??? 04/19/06 17:27 Read: times |
#114552 - It's been a few years ... Responding to: ???'s previous message |
I had some ARM experience about three years back, which was shortly after Samsung released a series of ARM7-based microprocessors. These all included SDRAM and FLASH/SRAM channels and ethernet, async serial, and various other channels of I/O, along with some parallel I/O. They came on moderately priced boards completely supported under LINUX, which included monitor and driver, as well as OS kernel sources for all their internal peripherals. Linux, of course, provides all the tools under GNU license, though there are others. That ARM version was priced at about $7 in quantity, and the EVK cost about $200US.
I didn't do any of the software development, but, given that the S/W guy had LINUX up and running on the day the EVB arrived, I have to say it must have been easy. Along the way, the boss decided on a switch of CPU's, so the next EVB was brought in, and, again, LINUX was running within a day or so. This one was another Samsung version, with more of the hardware we needed, and cost about $9US in quantity. The EVK cost about $200. Eventually we made a switch to a different mfg's ARM-9 product, because it had hardware that we'd otherwise have had to provide externally, though the CPU cost about $18 in quantity, but the EVK cost about $200US. The conclusion I drew from this exposure was that the ARM is well supported with a generous array of tools, and is easy to develop. The CPU's cost very little, and their performance is good. I'd certainly consider off-the-shelf ARM's for any application that required external memory. I would take a careful look at the spec's for any ARM-based product, however. My experience with the US-made ARM-9 part was that the mfg didn't know enough about the internal workings to tell me the relationship between its clock edges and its external memory interface strobes. While their SDRAM channel worked just great, they quite freely admitted that they'd never attempted to use external SRAM with it, and, since the bus cycle was only 20 ns long, it would have been nice to know on which edge of the clock the strobes became valid and later became invalid, as this impacted external memory subsystem design. Apparently, they'd bought the branch that produced this CPU and that, by then, subsidiary had engaged a consultant who was no longer available to fill in the missing details. That was unpleasant. There are single-chippers based on the ARM core, too, though I've not used them. They'll certainly get a good, hard, look from me next time I have to come up with a 16- or 32-bit design, becuase of the rich tool set that's available. RE |