits of a handful of individuals are easily exceeded. The founders of Geeksphone one stated that getting their company to the point at which they started shipping the One (their first device) required investments which roughly equaled the price of a small house. The GTA02 benefitted from the involvement of FIC, a manufacturer of computer parts. The challenge is to find a company who has a commercial interest in the project.

Of course, a larger company (or call it funds or investor) could change things over night (like Canonical tried) if they subsidize such a project.

But let's discuss.

Firstly, getting a professional company to produce in professional quantity is no problem - if you pay them in advance (crowd-funded?). But they can't do magic in getting the price down just because they are a company and not a community. They also need high quantities to purchase components or to design a nicer form factor or whatever. So unless there is enough demand, there is no interest to manufacture.

So why should they do it? No company or investor is doing investments for fun or education or to the benefit of mankind or for freedom. They always want something back, i.e. earn money or at least reputation.

For that purpose, they do a market analysis and find exactly the hurdles we see. Especially that it is not possible to get people to pay a higher price. Because there is not enough enthusiasm in the market. So they can't earn money nor reputation (because only a very small percentage of mankind will recognise it).

Additionally, they observe that even big companies like Sony and RIM struggle in the smart phone market and take this as an indicator how difficult it is to get a return on investment. Unless this changes (and I don't see any indication that it will), there will be no big company or investor.

Now you may ask why Golden Delicious Computers is still actively supporting the GTA04? The answer is the same. It is not for profit but for self-marketing purposes and based on donations. And fun and a rebel attitude like Asterix' village :)

So my conclusion of this situation is that we simply have to accept to pay a higher price tag to get something non-mainstream that is more free and open. Unless this happens, it will remain wishful thinking that we can get all the quality, appearance and low price tags of closed devices, but with open devices. We always have to pay more for an additional feature. And as long as the group who wants this additional feature is small, we have to pay much more.

Or (as the big alternative to building open hardware) we have to invest time and enthusiasm into reverse-engineering such devices and providing alternate ROMs. But then we must be very fast - before the device is obsolete and no longer available to anyone else.

> 
> 
>> ; I know there are some built specifically for Android (CuBox, Pandaboard, etc.).  I like the idea of Replicant in the tablet world (If you can roughly consider Replicant/Android on mini PCs akin to tablets).

BR,
Nikolaus




More information about the Replicant mailing list