Panda1 support board

[Wayne C. Gramlich]

All:

My nephew and I have been making slow but steady progress on providing
a support board for the PandaBoard. The project is located here:

http://gramlich.net/projects/robus/panda1/index.html

The boards were fabricated by SeeedStudio in China for about $1/board.

To date, one board has been populated and successfully programmed
with a "Blinky LED" program. (There is no LED on the board, so I
used a scope probe instead.) There is at least one major design
flaw (pins for Pololu voltage regulator are reversed) so there will
be at least a Rev. B.

The next major task will be to figure out how to get the PandaBoard
to talk to the support board. The support board supports, SPI,
I2C and serial. I'm tempted to try SPI first, although I could be
coaxed into another direction.

Cheers,

-Wayne

[Ryan Hickman]

That sounds awesome!

Got a picture of the completed board?

[Ralph Gnauck]

Wayne,

Looks good, is there any chance you could post a PDF of the schematic, I don't have the tools to open your files.

 

Ralph

---

From: Wayne C. Gramlich <wayne.gra...@gmail.com>
To: RobotRefPlatform <robotrefere...@googlegroups.com>
Cc: Wayne C. Gramlich <wayne.gra...@gmail.com>
Sent: Wed, November 2, 2011 9:22:52 PM
Subject: [RobotRefPlatform] Panda1 support board

[Wayne C. Gramlich]

Ralph:

I've been having some problems with .pdf files. I've generated one
for you that should work. Most web browsers now support .svg format,
so you should be able to see that as well.

Ryan:

I've added some pictures.

-Wayne

On 11/03/2011 09:46 AM, Ralph Gnauck wrote:
> Wayne,
> Looks good, is there any chance you could post a PDF of the schematic, I
> don't have the tools to open your files.
> Ralph
>

> ------------------------------------------------------------------------
> *From:* Wayne C. Gramlich <wayne.gra...@gmail.com>
> *To:* RobotRefPlatform <robotrefere...@googlegroups.com>
> *Cc:* Wayne C. Gramlich <wayne.gra...@gmail.com>
> *Sent:* Wed, November 2, 2011 9:22:52 PM
> *Subject:* [RobotRefPlatform] Panda1 support board

[wally_666]

On Wednesday, November 2, 2011 11:22:52 PM UTC-5, wayne gramlich wrote:

> The next major task will be to figure out how to get the PandaBoard
> to talk to the support board. The support board supports, SPI,
> I2C and serial. I'm tempted to try SPI first, although I could be
> coaxed into another direction.

My experience with SPI on the Pandaboard has been less than pleasant. The userspace SPIDEV driver is the easiest to use, but you have to patch and recompile the kernel to get it to work. This should get you started:

http://omapedia.org/wiki/PandaBoard_SPI

Building an out of tree SPI driver module is possible, but if you need anything other than spi CS0 you have to recompile the kernel to remove the pinmux conflicts with UART1, instructions for doing so are less than clear. I've posted what worked for me on the Googlegroups Pandaboard forum.

But I'm here trying to get the real time kernel running, as I couldn't meet my timing constraints for SPI updates.

Although its not for the Pandaboard I found this guide very helpful in my attempt at an interrupt driven spi driver:

http://www.jumpnowtek.com/index.php?option=com_content&view=article&id=57&Itemid=62

I was successful enough to prove it wouldn't solve my latency issues :(

[Bob Smith]

wally_666 wrote:
> But I'm here trying to get the real time kernel running, as I couldn't meet my timing constraints for SPI updates.

Out of curiosity, what SPI device are you reading and at what rate?

thanks
Bob Smith

[wally_666]

We are using an Analog Devices 8-channel D/A converter with 1000Hz sample interval (1mS SPIDEV updates), we are writing not reading, but that should make little difference, although I can see how missed deadlines on reading could mess up the SPI device if it needs some kind of acknowledgment.  Can't really generalize too much about SPI interfaces.

Works well except for the occasional 5+ mS latency :(

Hence I'm here trying to get a real-time system with under 100 uS latency.

Have compiled the same code for the Beaglebone, the long latency are not as common, but they are still there.

Having a GUI on the Pandaboad  was very helpful getting the basic code working, but now we are looking to solve the latency issues and make a stand alone box talking over WiFi.  Either Panda or Bone could work with a real-time enough OS.

At the end of the day, the PandaboardES with 12.04 desktop has been the worst Linux system I've ever actually tried to use going back to RedHat 2.01.  Switching to LxDE and the TI PPA kernel helped and got it to where it'd usually run a few days without crashing.

Pandaboard has got us to the proof of principle stage reasonably quickly, but we're at this point pretty far from being able to validate function because of the latency issues.

[Bob Smith]

wally_666 wrote:
> We are using an Analog Devices 8-channel D/A converter with 1000Hz
> sample interval (1mS SPIDEV updates)

> Works well except for the occasional 5+ mS latency :(
> Hence I'm here trying to get a real-time system with under 100 uS latency.

Is the 1 ms update in a control loop where you do not
know the value until 500 uS before the update? Or, is
it part of a waveform generator where you know the values
to write several milliseconds before the actual output?
If it is a pattern generator, then maybe you could use a
hardware FIFO chip to store the SPI values and clock them
out at the 1 mS rate.

I solved a similar problem on my FPGA peripheral card.
We have tone generator that plays simple melodies. Notes
can be as short as 10 ms which would have made it tricky
to do from the PC. I build a 16 note deep FIFO in the
FPGA and use the FPGA clock to play out the notes at the
right times. This approach worked pretty well.

> At the end of the day, the PandaboardES with 12.04 desktop has been the
> worst Linux system I've ever actually tried to use

Wait, that's not fair. What about 12.04 on a PC? That
is a pretty bad system too, you know.


I tried using a Beaglebone and had a lot of problems with
it and finally just gave it away. The USB subsystem has
some serious problems and TI has specifically said they
will not fix them. In my case an FTDI USB-serial port did
not work.