Well, a bootloader, though I doubt it would be necessary, could function over the programing interface just as easily as any other. Does your customer really care what the nature of the host attachment or device attachment is? Does he even know?



Erik,

If there were ONE and only ONE, programming adapter for ALL field-programmable devices, would it matter at all what's going on? The host computer, which is what the customer would attach in order to perform the field upgrade, talks through some channel. What does it matter how the channel is attached? Whether the host adapter is attached to a USB port or any other port, the adapter's driver software deals with that, and transfers the file following a rigorous protocol. Your customer doesn't know or care about the protocol. He just plugs in the adapter, runs the application, and waits for it to say "done."

It could, in fact, even be done over a USB-like channel, transferring one byte at a time and echoing it back, since it's half-duplex and very fast. I would NOT recommend the USB protocol, however, because it's WAY too complex.

As for the small packages, ... adding four pins wouldn't matter a bit, since you would simply put the device in a different package with more pins, and on a finer pitch if necessary. Some devices that have been around for decades have had programming protocols that used other pins than what one would normally imagine, just to tell it that the pins are now used for programming. Remember, too, that this 4-programming-pin mandate is not for existing devices. If you use a small package, what does it matter whether it has 8 pins on 1.5 mm pitch or 20 on 0.5 mm pitch, so long as the board area isn't increased? Nobody see's 'em anyway since they're on the underside. I don't like CSP's and BGA's, but they're what's "in."

RE