| ??? 09/17/09 08:38 Read: times |
#168929 - Still, majority of embedded systems are small Responding to: ???'s previous message |
You are forgetting yourself.
You ask: "What AVR/805x board can provide all that for comparable cost?" Irrelevant if my goal is to create a keyboard controller. I want a final product. An evaluation board with a gazillion features are of no use if the target hardware will never use any of them. Among other things, I work with CAN. What use is then that evaluation board if it doesn't have CAN? Testing is what you do all the way through the development of your firmware or hardware. It is not just something you do when you have a release candidate of hardwaer and/or firmware. About the board you did suggest. Throw away the LCD and the beeper, but add in CAN and some other features, and you will have something similar to one product I have developed the software for. Yes, the product do run Linux. But no, I'm not interested in running any development tools on it. No need to. TCP or BDM can be used if I want to debug in the hardware. Much of the software can be tested directly on a PC. Coming from a 8051, that development board may look like "wow". It normally isn't. The development board is only useful if I am going to use a processor from that specific family and will use that processor with a reasonable amount of the peripherials that the board has. If not, then I may just as well get a reference design for the real processor and drop in the peripherials I need and order a couple of prototypes. If debugging, I would prefer to have gdb running on a PC, and not on the target. When compiling, I would prefer to have gcc compile on the PC, and either copy the binary to the target system, or let the target system export or import a file system. Richard said:
Now, I don't sell these, nor do I even hold stock in the company, but it looks like a pretty inexpensive way to "get something out the door. It is SOP to start from a reference design, if the resulting product is intended to be running Linux. And in that case, this board is just one out of hundreds of boards. For very small series, you may ship this kind of board in your product. For larger quantities, you want your own hardware that is custom-designed. And you want prototypes of that hardware as soon as possible. Quite often, these reference designs may contain peripherial components that have become obsolete, so if a product is intended to ship for a number of years, you want 100% control. One important thing is that the market is way smaller for high-end solutions. For each embedded system running Linux, you may find 10 or 100 or 1000 systems that are too small. There are a lot of situations when the total manufacturing cost has to be $10. This is doable with an ARM chip, but obviously not by basing the design on such a board. That board may be very funny for a hobbyis, or for doing school projects, but a large percentage of commercial projects has completely different needs. And using cross-compilers really is no problem. It is normally an advantage. I don't really get why you in some threads can spend hours bashing C, because you are so convinsed that a program written in C will require one step larger processor, adding cost and having your customer thinking you are incompetent. In a completely different thread - discussing the slowdown of this forum - you are ignoring comments that ARM9 or ARM11 or whatever are nice but a lot of the ARM work is with smaller systems. Now you seem very busy explaining why complete embedded Linux systems are a good way to develop our products. A huge percentage of ARM chips sold, are not capable of running Linux. And they are selected for the simple reason that they are cheaper. Way cheaper. That is money on the bottom line. |
