| ??? 04/10/08 16:39 Read: times |
#153119 - I fear you've missed my point Responding to: ???'s previous message |
Maybe this belongs in a separate thread, as I fear it's no longer germane to the original subject ...
Erik Malund said:
in his post entitled the problem with
... JTAG is that you need to write a bootloader for on-site programming Why? I've never had to use a bootloader with JTAG. No JTAG programmable device I've ever used has required one. Bootloaders are use in place of more flexible hardware, I believe. ... serial ISP is that you are stuck if your thingy does not use serial otherwise, or, for that matter, is not 232 Well, how are you supposed to program a device that has no async serial facility? Remember, I'm saying that ALL field-programmable hardware devices (limited to IC's, of course) should be programmable with exactly the same host interface. Very few of my devices require async serial communication, and, of those, very few need RS232 levels or timing. ... SPI ISP is that everybody and his brother screw up something and post it as a "working" programming circuit Are you saying that if it's possible to screw something up, no one should use it? I suggested the JTAG host-interface hardware because it's so widely used with other programmable hardware. ... all of the above is that too many refuse to pay the nominal charge (SILabs $39, Atmel $25, ...) for the real thing. It is unfortunate that many think they can 'play' with uCs for a $5 investment. This, of course, is aggravated by the fact that this critical circuit usually is the first thing tackled. Whether one builds or buys is a decision, not an indictment. Your perception is probably correct, though. What I'm saying is that there should be one and only one host interface to an ISP circuit. I've said nothing about the details of the circuit, though it would undoubtedly have to be built into every ISP-capable device, which means that some 2-IC boards would suddenly have more ISP-related parts than otherwise functional ones. rather than dozens of versions having dozens of slight differences Richard, I am sure both you and I have finished and released something and then said "I should have done it this way". I guess that, maybe, 'new' silicon technology has something to do with it as well. I don't doubt it for a minute. However, once you release something, that's what you've got to "live" with. Changing it next time won't help if what you need is a "standard." If it was properly specified at the outset and your design meets the specifications, it is perfectly adequate. Erik All I'm advocating is that all manufacturers of programmable IC's ranging from MCU's to memories, should have the same 4-wire interface for programming. Further the timing and electrical specifications should be identical for all devices, no matter how simple or complex. One host interface should handle all field-programmable IC's. Back when JTAG was a pup, I got several host interfaces, all free from the IC vendor, and all different, from various manufacturers. It probably wasn't easy for them to arrange the four signals in a different yet convenient (for them) way. Each one had to "protect" his IP, doncha know! I promptly built a single one that would handle 'em all, using a small crosspoint switch that allowed me to set, manually, the associations between the vendor-supplied software and the host port hardware. The result was that, if I needed multiple vendors to be used, I could program each of them with their own software, yet not have to swap hardware all the time. Though one would have only a JTAG-type 4-wire interface and one of these, http://www.emulation.com/marketing...probe.html it would be easy enough, with the aid of a CPLD and a small PCB, to make popular FLASH memory devices field programmable without removing it from the circuit board, though some advance accommodation might have to be provided on the board. Keep in mind, please, that what I'm suggesting is that there be ONE and only ONE host-adapter, and ONE and only ONE device attachment allowing those same four signals, for all new field-programmable devices. What goes in the middle can be quite arbitrary. Adapters will surely appear for currently existing devices once the concept is proven. RE |
