Steve,

SATA works very well, and is the natural descendant of the parallel version that was popular up to a few years ago.

However, if someone comes to you with the old box, containing an 8" Winchester drive, which is what I described, it doesn't have the SATA interface. You're then forced to work out a way of transferring data over the available interface. This means, you have to build a "bridge."

Even if the serialized interface operates at 10 GHz, there's latency that doesn't help you. If you can avoid it, it's easier to manage the other timing constraints. The sort of "box" I described uses a parallel interface. It's close to what's used on a number of others, but not exactly. The particular type of device I had in mind uses "SASI" (Shugart Associates System Interface," which is a precursor of SCSI. It won't work on a SCSI interface, and it certainly won't work on ATA or SATA. People have data on these old boxes, and, when their PDP-11 breaks down and is finally hauled off, they have to have a way to retrieve the data. I've had others that worked similarly, but also weren't "exactly" the popular interface that they resembled. Medical equipment is still in use, for example, that uses 8" floppy disk drives. Now, these are going away at a fast pace, but I've designed and built interfaces between these boxes and PC's for the purpose of transferring the records stored in that way to common media. 8" floppy disk was once the "common medium."

What Jan just pointed out, and, essentially, what I've been saying for years, is that, "Where you sit determines what you see." I've seldom accused Erik of being totally wrong, at least up to the point of nomenclature with respect to 805x architecture. However, he necessarily sees things with different eyes than I do. He designs and builds boards that address his requirement sets, while I most frequently adapt COTS (Commercial Off The Shelf) boards to my ultra-low volume requirements. I've done what he does, but for the past few years have focused on these low-volume, highly specialized tasks. Microcontrollers are VERY well suited for them, as one generally can implement the "specialized" hardware in the prototype area on such a board, and handle the bulk of the task in firmware.

What I've been trying to point out to everyone is that, while a potential solution to any problem may exist, it may not be readily implementable.

If I have a task of producing one or one plus a spare, of a given interface, typically a bridge, it's not realistic for me to use a specialized microcontroller unless it very directly addresses my need. Unless a particular device physically fits my applicaiton circuit, which SMT parts seldom do, it's not worth my time to consider it. There are cases where there's sufficient need to warrant a new PCB for a given application. Most often, though, I use a board with an MCU and an adequate prototyping area to accomodate the specialized hardware I need. Sometimes, rather than adding programmable logic to the MCU card, I use a programmable logic board with sufficient prototyping area to permit adding the MCU. I like doing this sort of thing, and I like addressing these challenges in this way.

If I were building hundreds per day of each of my designs, which I've certainly done often enough, my approach would be different. I suspect, however, that my perspective aligns more often with the folks who come to 8052.COM with their questions than does that of the guy who's ordering his components in 10K/week quantity. Some of what's been discussed here has been intended for large volume. However, what works for me will work in the large volume environment as well. It just works better with considerations closer to what Erik proposes when you're working at high volume.

RE