Three questions

Gerald,

I have four questions, if you don't mind.

I'm designing the Bone into a fairly expensive piece of equipment. The
price difference between the two boards isn't that much of an issue. I
would like your thoughts on three things.

1) Is the white going to stay in production?

2) Are there enough leads brought out on the black to do 3 SPI ports?

3) Will code that runs on the white drop into the black? I've had a
custom kernel module written to get the SPI to do what I want. Should
it work on the black without modification? Being the same chip, I would
think so.

3) If you were in my shoes, which board would you use? All I need are
the ethernet port, some GPIO pins and 3 SPI ports.

Thanks,
John

White will stay in production as long as people buy them. We still make the original BeagleBoard you know!

Black is the same as White in respect to the SPI.

You tell me which one you need. Do you need 1GHZ? What features are missing from Black that you need? If there are none, then you can go either way. eMMC may offer some advantages over uSD, but that is up to you if you need that.

Gerald

I'm designing the Bone into a fairly expensive piece of equipment.
The price difference between the two boards isn't that much of an
issue. I would like your thoughts on three things.

1) Is the white going to stay in production?

The full set of design files can be gotten. You can get the boards
manufactured anywhere you like even if CircuitCo stops. I wouldn't
worry about it.

am335x is going to be around for a long time. Don't worry about that,
either.

2) Are there enough leads brought out on the black to do 3 SPI ports?

If you mean 3 chip selects, yes. There's only 2 SPI ports on the
am335x, 3 of the chip selects are in the big headers. If the locations
of the SPI pins conflicts, that's dependent on your application of the
board and peripheral usage.

3) Will code that runs on the white drop into the black? I've had a
custom kernel module written to get the SPI to do what I want. Should
it work on the black without modification? Being the same chip, I
would think so.

If you can port your module to Koen's 3.8 (or newer) kernel, yes, it'll
work fine. Might even work better :slight_smile:

The silicon changes do impact some kernel code. The 3.2 PSP kernel
tree shouldn't be relied upon for the black or any board running rev
2.0 silicon. Use Koen's tree till everything gets mainlined.

3) If you were in my shoes, which board would you use? All I need are
the ethernet port, some GPIO pins and 3 SPI ports.

Do you need to run at any extreme temperatures? If so, avoid XAM
parts, they're preproduction and not temperature rated. Get CircuitCo
or your favorite board house to build blacks with rev 1.0 silicon (it's
possible, just ask and pay) which can be had with temperature ratings.

If you want to run mainline Linux, give the am335x a little more time.
It's going to happen real soon now. AM180x would be a decent chip,
though (and has a lot more SPI) if you don't need Cortex-A8 features /
speed and you want to stick with TI parts. If you don't want to stick
with TI parts, there's a whole lot of choice out there.

If cost isn't an issue, why not use a SoM from a vendor? You can get
good stuff that's supported in mainline Linux / u-boot for $100 per
unit and that will stay in production for a long time. The carrier
board would be slightly more complex than a cape, but not by much.

-Andrew

White will stay in production as long as people buy them. We still make the
original BeagleBoard you know!

Black is the same as White in respect to the SPI.

Good.

You tell me which one you need. Do you need 1GHZ? What features are missing
from Black that you need? If there are none, then you can go either way.
eMMC may offer some advantages over uSD, but that is up to you if you need
that.

I would love to have the 1 GHz but wasn't there a discussion yesterday
that 1 GHz is overclocking the chip? In my application (control of an
induction heater), the application programs won't write any logs or
anything like that until and unless I put it in diagnostic mode. I
think the eMMC might be worth it, even if only because there are no
longer uSD card contacts to corrode.

Sounds like I need to get some Blacks and start experimenting.

Thanks,
John

I agree with Andrew Bradford - use a SoM from any vendor close to your
location and forget about the headache

I agree with Andrew Bradford - use a SoM from any vendor close to your
location and forget about the headache

I'm not sure using a custom vendor board, with yet another deviation of the Kernel is the right answer for everyone. The beaglebone is essentially a SoM, just in a slightly different form factor, but with an active community to help and a push to getting the support into the mainline kernel.

If a vendor gives you 3.2 and refuses to support anything else, you use 3.2 or go it alone, in which case you may as well be on the best documented device.

There's a bunch of vendors out there selling AM335x SoMs. The SoC is
the same, the PMIC is often the same, the RAM is often very similar.
Why wouldn't the beaglebone (and eventually mainline) kernel run on
them?

Yes, you'll need to do a little more work, like device tree, and maybe
some customization of your kernel config, maybe some small
modifications to u-boot or kernel configs to deal with a different
EEPROM data, but that's mostly trivial. Many SoM vendors also already
contribute and can show you how to do this.

I know that Phytec have done quite a lot of upstreaming on barebox and
have some nice docs for their AM335x SoMs.

-Andrew