The simplest tool I can think of that appears as an SPI slave and allows unrestricted access to all the bits to completely determine an arbitrary mode would consist of a header with 120 jumper positions (240 pins) and a parallel-load shift register 128 bits long. (The extra 8 shift register bits are for the fixed header byte that tells the master that this is a video mode initialization tool, and not some other kind of SPI device.) There would also need to be a 6-pin header for the ribbon cable that connects it to the graphics board where the SPI master resides (the SPI port on the OGA ASIC or OGD1 emulation). Possibly we might need to add a few gates to emulate the behavior of an SPI EEPROM; this adds very little cost. (We'd need at least one tri-state or open-drain gate, to turn on the data output upon receiving the NOT CS signal from the master.)
The shift register could be implemented with sixteen 74HC165s. Digi-Key lists it for 11 cents at 2500 quantity, tape-and-reel, so this item comes to $1.76. We would need six 40-pin headers to accept the jumpers. On a quick search I found a Molex 40-pin header priced at $1.95 at the 500 piece level; this contributes $11.70 to the parts cost. A 6-pin shrouded header for the ribbon cable is 60 cents. We need a bag of jumpers, 7.8 cents each at quantity 50, 120 required, $9.40. The price of the PC board will vary greatly, depending on quantity. If it's implemented as a low-density two-sided board with plated-through holes, there are vendors advertising in EE Times that can make two for $33 each. For a modest run of 100, I'd guess we could get them for $10 with careful shopping around. I'm not allocating cost for a case. I don't think it makes a lot of sense, for a board with exposed jumpers. We might want four snap-in plastic feet. With the above assumptions, I come up with a parts cost of $33.50 in moderate production quantity. With DIP switches substituted for headers and jumpers, we'd pay about 9 cents per switch position in quantities of 100 switches, so 120 switch positions would cost $10.80 and replace $21.10 worth of headers and jumpers. Hmmm, that's unexpected. The only downside is the increased board area, but that doesn't matter in a tool. Now the parts cost is $23.16. Phooey, I forgot the pull-up resistor packs. We need 12 SIP networks at 15 cents each. That's another $1.80. We'll want a few bypass capacitors, maybe an overvoltage protection diode across the power pins. Maybe a buck worth, all together. I don't have a good handle on what a contract assembly house would charge to assembly a batch of boards and package them for shipment, though. We'd need to shop around for a small-scale vendor. Other options... The ICs could be socketed, so that they could be replaced quickly if one get blasted by ESD. Sockets cost more than the ICs, but they're still pennies. The sockets could be auto-inserted on an automated board line, but the ICs would have to be plugged in by hand, probably by the customer. Traversal might want to offer the bare board for sale, but a kit would make no economic sense unless the quantities are too low for the up-front cost of machine assembly. A pick-and-place machine can assemble the board for less labor than making up individual kits. A second question you didn't ask is what a more capable tool with a microcontroller would be like. A 68HC908AB32 costs about $6, as I recall. Add an RS-232 interface IC and required glue logic, and we're probably in the $15 range, plus the bare board. I'm guessing $25, maybe $35 with a case. The downside to OGP is that it would require some on-board assembly code to make it do anything, which would have to be specified and written. Also, it wouldn't be a substitute for the simple set-the-power-up-mode tool, because it would require something on the other end of the RS-232 cable to tell it what to do. -------------- Original message ---------------------- From: "Timothy Miller" <[EMAIL PROTECTED]> > On 8/1/06, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote: > > > Now suppose that some SPI device on an external board is plugged into > > the > header. It might be a serial EEPROM programmed by some lab tool, it might be > a > plain shift register connected to a bank of DIP switches or jumper headers, > or > it might be a more sophisticated test tool with a microcontroller in it. > Regardless, it responds when selected the same way a serial EEPROM would, and > drives the data input to the TRV10. > > > Can you describe the least complex type of equipment that one would > need in order to inject an SPI signal in this way? And how much does > that cost? _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
