Re: Whither open hardware ?
Dave Ball wrote: What's the yard stick for measuring against here? I.e. are we talking about one-off from digikey/farnell, samples direct from the manufacturer, or limited-run (couple of hundreds) quantities? For the full process from RD to mass production, you need to have channels for small, medium, and large quantities. First just a few to figure out if and how the thing works. Then hundreds for the prototypes, and finally thousands for mass production. If a part is available from Digi-Key or a similar distributor, this helps enormously to accelerate RD and it can also help when in a pinch in a prototype run. - what are the integration costs ? is this things like placement of awkward (small pitch etc.) parts, FPC's etc., or ancillary parts such as partner chips? Depending on the part, it can be all of this and more. Requirements on the PCB and the SMT process, partner chips, extra voltages, mechanical and thermal issues, drivers, and so on. Examples: - if a new component reduces the minimum pitch or has a higher pad density than the rest, your PCB may get more difficult to make, possibly resulting in higher cost, a smaller choice of companies that have the technology, higher lead time, and so on. - some components have an unusual reflow profile, e.g., batteries (don't like the heat) or complicated BGAs (have to make sure even the most inaccessible ball reflows correctly). - CPUs have long lists of requirements on their power supplies and their sequencing. E.g., it was quite a puzzle to figure out how to make the 2442 with with the 50633. - extra voltages: some chips inexplicably want something slightly different from 3.3 V. There goes another LDO. - mechanical: need to find suitable space. Electromechanical components also need to interface mechanically, which may affect the shape of other elements. - thermal: don't cook your neighbours and don't be cooked by them. Also, some special layout may be needed to get the heat away from the chip. - let's not forget the software. If a chip needs a driver, that one has to be written, debugged, and so on. This also implies what one needs sufficiently open documentation, which can require a great deal of negotiation. Is the normal route of sourcing via a factory (even for prototypes etc.)? From a few searches it seems that getting hold of some parts (i.e. screens / touch layers) is incredibly difficult for one-offs. For the easily obtainable parts, you have many choices. Small quantities you get from Digi-Key, even if it's expensive per piece. Perhaps even medium quantities, if you don't already have a better channel. Large quantities, you get from the official distributor. If you're willing to take some (small) chances, you can also use other channels, e.g., to bring down lead time. Parts that are hard to get require contacts, muscle, illusions, or someone who can lend you some of these. You normally negotiate the whole package, so you don't only get samples but you also at least talk about the larger quantities you'll need in the future. For prototypes you want fast turn-around times, so involving a mass-production factory may not be such a good idea. They can do sourcing, but their mode of operation may not include quick changes and such. (E.g., GTA01 was prototyped in Taipei, GTA02 had many prototype runs at the MP factory, a process that was agonizingly slow and had a huge overhead, GTA03 was prototyped again in Taipei.) - Werner ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Whither open hardware ?
Hi Werner, Werner Almesberger wrote: There are many component choices for future phones. Things to consider when choosing chips include: snip - are they available (to us) ? What's the yard stick for measuring against here? I.e. are we talking about one-off from digikey/farnell, samples direct from the manufacturer, or limited-run (couple of hundreds) quantities? - what are the integration costs ? is this things like placement of awkward (small pitch etc.) parts, FPC's etc., or ancillary parts such as partner chips? - do they work as intended ? Hehehe. :-) Is the normal route of sourcing via a factory (even for prototypes etc.)? From a few searches it seems that getting hold of some parts (i.e. screens / touch layers) is incredibly difficult for one-offs. Dave ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Whither open hardware ? (was Re: Quick e-mail poll: Still using your Freerunner?)
Ken Young wrote: My two cents: If I were dictator of the gta02-core team (instead of someone who doesn't even contribute), I would repurpose the device as a GPS PDA. I would remove all the radio components except for the WiFi, and try to optimize for the longest battery life possible. Companies who were looking for a device for some project often asked Openmoko Inc. if they could have a GTA02 with some features removed or with other - often small - changes. Unfortunately, Openmoko Inc. did not have the resources for making such derivatives. However, this is a promise the approach chosen for gta02-core holds: with the whole design out in the open (Open Design Hardware [1]), anyone can independently define, implement, and produce derivatives. [1] http://people.openmoko.org/werner/openness/odhwdr-v1.pdf This doesn't mean that everyone is forced to fight all alone. To the contrary, there are many possible synergies along the way that are not visible as such in the traditional product development process, such as shared sourcing or shared manufacturing. For example, if you want to make your GPS PDA, you may choose a set of changes that fits your budget, e.g., by staying with the overall platform and physical shape but removing subsystems you don't need. When it then comes to sourcing components and manufacturing, the same facilities used for making the phone could offer their services also to your project. The incremental cost for them would be very small, much smaller than running a completely different product of similar complexity. Also the core company (or whatever form of organization) in charge of the base design would benefit. If it has spare engineering resources to put into derivative projects, it can choose to do so, favouring projects that best suit its agenda. If not, others can help out. Thus, the business opportunity is not wasted - it only goes to someone else you could think of as an ally. Even better, resources that can be shared contribute back to the whole ensemble of projects. E.g., if your PDA is wildly successful, sourcing may be able to get much getter conditions for parts than they did with just the phone. Or a new type of subsystem gets researched and is then available as a possible building block for the entire architectural family. Thus also the phone benefits. Now some may say that this is crazy and that anyone handing out designs so liberally would be robbed by competitors. In my experience, it's surprisingly hard to get people to steal your cool new ideas. Eventually, the thieves and parasites will show up, but you have to be successful for an awfully long time before they even notice you. [...] but rather to point out that there is really no hope at that a group of people such as the gta02-core team, working part time with no large corporate sponsor, will ever produce a product with hardware on a par with what the big players are contemporaneously offering. I agree on the point that there's no hope to mass-produce a phone without suitable resources. The resources don't have to be in one hand (e.g., you could have a consortium of entities each contributing their own capabilities and splitting the proceeds), but they have to be available. However, I don't think it's necessary to compete on leading edge technology. Often enough, less advanced components will yield an equally satisfying product. Besides, companies that don't have the sexiest product in their sector of the market are often much friendlier towards openness than those who do. Please don't take the poor performance of GTA01 or GTA02 as too much of an indicator of what second best can do. Both are based on very conservative designs (e.g., no DDR) and GTA02 has two thirds of its high-throughput peripherals share an incredibly slow bus. (Think of a first-generation PCI-based PC where someone chooses to use ISA cards for video and the SCSI controller. Would such a system properly represent the typical performance of the PCI architecture ?) The 2442 is now about five years old, and it shows all over the place. In an updated but similar design, unlike gta02-core suitable for mass-production, I would use something like the 2450, which has high-speed USB, 2D acceleration, and other goodies. In contrast, I think there still might be an unexploited niche in the GPS-PDA arena. I think there is a whole universe full of unexplored niches. It's hidden from us by a tall wall called high cost of entry. If we can find ways to lower that wall, a lot of interesting things should happen. - Werner ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Whither open hardware ? (was Re: Quick e-mail poll: Still using your Freerunner?)
On Sun, Jan 3, 2010 at 3:47 PM, Werner Almesberger wer...@openmoko.org wrote: The 2442 is now about five years old, and it shows all over the place. In an updated but similar design, unlike gta02-core suitable for mass-production, I would use something like the 2450, which has high-speed USB, 2D acceleration, and other goodies. There is also the openpandora project: http://www.open-pandora.org/index.php?option=com_contentview=categorylayout=blogid=2Itemid=2lang=en Is it unsuitable for a phone because of power inefficiency? Can be the ARM Cortex-A8 600Mhz used in a future phone? Laszlo ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Whither open hardware ?
Laszlo KREKACS wrote: There is also the openpandora project: http://www.open-pandora.org/index.php?option=com_contentview=categorylayout=blogid=2Itemid=2lang=en Is it unsuitable for a phone because of power inefficiency? Can be the ARM Cortex-A8 600Mhz used in a future phone? The pandora (and beagleboard) use the OMAP3530 which (afaik) is just a retail package of the (oem only) OMAP3430 used in the palm pre and motorola droid. [1] The docs are open [2] (except the power VR 3D subsystem), and from first looks it should be fine in a future phone - though it would be a radical departure from our existing designs. Dave [1] http://en.wikipedia.org/wiki/Texas_Instruments_OMAP [2] http://focus.ti.com/docs/prod/folders/print/omap3530.html ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Whither open hardware ?
On Sun, 2010-01-03 at 20:49 +, Dave Ball wrote: Laszlo KREKACS wrote: There is also the openpandora project: http://www.open-pandora.org/index.php?option=com_contentview=categorylayout=blogid=2Itemid=2lang=en Is it unsuitable for a phone because of power inefficiency? Can be the ARM Cortex-A8 600Mhz used in a future phone? The pandora (and beagleboard) use the OMAP3530 which (afaik) is just a retail package of the (oem only) OMAP3430 used in the palm pre and motorola droid. [1] The docs are open [2] (except the power VR 3D subsystem), and from first looks it should be fine in a future phone - though it would be a radical departure from our existing designs. There is one...it's called omap zoom II but it's not cheap If I understood well the GSM modem is optionnal Maybe I should have bought one instead of an htcdream (better buy now when the dollar is cheap) Denis. ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Whither open hardware ? (was Re: Quick e-mail poll: Still using your Freerunner?)
Hi, I would love to have an open phone with OMAP processor like OpenPandora and BeagleBoard. Both have the OMAP 3530 with 3D acceleration and DSP. Maybe a phone doesn't need the DSP. But it is amazingly powersufficient, IMO. The phone would be up to date from processor power, and there should not be a resource problem so soon, I guess. BeagleBoard and OpenPandora are both well supported in OE, so from SW-side there should not be so much effort (in theorie, you kow...). From HW-side another nicy is the POP (package on package). In the BeagleBoard they are using a variant, where the RAM and NAND (both in one Chip) is soldered on top of the processor, so the high speed lines to DDR-RAM are not going through the PCB (on the other of course, you have the process of soldering...). ... just dreaming ... Joerg On Sun, 2010-01-03 at 20:39 +0100, Laszlo KREKACS wrote: On Sun, Jan 3, 2010 at 3:47 PM, Werner Almesberger wer...@openmoko.org wrote: The 2442 is now about five years old, and it shows all over the place. In an updated but similar design, unlike gta02-core suitable for mass-production, I would use something like the 2450, which has high-speed USB, 2D acceleration, and other goodies. There is also the openpandora project: http://www.open-pandora.org/index.php?option=com_contentview=categorylayout=blogid=2Itemid=2〈=en Is it unsuitable for a phone because of power inefficiency? Can be the ARM Cortex-A8 600Mhz used in a future phone? Laszlo ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: Whither open hardware ? (was Re: Quick e-mail poll: Still using your Freerunner?)
Laszlo KREKACS wrote: Is it unsuitable for a phone because of power inefficiency? Can be the ARM Cortex-A8 600Mhz used in a future phone? There are many component choices for future phones. Things to consider when choosing chips include: - do they fit the intended purpose ? - are they open enough for our purposes ? - are they available (to us) ? - will they be available as long as we need them ? - are they affordable ? - what are the integration costs ? - what are the opportunity costs ? - do they work as intended ? - how do they fit our technical capabilities ? - what legal exposures do they cause ? Of course, you don't see companies advertize much on the issues listed above. Quite to the contrary - how often does one see a feature presented as patented proprietary technology as if this was a good thing ? - Werner ___ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community