> 12) Remove the 2.54mm 40 pin connector for the LCD, leave the 40 pin ZIF-FPC
>
> 3) Group the two 40 pin connectors together, to make an 80 pin block: 4x20
> pins = 40 Add-On + 40 Power/GPIO
:*
M
> 4) Instead of having the SDcard on the SPI interface, remove USB 2 from
> the add-on connector and connect it to a card reader IC, MOAI MA8121 to
> obtain a higher data transfer speed, more reliability and a reduction of its
> CPU usage from 55% to 10%.
If we use USB0 for the SD card converter, the SD/MMC card will always
appear as /dev/sda, also when other USB storage media are plugged in.
If we put it on USB2, it will move from sda to sdb (or c etc) at reset
if, say a USB pendrive full of data has been plugged in, which makes
it impossible (or difficult/complex) to have a fixed U-boot parameter
list specifying the root filesystem.
Is it too late to swap these, Giuseppe?
M
On Saturday 11 December 2010 22:45:43 Sergio Sorrenti wrote:
> Hi,
>
> we had an internal discussion in how to improve current Sim.One v1.3
> design, to re-route the board with a better open source tool this time,
> Kicad<http://www.lis.inpg.fr/realise_au_lis/kicad/>
> ,
> that we already used for all the boards of
> Mizar32<http://www.simplemachines.it/index.php?option=com_content&view=arti
> cle&id=13&Itemid=24>project.
>
> Now this list of proposed modifications (15 points) to develop v1.4 of
> Sim.One,
> is adressed to all the Sim.One beta testers.
>
> Please comment, ask for clarifications,
> or suggest your own modification to the current revision that you have in
> hand.
>
> Have a look at the two attached images, they will clearish a lot
> the new design direction.
>
> Here it's the list...
>
> 0) Divide the schematics into two boards: a CPU/SDRAM/Flash module on a
> 240-pin SO-DIMM (like DDR3) and a base board with the rest of the
> components and the connectors. This way, instead of having to produce the
> whole thing with an 6 layer process, only the CPU module needs 6 layer,
> while the base board can be made with just 2 layers, allowing both
> Simplemachines and users to revise and/or adapt the base board more
> easily.
>
> 1) New Mechanichal Dimensions according to Innocenzo's drawing: a base
> board 3.5mm with connectors on the rear, not on the left side.
>
> 2) Made the add-on connector stackable, allowing stackable boards to
> interface with the base board. Add holes on the PCB for stacked PC104
> boards, like the
> ICOP<http://www.icop.com.tw/DB/upload/manual/VDX-6329_UM_v1r1A.pdf>(page
> 8) and
> ADVANTECH<http://downloadt.advantech.com/download/downloadsr.aspx?File_Id=1
> -FQZ2Z4>(page 20) 3.5'' boards.
This is a very good idea.
>
> 3) Group the two 40 pin connectors together, to make an 80 pin block:
> 4x20 pins = 40 Add-On + 40 Power/GPIO
You are going to please robotics people with that change, I am all for it.
>
> 4) Instead of having the SDcard on the SPI interface, remove USB 2 from
> the add-on connector and connect it to a card reader IC, MOAI MA8121 to
> obtain a higher data transfer speed, more reliability and a reduction of
> its CPU usage from 55% to 10%.
>
> 5) Re-Route the SPI and the GPIO freed from the SDCard to the 80 pin
> stacked Add-On block.
>
> 6) Add I2C (that currently ends with clock chip) to the 80 pin Stacked
> Add-On block.
Cool.
>
> 7) See if we can connect a PSP 4.3'' LCD, to be fitted on the back of the
> base board, and include the electronics needed to power the display.
>
> 8) Change the RJ45 serial connector to a standard DB9 RS232 connector for
> the serial port.
Both DB9 and RJ45 are pretty convenient for RS-232 console cables in my
opinion.
>
> 9) Use 5 GPIO for the 5 keys instead of 3 so that Linux can drive these
> keys with its existing drivers.
>
> 10) Earth the USB connector shells
You mean Ground ?
>
> 11) Fix the footprint of the audio jacks and use 3 pins jacks for the
> Microphone and the Headphone.
>
> 12) Remove the 2.54mm 40 pin connector for the LCD, leave the 40 pin
> ZIF-FPC
>
> 13) Change the character LCD connector to a 2x10 2.54mm header to be used
> with a cable.
That would be more standard, thanks!
Yes. It was one of the theories for the USB problems. It wasn't that,
but it does no harm.
M
Some thoughts:
Silkscreen:
Labeling of jumpers could be made easier to understand. E.g. instead
of having a jumper labeled "JP2", the silkscreen could say "BOOT1", or
"JP2/BOOT1", making it very clear what the jumper is used for without
having to look in the manual. Of course it might be difficult to fit
all the labels on the silkscreen. Maybe a lookup table printed on the
backside of the board as I have seen on some other boards.
USB:
What are the possibilities for using MA8121 and a passthrough (maybe
by jumpers) for SPI to the SD card connector?
Will MA8121 fix the USB disconnect problems or cause this type of
problem for SD card too?
Maybe this will not be an option with the MA8121 SD card converter.
What I've been thinking of earlier is:
Normal USB connectors for all three USB ports. And one USB routed to
the ADDON-connector, as in v1.3.
Additional mechanical stability to the USB connectors could be gained
from soldering a piece of metal between the USB shells. However, from
what I have seen on my v1.3 board, this have not been necessary. But
I'm quite careful with it.
About U-boot parameters and the USB0/USB2 connected to the SD card
converter that Martin mention in his post: if rerouting to USB0 is
impossible, maybe the enumeration in the linux kernel could be patched
to always set USB2 as /dev/sda? I have not looked into the details so
I dont know if this is possible or causes problem.
RS232:
A "normal" connector instead of RJ45 is a good idea to me.
Connecting/Disconnecting the serial cable a 100 times make the RJ45
plastic wear out.
I dont know if it would make sense to save base board area, due to the
bulky RS232 connector, and use a stacked serial addon board instead.
CPU:
I understand a lot of good work have been done by Martin and the rest
of the team with software for the EP9307. The MaverickCrunch work was
amazing. But as there are some shortcomings with the CPU hardware bugs
and no new CPU revision will be made AFAIK, maybe it would be an
option to look at other CPUs? It would be ultimately cool if the base
board could be design for use with several different CPU boards.
About new EP9307 revisions, what is the latest you have heard?
Mechanical:
The base board and CPU board should optionally be possible to fasten
very securely to each other to protect the system against mechanical
vibrations, shocks etc. A suggestion would be screws through both the
boards and a distance spacer between where necessary.
Reset button etc:
Make every surface that you can possibly touch during normal handling
of the board safe to touch without causing resets or short-circuits.
Well, as far as possible anyway.
What would the minimal case dimensions be for base board + CPU board?
Rerouting of SPI and I2C to GPIO connector:
Very good idea.
Regards,
Marcus Jansson
A printed table is the better of the two, I think, but it impacts on
the elegance.
Yes, I would have found cleared connector labelling a help. I think I
still don't know what some of the sim1 physical connectors are, just
by looking at the board.
> Normal USB connectors for all three USB ports. And one USB routed to
> the ADDON-connector, as in v1.3.
EP93xx have three USB total, not four.
> the enumeration in the linux kernel could be patched
> to always set USB2 as /dev/sda? I have not looked into the details so
> I dont know if this is possible or causes problem.
We can still reroute the USBs, so can the simpler idea.
One of the constrains on the new design is that the existing software
runs on both.
Having the machine in the mainline linux kernel is a privilege; having
two incompatible hardware versions is embarassing.
> CPU:
> option to look at other CPUs? It would be ultimately cool if the base
> board could be design for use with several different CPU boards.
See above: Sim.One v1.4 is a new layout for the same computer.
The computer you're talking about is the Sim.Two :)
M
Or maybe you mean alternative CPU SO-DIMM boards?
Well, if they can work with the same bus signals that the EP93xx
family needs between the SoC/RAM/Flash units and all other
peripherals, I don't see why not.
M
About USB0, it is not better just wire it out on the board and provide
usb-microsd adapter? MA8121 solution would give better performance?
federico
Before Sim.One I did not had RJ45 serial port. Good that cable was provided,
otherwise I would have to dig for schematics. Using of DB9 allows to grab
serial cable from any devboard kit.
The question is male or female DB9... Male one allows to direct use nullmodem
cables which are most popular. Female allows to directly connect USB-serial
adapters. But gender is a question of target use - for me it can even be
FTDI2332L + miniusb connector ;D
Regards,
--
JID: h...@jabber.org
Website: http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz
Ideally, you could do something ala Sheevaplug where the same FTDI232 chip is
providing both UART and JTAG. That's very handy for a board.
--
Florian
2010/12/13 Marcin Juszkiewicz <mar...@juszkiewicz.com.pl>:
> Dnia niedziela, 12 grudnia 2010 o 14:16:07 Florian Fainelli napisał(a):
>> > 8) Change the RJ45 serial connector to a standard DB9 RS232 connector
>> > for the serial port.
>>
>> Both DB9 and RJ45 are pretty convenient for RS-232 console cables in my
>> opinion.
>
> Before Sim.One I did not had RJ45 serial port. Good that cable was provided,
> otherwise I would have to dig for schematics. Using of DB9 allows to grab
> serial cable from any devboard kit.
>
> The question is male or female DB9... Male one allows to direct use nullmodem
> cables which are most popular. Female allows to directly connect USB-serial
> adapters. But gender is a question of target use - for me it can even be
> FTDI2332L + miniusb connector ;D
+1 for use a DB9 Male.
I like to think that Sim.One can be used as DTE.. :D
federico
That's an idea, though being able to put the root filesystem on an SD
card is physically more compact and seems more elegant.
> MA8121 solution would give better performance?
Much better than SPI, but I think you mean "better than separate
USB/SD adapter" - in that case I would guess that the performance
would be identical.
At the Linux software level, both solutions look the same to Linux
(except, I suppose that a user might put the SD adapter and data USB
pendrive in the wrong sockets and so the roofts would become /dev/sdb
and the board wouldn't boot, while rootfs on SD card USB0 cannot go
wrong)
What's the difference in cost and in technical difficulty for putting
a chip on the board versus putting a 3rd USB socket?
M
+1 for male, to be as similar as possible to the rest of the world's
computer serial ports, so that more standard hardware will be possible
to use in the same way as with other computers.
M