[Community] Releasing GTA04 hardware source files
Dr. H. Nikolaus Schaller
hns at goldelico.com
Sat Aug 24 20:30:45 CEST 2013
Am 24.08.2013 um 17:03 schrieb Sven:
> Hello Nikolaus,
> thank you for being such straightforward clear and honest. That's the right basis to continue the discussion.
> You know that I'm an employee of a medium-sized company that produces microelectronic boards for industry - note well in Germany. For a long time we did the same as you say: Selling boards together with printed schematics and extensive manuals. But recently we started to also provide sources of schematics and layouts at least for some special socalled OEM-boards. Meanwhile we do offensive marketing for that. When I asked the manager why on earth he made this decision, he switched on his beamer and showed me that meanwhile there're tools online in the internet with which you're able to click together even highly sophisticated circuits with only some mouse clicks. Simply put a processor together for example with some interfaces for ethernet, USB and SD card, click on a button like "finished" and you'll get the whole sources of a complete single board computer ready for download and without any costs, ready for production.
I think this needs a lot of preparation and standardization to be possible by some "system compiler". If we have the target to integrate as much as possible into given space (GTA02 case) it might fail. It might work in your case because you have less restrictions.
> So please let me be the one who rises his finger in order to explain you, why Bobs idea is the one for which you're asking:
> Based on the sight of someone like me, who doesn't have time and due to this reason also not the knowledge to contribute the GTA04 project by development, the progress of GTA04 project currently got stuck at two fronts: Hardware and Software.
> Software: The most advanced software for GTA04 is QtMoko v55.
> The main person contributing this made the strategic decision to focus on Replicant now. For a long-term-view this decision might be right, because it ties in more developers into the whole project.
Yes, I have the same view that it might be better to get Replicant onto the GTA04 with millions of application developers. Better than continuing a really nice niche OS like QtMoko, but where essentially nobody is developing applications for.
This is why I did discuss a lot with Paul about what we require to get a 3.11 based kernel that supports Replicant 4.x and is feature complete and a good basis for fixing all the issues discussed elsewhere and listed in the issue tracker (e.g. power management, 1GHz operation).
Work on this will speed up a little in the next weeks, after end of the holiday season.
> But the short time result is, that since nearly half of a year there's been no new release of QtMoko. Especially we don't have a v56 with working camera. This extenuates desires of non-developers very much, resulting in very few preorders for you. This problem cannot be solved, because I think Radek refuses to be cloned. ;-))
Well, we should send him a "clone" message :)
Anyways this is my current strategy for software:
1. get a kernel that works
2. is up to date and ideally can be downloaded from kernel.org
3. fully supports Android/Replicant
4. has many applications and
5. many application developers who
6. need devices.
> Hardware: This problem is well known by you. Every aspects have already been discussed largely within this list. But I disagree heavily with your claim, that you will not have any benefit when you share your sources. Would we ever have Linux if Linus would have argued the same way? No!
We would have GNU/Hurd. It was already in development back then and Linus came with a different solution that did get more traction for reasons that I don't know. But definitely it was not openness which differentiated Linux from Hurd kernels.
> Never! We all know that Linus up to now states that the best decision in his life was to put his first sources under the GPL. And how great was/is his success!
Yes, this is a great success.
> So I tell you: What Linus did with software, please follow him with hardware! That's the solution for the problems! It even will fill up your refrigerator the same way, because you're still the one who has all the knowledge about production of someone, who is and will still keep a member of our community! Or in other words: If for example I would have the time again like more than 20 years ago to grub through all this stuff and then develop my own ideas, who would be the person/the company to which I want to commit my results? Of course you and GD!
> Of course it would be up on you to live with the same problems that Linus has: The commits offered to you will not necessarily be of the same quality you expect. But that's your chance to develop yourself and GD: Find a way to manage this!
> I even offer you the first step. Please make the following intellectual game: You're the manager of TI in the early 1980ies and be proud of offering a computer named TI-99/4A:
> Now I come to you and proudly present you my perfect working further development of your machine. I even gave it a name. I call it TI-99/Proton:
> What will you do? How will you react? Will you still insist on your claim that it's better to do all hardware development by yourself? Or will you be able in 1985 to prevent TI from having to make the decision to cancel the whole TI-99 project?
Hm. This makes me a little puzzled since it mixes two things. It appears that you make the assumption that not providing the project files is identical to not accepting proposals for improvements from the community.
We are really open to any hint to improve something, and indeed there have been proposals on these lists and by private mails in the past. None of them were simply rejected but discussed. They all were evaluated and some were accepted because they were better. By this process we did get from the A3 board to A4 and to A5.
And it was never a problem that such proposals were written in plain text e-mails and not a ready-to-apply patch for some schematics file. And they did come from people who did not have access to the PCB layout files etc.
In fact this way of work even lowers the barrier for others to make contributions since they can simply describe what they want to change. Doing the change in the specific tools is done by those who are experienced in operating the tools.
So again I conclude that publishing all the project files does not solve an existing problem.
The real hardware problem is completely different: it is a mismatch of the sum of the component prices with the total price people are accepting for the sum of features we get, i.e. the perceived value vs. the real cost.
If we could offer exactly the existing and unmodified GTA04A5 for let's say 149 EUR we would sell more than Ubuntu Edge did have supporters.
To give you some background:
we can get single DM3730CBP100 chips e.g. from Mouser @40 EUR (roughly). To this we have to add memory chips, glue chips, some expensive capacitors, many cheap ones and resistors, connectors, PCB, Modem and all the other things. This sums up to a big amount of money. I don't know the current values, since they change daily (USD vs. EUR exchange rate) and strongly depend on the number of chips we buy and how good we negotiate in a specific situation. E.g. for the DM3730CBP100 you can look at Mouser's catalog how prices go down for 10, 25, 100, 500, 1000 chips. Well, we can even use a different source (Avnet Express, or some others) and they may be cheaper, but it does not change the problem significantly.
After adding production cost, testing, some savings for warranties and other defects, WEEE, VAT, and some savings to finance future developments the price simply sums up to 599 EUR if we build 200 units.
Now people say: but at 599 EUR I want to have an Überphone feature set which they don't get with a GTA04 based design.
Or they say that they can get the same feature set from some unknown chinese fab for 129 EUR. Well they don't have the schematics and the kernel source, but at 129 EUR vs. 599 EUR they can trash it asap...
This is the real problem we will have with any hardware (being it a GTA04, a GTA05 or anything else) where I don't know any solution.
And there have been many proposals to address the production cost:
* manufacture in China - well, they don't even start to discuss for 200 units
* take cheaper components - there is only one DM3730CBP and no cheaper variant; some cheaper components are too big to fit them into the GTA0x case
* leave out some less important features - if we remove e.g. GPS, WLAN, Bluetooth, Sensors etc. the total cost goes down to approx. 80%, i.e. 479 EUR. But for a smartphone w/o WLAN, GPS, Sensors etc.! I.e. the feature set is going down to 50%
* use an OMAP4 or OMAP5 - those are not available in small quantities at catalog distributors and even more expensive
* use a cheap SoC module - almost all of them are too big to fit into a smartphone case
The only "solution" is to convince people to pay the cost it will have to produce more boards. Or there will be no GTA04A5 :(
So in my view it is a set of equations with zero solutions and I would be very happy if someone can point us to another approach for the production cost vs. perceived value problem.
And this is an area that has nothing to do with design tools, design files etc. and I have no idea how opening them more would change anything in the equation system.
Since somebody mentioned in this discussion (maybe it was on the openmoko community list): yes, this is a slightly negative view. But it is based on experience and about 2 years trying to find a solution. But it is not so negative that I reject any improvement. It is more like in scientific research where you may have to run experiments for several years until you see if your theory is right or not. Or even someone else is pointing out that you are right or not. And I am more or less waiting for such an event that someone is coming with a really brilliant idea.
More information about the Community