Once again I find myself contemplating using 0402 sized components. The 0603s just look so big compared to the FPGA pin spacing...
With the inner two layers under and near the FPGA taken by the +3.3V and Ground planes, and the top layer broken up by the keypad footprints and connecting matrix, I'm having to do a lot of routing on the bottom. The decoupling capacitors cover a lot of the periphery of the FPGA because they can't be mounted on top, so I think I'm going to use 0402s for the 0.47uF (high-frequency) capacitors that should be close to the chip. The 4.7uF (low-frequency) decoupling capacitors can be placed farther out, so they'll stay 0603s.
Watch this blog when I start assembling these boards for "Why do I do these things to myself?" type postings as I try to place these flakes of pepper on my board.
Experiments with the world's first microprocessor
and other electronic explorations
Thursday, May 10, 2018
Wednesday, May 9, 2018
P170-DH Replacement Board layout progress
It may look like someone threw a handful of colored sticks on a black table, but this is the PC board that will go into my gutted Canon P170-DH calculator:
The mechanical outline, mounting holes, vacuum fluorescent display pads, and positions of the keypad contacts should be correct within a half-millimeter. The heavy gray lines that appear to partition the board are the locations of the support ribs on the bottom of the calculator shell that help support the PCB against the pounding of the keypad buttons.
The green square in the middle of the board is the Xilinx Spartan-6 LX9 FPGA; green pads mean it's on the back of the board. The 6x1 connector footprint at the top left is for the JTAG programming interface to the FPGA. The 20x2 connector is intended to allow me to plug my logic analyzer into the board through a ribbon cable; all free (and some semi-free) pins will be wired to this connector to maximize my debugging capabilities. The alternative would be the Xilinx ChipScope debugger which works through the JTAG, but I don't have the budget for that.
Clearly this isn't complete, given that there are parts that haven't been placed yet. Some circuitry hasn't even been added to the schematic, like the external serial interface. There's plenty of room for it, as long as I'm careful to avoid putting a via through a keypad footprint or a component under a support rib.
I figure I'm about a month away from sending this out for fabrication. When it's ready I'll probably use JLCPCB in China. I can get 5 boards from them for less than half the cost of one through US or European fabricators. If it was a matter of saving $5 or even $10, I'd buy locally (OSH Park does a great job), but when we're talking over $200 I'm going with the least expensive place with a good reputation.
BTW, this is all being done with the KiCad v5.0.0-rc2-dev pre-release. There are a few bugs still on the critical list, but it's remarkably stable. The v5 release fixes many of my biggest complaints about KiCad, and the developers are being very proactive about fixing problems as they learn of them. I've submitted about 25 bug reports thus far, most fairly minor (though one was graded a medium priority) and many were fixed within a day or two of being reported.
The mechanical outline, mounting holes, vacuum fluorescent display pads, and positions of the keypad contacts should be correct within a half-millimeter. The heavy gray lines that appear to partition the board are the locations of the support ribs on the bottom of the calculator shell that help support the PCB against the pounding of the keypad buttons.
The green square in the middle of the board is the Xilinx Spartan-6 LX9 FPGA; green pads mean it's on the back of the board. The 6x1 connector footprint at the top left is for the JTAG programming interface to the FPGA. The 20x2 connector is intended to allow me to plug my logic analyzer into the board through a ribbon cable; all free (and some semi-free) pins will be wired to this connector to maximize my debugging capabilities. The alternative would be the Xilinx ChipScope debugger which works through the JTAG, but I don't have the budget for that.
Clearly this isn't complete, given that there are parts that haven't been placed yet. Some circuitry hasn't even been added to the schematic, like the external serial interface. There's plenty of room for it, as long as I'm careful to avoid putting a via through a keypad footprint or a component under a support rib.
I figure I'm about a month away from sending this out for fabrication. When it's ready I'll probably use JLCPCB in China. I can get 5 boards from them for less than half the cost of one through US or European fabricators. If it was a matter of saving $5 or even $10, I'd buy locally (OSH Park does a great job), but when we're talking over $200 I'm going with the least expensive place with a good reputation.
BTW, this is all being done with the KiCad v5.0.0-rc2-dev pre-release. There are a few bugs still on the critical list, but it's remarkably stable. The v5 release fixes many of my biggest complaints about KiCad, and the developers are being very proactive about fixing problems as they learn of them. I've submitted about 25 bug reports thus far, most fairly minor (though one was graded a medium priority) and many were fixed within a day or two of being reported.
Tuesday, May 8, 2018
More info on the SST21x MOSFETs
When I searched for a MOSFET for use in my i4004 reconstruction, the only NMOS device I found that met my requirements was the Phillips (now NXP) BSS83. Production of the NXP BSS83 ceased in 2015.
In 2013 I found the SST211, SST213, or SST215 by CaLogic might be a suitable replacement, though the cost was significantly greater. Ouch.
When I search for the SST21x, I see four manufacturers: Siliconix, Vishay, Linear Systems, and CaLogic. Tonight I came across a press release from 2001 that explains some of my confusion over who makes these chips. It looks like the SST21x series was created by Siliconix. In 2001, Vishay, which by then owned over 80% of Siliconix, sold the the technology to manufacture the SST21x series to a company known as Linear Systems. So really the Siliconix, Vishay, and Linear Systems SST21x parts are the same. CaLogic claims their SST21x chips are a "proprietary design" so their FETs may be different but with nearly identical specs.
With the NXP BSS83 out of the picture, the SST211 or SST215 seems to be the best bet for a replacement. If you can afford to pay about $1/ea in 100+ quantity.
In 2013 I found the SST211, SST213, or SST215 by CaLogic might be a suitable replacement, though the cost was significantly greater. Ouch.
When I search for the SST21x, I see four manufacturers: Siliconix, Vishay, Linear Systems, and CaLogic. Tonight I came across a press release from 2001 that explains some of my confusion over who makes these chips. It looks like the SST21x series was created by Siliconix. In 2001, Vishay, which by then owned over 80% of Siliconix, sold the the technology to manufacture the SST21x series to a company known as Linear Systems. So really the Siliconix, Vishay, and Linear Systems SST21x parts are the same. CaLogic claims their SST21x chips are a "proprietary design" so their FETs may be different but with nearly identical specs.
With the NXP BSS83 out of the picture, the SST211 or SST215 seems to be the best bet for a replacement. If you can afford to pay about $1/ea in 100+ quantity.
Subscribe to:
Posts (Atom)