Ever thought about adding more memory to your 328 based design but didn't want to step up to a new board?
I'm thinking of making a 1284 to 28 pin DIP adapter to do that.
I think this use of pins makes sense to do that.
Routing is turning out to be tricky tho.
1284-28pin adapter.sch (98.2 KB)
1284-28pin adapter.brd (43 KB)
why not make a larger board that overhangs one side
and no I just made a new board, its pretty simple on perf and the 40 pin dip ]
Hi Bob
Would that be a simplified path to standalone if you use a Goldilocks for the prototype? If so, any idea what kind of cost you're going to be able to supply these for?
Cheers ! Geoff
Is there space vertically? Seems like such a board may be a bit too tall.
strykeroz:
Hi BobWould that be a simplified path to standalone if you use a Goldilocks for the prototype? If so, any idea what kind of cost you're going to be able to supply these for?
Cheers ! Geoff
Goldilocks is not due out until July. It is $45 and an additional $10 to ship it here to the US.
Crossroads can turn the board around in far less time at a lower price and it will have the advantage of being able to be plugged into a present design.
They are 2 different approaches.
I like the idea.
I already have Goldilocks on order, but that's for another need. I see what you mean though - the additional pins of Goldilocks won't be addressable by this so it's really a RAM/FLASH/EEPROM upgrade for Uno.
Will you be using the on-Uno resonator or adding a crystal?
What I am attempting is a card that can replace the 28 pin DIP for a RAM/FLASH/EEPROM upgrade for Uno.
You unplug your chip, plug this in instead.
It'd also be simple to add a uSD socket to a Bobuino, but I wasn't really after a new development type board.
Using a 4 or 6 layer board would help with the routing, but make it bit more costly.
Good idea though.
Bob, what kind of pins would be connected to the pcb to plug into the 328 socket?
That's always a problem with this sort of mezzanine board. Do the pads on the pcb
need to be so long?
Also, if you try rotating the 1284 chip around different ways, you'll usually find one
orientation that makes the layout a lot simpler - but you've probably tried that
already, I imagine.
Also, you might think up ways to use the extra pins on the 1284 chip. Some people
of course just add pads so you can mount additional header pins, but those would
interfere with mounting shields. You might double check what the vertical clearance
of this board would be when mounted on a 328 socket. It would be close.
FWIW, if I were doing this, I would try adding a layout for a SOIC8 SRAM chip to connect
to the extra 1284 pins. You could use shiftin(), shiftout() to interface to unused pins,
rather than the usual SPI pins. Just a thought. This operates at both 5V and 3.3V,
I want to use SMD pins on the bottom. Not sure what those will be yet, need something that is IC leg size to fit in a regular socket. For now, I just grabbed a part that had pads at the right pitch and placed them DIP leg-width apart. Can make a real symbol up. Could help with routing too.
None of the orientations seem to work out well, even the 45 degree ones.
Maybe more layers will be the solution. Or letting the board width increase - keep it narrow enough to fit the UNO and not interfere with the headers on one side and the ICSP header on the other.
With the small SMD package, vertical height will not be an issue.
DIP package is 4.8mm thick, the 44M1 package is just 1mm, standard PCB up to 1.6mm.
I can consider the extra SOIC chip once I get the rest to work.
Something like this perhaps?
It is wider than a DIP IC, but I checked the size and it appears it would fit in the gap on an arduino. Don't know about height wise, but with the right connectors, perhaps.
I used a TQFP variety as they are nicer to work with.
I've added the footprint of a SOIC8 memory IC. This would work for SPI Flash, SPI SRAM, or SPI EEPROM, whichever is desired. The write protect pin of the flash or eeprom would need to be connected to +5v using the solder jumper.
In order to make it fit, I have however had to use an I/O pin to supply power to the memory. Most flash/sram draw less than 20mA, so it should be safe. Plus if you don't want to use it you can just make the pin an input or a low output to disable the chip.
I've also wired it up so that if you don't want the memory, you could solder on wires, or two right angle connectors to break out the chips JTAG connections.
It may also be a good idea to add some additional decoupling capacitors on the board.
1284-28pin adapter.brd (66.8 KB)
1284-28pin adapter.sch (102 KB)
Slightly more interesting shape - just for fun :).
Perhaps with a 0.8mm pcb and short headers, you should be able to fit it under a standard sheild. I would be a touch worried about pins sticking down from the sheild above if there were THT components on said sheild.
1284-28pin adapter.sch (102 KB)
1284-28pin adapter.brd (67.3 KB)
I've changed it some. I'd rather see actual SPI hardware driving an SPI device, I'm not into bit banging that stuff out. I am here for speed!
Along those same lines, everyone sees discontented with the the onboard resonator - so lets make a crystal an option.
I arranged a thru hole part so the leads would not interfere with any legs underneath, and made connection to the legs optional via solder jumpers underneath.
A 3.5mm high crystal
http://www.digikey.com/product-detail/en/HC49US-16.000MABJ-UB/300-8499-ND/862775
on a 1mm thick board
would be the same thickness as the DIP part it is replacing.
I also measured the UNO area with a caliper, there is room to the right for the board to be 0.3" wider on that side as thought.
1284-28pin adapterR1.brd (73.3 KB)
1284-28pin adapterR1.sch (138 KB)
Ah - just dawned on me, any pins selected would likely be on a right angle plastic holder, taking up some room. So a SMD crystal may be needed after all.
I mentioned using s.w. SPI for an SRAM chip since some of the UNO sized shields,
eg some of the ethernet boards, apparently won't release the SPI port for other
devices to use. I haven't clocked it out but my guess is shiftin/shiftout will probably
run about as fast as the SPI port, which I believe is set to 2 Mhz by default. Pretty
slow. The SRAM chips can handle 20 Mhz.
I still think the first and most serious problem is which pins to use on the mezzanine
board. What will plug into a regular 28 pin DIP socket without destroying the socket
dual-wipe springs in the process? The pins have to be flat and oriented longitudinally
like on an IC.
Machined header pins like these fit.
http://www.dipmicro.com/store/HDR20MR
Have to see if a lower profile version is available somewhere.
How does a slave device not release the SPI port? That's poor design. The part should get off the MISO/MOSI lines when slave select is not enabled.
I do have some of the following in the drawer that I've used on various projects. They
fit tightly into machine pin sockets, but I've always felt they were a little thick for
dual-wipe sockets, and might bend out the springs. Plus they're terribly expensive.
Eg, type A, L, etc.
http://www.mouser.com/MobileCatalog.aspx?page=1754
The business with the ethernet shields hogging the SPI port is something I remember
reading and made a mental note of, but can't attest to any in particular. Maybe only
on some older boards, ??? Maybe something to check out.
You could try something like this:
http://uk.farnell.com/wearnes-cambion/460-3231-01-03-00/pin-0-64mm-1-30mm-l-gold/dp/2066559
They have a 0.64mm diameter, 3.18mm long pin which is perfect for a DIP socket.
Then there is a 0.64mm long collar followed by a 1.3mm top section. That means the circuit board would be approx 2mm above the DIP socket. Add on 0.8mm for the board and you are left with 1.2mm for the components.
Soldering them on would be an interesting challenge, but you could still use the staggered SMD footprint, just solder them on on alternate sides - maybe use a socket to hold the pins while they are soldered to keep them aligned. The only thing you'd need to be careful of is that the vias are either covered by solder mask, or move them so there are none on the centre lines of the two connectors.
